When parasites persist: tapeworms survive host extinction and reveal waves of dispersal across Beringia

Abstract

Investigations of intercontinental dispersal between Asia and North America reveal complex patterns of geographic expansion, retraction and isolation, yet historical reconstructions are largely limited by the depth of the record that is retained in patterns of extant diversity. Parasites offer a tool for recovering deep historical insights about the biosphere, improving the resolution of past community-level interactions. We explored biogeographic hypotheses regarding the history of dispersal across Beringia, the region intermittently linking Asia and North America, through large-scale multi-locus phylogenetic analyses of the genus Schizorchis, an assemblage of host-specific cestodes in pikas (Lagomorpha: Ochotonidae). Our genetic data support palaeontological evidence for two separate geographic expansions into North America by Ochotona in the late Tertiary, a history that genomic evidence from extant pikas does not record. Pikas descending from the first colonization of Miocene age persisted into the Pliocene, subsequently coming into contact with a second wave of Nearctic colonists from Eurasia before going extinct. Spatial and temporal overlap of historically independent pika populations provided a window for host colonization, allowing persistence of an early parasite lineage in the contemporary fauna following the extinction of its ancestral hosts. Empirical evidence for ancient ‘ghost assemblages' of hosts and parasites demonstrates how complex mosaic faunas are assembled in the biosphere through episodes of faunal mixing encompassing parasite lineages across deep and shallow time.

1. Introduction

Beringia, which spans eastern Siberia and northwestern North America, played a central role in shaping species distributions and patterns of genetic diversity across the Holarctic. In the time since the northern continents first separated during the Neogene, Beringia has intermittently served as either a terrestrial corridor and high-latitude refugium during glacial maxima, or a barrier to gene flow during interglacial periods [1,2]. This complex history of fluctuating interconnectivity, mediated by climatic oscillation, served as a driver for recurrent episodes of geographic range expansion and faunal mixing across Beringia among various northern species [3,4], with far-reaching consequences for the assembly of Holarctic biotas [5].

As animals expanded into and across Beringia during past glacial periods, they were accompanied by a diverse assemblage of parasites. A growing body of work on small mammals and their helminth parasites is revealing a complex history of faunal assembly that includes cyclical waves of transberingian dispersal, regional extirpation of host and parasite lineages, and pervasive host colonization by parasites [3,6,7]. For example, species complexes among cestodes, such as those represented by the genera Arostrilepis [8,9], Paranoplocephala [10] and Anoplocephaloides [11], are diverse and abundant parasites of Holarctic rodents that have undergone numerous events of transberingian dispersal between Eurasia and North America. These events were mediated by episodic interactions among numerous host species representing several major lineages of arvicoline rodents. Climatic oscillation caused cycles of environmental perturbation and stability through the late Pliocene and Quaternary, driving biotic range fluctuations that led to ecological mixing and the formation of new faunal assemblages. Geographic dispersal by parasites created opportunities for novel host–parasite interactions, which facilitated host colonization through ecological fitting [12], and served as a driver for downstream parasite diversification [13].

As our understanding of host and parasite biogeographic histories across the Holarctic improves, we find that parasites enrich our understanding of complex host histories [13–15]. For example, parasites of the two extant species of pikas (Lagomorpha: Ochotonidae) in North America revealed phylogenetic patterns that inverted traditional biogeographic interpretations of host history [16]. North American pikas are sister [17], arising from a single ancestral colonization across Beringia [18]. The sister relationship between the pika species does not discriminate between alternative histories of dispersal that underlie their current geographic ranges. A parsimonious explanation for their history is that the northern Ochotona collaris (Nelson, 1893) (collared pika; distributed in Alaska and northern Canada) represents the ancestral population of pikas that first colonized North America via Beringia. Subsequent expansion southward would have led to the origin of Ochotona princeps (Richardson, 1828) (American pika), which is distributed at lower latitudes across North America's Intermountain West. By contrast, comparative phylogeographic data from a suite of pika-associated parasites showed that the reverse is more likely, with O. collaris descended from O. princeps [16]. This history was revealed by shared patterns evident in multiple independent parasite lineages, in which northern parasite populations in O. collaris rendered southern populations paraphyletic.

Evidence that parasites can reveal unanticipated complexity in historical patterns of temporal and spatial persistence of hosts raises questions regarding the limits of biogeographic resolution that can be achieved by refining our understanding of parasite histories [13,19]. To what extent do parasites record the deeper history of North American colonization by pikas, which has been largely lost from the hosts themselves via population extirpation and lineage sorting? One parasite lineage, the cestode genus Schizorchis Hansen, 1948, records the episode of northward expansion that led to the origin of O. collaris and its endemic parasites, but also retains more deeply divergent clades that each includes North American and Asian species [16]. Increasing resolution of the Schizorchis phylogeny may refine our understanding of the complex dynamics of pika-parasite biogeographic histories across a deeper temporal scale than previously recovered.

Here, we develop a time-calibrated phylogeny for Schizorchis using a DNA sequence dataset of a scale that is rarely applied in studies of cestode phylogenies. These data allow us to date biogeographic events in the history of the parasite with unprecedented resolution. We address the following questions. (i) How many times did Schizorchis lineages traverse Beringia? (ii) Were geographic expansion events restricted to the Pleistocene, or did faunal assembly involve dispersal during the late Tertiary? (iii) Were separate transberingian dispersal events synchronous (occurring during the same glacial period) or asynchronous (occurring during separate glacial periods)? (iv) How does the intersection in timing between parasite dispersal history and the host's fossil record reveal the origins of and interactions between North American pikas?

2. Methods

(a). Study system

Schizorchis is the dominant genus of cestodes associated with pikas globally and the only known tapeworm lineage in North American Ochotona. There are seven nominal species in the genus, with five Palaearctic endemics (Schizorchis altaica Gvozdev, 1951; Schizorchis mongoliensis Tinnin, Gardner, & Ganzorig, 2008; Schizorchis nepalensis Rausch & Smirnova, 1984; Schizorchis ryzhikovi Rausch & Smirnova, 1984; and Schizorchis yamashitai Rausch, 1963) and two Nearctic endemics (Schizorchis caballeroi Rausch, 1960; and Schizorchis ochotonae Hansen, 1948). Schizorchis is solely associated with pikas [20] and free-living oribatid mites, which serve as intermediate hosts [21]. The latter probably do not contribute greatly to long-range dispersal of the parasite [22]. Consequently, the biogeographic history of Schizorchis is expected to reflect that of Ochotona, without other confounding influences. Two deeply divergent Schizorchis lineages occur in North America. One lineage (hereafter the ‘ochotonae-like' lineage) is widespread among populations of O. princeps and O. collaris, and includes all named North American species as well as lineages tentatively identified as Schizorchis species 2 and Schizorchis species 3 [16]. The other (hereafter the ‘yamashitai-like' lineage) is known from just three populations of O. princeps and is represented by a clade identified as Schizorchis species 1, which is sister to S. yamashitai from Eurasia [22].

(b). Data collection

Extracted DNA from a total of 87 individual cestodes used for previous investigations of Schizorchis diversity [16,22] provided a primary resource for DNA sequence data collection during this study. Vouchers for parasite specimens are archived at the United States National Parasite Collection and the University of New Mexico's Museum of Southwestern Biology (electronic supplementary material, table S1).

We accumulated mitochondrial sequence datasets that included an 836 base pair region spanning the 12S and 16S ribosomal genes (rDNA; 6 specimens) and a 546 base pair section of the cytochrome b (CYTB; 82 specimens) gene. PCR details are described in electronic supplementary material, Methods. To these data, we added 81 rDNA sequences from prior studies [16,22]. All sequences are deposited in GenBank (electronic supplementary material, table S1).

We used target gene enrichment to acquire nuclear sequence data from 12 Schizorchis specimens (electronic supplementary material, table S1) that were selected to represent eight major lineages documented previously [16]. This sequencing approach is described in detail elsewhere [23]. Loci were screened to maximize data quality and coverage across species and eliminate putative recombinants and paralogs (see electronic supplementary material, Methods). Ultimately 150 loci were included for analysis, as tests of subsets of the data (50 and 100 loci) confirmed that phylogenetic inferences were stable regardless of the number of loci, and the computational difficulty of the analysis increased considerably as loci were added. Sequences are accessioned in GenBank (numbers MT563532-MT565284).

(c). Phylogenetic analysis

To evaluate relationships among species and lineages of Schizorchis, we generated separate mitochondrial and nuclear phylogenies. For the mitochondrial datasets, we concatenated rDNA and CYTB sequences from the same individuals and established separate data partitions for each CYTB codon position and the rDNA. These data were analysed under maximum-likelihood and Bayesian optimality criteria using IQ-TREE v. 2.0 [24] and MrBayes v. 3.2.7 [25], respectively (see electronic supplementary material, Methods for details). We did not include an outgroup in either analysis, but rooted the phylogenies at the node previously inferred to be basal for Schizorchis [16], which is confirmed by our multi-locus nuclear analysis.

To assess phylogenetic relationships based on the nuclear genomic dataset, we first concatenated the 150 nuclear loci to form a single data matrix 79 561 nucleotides in length. We used IQ-TREE to select an appropriate partition scheme and suite of nucleotide substitution models, and to conduct a maximum-likelihood tree search with 1000 bootstrap replicates [26].

Because concatenation of unlinked loci fails to take into account independent genealogical histories among loci that can yield insight into the true species tree, we applied the time-calibrated multi-species coalescent approach implemented using StarBEAST2 [27] in BEAST v. 2.62 [28,29]. Individuals were assigned to lineages based on mitochondrial identity [16], with the exception that individuals representing two divergent clades of Schizorchis sp. 1 were treated as separate taxa. We applied a strict molecular clock and linked the clock model among loci, but allowed the site models and trees to vary among loci. We applied analytical population size integration and used bModeltest to select appropriate models of nucleotide substitution for each locus [30]. We fixed nucleotide frequencies to empirical values after initial runs revealed erratic and poorly converging estimates for these parameters when allowed to vary freely. The calibrated yule model was selected for the tree prior. To time-calibrate the multi-locus species tree, we established a prior on the age of the common ancestor of S. caballeroi and S. ochotonae based on the inferred timing of the split between O. collaris and O. princeps [31]. This was defined by a normal distribution with mean 2.562 (σ = 0.34) and is discussed further in electronic supplementary material, Methods.

The StarBEAST2 analysis was run twice in separate runs of greater than 9 billion generations each to confirm convergence on similar results. Samples were drawn every 1 million generations and stationarity was monitored using Tracer. Burnin length varied between the runs (1 billion and 3 billion generations, respectively). Tree topology and parameter values were calculated based on a combined sample drawn from 14.5 billion post-burnin generations. A portion of the StarBEAST2 analysis was completed using tools implemented in the CIPRES science gateway [32].

To estimate the mitochondrial nucleotide substitution rate for Schizorchis, we repeated the StarBEAST2 analysis on the mitochondrial dataset. Parameterization of the calibrated yule model followed that of the nuclear analysis, except that appropriate ploidy was assigned for the mitochondrial genome, data were partitioned as described for the traditional phylogenetic analyses of the mtDNA, and trees were linked between partitions. The analysis was run twice for 300 million generations each, with sampling every 250 000 generations. Stationarity and convergence were assessed in Tracer (all parameter ESS values greater than 800). Results of the two runs were combined after discarding 10% of each run to calculate the final tree topology and parameter estimates.

(d). Biogeographic analysis

To estimate historical biogeographic dispersal events, we used BioGeoBEARS v. 1.1.2 [33,34] as implemented in RASP v. 4.2 [35]. We conducted the biogeographic analysis on the nuclear StarBEAST2 consensus tree and assigned each taxon in the tree to one of three geographic regions: Palaearctic, Nearctic North (Alaska and Northwest Territories) and Nearctic South (Intermountain West). We constrained the basal node to the Palaearctic and excluded the possibility of ancestral ranges that spanned both the Palaearctic and Nearctic South, but permitted ranges that spanned adjacent areas. We calculated corrected AIC scores for dispersal extinction cladogenesis (DEC) [36], DIVALIKE (likelihood implementation of dispersal-vicariance) [37], BAYAREALIKE (likelihood implementation of BayArea) [38] and the + j formulation of each of these models, which incorporates jump dispersal into biogeographic estimations. We also used likelihood ratio tests to test the hypothesis that each pair of biogeographic models with and without the j parameter yields statistically equivalent likelihood scores.

3. Results

(a). Phylogenetic analysis

Phylogenetic analyses of the mitochondrial dataset were strongly concordant (electronic supplementary material, figure S1) and corroborated previously published phylogenies based solely on rDNA [16,22]. The ochotonae-like and yamashitai-like lineages of Schizorchis are both evident, each of which include both North American and Asian species. The addition of CYTB improved resolution of key relationships in the mitochondrial tree, reinforcing the presence of five discrete lineages associated with O. princeps in North America's Intermountain West. Three of these form a cluster that is rendered paraphyletic by S. caballeroi, which is associated with O. collaris in Alaska and northern Canada.

Multi-locus nuclear analyses robustly confirmed the interspecific phylogenetic relationships inferred from the mitochondrial dataset. Maximum-likelihood analysis of the concatenated nuclear loci yielded an identical tree topology to that of the multi-species coalescent analysis, and both trees had high nodal support for all internal nodes (figure 1). In the StarBeast2 analysis, all parameters achieved ESS values greater than 1000 except for the calibrated yule speciation rate, which stabilized around the same relatively narrow range of values in both runs (95% highest posterior density 0.04 to 0.1), but oscillated gradually within this range. Evaluation of results isolated from different sections of the MCMC chain representing either high or low speciation rate values, respectively, indicated that tree topology and nodal support were not influenced by this parameter. Node age varied subtly, with the lowest speciation rate values associated with less than 1% older node ages relative to the highest rates. This variation is largely encompassed within the 95% highest posterior density ranges for node ages (figure 1) and does not affect our major conclusions.

Figure 1.

Time-calibrated multi-species coalescent-based phylogeny of Schizorchis based on 150 nuclear loci. Coloured circles denote geographic associations for each tip and node inferred using the BAYAREALIKE + j (main phylogeny) or DEC + j and DIVALIKE + j (inset) biogeographic models. Geographic dispersal events are denoted with arrows that are coloured and oriented to denote directionality of colonization. All nodes are supported by Bayesian posterior probabilities and maximum-likelihood bootstrap values of 1.0 and 100%, respectively. Shaded bars on nodes span 95% highest posterior density intervals for node age. (Online version in colour.)

Estimates for node ages indicate that the origins of Schizorchis date to the late Oligocene or early Miocene (28 to 15 Ma), with most extant diversity arising within the last 5 Ma. Non-overlapping divergence time estimates for the basal nodes of the yamashitai-like and ochotonae-like lineages indicate asynchronous speciation events initiating the origin of these clades. The StarBEAST2 analysis calculated a mean nucleotide substitution rate across the nuclear loci of 1.38 × 10⁻³ substitutions per site per million years (1.0 × 10⁻³ to 1.8 × 10⁻³ 95% HPD), while the rate of nucleotide substitution across the mitochondrial markers was 4.26 × 10⁻³ substitutions per site per million years (2.75 × 10⁻² to 6.34 × 10⁻² 95% HPD).

(b). Biogeographic analysis

Pairwise likelihood ratio tests of biogeographic models favoured inclusion of the j parameter in all cases (p < 0.0001). However, likelihood scores for the + j formulations of all biogeographic models were identical, precluding selection of an optimal model under corrected AIC. We therefore report results for all three models (figure 1) and note that they present nearly identical biogeographic interpretations. All models reveal two independent events of geographic colonization into North America, respectively, near the base of the yamashitai-like and the ochotonae-like clades. All models also show a basal association with the Intermountain West among North American ochotonae-like Schizorchis, with subsequent geographic expansion northward into Alaska. The only distinction evident is that the BAYAREALIKE + j model estimates the common ancestor of S. caballeroi and the clade that includes S. ochotonae and Schizorchis sp. 3 to have been in the Intermountain West. The DEC + j and DIVALIKE + j models estimate a higher probability for the colonization of Alaska to have occurred on the branch leading to this node. We favour the more parsimonious interpretation of the BAYAREALIKE + j inference and note that all estimates indicate historical expansion northward from the Intermountain West.

4. Discussion

Schizorchis has a biogeographic history in North America that is deeper and more complex than that of its extant hosts, evidently arriving in North America via two independent expansions from Asia. The first geographic colonization, probably dating to the late Miocene, led to the establishment of the yamashitai-like lineage across western North America, which subsequently declined, persisting in just a few relictual localities [22]. A later colonization by the ochotonae-like Schizorchis is dated to the middle Pliocene. This second wave expanded widely across the Intermountain West and dominates contemporary diversity. The current parasite diversity therefore is a mosaic reflecting discrete events of expansion during the late Tertiary.

The conclusion that there were two temporally disjunct expansions out of Asia relies on accurate estimates on the timing of colonization. Within the ochotonae-like lineage, the timing of transberingian dispersal is constrained by the origin of Schizorchis sp. 2. However, within the yamashitai-like lineage, our reconstructed phylogeny reveals a long branch subtending Schizorchis sp. 1, leaving open the possibility that colonization occurred more recently than implied by our phylogeny (e.g. coincident with that of Schizorchis sp. 2). With three known Palaearctic species missing from our tree, the timing of this earliest transberingian colonization by Schizorchis could yet be revised and resolved. Surveying poorly sampled regions of Eurasia for all extant mammal and parasite diversity is critical to refine our understanding of the timing of past intercontinental dispersal events [6,7]. However, we deem it unlikely that both the ochotonae-like and the yamashitai-like lineages of Schizorchis entered North America during a single expansion by an ancient species of Ochotona. For this to occur, both cestode lineages would have probably had to coexist within the same population as they expanded across Beringia. Available evidence suggests that Schizorchis species rarely, if ever, co-occur within populations [22]. Indeed, the relictual distribution of the North American yamashitai-like Schizorchis may be a consequence of secondary contact and range retraction in the face of competition from more recently arrived ochotonae-like species. Temporal concordance between the base of Schizorchis sp. 1 and the apparent Nearctic arrival of ochotonae-like Schizorchis is consistent with the hypothesis that population fragmentation by the former coincided with geographic expansion by the latter.

The probability of both Schizorchis lineages jointly arriving in North America within the same host population is further reduced when we consider that long-distance dispersal generally favours the loss of parasite diversity. Similar to founder event bottlenecking of genetic diversity, parasites can ‘miss the boat' when dispersing juveniles fail to transport the full complement of parasites present in the source population. The biogeographic consequences of this are demonstrated in the lack of parasite diversity in O. collaris relative to O. princeps. When pikas apparently dispersed northward from the Intermountain West and established the population that became O. collaris, a history that is corroborated by the results of this study, they failed to bring several parasite lineages with them, including the yamashitai-like Schizorchis [16].

Separate waves of transberingian dispersal by Schizorchis are consistent with the deeper palaeontological record for Ochotona. The genus first appears in the Asian fossil record during the Late Miocene (ca 10 Ma [39,40]), and it arrived in North America by the Hemphillian land mammal age (ca 6 to 7 Ma [41]). That early arrival, represented by Ochotona spanglei, established a range across western North America that extended from Oregon to Nebraska [42]. Our finding that the yamashitai-like Schizorchis lineage may have entered North America as early as the Late Miocene (ca 7 to 13 Ma; figure 1) raises the possibility that it arrived in association with the ancestors of O. spanglei. Though estimates for the timing of North American colonization are not identical between the Ochotona fossil record and the Schizorchis molecular clock, we note that the former represents a minimum age while the latter represents a maximum, so the discovery of older fossils or additions to the Schizorchis phylogeny could narrow the gap between them.

Alternatively, the colonization of North America by the yamashitai-like Schizorchis could have followed another ancient lineage of ochotonid. Living pikas reflect a fraction of past ochotonid diversity, and given that the root of the Schizorchis phylogeny pre-dates the appearance of Ochotona in the fossil record by roughly 10 million years, we assume that archaic ochotonids of other genera were competent hosts for these cestodes. During the Miocene, several genera of ochotonids, now all extinct except for Ochotona, occupied diverse habitats across Eurasia and North America [43]. However, the major transberingian dispersal event that brought archaic ochotonids to North America occurred at the start of the Hemingfordian land mammal age (ca 19 Ma [44]), significantly pre-dating our estimate for the arrival of Schizorchis. It is therefore unlikely that this was the source of the Nearctic yamashitai-like Schizorchis.

If yamashitai-like Schizorchis first crossed Beringia in association with the ancestor of O. spanglei, their existence in modern pika populations has implications for a poorly understood period in the history of North American pikas. A lack of fossil evidence for Ochotona in North America during the Pliocene has been interpreted to indicate that the Late Miocene pika colonization failed [41]. However, the persistence of the yamashitai-like Schizorchis to the present raises the possibility that O. spanglei or its kin survived long enough to encounter and exchange parasites with the new wave of Ochotona from Asia that subsequently invaded North America. This contact could have occurred during the Early Pleistocene, when modern North American pikas first appear in the fossil record and spread widely across the continent. These included small pikas morphologically similar to O. princeps and O. collaris along with a larger-bodied species, Ochotona whartoni [45,46]. Both forms survived through much of the Pleistocene, with broad geographic distributions that extended to eastern North America where pikas no longer occur [47]. Though the fossil record for Ochotona is relatively shallow in North America's Intermountain West, beginning around 35 ka [48], there is genetic evidence for much deeper occupancy of that region spanning much of the Pleistocene [49]. Given the temporal and spatial patchiness of the pika fossil record, details of the timing and pattern of geographic expansion are not well resolved, though diverse Ochotona species were widespread in Asia during the Pliocene and Pleistocene and undoubtedly served as the source for recent North American pikas [45].

The arrival of the ochotonae-like Schizorchis in North America presumably accompanied the second wave of Ochotona to cross Beringia, but there is a discrepancy in the timing of this event based on the fossil record (Early Pleistocene [45,46]) versus our molecular clock-based point estimates (3.5 to 4.4 Ma). Certainly, the North American fossil record may be incomplete and therefore might fail to record the leading edge of this latest transberingian dispersal by pikas. However, we also acknowledge that molecular clocks must be interpreted cautiously under the best of circumstances, and our analysis by necessity has limitations. With no fossil record available for cestodes, it was necessary to calibrate the Schizorchis phylogeny based on an inferred relationship to the host's phylogeny, which in turn was time-calibrated based on a single molecular marker and a probable linkage between population demographics and climatic history [49]. This chain of inference undoubtedly compounded errors, which are to some degree accounted for by the 95% HPD intervals surrounding the key splits at the base of the ochotonae-like Schizorchis clade. These conservatively place the colonization of North America by this group between 2.5 and 5.7 Ma, bolstering an emerging picture of Pliocene-age colonization by the ancestor of O. princeps and O. collaris.

Though not perfectly synchronous with fossil data, conclusions drawn from our molecular clock analysis are largely consistent with palaeontological interpretations of the history of North American pikas. This is not the case for molecular clock-based assessments of pika diversification, which have yielded estimates for the timing of Nearctic colonization that largely pre-date all direct fossil evidence of Ochotona in North America [17,50]. Using external calibration points associated with splits between ochotonids, leporids and rodents, the age of divergence between the North American pikas and their Palaearctic relatives has been estimated to fall between 4.7 and 15.7 Ma, contradicting the fossil evidence for separate early and late Nearctic colonizations by Ochotona [45]. This disagreement between molecular and palaeontological data could in part reflect uncertainty in the calibration points themselves (multiple calibration schemes were considered [17,50]), but probably also time-dependent biases in rate estimation. Evolutionary rates scale negatively with the age of the clock calibration [51]. The use of ancient divergences to calibrate the pika phylogeny probably resulted in relatively low rate estimates, which when applied to shallower nodes would have overestimated divergence times. Calibrations that are closer in time to nodes of interest are more likely to provide appropriate rate estimates [52]. Thus, our use of a mitochondrial rate calibrated against events of the Late Pleistocene [49] may provide better accuracy for relatively shallow events, such as the split between O. collaris and O. princeps.

Advancements in molecular dating of parasite phylogenies will be critical for interpreting the Holarctic histories of parasite lineages. Calibrating molecular clocks for parasites is notoriously difficult given the near-complete lack of a parasite fossil record. Fossils of parasites are rare and restricted to resistant life stages (helminth eggs, protozoan cysts) that are unlikely to be identifiable below the family level (e.g. [53]). Attempts have been made to estimate clocks for helminth phylogenies by extrapolating rates from related lineages that have a more complete fossil record (e.g. insects and nematodes [54]), or by assuming that parasite lineages arose in tandem with host lineages (e.g. elasmobranchs and cestodes [55]), an assumption that has largely been refuted as a general expectation for host–parasite coevolution [13,56]. For shallower time frames, identifying internal calibration points based on biogeographic and host events that fall within the period of interest is likely to be more informative, which is the approach that we employed. Though not without its weaknesses, our analysis is among the first to provide plausible nucleotide substitution rate estimates for cestodes at both the mitochondrial locus and across a large sample of nuclear genes. This foray into genome-scale investigations of cestode phylogeny and biogeographic history underscores the potential for more expansive genomic inquiries into host and parasite demographic history, adaptation, gene flow and coevolution to yield insights into the structure of the biosphere [57].

(a). Ghost assemblages reveal hidden host histories

Our results demonstrate how parasites can refine the understanding of host biogeography and illuminate past ecological interactions that determined current host–parasite community structure [15]. Because Schizorchis lineages apparently survived from the Miocene to the present, even as some of their earliest hosts dwindled to extinction, the record of episodic geographic expansion by pikas from Asia into North America, and subsequent colonization by parasites from one host lineage to another, has survived. In this context, the yamashitai-like Schizorchis lineage represents an ecological relict [58,59] that reveals the former presence and persistence of an ancient assemblage of pikas and parasites that no longer exists. The parasite remnants of this ‘ghost assemblage' contributed to the modern pika-parasite fauna through colonization of secondary hosts, creating a mosaic of temporally deep and shallow parasite lineages occupying the relatively young modern North American pikas. Pikas support a rich diversity of parasites, which provide additional opportunities to test the generality of ghost assemblages in shaping the structure of modern parasite faunas [16].

Ghost assemblages, revealed through lineage persistence linked to colonization and host-group extinction, have been predicted but rarely demonstrated empirically until recently [13]. Patterns of lineage persistence through colonization and subsequent diversification were initially observed via phylogenetic inference over deep time among assemblages of cestodes associated with marine and terrestrial communities that underwent restructuring during global-level extinction events (e.g. [60,61]). The frequency of cycles of episodic ecological disruption, geographic expansion and population isolation extending from evolutionary through ecological time suggests a potentially pervasive role for this mechanism in structuring patterns of faunal diversity [5,13]. Now understood under the Stockholm paradigm [62–64] to be a consequence of episodic ecological disruption, ecological fitting and the variable capacity of parasites to use widespread host-based resources, such events of host colonization are predicted to be common across spatial and temporal scales.

Ghost assemblages are implicated in other host–parasite systems that have undergone episodic biotic expansion, extirpation and host colonization. Across the complex insular landscape of the Alexander Archipelago in southeast Alaska, for example, populations of the nematode Soboliphyme baturini were structured through repeated waves of geographic expansion by their marten hosts (Martes americana and Martes caurina) during the late Pleistocene [65]. Endemic island populations of S. baturini established in association with M. caurina survived the extirpation of this original host by colonizing M. americana that invaded from the mainland. Similarly, repeated waves of geographic expansion and contraction by arvicoline rodents across Beringia have driven instances of host colonization by diverse cestodes [9,11], which have allowed relatively ancient parasite lineages to persist despite local extirpation of ancestral hosts.

5. Conclusion

Sequential dispersals by pikas into North America exemplify episodes of climate-driven geographic expansion that represent taxon pulses [66], which set the stage for new diversity to arise and ecological interactions to be established. These events created host colonization opportunities when the second wave of Ochotona encountered the yamashitai-like Schizorchis, which had the capacity to take advantage of the new host resource due to ecological fitting [12,62]. These phenomena are core elements of the Stockholm paradigm [64], interacting through deep time to produce complex spatial and temporal mosaics of host and parasite diversity across the Holarctic [5,6]. To resolve the biogeographic, ecological and evolutionary histories of parasites and their hosts, it will be necessary to expand the specimen-base using integrated host–parasite sampling protocols (e.g. [67]) to document cestodes from poorly sampled geographic regions and host taxa [7] and to develop a holistic view of the biosphere linked to archives accumulated over time [68].

Ethics

No new fieldwork or specimen collections were done to complete this study. All DNA extracts used in this study were acquired from museum specimens collected under appropriate IACUC oversight prior to 2006. The majority of host specimens were originally collected under IACUC protocol 05–11 (Cornell University). All host and parasite specimens were collected with required permissions from all relevant state, provincial and federal agencies. Specimens collected internationally were exported with permission from Russian, Chinese and Canadian wildlife and customs enforcement agencies as appropriate, and in accordance with all international treaties. Specimens were imported to the United States following US Fish and Wildlife and Department of Agriculture policies.

Data accessibility

All DNA sequence data used in this study are deposited in GenBank and specimens are deposited in accessible museum collections. GenBank accession numbers and museum catalogue numbers are listed in electronic supplementary material, table S1.

Authors' contributions

K.E.G. conceived of the study, conducted extensive data analyses and drafted the manuscript. H.M.T. collected a portion of the genetic data and contributed to the data analysis and drafting of the manuscript. C.L. collected the multi-locus nuclear dataset. E.P.H. contributed to the development of the study and preparation of the manuscript. All authors gave final approval for publication and are accountable for the work described therein.

Competing interests

We declare we have no competing interests.

Funding

Some specimens were collected through the Beringian Coevolution Project led by Joseph Cook and EPH (NSF DEB 0196095, 0415668, 1258010). Aspects of this work were supported by grants from the National Science Foundation (DEB 0506042, 1256943), Northern Michigan University, Sigma Xi and the American Society of Mammalogists.