Discussion
In this study, we show a new pattern of population structure that can arise at the secondary contact zone (SCZ) due to different courses of evolution in hosts and parasites. The main signature of this pattern is a conflicting arrangement of mitochondrial markers in the host and the parasite at the SCZ (Figure 2). The host’s mitochondrial lineages, coming from different refugia, mix across the whole area of the host species’ final distribution and re-establish a panmictic population. In contrast, the parasite’s mitochondrial lineages stop their dispersal at the SCZ. In our model, the re-established panmixia of A. flavicollis is strongly suggested by a previous study on mitochondrial and microsatellite markers (Martinů et al., 2018) and further corroborated by preliminary results of our recent rad-seq analysis (in prep). For the louse P. serrata , we detected a sharp geographic division between the SE and SW lineages using short (379 bp) cytochrome oxidase I (COI) haplotypes sampled across Europe (Martinů et al., 2018). To obtain a more informative comparison of genetic distance within and between the SE/SW clusters, in the present study we demonstrate this split on near-complete mitochondrial genomes from 13 samples collected across the SCZ (Figures. 3, 4). From a strictly theoretical point of view, the pattern produced by the mitochondrial data can be explained by several scenarios. The first explanation is based on the strong presumption that louse population structure will be determined entirely by the hosts’ migrations, given that the lice are highly host-specific and intimate parasites. Consequently, the discrepancy shown in Figure 2 would be a sampling or methodological artifact. However, considering the geographic extent and the number of samples in our previous study (Martinů et al., 2018), we believe that a methodological artifact is a highly implausible explanation. This view is further supported by the present study of the complete mitochondrial sequences and the same genealogical pattern obtained for 12 complete genomes of the maternally inherited symbiont L. polyplacis (Figure 4).
A second theoretical possibility assumes that the lice speciated during their separation in refugia before secondary contact of their hosts, due to their shorter generation time. A similar case was reported by Hafner et al. (2019) for a recent secondary contact of two subspecies of pocket gophers and their lice. While the gophers established a HZ, their lice had already speciated and their contact resulted in “competitive parapatry”, with one louse species replacing the other. The authors also pointed out that the distribution data on the pocket gophers and their chewing lice indicate many instances of range overlap, potentially representing zones of competitive parapatry or species replacements. There are two strong arguments against applying similar scenarios to our system. A theoretical objection is that since the two A. flavicollis mtDNA lineages do not create a SCZ or HZ, but intermix across Europe, it is difficult to envisage a mechanism which would prevent dispersion of the two new louse species across the SCZ. Since both louse mtDNA lineages share the same host species and live in identical ecological environments (as evidenced by sampling both lineages even from the same host individuals), their mutually exclusive distribution is obviously not due to different adaptations (i.e. different host/environment specificities). Also, competitive exclusion is a very unlikely cause as demonstrated by the frequent coexistence of the S-lineage and N-lineage (Martinů et al. 2018). An empirical argument rests on the comparison between the mtDNA and SNP data. If the two mitochondrial lineages were fully isolated non-interbreeding species, we would expect to see the same pattern (i.e. two clearly separated and distant clusters) for both the mtDNA and the SNP sets. However, the comparison in Figure 4 shows that the two sets of data provide very different pictures, where the wide genetic distance between the mitochondrial markers is not matched by similar patterns in the nuclear markers. Although more extensive sampling across the SCZ will be needed to further study the HZ structure in more detail, the visualization based on our 13 SNP sets can be well interpreted. In contrast to the mtDNA, the two SNP sets do not create two distant clusters with the inner genetic diversity negligible when compared to the inter-cluster distance, as would be expected for long-term genetic isolation (Figures. 4, S4, S5). The positions of the individual samples in the PCA plot of the SNPs are determined partly by their geography and partly by their origin within the SW/SE lineage. As a consequence, the SE individuals, sampled at several close localities, form a tight cluster while the SW individuals, sampled from a considerably larger area, are scattered in the PCA plot and their distances are comparable to (or even larger than) the distance between the two clusters. Moreover, the SW sample 98d_Pro_SW collected from the same locality (and same individual mouse) as some of SE samples was genetically closer to the SE population than to any of the SW samples.
The third hypothesis assumes that during their separation, the two parasite lineages reached a high degree of genetic differentiation resulting in a strong but not absolute postzygotic barrier, whilst lacking an efficient prezygotic barrier preventing them from mating. As a consequence, upon encountering each other they formed an extremely narrow HZ in which the majority of the inter-lineage mating’s fail. In this case we would expect a sharp geographic division between the SW and SE population and genetically close or overlapping clusters of the nuclear markers around the SCZ. Based on the data presented in this study and the previous extensive analysis of mtDNA (Martinů et al., 2018), we consider this hypothesis to be the best explanation of the observed patterns. A decoupled genetic structure of a host and its parasite(s) is not exceptional. It has been reported in various host-parasite associations and caused by different biological and/or environmental circumstances (e.g du Toit, van Vuuren, Matthee, & Matthee, 2013; Hafner et al., 2019). However, to our knowledge, theApodemus -Polyplax association presented here is the first known example of genetic structuring caused by a parasite’s HZ created in the absence of the reciprocal host’s HZ. There are several possible factors behind the lack of evidence for similar patterns in nature. Firstly, only a few studies have dealt with HZ in parasites, and they were usually approached in relation to their hosts’ HZ (e.g. Theodosopoulos et al. 2019). This is understandable considering the prevailing view of parasites’s evolution being predominantly determined by their hosts. Secondly, it is likely that emergence of this pattern during secondary contact requires “well-tuned” ratios of genetic diversification between host and parasite populations. If the diversification is too strong, it may either result in speciation of both counterparts (i.e. classical cospeciation Page, 2003), in speciation of the parasite and emergence of HZ in the host (Čížková et al. 2018; Hafner et al. 2019) or in HZ for both counterparts (e.g. de Bellocq et al. 2018). On the contrary, if the diversification is too weak, both counterparts will re-establish panmictic populations. This only leaves a narrow window of time for hosts’ panmixia vs. parasite’s HZ. Yet, such cases do not necessarily have to be rare in nature, they may just be understudied or unnoticed due to the a priori view that evolution in host-specific parasites is linked to their hosts. The case we present here shows that one possible indication of a decoupled pattern is strong mtDNA structure in a panmictic host population.
Genetic incompatibility between two populations at the SCZ can be caused by various mechanisms. Apart from the differences accumulated in the nuclear genetic information, interbreeding can also be prevented by incompatibility of mitochondrial and nuclear genetic information (Wolff, Ladoukakis, Enríquez, & Dowling, 2014). In our system, the lice are known to have their mitochondrial DNA split into several circular minichromosomes (Cameron, Yoshizawa, Mizukoshi, Whiting, & Johnson, 2011; Song et al., 2019). The distribution of mitochondrial genes among the minichromosomes is not entirely conserved - there are several differences in the gene arrangements when comparing the speciesPolyplax spinulosa and P. asiatica (Dong et al., 2014). To address the possibility of mitochondrial gene rearrangement as the barrier between the SW and SE lineages, we reconstructed full minichromosomes (their coding part) from all sequenced samples. In all cases, we found the same gene arrangement, indicating that mitochondrion-nucleus incompatibility is probably not causing the SE-SW isolation (Table S3). Although we cannot exclude existing differences in the non-coding part of the minichromosomes, available information suggests a high level of conservation between minichromosomes within and between related species (Dong et al., 2014). Thus, we assume no mito-nuclear incompatibility to occur in the noncoding part either. In a similar way, we were not able to detect any significant difference between the L. polyplacis genomes from the SE and SW lineages. While, strictly speaking, the high similarity of the SE and SW in respect to the gene content of minichromosomes and symbiotic genomes is not a refutation of the idea that mitochondrial-nuclear and/or symbiont-nucleus incompatibility could cause the barrier, it makes it at least unsubstantiated.
It would be speculative to infer other genetic sources of incompatibility between SE and SW lineages without a detailed study of the louse nuclear genome and more extensive sampling in the SCZ, which is beyond the scope of the current study. Nevertheless, based on evidence collected from three genetic resources, the two maternally inherited markers (Legionella and mtDNA) and nuclear SNP diversity, we were able to unambiguously distinguish between the three possible scenarios of host-parasite incongruence. We propose a new mechanism in host-parasite co-evolution, where a narrow HZ is present in the parasite without a corresponding break in the genetic structure of its host. In this way, the panmictic population of the host is “cleaned” of the parasite lineage present on one side of the parasite’s HZ and replaced by a different parasite lineage on the other side. Given that this evolutionary scenario can easily pass unnoticed (due to the lack of structure in the host) we hypothesize that “host-cleaning filters” may be more common than is currently known, particularly in highly host-specific parasites.