Correspondence Robert Jehle, School of Environment and Life Sciences, University of Salford, Peel Building, Salford Crescent, Salford, Manchester M5 4WT, UK. Email: firstname.lastname@example.org
Encountering cryptogenic populations that are either native or introduced is a common but underreported phenomenon in field biology. Such local species' occurrences of unknown origin hamper our understanding of species' natural distribution ranges, and pose a problem to conservation management decisions. Genetic tools are frequently used to infer the ancestry of natural or invasive populations based on spatial geographical variation. Here we describe the occurrence of cryptogenic crested newts Triturus cristatus c. 25 km away from their main range and situated in an area where an introduction has taken place half a century ago. We first verify the suitability of a coalescent-based analysis that reconstructs the number of founder individuals of a putative propagule from a source pool, based on previously published data documenting a well-known introduction (the Laysan finch on the Hawaiian archipelago). After showing the validity of the approach, we apply this analysis to the case of T. cristatus, revealing that the number of effective founders would have been of the same order as the current effective carrying capacity of the population. We thus argue that the local T. cristatus occurrence is in fact of natural origin, with the documented introduction having little or no impact to the current gene pool. The result has major implications to our understanding of the species' habitat requirements in a zone of parapatry with a closely related species.
Animal and plant species that perpetuate in areas beyond their natural ranges are of concern, because they can cause considerable ecological and economic damage, including competition with native species, species extinctions, food-web disruptions, community alterations, ecosystem conversion, fisheries collapse, agricultural loss and disease epidemics (Davis, 2009; Kraus, 2009). The detection of populations introduced into the wild is either based on reported incidences of introductions, or deduced from the observation that specific populations are unconnected to the remainder of the species' gene pool at a distance that is impossible to be covered by natural dispersal. However, apart from human-facilitated invasions, isolated occurrences could also be explained by, for example, a regressing range border under the influence of environmental change, leaving behind ‘island’ populations that persist in locally favourable conditions. In cases where the study organism is particularly mobile, or where the potentially allochthonous population is, or has been, within reach to the species' natural range, the distinction between native and introduced populations is thus often obscure, a case which can be described as ‘cryptogenic’ (Carlton, 1996). In aquatic ecosystems, cryptogenic species can exceed the number of known introduced species at a given site, and their commonness can lead to an underestimation of the impacts of biological invasion as well as a misleading capability of statistical models to predict the spread of unwanted species.
To better understand the causes and potential consequences of an invasion, it is important to determine the source of the introduced individuals as well as the number of founders that formed the spreading population (Sakai et al., 2001). To reveal these parameters when historical information about the introduction is unavailable, data derived from population genetic markers can be used to model the past demographic history of an invasive species (e.g. Ross & Shoemaker, 2008). An introduced population is also expected to have a genetic makeup that supports its split from the putative deme of origin, and would have undergone a genetic bottleneck that coincides with the timing of the introduction, with the number of introduced individuals determining bottleneck severity. However, caution is required when applying population genetic models to such cases. Following introductions, temporal changes in genetic variation do not always adhere to predictions made by neutral theory, as different selective regimes and strong stochasticity can act on small demes outside their natural range (Noor, Pascula & Smith, 2000; Roderick & Navajas, 2003).
Triturus cristatus and Triturus marmoratus are tailed amphibians (so-called large-bodied newts) with an c. 300 km wide zone of overlap in France (Fig. 1a). Where both species co-occur, a mosaic distribution is found showing, to varying degrees, a mixture of sympatry and parapatry (Arntzen & Wallis, 1991). An example of a local cryptogenic population is the presence of crested newts T. cristatus near the town of Pré-en-Pail (PP) in the north-east of the ‘Département’ Mayenne. In this area, T. cristatus occurs in an enclave 20–30 km away from the main distribution, a distance that exceeds the species' natural maximum dispersal by at least one order of magnitude (Arntzen & Wallis, 1991) (Fig. 1b and c). We are particularly concerned with the origin of this population, because the distinction between natural occurrence and anthropogenic introduction would give rise to different historical reconstructions of the locally parapatric distribution, impacting on our understanding of range border dynamics as a function of the environment (Arntzen & Espregueira Themudo, 2008). Whereas an advance of T. marmoratus over T. cristatus might have led the T. cristatus population in PP to become disconnected from the main stock, this would contrast with our inferences further south in the study area, where T. cristatus has locally superseded T. marmoratus (Arntzen & Wallis, 1991). However, the alternative possibility of an anthropogenic origin of the T. cristatus enclave cannot be excluded. The French researcher Louis Vallée, whose decade-long field work is responsible for the detailed distribution data from the 1940s and 1950s (Fig. 1b), reported on ‘the appearance of T. cristatus×T. marmoratus hybrids after the release of T. cristatus’ near the presently observed PP-enclave (footnote to table 2 in Vallée, 1959, pers. comm.).
In the present study, we perform two analyses. First, we ask whether a relatively newly developed coalescent-based approach that estimates the propagule size for a population that recently formed from a known origin (Anderson & Slatkin, 2007, implemented with the software NfCoNe) is able to accurately predict the number of founder individuals in biological introductions. To achieve this, we use published allelic data from a population where the time, source and number of introduced individuals is known (a study on the Laysan finch introduced to the Hawaiian archipelago, Tarr, Conant & Fleischer, 1998), and compare these figures with the estimates obtained by NfCoNe. Having confirmed the validity of the approach, we then use NfCoNe to investigate whether the cryptogenic T. cristatus population of the PP region is the result of an introduction or whether it is of natural ancestry. Specifically, we test whether the history of the population is consistent with an anthropogenic introduction in recent historical time. As we estimate that the population has probably been of approximately constant size since the time of the putative introduction, we argue that our cryptogenic population is in fact an isolated natural occurrence.
Materials and methods
Anderson & Slatkin (2007) describe a maximum-likelihood method based on the coalescent principle that is implemented in the software NfCoNe (downloadable at http://users.soe.ucsc.edu/~eriq/dokuwiki/doku.php?id=software:softidx). NfCoNe estimates the number of founding chromosomes of a population that is known to have been established at a specific time in the past, and that has received no immigrants afterwards. The colony is established by an unknown number of diploid founders (c/2) derived from a known source population T generations in the past. Because the true demographic history of the colony is not known, it is assumed that the population size follows a logistic rate of increase (r) with a carrying capacity NK (Anderson & Slatkin, 2007). Although this approach has been already used to infer the number of founders in an introduction (Ross & Shoemaker, 2008), we here validate its power before applying it to determining the origin of a cryptogenic population.
The Laysan finch Telespiza cantans is an endangered honeycreeper (Drepanidinae: Fringillidae) from the Hawaiian archipelago. The species underwent a population bottleneck on Laysan Island in the early 1900s, but following the eradication of introduced rabbits responsible for vegetation loss, the finch population recovered and has fluctuated around a mean of about 10 000 since 1968. In 1967, 108 finches were translocated to South-east Island. After an initial decline to 30–50 individuals, finch numbers rapidly increased and have stabilized at about 500. Genetic changes in the finch populations were examined at 29 alleles over nine microsatellite loci (Tarr et al., 1998, and references therein). Population demographic parameters are documented as follows: c=60–100, namely two times the number of surviving adults; three surviving chicks per pair of parents equals a rate of increase 1.5 per generation (Noor et al., 2000) which equals a logistic growth rate of r=0.1; T=5, because the time since introduction is 20 years and generation time is 4 years and NK is 500 (Tarr et al., 1998; Noor et al., 2000). The introduced 30–50 individuals represent an Ne of 21–33, so Ne/n is in the 0.4–1.0 range, and NK is in the range of 200–500. However, the Ne/n ratio is not necessarily linear, with small populations having a larger proportion of breeders (Ardren & Kapuscinski, 2003; Jehle et al., 2005; Ficetola et al., 2009) and, as noted, Ne estimates are accompanied by large confidence intervals. Because of the uncertainty involved, we evaluated the effect of NK on the estimate of c over a large range (20<NK<500).
Triturus cristatus source and colony populations
Tissue samples for microsatellite genotyping were obtained in 1997 from all populations under consideration, by collecting embryos raised in water-filled containers until hatching before sacrifice and storage in 96% ethanol. The source population of L. Vallée's introduction went unrecorded, but is most likely to have been around the city of Laval, where he lived (Vallée, 1959; Fig. 1b). We chose to sample pond 314 close to Laval (n=40 individuals) as a potential source pond. We also also sampled a population near Chateau-Gontier in the south of Mayenne (pond 431, n=45). In the 1940s, only T. marmoratus occupied this area, with the current T. cristatus population of Chateau-Gontier thus representing the general invasion of this species into central Mayenne (Fig. 1b and c). Both populations were pooled for the NfCoNe analysis (n=85). The PP-area where the introduction occurred and where we sampled the putative colony population is c. 65 (pond 314) and 90 km (pond 431) north-east from its source; no further records exist about the T. cristatus populations described in Vallée (1959) until Schoorl & Zuiderwijk (1981). The original introduction site (pond 101, Vallée, 1959) had by 1997 become unavailable for Triturus newts due to desiccation, so we sampled from the nearest extant population at a distance of c. 700 m (pond PP63, sample size n=58).
Microsatellite genotyping and data analysis
We used the eight polymorphic microsatellite loci and associated genotyping procedures outlined in Krupa et al. (2002) and Jehle et al. (2005) to genetically characterize our study populations. Polymerase chain reactions (PCRs) were performed in Hybaid thermal cyclers (Hybaid Ltd, Basingstoke, UK), and PCR products were separated on an Applied Biosystems model 377 semi-automated sequencer (Foster City, CA, USA). Alleles were scored using the software genescan and genotyper (Applied Biosystems). Tests for Hardy–Weinberg and linkage disequilibria were performed using genepop'007 (Rousset, 2008), which was also used to calculate pairwise genetic distances between populations (Fst). Measures of allelic richness were obtained with fstat (Goudet, 1995).
In order to estimate c (the number of chromosomes present among the colony founders) for the PP population in the analysis implemented in NfCoNe, we assume the following values for the independent variables:
r–The intrinsic rate of growth of the colony population. An estimate for an expanding T. cristatus population is available from a newly created pond in Pas-de-Calais, France (Arntzen & Teunis, 1993). Over 3 consecutive years, r was 0.56, 0.48 and 0.72. We apply r=0.56 as the most likely value and evaluate r over the 0.32–1.00 range.
T–The number of generations since the introduction occurred. Vallée's (1959) fieldwork was conducted between 1945 and 1955. Taking the mean (1950) for the observation of hybrids in the colony area, and taking into account that hybrids require 3 years for maturation, 1947 can be taken as the best approximation for the year of introduction. Our sampling was 50 years later (1997). The generation time of T. cristatus is 2 years in males and 3 years in females (Arntzen & Hedlund, 1990). Taking the mean value gives T=20, whereas the maximum is T=26 (1945–1997, i.e. 52 years at a generation time of 2 years), and the minimum is T=14 (1955–1997, i.e. 42 years at a generation time of 3 years).
NK–The effective carrying capacity of the colony population as the number of diploids. The population census size (n) of the colony is unknown, but estimates have been made in two similar field ponds in the same study area over the last three decades (ponds 278 and 2D5; see fig. 2 in Arntzen et al., 2009), where large-bodied newts population size (including T. cristatus and T. marmoratus) has been fairly stable in the range of 200–300 adults (Jehle et al., 2005; J. W. Arntzen, unpubl. data). Effective population sizes (Ne) in total six ponds are more variable, with Ne/n ratios ranging between 0.07 and 0.51 buffered against census sizes (small populations have a higher Ne/n ratio than larger populations, Jehle et al., 2001, 2005). Because of the uncertainty involved (Ne estimates are accompanied by large confidence intervals), we allowed for a large range (50<NK<5000).
In sum, we estimated propagule size for an array of 0.32<r<1.00, with increments of 0.04; 14<T<26, with increments of 2; and NK at the values 50, 70, 100, 140, 200, 300, 500, 700, 1000, 1400, 2000, 3000 and 5000. The number of Monte Carlo replications was R=100 000 in the subroutine ‘coalit’ and m=2000 in ‘nfcone’. The best supported parameter set (r=0.56, T=20 and NK=100) was run in duplicate, with 10 times these numbers of replications. We assumed no genetic drift in the source population since the time of colony founding, that the loci constituting the genetic profiles were unlinked and that the mutation rate was negligible.
Laysan finch benchmarking study
If the carrying capacity is taken as matching the long-term census population size NK=500, the number of founding chromosomes c is estimated at 45 (95% confidence interval 26–81; Fig. 2). At NK=50, c is estimated at 50 (95% confidence interval 28–95). This translates into a propagule size of maximally c/2 (considering that females may have been inseminated), or 24 individuals (95% confidence interval 13–47). Therefore, the genetic estimate from NfCoNe based on nine microsatellite loci largely overlaps with the demographically determined number of 30–50.
The propagule size of the cryptogenic T. cristatus population
We observed 56 alleles among the eight polymorphic microsatellites (four to 11 alleles per locus) in the putative source and colony populations of T. cristatus (Table 1). Contrary to expectations, allelic richness but not observed heterozygosity was higher in the PP population than in the main species' range. We observed significant deviations from Hardy–Weinberg equilibrium at some of the loci, but no significant linkage between loci. The genetic distance (Fst) between the pooled sources and the colony was 0.097, while the respective values for the sources treated separately were 0.089 (pond 314) and 0.132 (pond 431).
Table 1. Population genetic data for two native and a cryptogenic population of crested newts Triturus cristatus based on eight polymorphic microsatellites
Loci out of Hardy–Weinberg equilibrium
Mean allelic richness
Source refers to the pooled populations 314 and 431 and colony refers to the sampled population in the Pré-en-Pail enclave. Allelic richness is calculated against the minimum sample size (n=40).
At r=0.56, NK=100 and T=20, the estimated number of founding chromosomes was 116 (95% lower bound of the confidence interval is 37; note that we are not interested in the upper bound which is more tedious to calculate). Figure 3 shows that the parameter space of the estimated propagule size varied more strongly over the 14<T<26 range than with the observed values of the logistic growth parameter 0.32<r<1.00. At NK=200 and beyond, c increased steeply (and actually became immeasurable, due to extremely long computer runs with increasingly unstable results). The best estimate, reflecting a propagule size of 58 adults (c=116, assuming the absence of inseminated females in the propagule), was situated close to the parameter space where the inferred propagule size exceeded the carrying capacity of the isolated population (c>2NK; shaded areas in Fig. 3a and c). Conversely, parameter values consistent with an introduction propagule size of 10 diploid adults were only found towards the 95% lower bound of the estimate, and when NK>400 (stippled areas in Fig. 3b and d).
In this paper, we show that the coalescent approach developed by Anderson & Slatkin (2007) is well-suited to infer the number of founding individuals in species invasions, which is an important parameter in, for example, predicting the future spread of an invader (Lockwood, Cassey & Blackburn, 2005). We also show that the same approach can be used to determine whether a cryptogenic population is derived from an introduction or whether it is in fact of natural origin. If a natural population is situated at the edge of a species' range, the bottleneck caused by its foundation a relatively long time ago is not traceable through demographic simulations that look back at a timescale only covering the potential recent introduction.
Amphibians are relatively immobile organisms, and understanding the origin of isolated populations is vital for their effective conservation. Among examples of isolated populations, the formation of enclaves in the yellow-bellied toad Bombina variegata in mountain ranges surrounded by the red-bellied toad Bombina bombina (Arntzen, 1978) has been attributed to the regression of the range of the former in parallel with the expansion of the range of the latter. A similar process of enclave formation in parapatric species has also been proposed to explain the fragmented distribution of various Triturus species in northern Serbia (Wallis & Arntzen, 1989) and in western Portugal (Espregueira Themudo & Arntzen, 2007). In amphibian species that have been clearly introduced into non-native areas, genetic markers have been particularly successful in tracing a detailed invasion history and route (Estoup et al., 2001; Zeisset & Beebee, 2003; Ficetola, Bonin & Miaud, 2008). Hypotheses about the origin of populations without prior historical knowledge of their spread can also be tested by investigating patterns of genetic variation (e.g. Vences et al., 2004; Correa et al., 2008; May & Beebee, in press). We here used a temporal line of arguments derived from the coalescent theory to investigate whether an isolated occurrence of the newt T. cristatus can be attributed to a documented local introduction, or whether it was the result of a moving range margin bordering the closely related T. marmoratus. In the latter case, the documented introduction would, following the terminology of Dodd & Seigel (1991), have constituted a ‘repatriation’ (a boost to an already existing, unknown population).
Introduced populations are characterized by a genetic bottleneck at the time of founding followed by a size expansion, as well as a demographic link to their population of origin (Lee, 2002). This appears to be a very suitable scenario in which to estimate the number of potential founders using the coalescent approach developed by Anderson & Slatkin (2007). However, empirical populations do not always conform to theoretical expectations, because simplifying assumptions such as selective neutrality may not be met (e.g. Noor et al., 2000; Kolbe et al., 2004). Thus, it is useful to validate the estimated effective number of founding chromosomes with an actual number of known introduced individuals. In the case of the Laysan finch in the Hawaiian archipelago (data from Tarr et al., 1998), the size of the propagule reconstructed by NfCoNE (Anderson & Slatkin, 2007) was in remarkable agreement with the documented size of the propagule pool, demonstrating the method's precision and robustness against varying parameters.
In the case of T. cristatus in the PP-area of Mayenne in France, we investigated a scenario with very similar statistical properties to the benchmark study of the Laysan finch (comparable current population size, numbers of generations since the introduction took place and genetic marker resolution), but with the origin and history of the population being under question. Our temporal genetic analysis provides strong evidence that the propagule size of the investigated occurrence of T. cristatus is about equal to the probable effective carrying capacity in the current population. The implications of this finding for two possible scenarios on the origin of the population are discussed below.
In the first scenario, we consider if the T. cristatus introduced in the area by L. Vallée were the ancestors of the current population, involving an estimated effective propagule size of 116 chromosomes. A single inseminated female (newts are characterized by internal fertilization) would contribute four chromosomes into the newly founded population, and multiple matings (a regular occurrence in newts, Halliday, 1998) could increase this number. On the other hand, it is likely that not all introduced individuals would have reproduced (evidenced through Ne/n ratios for large-bodied newts being well below unity, Jehle et al., 2005), and the census propagule size would probably have been larger that the estimated number of 58 (116/2) diploid individuals. Louis Vallée's introduction was presumably not a deliberate attempt to establish a new T. cristatus colony, but rather the semi-accidental release of a few animals redundant from morphological studies. On balance, we therefore consider it unlikely that a sufficiently large number of adult individuals was introduced to explain the genetic variation that we observe in the PP-area today.
In the second scenario, the observed T. cristatus occurrence is of natural origin, with the possibility of supplementation by an additional introduction of a small number of individuals. Clearly, a propagule pool of a size expected to match current Ne indicates that this is the most likely case. Moreover, it indicates relative demographic stability over 20 generations, with comparable population sizes in the 1950s and 1990s. With regard to the documented introduction, this would imply either that the introduced individuals were unsuccessful in reproduction, or that the gene pool of a small number of introduced individuals was absorbed into the larger gene pool of an existing population. Interestingly, admixture with a small number of introduced individuals might explain the unexpectedly high allelic richness, at still lower heterozygosity than in the main range (see also Gillis et al., 2009, who found higher nucleotide diversity but lower numbers of haplotypes due to an invasion resulting in admixture). In the case example provided by Anderson & Slatkin (2007), it was also unknown whether the propagule pool was of natural origin or derived from an introduction, although in contrast to our case the latter was assumed throughout the analysis. However, no data are provided on allele frequencies of the source and the propagule, leaving it open whether biases could arise through alleles present in the propagule but absent in the source, as is to a low degree the case in our study.
In this paper, we focus on one population in the PP-area only, although we have documented 17 T. cristatus ponds in the enclave (Arntzen et al., unpubl. data.). However, the investigated PP63 is one of the two largest demes (in most other cases only a small number of newts were found), and the nearest to the location of the original introduction in a now-disappeared pond. An increased effective carrying capacity (NK) would lower the estimated number of founder chromosomes but not to an extent that would change our main conclusion (Fig. 3). Genetic drift in the source population would overestimate the precision of NfCoNe, but not alter its main estimate of c (Anderson & Slatkin, 2007); by pooling two ponds from the most likely founder area we also ensure that as many representative chromosomes as possible were sampled.
In summary, we have shown that an enclave occurrence of T. cristatus is likely to be of natural origin, in spite of a documented introduction of an unknown number of individuals. This is of vital importance to improving our general understanding of ecogeographical dynamics of species range edges (e.g. Arntzen & Espregueira Themudo, 2008). The current interpretation of a natural basis of the PP-enclave is further supported by the disappearance of T. cristatus from the central-north of Mayenne, contrasting with the more general trend in the south, where T. cristatus has taken over from T. marmoratus (Fig. 1). This suggests a more complicated biogeographical history of the two species than a simple range expansion of T. cristatus to the cost of T. marmoratus (Arntzen & Wallis, 1991), with perhaps recurrent episodes of range expansion and contraction.
We thank Andy Krupa for help in the laboratory and Eric Anderson, Marieke van Erp and Alain Frantz for help with running the NfCoNe software. Tissue samples were collected using the permit 97/204 of the of the Ministére de l'Environment, Paris.