A hybrid zone of two toad sister species , Rhinella atacamensis and R . arunco ( Anura : Bufonidae ) , defi ned by a consistent altitudinal segregation in watersheds

Delimiting the spatial extension of a hybrid zone is essential to understand its historical origin and to identify the geographical and/or environmental factors which delimit it. Rhinella atacamensis and R. arunco are two sister species which together inhabit Chile between 25° and 38° S. Their distribution limits coincide at about 32° S, where recently it was reported that they hybridize in a small watershed (Pupío creek). Although the genetic evidence suggests that these two species form a hybrid swarm, they are not mixed homogeneously in the entire watershed, but rather are spatially segregated: R. arunco is found in the lower part of the creek and R. atacamensis in the higher part. An extensive exploration north and south of 32° S revealed other instances of hybridization, with the same pattern of spatial segregation within other watersheds. This study describes the hybrid zone combining mitochondrial sequences and nuclear AFLP markers. In the northern part, the hybrid zone is a narrow strip which crosses several watersheds and extends more than 130 km from NW-SE, so that R. atacamensis is found at higher altitudes towards the south. However, two points south of this strip show that the hybrid zone is more extensive and complex, and probably extends along the entire border of the mountain chains which form the watershed of the Aconcagua River (32°30’ 33° S). We propose an explanation for the origin of this hybrid zone considering paleoclimatic and orographic information, and briefl y discuss the taxonomic implications of these results.


INTRODUCTION
Hybrid zones (relatively narrow areas in which two genetically different populations, whether of the same or dif ferent species, meet and produce hybrids) are highly idiosyncratic phenomena, since they involve diverse evolutionar y and ecological processes which interact at different spatial and temporal scales (Barton & Hewitt 1985, Arnold 1997, Howard et al. 2004).Examination of historical and current distributions of various taxa which co-exist over wide geographic areas has led to suggest that hybridization has been important in the diversifi cation and current composition of those biotas (Swenson 2010, Nieto Feliner 2011).More restricted studies have demonstrated that this phenomenon is frequent in a great variety of taxonomic groups (e.g., Willis et al. 2006, Grant & Grant 2008, Lepais et al. 2009, Fontenot et al. 2011).Although these examples can be considered as exceptional since they are from well-known taxonomic groups or regional biota, they show that natural hybridization is a common and widely distributed phenomenon geographically and taxonomically.
In order to identify the processes involved in the origin and persistence of hybrid zones it is essential to defi ne their spatial extension, which can now be done with high precision using molecular markers.A detailed study of the geography of these zones may contribute to elucidate the historical factors or population events which originated them (e.g., Hofman et al. 2007, Hird et al. 2010, Edwards et al. 2011), and to identify the current physical and environmental conditions which determine their location and extension (e.g., Buckley et al. 2003, Yanchukov et al. 2006, Shields et al. 2010, Hapke et al. 2011).
This study illustrates how an extensive e x p l o r a t i o n a n d t h e c o m b i n e d u s e o f mitochondrial and nuclear markers may reveal the complex geography of a hybrid zone, and how its location and confi guration allow the generation of an explanation for its origin.The taxa involved are two sister species of toads, Rhinella atacamensis (Cei, 1962) and R. arunco (Molina, 1782) (Méndez 2000, Pramuk 2006), endemic to nor th-central Chile, whose joint distribution ranges extend from approximately 25º S to 38º S (Cei 1962, Veloso 2006, Correa et al. 2008).Both species are nocturnal and can be found in borders of rivers, creeks, lagoons and other water bodies, including canals, dams and other man made environments (Cei 1962).Until recently, it had been assumed that these two species had allopatric distributions, with their limits located around of 32° S.However, populations of the Pupío watershed (a small creek located at 32° S) were recently described as a possible hybrid swarm of R. atacamensis x R. arunco (Correa et al. 2012).One of the main fi ndings of that study was that a spatial segregation exists within the Pupío watershed: R. atacamensis occupies the higher part of the watershed, whereas R. arunco inhabits the lower part.At mid elevations, they hybridize.
The objective of this study was to defi ne the spatial extension of the contact zone of R. atacamensis and R. arunco, for which we per formed an intensive fieldwork nor th and south of the Pupío watershed.Using a combination of mitochondrial (sequences of the control region) and nuclear (AFLP) markers we mapped the hybrid zone, which allowed us to evaluate whether the two species maintain the same spatial segregation in other watersheds and to identify the areas which need to be sampled to defi ne the zone with greater precision.We also suggest a biogeographical explanation for the origin and present conformation of the hybrid zone, considering the paleoclimatic history and the geographical relief of the study area.

Material and study sites
Between years 2007 and 2011 we collected a variable number of samples per locality from 43 localities which represent almost the entire known distribution ranges of R. atacamensis (25° to 32° S) and R. arunco (32° to 38° S), except for 120 km in the southern range of R. arunco (Fig. 1, Table 1).Sampling was more intense between 31° and 33° S, around the Pupío watershed (32° S) in which hybridization was originally described (Correa et al. 2012;Fig. 1).The individuals used were mainly tadpoles, postmetamorphics and juveniles.We also included a few adults, most of which were sampled by excising a small piece of interdigital membrane and released in the place they were collected.The specific identity of the specimens from north of 31° S and south of 33° S can be unambiguously established because only pure populations of each species have been described beyond these limits and both species can be distinguished without diffi culty by their coloration patterns.Adults of R. atacamensis have sexual dimorphism in background color (whitish in females and yellowish in males) and small reddish spots on the dorsum or on the eyelids, while the background color of R. arunco varies from light grey to dark brown in both sexes (see more details in Correa et al. 2012).Postmetamorphics and juveniles of both species also can be distinguished for their color patterns.The permits for the capture and collection of the animals were provided by the Servicio Agrícola Ganadero (SAG) (resolutions 3085/2000, 2105/2004 and 13/2006).All collected individuals and tissues were deposited in the herpetological collection of the Departamento de Biología Celular y Genética of the Universidad de Chile (DBGUCH).

Obtaining DNA and mitochondrial sequences
We obtained sequences of the mitochondrial control region for 359 individuals from 43 localities (Table 1).The DNA was extracted principally from muscle tissue: from the thigh of adults, the tongue of juveniles and postmetamorphics, and from the tail of larvae.Occasionally we used liver, digit or interdigital membrane; the last only in those adults individuals which were returned to their habitat.DNA was extracted using a modification of the salt method of Jowett (1986).The mitochondrial fragment sequenced included the 3' extreme of the cytochrome b gene and approximately 850 bases of the contiguous extreme of the control region (noncoding).The primers used to amplify this fragment were CytbA-L (5'-GAATYGGRGGWCAACCAGTAGAAGACCC-3') and ControlP-H (5'-GTCCATAGATTCASTTCCGTCAG-3'), designed by Goebel et al. (1999).The PCR protocol is the same used in Correa et al. (2012).

Obtaining AFLP markers
We obtained AFLPs for a representative subset of 205 individuals (among the 359 ones used for obtaining sequences) from 27 localities (Table 1).Details for obtaining (digestion, ligation, pre-selective and selective PCRs steps), genotyping and coding AFLPs are given in Correa et al. (2012).The only difference is that in the present study we used four combinations of selective primers, three of which were used in the earlier study.The combinations of selective primers were MseI-CAC/6FAM-EcoRI-ACT, MseI-CAA/VIC-EcoRI-ACC, MseI-CAT/NED-EcoRI-ACG and MseI-CAC/PET-EcoRI-ACA.
As a simple way of identifying hybrid individuals, a set of diagnostic markers (i.e., markers present in 100 % of the individuals of one species and absent in 100 % of Fig. 1: Location of the 43 localities of Rhinella included in this study, numbered from north to south (see Table 1).Squares are localities from which both mitochondrial sequences and AFLP markers were obtained; only mitochondrial sequences were obtained from the localities with circles.The map at the extreme right is an amplification of the zone where sampling was more intense (31°30' -33° S).Continuous thin lines in this map indicate the limits of the watersheds.Coordinates and altitudes of the 43 localities of Rhinella atacamensis, R. arunco and the hybrid zone included in this study, ordered and numbered from north to south (see map in Fig. 1).The number of individuals of each locality used to obtain AFLP markers and mitochondrial sequences is also indicated.

Locality
Latitude the other) was defi ned using only individuals from the watersheds in which only haplotypes of one species were observed.The diagnostic markers of R. atacamensis were defi ned based on fi ve localities north of the Choapa River watershed (83 specimens) and those of R. arunco from four localities south of the Aconcagua River watershed (16 specimens) (Fig. 1).

Phylogeographic and genetic analyses
Mitochondrial sequences were edited with BioEdit v7.0.7 (Hall 1999).We performed an initial alignment with ClustalX v2.0.12 (Larkin et al. 2007) of the fi rst sequences obtained for the two species; the rest were then added manually.Haplotypes were generated with DnaSP v5.10.01 (Librado & Rozas 2009), including the sites with gaps.The phylogenetic relationships among the haplotypes of the two species were estimated by constructing a haplotype network with the method of median joining using the program Network v4.610 (Bandelt et al. 1999).Default parameters were used for obtaining the network, which was also used to visualize the mitochondrial genetic divergence at the intra-and interspecific levels (as mutational steps) and the relationships among the haplotypes present in the hybrid zone.
The presence of hybrids in Rhinella localities was investigated with NewHybrids v1.1beta3 (Anderson & Thompson 2002).This program uses a Bayesian framework to calculate the posterior probability that each individual of a sample belongs to one or more pre-defined categories of hybrids or to one of the parental species.We specified six categories: pure R. atacamensis; pure R. arunco; fi rst generation (F1) hybrid; second generation (F2) hybrid; backcross between F1 and R. atacamensis; and backcross between F1 and R. arunco.We did not incorporate the species as additional information for this analysis.The length of the Markov Chain Monte Carlo procedure was 1000000 iterations, with the fi rst 100000 discarded as burn-in.We performed various replicas in order to judge the consistency of the results.The location of the hybrid zone was defi ned considering the geographic distribution of the haplotypes of the control region and the individuals classifi ed by Newhybrids.

Variation of the mitochondrial sequences
We obtained an alignment of 921 sites, of which 170 were variable among the two species (12 sites with gaps).We found a total of 121 haplotypes for the two species (including those defi ned by the gaps): 58 corresponding to R. atacamensis and 63 to R. arunco (Fig. 2).Although the number of individuals with haplotypes of R. atacamensis was greater (185), the observed level of intraspecifi c divergence (estimated as mutational steps) was greater in R. arunco (22 steps vs. 19 in R. atacamensis).The intraspecific divergence was much less than that obser ved between the species (75 mutational steps).
In four localities, Puente Pupío, Pupío Medio, El Sobrante and Las Chilcas (Fig. 1 Figure 2 illustrates the relationships among the haplotypes of these localities.The haplotypes are not directly related, but are dispersed within the networks of each of the species.This pattern was more accentuated in the haplotypes of R. atacamensis.The sequences were deposited in GenBank with accession numbers AY818062, AY818063, HQ132482-HQ132670 and KC778198-KC778365.
we found individuals with a mixture of the diagnostic markers of the two species, including the four in which there were mixed haplotypes.In seven other localities between those latitudinal limits we found only individuals with markers of one or the other species (see below and Fig. 3).
The Newhybrids analysis agreed completely with the distribution of the diagnostic markers.As expected by their phenotype (color patterns) and region control haplotypes, all individuals from localities north of 31°30' S and south of 33° S were correctly identifi ed as belonging to their respective species.Also, according to their color patterns, the individuals of the localities of the upper zone of the Choapa River (Canela Alta, Los Perales, Palquial) and the two from Quebrada Seca were classifi ed as R. atacamensis (Fig. 3), whereas the individuals from the lower sectors of the watersheds south of the Pupío watershed (Quilimarí, Los Molles, El Trapiche, Illalolén) were classified as R. arunco.Only in three localities individuals were found which could be considered fi rst or second generation hybrids, while in another fi ve only some individuals classifi ed as backcrosses were present (Fig. 3).  1 and Table 1).The probable location and extension of the hybrid zone is indicated with a hatched area and discontinuous lines on the map, which also has question marks in zones which need to be explored.The localities indicated with squares are those included in the Newhybrids analysis; circles represent the localities with only haplotype data.As in Fig. 1, continuous thin lines in the map indicate the limits of the watersheds.

Delimitation of the hybrid zone
The limits of the hybrid zone were drawn to include all localities in which mitochondrial haplotypes of both species were found and/ or individuals were classified in one of the last four hybrid categories of Newhybrids mentioned above.In its northern part, the zone appears to have a regular form, crossing several watersheds in a NW-NE orientation (Fig. 3).However, in the southern par t, two points located in the Aconcagua River watershed, Quebrada Seca and Las Chilcas, suggest that the geography of the zone is more complex.For this reason, this part of the hybrid zone was drawn with discontinuous lines, one which surrounds Quebrada Seca, which would be an isolated locality of R. atacamensis, and the other extends in Los Andes foothills on the border of the Aconcagua River watershed and includes Las Chilcas (Fig. 3).We also added question marks in the Fig. 3 to indicate those areas which should be sampled to defi ne the limits of the hybrid zone with greater precision.
The zone described in this study extends from the locality of Huentelauquén to Las Chilcas, a latitudinal distance of about 150 km (Fig. 3).Examining the distributions of the two species, R. atacamensis was found at increasingly higher altitudes towards the south, although the locality that until now defi ned the southern limit of its distribution (Quebrada Seca) appears to be isolated from the rest of the populations of Los Andes range.Inversely, R. arunco was found only in the lower part of the watersheds to the north of the Aconcagua River watershed, reaching the coastal strip at the mouth of the Choapa River (Fig. 3).The locality of Las Chilcas is an exception to this pattern, since there R. atacamensis was found at less than 1000 m in the Aconcagua River watershed; in the other two localities of this watershed, Resguardo Los Patos and Quebrada Seca, it was found at 1211 m and 1693 m, respectively.DISCUSSION T h i s s t u d y r e v e a l e d t w o i m p o r t a n t biogeographic aspects of the hybrid zone of R. atacamensis and R. arunco: fi rst, the zone has a relatively reduced extension compared to the distribution of both species, and second, there is a consistent altitudinal segregation between them.The estimated extension of the zone (150 km in latitude) is wide in comparison with the distribution of each species, but is reduced in relation to the combined distribution (about 1450 km length).Additionally, the two species are not homogeneously distributed in the watersheds; thus the hybrid zone occupies a reduced fraction of each and may be represented as a narrow strip which crosses them.The second aspect, the altitudinal segregation, was obser ved by Correa et al. (2012) in one of the watersheds of the zone (Pupío creek).However, the present study shows that it is a consistent pattern in all the area of contact of the two species.In animals, this kind of hybrid zone, where the parental species are consistently separated along elevational gradients, has been scarcely repor ted in the literature (Culumber et al. 2010).
A factor that limits the spatial extension of hybrid zones is the extent to which the species are mixed and are permeable to the gene introgression.If there exist strong reproductive bar riers (prezygotic and/or postzygotics), hybrids are scarse and the hybrid zone is relatively narrow or almost nonexistent, when the surrounding areas where introgression has been detected are not considered (e.g.Colliard et al. 2010, , Taylor et al. 2012, Miraldo et al. 2013).Correa et al. (2012) provided genetic and reproductive evidence showing that the populations of R. atacamensis and R. arunco of the Pupío creek (32° S) conform a possible hybrid swarm, suggesting the absence of reproductive barriers.This study reinforces and spatially extends that initial obser vation, showing also no evidence of mitochondrial and nuclear introgression outside of the defi ned hybrid zone.This apparent absence of reproductive barriers between R. atacamensis and R. arunco (which allow us to dismiss a priori a process of parapatric speciation), the large genetic divergence compared to the intraspecific variation (which is evident in the region control network) and the reduced extension of the hybrid zone detected so far suggest that the hybrid zone originated by secondary contact, which implies that one or both species extended its distribution.The paleoclimatic and orographic information given below allows us to hypothesize that one species expanded its range (R. arunco), displacing the other, which would explain the geographic conformation of this hybrid zone.
In Chile, the changes in the distribution of the flora due to the climatic fluctuations produced by the Pleistocene glaciers and during the Holocene are well known (Villagrán & Hinojosa 1997, Villagrán et al. 1998, Villagrán 2001).Due to the geographic conformation of the countr y, vegetation displacements have mostly been north and south.However, these changes were modified by the geographical relief, thus these displacements have reached dif ferent latitudes depending upon their altitude.Thus the Mediterranean vegetation (mainly sclerophyllous shr ubland) reached a lower latitude in the last glacial maximum, mainly in the coastal zone (Villagrán 1995, 2001, Villagrán & Armesto 2005).Considering that R. arunco is an endemic species of the Mediterranean ecoregion (central Chile), the northward expansions of the Mediterranean vegetation during the glacial cycles of the Pleistocene may explain the presence of this species mainly in the coast between 31°30' and 32°30' S (Fig. 3).If in this area there were glacial refugia for this or other species, they have not been described.Moreover, although the coastal plain is currently narrow, it was wider during the last glacial maximum and could have served as a corridor which allowed the northward expansion across the successive watersheds.These same mechanisms have been suggested to explain the colonization of south-central Chile by a sigmodontine mouse and a continental fi sh (Unmack et al. 2009, Palma et al. 2012).A range expansion of R. arunco through areas of low altitude and slight slopes (i.e., the valleys) would also explain the presence of R. atacamensis at increasingly higher altitudes towards the south, including a high locality, Quebrada Seca, apparently isolated south of 32°30' S (Fig. 3).A possible explanation for the formation of this distributional pattern is a competitive displacement of the lowland populations of R. atacamensis by R. arunco, although if this or other ecological processes are involved need to be evaluated.
There are reports of hybrids zones closely associated to environmental transition areas or ecotones, where the hybridization is frequent, but limited by extrinsic and/or intrinsic factors (e.g.Yanchucov et al. 2006, Hapke et al. 2011, Chavez et al. 2011, Culumber et al. 2012).The altitudinal segregation which defi nes the hybrid zone of R. atacamensis and R. arunco suggests different environmental preferences, but this does not agree with the fact that both species occupy both lower and higher zones of the watersheds, in the areas of their distributions where only one species occurs.Moreover, there are no abrupt changes in environmental parameters along the courses of the rivers in which the hybridization occurs, or at least they are not apparent.In a wider climatic context, the zone is located in the xeric-oceanic Mediteranean bioclimate, which borders on the north with the oceanic-desert Mediteranean bioclimate and on the south with the seasonal rain-oceanic.The transition between these bioclimates is gradual and is closely associated with a nor th-south precipitation gradient (Luebert & Pliscoff 2006).Also this zone is within the area in which the transverse valleys (which are more or less perpendicular to Los Andes range) disappear and the Coastal range begins to appear; these are low, older mountains found from 33° to 41° S and are parallel to Los Andes range.In both cases, the transitions are gradual and do not coincide with the location of the hybrid zone, so we discount that the climatic conditions and/or geographic relief might be more relevant than a possible historic event (expansion of R. arunco) to explain its origin and current conformation.
Finally, we would like to consider briefl y the consequences of the existence of a hybrid zone of R. atacamensis and R. arunco for the taxonomy and conser vation of Rhinella in nor th-central Chile.The great majority of populations between 25° and 38° S may be assigned unequivocally to one or other species, by phenotype, mitochondrial and nuclear markers (diagnostic AFLP markers).However, populations of the hybrid zone may be considered as a fusion of both species (Fig. 3).The geographic location of the zone implies the presence of pure populations of both species in the main watersheds between 31°30' and 33° S, but they are separated by mixed populations whose exact extension is diffi cult to determine without estimations of gene introgression in these systems.Moreover, Correa et al. (2012) and the present study demonstrate that the majority of individuals from hybrid populations have dif ferent propor tions of the nuclear markers of both species, thus they cannot be assigned to one or another.Therefore, we suggest expanding the taxonomic defi nition of each species, including all the respective pure populations and those composed of a mixture of both.Thus, now the distribution range of R. atacamensis would be defi ned between 25° and 32° 52' S (Las Chilcas), whereas that of R. arunco would be defi ned between 31° 35' (Huentelauquén) and 38° S. In a conservation context, Allendorf et al. (2001) suggest this type of natural hybrid zone, where apparently the reproductive success of hybrids is similar to that of parental species (Cor rea et al. 2012), constitutes a eligible conservation unit, although in this case their exact geographic limits remain to be defi ned.These proposals would allow us to formalize the discover y of this hybrid zone, which also adds an interesting evolutionar y dimension to the study of the biogeography of the amphibians of Chile.

Fig. 2 :
Fig. 2: Haplotype network of the mitochondrial control region of Rhinella atacamensis and R. arunco, including all observed haplotypes (121).The amplifi cations of the networks for each species are shown on the right; the haplotypes of both species found mixed in four localities of the hybrid zone are indicated in gray: Pupío Medio (PM), Puente Pupío (PP), El Sobrante (ES) and Las Chilcas (LCh).The size of the circles is proportional to the sampling frequency of each haplotype and the length of connecting lines is proportional to the mutational steps that separate them.

Fig. 3 :
Fig. 3: Result from Newhybrids (left) and map of the hybrid zone of Rhinella atacamensis and R. arunco.Each colored bar represents one individual and the extension of the color indicates the probability of belonging to one of the six categories specifi ed below the bars according to the AFLPs analysis performed in Newhybrids.Numbers at the right of the bars indicate the localities of the individuals (locality numbers are the same as Figure1and Table1).The probable location and extension of the hybrid zone is indicated with a hatched area and discontinuous lines on the map, which also has question marks in zones which need to be explored.The localities indicated with squares are those included in the Newhybrids analysis; circles represent the localities with only haplotype data.As in Fig.1, continuous thin lines in the map indicate the limits of the watersheds.

TABLE 1 .
Continuationdetected mixtures of haplotypes of both species.