Stabilization of a salamander moving hybrid zone

Abstract When related species meet upon postglacial range expansion, hybrid zones are frequently formed. Theory predicts that such zones may move over the landscape until equilibrium conditions are reached. One hybrid zone observed to be moving in historical times (1950–1979) is that of the pond‐breeding salamanders Triturus cristatus and Triturus marmoratus in western France. We identified the ecological correlates of the species hybrid zone as elevation, forestation, and hedgerows favoring the more terrestrial T. marmoratus and pond density favoring the more aquatic T. cristatus. The past movement of the zone of ca. 30 km over three decades has probably been driven by the drastic postwar reduction of the “bocage” hedgerow landscape, favoring T. cristatus over T. marmoratus. No further hybrid zone movement was observed from 1979 to the present. To explain the changing dynamics of the hybrid zone, we propose that it stalled, either because an equilibrium was found at an altitude of ca. 140 m a.s.l. or due to pond loss and decreased population densities. While we cannot rule out the former explanation, we found support for the latter. Under agricultural intensification, ponds in the study area are lost at an unprecedented rate of 5.5% per year, so that remaining Triturus populations are increasingly isolated, hampering dispersal and further hybrid zone movement.


| INTRODUCTION
The interactions that determine a species' position in the ecosystem are many and include predation, competition, disease vectors, and many others. When species expand their geographical range, such as following a glacial period, they encounter new habitats and run into species previously unknown to them. If a species encounters another closely related species, the two may interbreed, possibly leading to offspring in a more or less narrow hybrid zone. At least 10% of animal species and 25% of plant species, mostly the youngest ones, are involved in hybridization and potential introgression with other species (Mallet, 2005). Zones where related species meet, mate, and hybridize are particularly frequent in areas of postglacial colonization (Hewitt, 1999;Taberlet, Fumagalli, Wust-Saucy, & Cosson, 1998).
Hybrid zones are important as a "natural laboratory" for speciation research and serve as windows on evolutionary process (Abbott et al., 2013;Harrison, 1990;Hewitt, 1988). Moving hybrid zones have an additional edge because, by moving, introgressed genes are continuously being tested in new environmental and genetic backgrounds.
To exploit this asset, it is important to improve our understanding of the spatiotemporal dynamics of hybrid zones. We here investigate a unique case of a hybrid zone that moved and then stabilized in historical times. Distinguishing between stable (tension) and dynamic (moving) hybrid zones has important implications for our understanding of the role of differential introgression and selection in shaping species boundaries.
Species hybrid zones frequently show a strong ecological component. One example is the genus Bombina in which the lowland red-bellied toad (Bombina bombina, with an Ukrainian glacial refugium) encounters the mountain dwelling yellow-bellied toad (Bombina variegata, with a Balkan glacial refugium) all along the lowland-mountain transition of Central Europe (Szymura, 1993;Vörös, Mikulíček, Major, Recuero, & Arntzen, 2016). This hybrid zone could have formed along the species ecotone where it remained in stable position, but another explanation is that it formed elsewhere and then moved till the lowland-mountain transition was reached. Biogeographical evidence supports the latter scenario (Arntzen, 1978). This author documented the presence of "enclaves" of B. variegata surrounded by B. bombina and argued that in low dispersal organisms, such as toads, the only reasonable explanation for enclaves is species displacement. The same line of reasoning has been applied to myobatrachid frogs in western Australia (Littlejohn & Roberts, 1975) and to Triturus newts in the Iberian (Arntzen & Espregueira Themudo, 2008) and Balkan peninsulae (Wielstra & Arntzen, 2012).
Another case of a moving hybrid zone is that of the newts Triturus cristatus (the northern crested newt) and Triturus marmoratus (the marbled newt). These species engage in a habitat patchwork in central France, with adult F 1 hybrids making up 4% of the total adult population (Arntzen, Jehle, Bardakci, Burke, & Wallis, 2009). Evidence for movement of the hybrid zone is threefold: first, through direct observation, involving the surveying of species and hybrids over a large area in ca. 1950 and in 1979. This showed the (northward) advance of T. cristatus, the regression of T. marmoratus, and the continued presence of hybrids (Schoorl & Zuiderwijk, 1981;Vallée, 1959) (Figure 1). Second, T. marmoratus is surrounded by T. cristatus in enclaves and other persisting occurrences in areas of species replacement (Arntzen & Wallis, 1991;Arntzen, 1996; see also Arntzen, Burke, & Jehle, 2010). Third, genetic variation thought to result from hybridization is significantly higher in T. cristatus than in T. marmoratus (Arntzen & Wallis, 1991). This observation is in line with the direction of hybrid zone movement.
We set out to survey the distributions of T. cristatus, T. marmoratus, and their hybrids over the French "département" Mayenne in 2014 and 2015. That is ca. 65 years after the first survey (Vallée, 1959) and 35 years after the second survey (Schoorl & Zuiderwijk, 1981) in the same area. The objective of this third survey is to document the present-day distribution of the two newt species and make comparisons to the two previous analyses of the hybrid zone, in particular to see whether (1) T. cristatus continued its advance over T. marmoratus, (2) enclaves persisted or dissolved over time, or (3) equilibrium conditions between T. cristatus and T. marmoratus have been reached. A further aim is to identify ecological correlates of the position of the hybrid zone in situations of stasis and flux.

| Fieldwork
Fieldwork was carried out over the département Mayenne in western  (Dejean, Miaud, & Schmeller, 2010;Johnson, Berger, Philips, & Speare, 2003;Schmeller, Loyau, Dejean, & Miaud, 2011). To obtain an impression of the possible decline of amphibian breeding localities over time that might be interfering with the monitoring of the position of hybrid zone, we went back to pond locations documented ca. 35 and 18 years ago ( Figure 1). We revisited the following: (1) the Triturus localities from the second survey, (2) the amphibian ponds from the second survey north of the N12 and D35 roads (AZ, unpublished data), and (3) the amphibian ponds from a survey in the Pré-en-Pail area in 1997 (JWA, unpublished data). We do not report on a few F I G U R E 1 The historical distribution of the northern crested newt, Triturus cristatus (hatched), and the marbled newt, Triturus marmoratus (shaded), over the département Mayenne (a) at the first survey at ca. 1950 (data from Vallée, 1959) and (b) the second survey in 1979 (data from Schoorl & Zuiderwijk, 1981). Pond localities investigated are shown by black dots. The asterisk in the left panel has initially been attributed to a T. cristatus introduction, but is likely to represent a natural occurrence (see text for details). The figure is reproduced from Arntzen and Wallis (1991), with permission ponds for which we were uncertain about the exact location or where access was denied.

| Molecular identification
DNA extraction followed the chelating resin-based procedure of Walsh, Metzger, and Higuchi (1991). Individual eggs were placed in 1.5-ml tubes with 0.4 ml of a 5% chelex resin solution (Chelex ® 100 sodium form, 50-100 mesh) and 5 μl proteinase K (ProtK, 20 mg/ml) and left overnight to lyse at 65°C. The dissolved tissue was heated at 95°C Because a complete match was observed between the two mtDNA markers, it was deemed unnecessary to study more than a single marker in the other material. Genotypes were called automatically by the module Kraken™ of LIMS controlling the LGC genomics SNP genotyping line, visually inspected and if necessary, manually corrected. An ambiguous signal with no call made precluded the identification of 41 eggs (3.5%).

| No numerical correction on species counts
Triturus cristatus and T. marmoratus are dissimilar species in many respects. Relevant for data interpretation are the parameters that could affect species counts, namely (1) the length of time the adults stay in the water when reproduction is over (long in T. cristatus and short in T. marmoratus, Bouton, 1986;JWA, unpublished (3) and (4) are in equilibrium, we determined the dry weight of 50 T. cristatus and 39 T. marmoratus eggs, collected from ponds where hybrids or the counterpart species were known to be absent. The mean dry weight of T. cristatus eggs was 1.81 ± 0.34 mg (standard deviation), and for T. marmoratus eggs, it was 1.75 ± 0.33 mg, with no significant difference between the species (t-test, p > .10). The energetic investments per egg appear similar for both species, and we assume that the high annual fecundity of T. marmoratus females (3) is offset by the skipping of annual breeding opportunities (4). While a significant reproductive asymmetry has been reported with hybrids predominantly transmitting the cristatus mtDNA haplotype (Arntzen et al., 2009), hybrids are relatively rare and the bias is minor. In all, numerical adjustments to the species counts were deemed unnecessary.

| Species distribution models
The dependent variable in the study is species composition, in which T. cristatus and mixed populations with >50% T. cristatus are contrasted with pure and otherwise mixed T. marmoratus populations.
The morphologically intermediate and easily recognizable T. cristatus × T. marmoratus F 1 hybrids were taken to represent both species.
Statistical evaluation was carried out with a logistic regression procedure in SPSS 20.0 (IBM Corp., 2011) in which (1) Table 1. The conditions they describe were extracted over a circular area with a radius of 250 m around the pond, for all investigated Triturus populations. This scale is a compromise between mapping accuracy, average interpond distance in the study area, and the distance that adult amphibians reside, migrate, or disperse from ponds (Semlitsch & Bodie, 2003;Smith & Green, 2005). GIS analyses were carried out with ILWIS 3.3 (ILWIS, 2005).

| RESULTS
A total of N = 183 adult Triturus newts were observed in 25 ponds.
One hundred thirty-five (74%) were T. cristatus in 20 ponds, 33 (18%) were T. marmoratus in nine ponds, and 15 (8%) were T. cristatus × T. marmoratus F 1 hybrids in three ponds. A total of 1,155 eggs from 97 ponds were identified to the species. Approximately two-third of the investigated eggs had the cristatus mtDNA haplotype, and one-

| DISCUSSION
The notion that hybrid zones may move over the landscape is relatively new to science (for a review see Buggs, 2007). The argument for movement is mostly inferential (Buggs, 2007; see also Carling & Zuckerberg, 2011;Charpentier et al., 2012;Gay, Crochet, Bell, & Lenormand, 2008;Krosby & Rohwer, 2009;Leafloor, Moore, & Scribner, 2013). Examples of direct observation on hybrid zone dynamics are rare, because of the time frame involved (Buggs, 2007; see also Engebretsen Triturus cristatus and T. marmoratus are hybridizing species in which the frequency of hybrids in the adult F 1 class averages at 4% (Arntzen et al., 2009;Vallée, 1959). The fertility of the F 1 hybrids is low, and interspecific gene flow is limited. The T. cristatus and T. marmoratus contact classifies as a traditional "mosaic hybrid zone" (sensu Harrison & Rand, 1989) on account of the following: (1) the bimodal (or "trimodal," Gay et al., 2008) genetic profile displayed, (2) the pronounced ecological differentiation expressed by the species, and (3) the patchy distribution pattern at which they engage. Following the initial survey on Triturus newts in Mayenne shortly after the Second World War (Vallée, 1959), T. cristatus has superseded T. marmoratus over a large portion of the département (Arntzen & Wallis, 1991;Schoorl & Zuiderwijk, 1981) (Figures 1 and 2). Hybrid zone movement and species displacement are thought to be triggered or accelerated by anthropogenic change, namely the removal of hedgerows in the postwar period that allowed the fairly aquatic T. cristatus to supersede T. marmoratus, which is a more terrestrial species with preferences for hilly and forested terrain (Schoorl & Zuiderwijk, 1981).
Triturus cristatus has a northerly European distribution, and T. marmoratus has a distribution in southern France and Iberia. In spite of this general pattern, the area T. cristatus took over is in the south of Mayenne, whereas the distribution of T. marmoratus over the northern part of Mayenne remained largely unchanged (Figures 1 and 2). The effective dispersal of T. cristatus was ca. 1 km per year (Arntzen & Wallis, 1991).
Triturus marmoratus was not wiped out completely from the south of the département but persisted in local populations at low frequency, often in syntopy with T. cristatus and with ongoing hybridization and introgression. The continued syntopy of the species is not surprising given the newts' longevity, with an observed maximum of 14 years in T. cristatus and T. marmoratus and 17+ years in F 1 hybrids (Francillon-Vieillot, Arntzen, & Géraudie, 1990). A wider analysis of the distribution of Triturus species and the landscape suggests that the dispersal route taken by T. cristatus is along the Loire river (JWA, unpublished) and that T. cristatus entered Mayenne from the east. An eastern point of entrance is supported by Vallée's (1959) distribution data (Figure 1a). From here, the species dispersed westwards and northwards, both prior to 1950. The northernmost T. cristatus locality near Pré-en-Pail was initially interpreted as an introduction (cf. footnote to Table 2 in Vallée, 1959), but molecular genetic data indicate that the local presence is natural. The arguments underlying this claim are as follows: (1) assignment tests indicate a closer genetic relationship with other northern T. cristatus populations than with the presumed source of the introduction near the city of Laval in the center of the département, and (2) the genetic variability of the Préen-Pail T. cristatus population is such that the propagule size of an introduction must have been large, which is unlikely (Arntzen et al., 2010).
The alternative scenario is the northward dispersal of T. cristatus to have reached the Pré-en-Pail area along the Mayenne river. The existence (or past existence) of this dispersal corridor is supported by observations to the northwest (first survey, Figure 1a) and northeast of Mayenne city (third survey, Figure 2). We assume that at the second survey, this connection remained unnoticed due to sparse sampling (Figure 1b).
The present-day distribution of species and hybrids is comparable to that of the second survey ( Figure 2). Southern localities with just T. marmoratus ("marmoratus enclaves") are rare. The local absence to interpret the current pattern (with more mixed populations in the southwest than in the southeast of the département) as reflecting the colonization and superseding process (Figure 2). It is puzzling though why the same pattern was not found already at the second survey ( Figure 1b). It is equally tempting to attribute the northward shift of the species border to climate warming, but this counteracts the fact that it is T. marmoratus from southern France and Iberia, presumably well adapted to high temperatures, that is the receding party (cf. Taylor, Larson, & Harrison, 2015; and other parameters appear to be involved. A remarkable feature is the natural occurrence of a syntopic/allotopic T. cristatus population in the northeast of the département. This occurrence was observed at all three surveys and goes to show that an isolated occurrence or enclave may actively be formed as an expansion of the range ahead of the main distribution. The study of a wide variety of environmental variables yielded no new insights into the ecological parameters that determine the mutual T. cristatus/T. marmoratus distribution in Mayenne, except for the density of water bodies that was identified as a significant contributor to the best distribution model, in addition to altitude and the density of hedgerows and forestation (cf. Arntzen & Wallis, 1991). Of the parameters that contributed to the species model, the density of hedgerows is most likely to change over time. Indeed, the typical "bocage" (dense network of mature hedgerows) landscape has largely been eradicated to allow for larger field sizes, in particular in the flat, southern part of Mayenne. For a dramatic illustration of the magnitude of the change, see Figure 4. The flat southern part of Mayenne with a high density of water bodies appears particularly suited for (the advance of) T. cristatus, the more aquatic of the two species. Its relative advantage over T. marmoratus, the more terrestrial species, was boosted by the removal  (Schoorl & Zuiderwijk, 1981), but parameters like these are difficult to model in a spatial context. In mixed ponds, the contribution of T. cristatus to the total population was negatively correlated with the density of hedgerows around the pond. Accordingly, the cutting of hedgerows may well have been the prime factor that caused the hybrid zone to move.
As noted, the changes we documented in the distribution of T. cristatus vs. T. marmoratus over Mayenne from the second to the third survey are relatively small. Triturus cristatus has not markedly continued its advance over T. marmoratus, and mixed populations are equally frequent now as they were 35 and 65 years ago. It is therefore fair to judge that the T. cristatus vs. T. marmoratus moving hybrid zone has come to a standstill.
Prime candidates for forming a barrier to the further advance of T. cristatus are the ecological variables that help to explain the current species distribution, namely altitude (p < .001), pond density (p < .01), and hedgerow and forestation (p < .05) (Figure 3). The transition from flat to hilly terrain at 113-161 m a.s.l. that sets the species apart also separates the south of the département from the north (Figure 3a). This parameter to some extent represents the other selected parameters because the amount of shelter provided by hedgerows and other small landscape elements (favoring T. marmoratus) is higher in hilly terrain than in flat areas and more ponds (favoring T. cristatus) are found in flat than in hilly areas.
To disentangle the relative contribution of altitude, forestation and pond density will require data at either a larger scale (e.g., countrywide), or at a smaller scale where species interactions are most intense. An alternative explanation for hybrid zone fixation is the pond loss that followed agricultural intensification and the change from pasture to arable land use. In Mayenne, the loss of Triturus ponds was 82% in the 1950-1979 interval (Schoorl & Zuiderwijk, 1981) and another 86% in the 1979-2015 interval. This amounts to a steady loss of 5.5% of ponds per year. This rate is substantially higher than the annual loss averaging at 0.8% in United Kingdom (Wood, Greenwood, & Agnew, 2003) and 3.5% in northwestern France (Curado, Hartel, & Arntzen, 2011). The decline will have compromised the dense network of ponds that is required for a healthy Triturus population network (Halley, Oldham, & Arntzen, 1996). Without ponds as "stepping stones," new areas cannot be colonized.