Using the available distribution data of I. elegans and I. graellsii in Spain, we constructed two geographic maps to show their overall distributional range (Figure 2A, Ischnura elegans; 2B, I. graellsii). Species overlap in north and central Spain (from west to east). In particular, I. graellsii is found all over the Iberian Peninsula, while I. elegans is very rare in southern Spain (Figure 2).

All population groups (Spanish I. elegans populations, European I. elegans populations excluding Spain and I. graellsii populations) exhibited a high degree of molecular diversity (see Table 1). Estimates of observed and expected heterozygosity were similar and ranged from 0.56-0.70 and 0.61-0.76, respectively (Table 1). In the Spanish I. elegans populations, we detected a total of 87 alleles, 11 less than in the other European I. elegans populations. In contrast to I. elegans, we found somewhat fewer (66) alleles for I. graellsii, perhaps due to the fact that the microsatellites were specifically developed for I. elegans which could cause an ascertainment bias. Estimates of allelic richness were comparable between the Spanish and the other European I. elegans populations (6.54 and 6.03, respectively), and were considerably higher than in I. graellsii (3.41; Table 1).

Analyses of the overall genetic structure showed that I. elegans populations were significantly differentiated from one another in Europe outside Spain (F_ST = 0.031, P < 0.0001) as well as within Spain (F_ST = 0.049, P < 0.0001). Moreover, populations of I. graellsii were also significantly differentiated from one another (F_ST = 0.029, P < 0.0001).

Four of 23 principal component axes accounted for a significant amount of genetic variation among samples, as indicated by a screen plot. The first axis contained 24% (F_ST = 0.024, P = 0.11), the second axis 17% (F_ST = 0.017, P = 0.08), the third axis 14% (F_ST = 0.014, P = 0.07) and the fourth axis 11% (F_ST = 0.011, P = 0.05) of the total variation (Figure 3). The first two principal components thus summarise almost half (41%) of the total variation inherent in all I. elegans and I. graellsii populations. The scores for the first principal component axis revealed a clear separation of I. graellsii and I. elegans populations, suggesting that major population differences were predominantly caused by species rather than geographic areas per se. Major population differences (nested inside species) were revealed by the second PCA axis, where two of the Spanish I. elegans populations (Louro and Laxe; Table 2 and 3) were situated in the same quadrant that was occupied by all I. graellsii populations (Figure 3).

Analyses of the population structure in STRUCTURE supported the molecular differentiation between the two Ischnura species, although the ΔK-method suggested three clusters as the most likely population structure (Figure 4). More specifically, the results showed that one of the genetic clusters highly corresponded to the I. graellsii group, while the other two genetic clusters were best represented by I. elegans genotypes (one mainly corresponding to northern, central and southern European populations, and one to populations in eastern Europe), with the Spanish populations showing an intermediate assignment to each of these three genetic clusters (Figure 4).

Based on these results, we included all populations of I. graellsii and I. elegans in the STRUCTURE runs (K = 2) to analyse the individual admixture proportions of I. elegans individuals in Spain (Table 4). These results showed that one genetic cluster clearly corresponded to I. graellsii, while the other cluster corresponded to I. elegans (European populations outside Spain; Table 4; Figure 5). The majority of individuals of the European I. elegans outside Spain (95%) and I. graellsii (88%) were assigned with a certainty of at least 90% to each of these clusters. However, only 27% of individuals of the Spanish I. elegans were strongly assigned to I. elegans (admixture proportions ≥ 90%) and one individual (from the sympatric population Alfaro) was assigned to I. graellsii (admixture proportion to I. elegans < 10%). The rest of the Spanish I. elegans individuals were intermediate between the two clusters (with a skew of admixture proportion towards I. elegans, see additional file 1), suggesting a significant degree of introgressed I. graellsii alleles (Table 4; Figure 5).

The admixture proportions of the artificial hybrids and backcrosses ranged between 11-89% (Table 5; Figure 5). The F₁ and F₂ showed admixture proportions to I. elegans between 67 and 21%; the 1 GB (first I. graellsii backcross) between 67 and 11%; and the four I. elegans backcrosses (1-4 EB) between 89 and 21% (Table 5; Figure 5). Based on these results, we defined three conservative assignment groups: backcrosses with I. elegans with admixture between 89-68%; backcrosses with I. graellsii ranging between 20-11%; and a mixed group of F₁, F₂ and backcrosses ranging between 67-21%. When the Spanish I. elegans individuals (N = 166) were allocated by their admixture proportions in one of these three groups, a total of 97 individuals (admixture proportions to I. elegans between 89-68%) were assigned to backcrosses with I. elegans (1-4 EB), whereas not a single individual was assigned to the backcross with I. graellsii (1 GB) (Table 5). Finally, the rest of the Spanish I. elegans (i.e. 23 individuals with admixture proportions to I. elegans between 67-21%), excluding the single individual from Alfaro that was assigned to I. graellsii, were assigned to the mixed-hybrid group, including F₁, F₂ and backcrosses with the parental species (Table 4).

The alignments for the Cytochrome C Oxidase I and II (COI-COII), Cytochrome B (CYTB), 12S rRNA (12S) and NADH Dehydrogenase 1 (ND1) fragments included 591 bp, 673 bp, 457 bp, 370 bp and 591 bp, respectively (Table 6). All new sequences were deposited in GenBank (accession numbers: HQ834794-HQ834810). COI fragment showed no polymorphic sites, revealing a unique haplotype (H1) that was shared by the two species. COII showed one polymorphic site, revealing two haplotypes (haplotype diversity, H = 0.409 ± 0.133; nucleotide diversity, π = 0.00122). Haplotype H2, the most abundant haplotype, was shared by the species in both allopatric and sympatric regions, while haplotype H3 appeared only in the three samples of I. graellsii from one allopatric region (Morocco). CYTB fragment showed two polymorphic sites revealing three haplotypes (H = 0.163 ± 0.099; π = 0.00053). Each species showed one unique haplotype (H5 and H6), in a single sample from one allopatric region (Greece and Morocco), respectively, while the rest of the samples of both species shared a common haplotype (H4). The 12S fragment showed three polymorphic sites revealing four haplotypes (H = 0.087 ± 0.047; π = 0.00024). Ischnura elegans showed three unique haplotypes (H7, H8, H9), each of which appeared in a unique sample (two haplotypes in samples from allopatric regions and one in a sample from a sympatric region), while the rest of the samples of both species shared the common haplotype H10. The last fragment, ND1, showed one polymorphic site, revealing two haplotypes (H = 0.250 ± 0.180; π = 0.00042). The most abundant haplotype (H11) was shared by both species (from allopatric and sympatric regions), while the second haplotype (H12) appeared only in one sample of I. graellsii (from an allopatric population in Portugal).

All populations showed all three female morphs, with the exception of Saïdia in northern Africa (where the aurantiaca morph was missing) (Table 7). In populations dominated by I. elegans from north-central, central and eastern Spain, androchrome, infuscans and infuscans-obsoleta frequencies were highly variable between populations (range of androchrome frequencies: 3.3-42.9%; infuscans: 6.7-68.8%; infuscans-obsoleta: 11.8-76.7%) which is consistent with what has previously been observed for the I. elegans populations from north-western Spain 21 . In allopatric populations of I. graellsii, androchrome frequencies were always below 20% (3.0-18.8%); with the infuscans morph being the most abundant morph type (64.9-89.7%). The sympatric populations, O Vilar and Xuño, which have previously shown very high frequencies of androchromes for both species 21 , also showed in our study high levels of androchrome frequencies (12.5-15.7% in I. graellsii and 40.0-47.6% in I. elegans). In addition, the sympatric populations Las Cañas and Alfaro from central Spain further showed high androchrome frequencies (13.0-17.1% in I. graellsii and 69.7-70.8% in I. elegans).

In vertebrates, hybridization is particularly common in fish, where several hundred interspecific and intergeneric crosses have been reported, and in birds, with roughly 10% of all species known to have bred in nature with another species e.g. 5 27 46 47 48 49 50 . In comparison, detailed genetic studies of hybridization and introgression, and thus the knowledge about these phenomena, are lacking for most odonates 12 13 . As mentioned above, three previous genetic studies failed in detecting extensive hybridization between different pairs of odonate species 14 15 16 17 . Our admixture analysis in STRUCTURE showed that the I. elegans populations Louro and Laxe in north-western Spain had the highest degree of introgression of I. graellsii alleles. This is not surprising, given that north-western Spain is the region where hybridization between I. elegans and I. graellsii was most recently detected 20 . In 1990, I. elegans, I. graellsii and hybrids were found in Foz (north-western Spain), which is geographically close to Louro and Laxe. However, after only ten years, the population was dominated by individuals that were phenotypically classified as I. elegans, although morphological intermediate and I. graellsii males were occasionally detected 21 . The population Laxe was visited for the first time in June 2001 (two visits) and only I. elegans were detected, although at a low densities (between 0.2 to 1.3 captured males/minute). At a visit in June 2002, I. elegans were found at a similar density (0.5 males/minute). Five years later (in 2007), we revisited the population and the density had reached the highest value in the region. A total of seven visits were conducted in 2007, between June and August, and the density ranged between 3.3 and 14.7 males/minute; and neither I. graellsii nor putative hybrids were detected. Nevertheless, the admixture analyses showed that none of the 14 examined individuals at Laxe could be genetically assigned to pure I. elegans status. Nine individuals were assigned to F₁, F₂, or backcrosses with I. elegans (admixture proportion between 67-21%), and five individuals were assigned to backcrosses with I. elegans (admixture proportion between 89-68%). In 1980, Louro was visited for first time by Ocharan 22 and mainly I. graellsii individuals were observed. However, Torralba and Ocharan 51 reviewed Ocharan's samples from Louro (sampled in 1980) and have now identified both species in the sample; nevertheless, I. graellsii species still remains the dominant species. However, on our first visit in 2000 and on subsequent visits, I. elegans individuals completely dominated in numbers (a couple of I. graellsii individuals were detected among hundreds of I. elegans). Only one out of the 15 Louro individuals that were molecularly examined in the present study was assigned to be pure I. elegans. In addition, the vast majority of the individuals (93%) showed an assignment proportion between 89-68%, suggesting that these individuals were backcrosses, and not F₁ or F₂ hybrids.

These results are corroborated by the PCA-analysis. Both Laxe and Louro were placed within the I. graellsii quadrant, indicating a significant degree of molecular similarity between these I. elegans populations and I. graellsii. Analyses of the remaining Spanish I. elegans populations showed that only 31% of the individuals were pure I. elegans, and that the remaining 69% showed admixture proportions expected for hybrids and backcrosses with I. elegans. In the Alfaro population in the north-central Spain where both species co-occur in equal numbers, one individual classified as I. elegans had a very high proportion of I. graellsii alleles. This either suggest that this individual was misclassified despite the fact that we only collected males to minimise misidentifications 20 , or that it is a backcross that has inherited a very high proportion of I. graellsii alleles at the few markers analysed.

Our suggest that hybridization between I. elegans and I. graellsii is asymmetric and largely unidirectional. This corroborates previous findings in the field 20 and laboratory 21 showing that I. graellsii has a mating disadvantage in sympatry with I. elegans. Heterospecific matings, and matings between the F₁-hybrids and the parental species, rarely take place with I. elegans females, but occur more frequently among I. graellsii females, I. elegans males and hybrids [Sánchez-Guillén RA, Wellenreuther M and Cordero-Rivera A: Strong asymmetry in the relative strengths of prezygotic and postzygotic barriers between two damselfly sister species, submitted]. The reason for the almost complete lack of hybridization between I. graellsii males and I. elegans females is that males cannot grasp the female by their protothorax 20 , a mechanical handicap that appears to be a very efficient prezygotic isolation mechanism. Previous work on plants 52 53 and animals 27 has suggested that unidirectional hybridization usually occurs between the females of the rare species and the males of a common species, but not vice versa. Wirtz 27 reviewed the factors promoting unidirectional or reciprocal hybridization and proposed a sexual selection hypothesis for unidirectional hybridization based on the fact that females normally invest more in offspring and, therefore, are more discriminating than males. When heterospecific males are less abundant than conspecific males, females rarely mate with heterospecific males. Consequently, under such a condition, the rare species is usually the maternal parent of the hybrids. However, this is not the case in our study where the more abundant species is initially I. graellsii. Ischnura elegans, on the other hand, appears to be the intruding species and is hence initially the rare species, which has been expanding its range in Spain and is now displacing I. graellsii from some populations and regions. Thus, the direction of hybridization between these two Ischnura species is opposite that expected based on the rare female hybridization hypothesis just outlined e.g. 27 , but follows the prediction proposed by Grant and Grant 28 , namely that when the rare species is also the bigger species (in our study I. elegans), hybridization can be impeded by mechanical incompatibility. Laboratory tests have detected that I. elegans females and I. graellsii males are mechanically impaired to form a tandem, preventing over 93% of all matings, while only 13% of the matings between I. graellsii females and I. elegans males are mechanically prevented [Sánchez-Guillén RA, Wellenreuther M and Cordero-Rivera A: Strong asymmetry in the relative strengths of prezygotic and postzygotic barriers between two damselfly sister species, submitted].

Johnson 33 proposed that the colour polymorphisms in odonates could be maintained due to hybridization between closely related species and balanced by differential predation pressure. According to this hypothesis, androchrome females benefit from avoiding matings with heterospecific males, while gynochromes females are involved in heterospecific matings (usually sterile); the relative fitness of the colour morphs is balanced by a higher probability of predation of the androchrome females. In a recent study on I. elegans and I. graellsii 21 , a role of hybridization for the temporal maintenance of contrasting androchrome frequencies in nearby populations in north-western Spain was suggested. The relatively low androchrome frequencies in I. elegans populations located close to I. graellsii populations, and the relatively high androchrome frequencies in I. graellsii populations adjacent to I. elegans populations, were hypothesized to be caused by the absorption of the, in other circumstances, typical morph frequency 21 . Ischnura elegans data from other sympatric regions in the Iberian Peninsula now shows that the substantial variation in female morph frequencies between different populations in north-western Spain is not unique to this region; also in other parts of Iberia morph frequencies are drastically variable between populations (androchrome: 3.3-70.8%; infuscans: 6.7-72.9%; infuscans-obsoleta: 3.0-76.7%). Furthermore, other recent studies have revealed that disparate morph frequencies are not restricted to Iberian populations 53 54 , although androchrome frequencies are typically more abundant in northern latitudes and show less variation between populations than in Iberia [with the exception of Iberia, 53]. In our study, the highest frequencies of androchromes were found in the four populations with both species present (Xuño, O Vilar, Alfaro and Las Cañas; Table 7); in I. graellsii androchrome frequencies ranged between 12.5 and 17.4% and in I. elegans between 40.0 and 70.8%. Nevertheless, the I. graellsii population Ribeira de Cobres had androchrome frequencies of 18.8%, for this species an atypically high level. However, this comparatively high androchrome frequency cannot be explained by the influence of nearby I. elegans populations, because this population is located in the allopatric region (southern Portugal). Furthermore, in I. elegans the lowest androchrome frequencies were found in north-central Spain (population named Arreo) where the nearest I. graellsii population (named Troi) had higher androchrome frequencies (14.8%) than the I. elegans population (3.3%), and in a coastal population near the Mediterranean coast (eastern Spain) (Amposta; 3.3% androchromes), but here we lack data of I. graellsii morph frequencies in nearby populations.

In conclusion, hybridization is likely to have important implications for the maintenance of multiple female morphs, but only during the short period when two species with contrasting morph frequencies start to hybridize.

The evolution, maintenance and adaptive function of genetic colour polymorphism have received very much attention in a broad range of organisms, e.g. in birds 55 , reptiles 56 and insects 57 . However, the role of hybridization in this context has received little attention, which may simply reflect the fact that only a few suitable study systems are available to study; one being the presently described Ischnura damselfly complex in Spain. Another case is the hybrid zone of the land snails Mandarina mandarina and M. chichijimana in the oceanic Bonin Islands 58 . In that system, the variability of the colour polymorphism in the hybrid population was substantially higher than that in the pure populations, suggesting that morphological variation was maintained by hybridization 58 . These results highlight the importance of hybridization as a source of morphological variation, diversity and evolutionary novelty.

We conducted a revision of the distribution data of the two species along the Iberian Peninsula from 1866-2008 using data from Baixeras et al. 60 , Jödicke 61 and Ocharan 22 62 , in the region around la Rioja using data from Tomás Latasa (personal communication) and along the Iberia using data from Jean Pierre Boudot (personal communication). Using DMAP (Distribution mapping software, Version 7.0) we constructed two geographic maps showing the distribution of both species in Iberia (Figure 2).

Samples of I. elegans and I. graellsii were collected from 26 populations from Europe and northern Africa (see Table 2 for details of sampling locations). In particular, all European populations, except the populations from Spain, were classified as allopatric I. elegans populations and included a total of 220 individuals from 13 I. elegans populations. In depth sampling of both I. elegans and I. graellsii was carried out in the sympatric region in Spain; in the north-western corner, the central parts and along the east coast (Figure 1). In these areas, nine sympatric populations (166 individuals) were sampled, and these were, with the exception of Alfaro, dominated by individuals that were phenotypically classified as I. elegans (see Table 3 for details of species proportions in the sampled populations). In addition, four allopatric I. graellsii populations (56 I. graellsii individuals) from the north, central and south Iberia, and northern Africa were sampled.

At each allopatric and sympatric I. elegans population (see Table 2), and at the four allopatric I. graellsii populations, a minimum of 20 adult males were collected during the flight season between 1999-2008 using hand nets. Captured individuals were stored in 100% ethanol until DNA extraction. Only males were sampled because the identification of male I. elegans, I. graellsii and hybrids is more reliable than that of females. Note that the few individuals classified as I. graellsii and hybrids in the Spanish I. elegans populations, mainly found in Alfaro in north-central Spain (Table 3), were not included in the genetic analyses because the aim of this study was to test for introgression in I. elegans populations.

To extract DNA from the samples, the head of each damselfly was removed, dried and then homogenized using TissueLyser (Qiagen). DNA was extracted by proteinase K digestion followed by a standard phenol/chloroform-isoamylalcohol extraction 63 . The purified DNA was re-suspended in 50-100 μl of sterile water. The genotypes of all damselflies were assayed at six microsatellite loci previously isolated for this species [I-002, I-015, I-041, I-053, I-095, I-134, for details see 33]. These loci did not deviate statistically from Hardy-Weinberg expectations and linkage equilibrium, and showed no evidence for presence of common null-alleles (using Micro-Checker; 64 , within populations of both species 34 . One primer of each pair was 5'-labelled with 6-FAM, HEX or NED florescent dyes. Polymerase chain reactions (PCRs) were carried out in 10 μL volumes on a GeneAmp PCR System 9700 (Applied Biosystems) and contained 4 pmol of each primer, 15 nmol MgCl₂, 1.25 nmol dNTP, 0.5 U Ampli-taq polymerase and 10-20 ng template. The conditions were: denaturation step of 94°C for 2 minutes, then 35 cycles at 94°C for 30 s, touch-down from 62-58°C for 30 s, 72°C for 30 s followed by 72°C for 10 minutes. Multiplex primer reactions were performed for combinations of primers with matching annealing temperatures but differing size ranges and dye labels, then mixed with a labelled size standard and electrophoresis was conducted on an ABI PRISM 3730 Genetic Analyzer (Applied Biosystems). GeneMapper 3.0 (Applied Biosystems) was used for fragment size determination and allelic designations.

The program FSTAT 65 was used to calculate several basic population genetic measures, namely, the expected heterozygosity (H_E), observed heterozygosity (H_o), number of alleles, and the allelic richness for each population. These aforementioned measures as well as the genetic differentiation between populations (F_ST) were also calculated for three regions, namely the Spanish populations of I. elegans, the European populations of I. elegans (excluding Spain), and the entire sample of I. graellsii.

Principal component analysis (PCA) was used to reduce the variation in the multivariate data set (consisting of 117 alleles at six loci) to two linear combinations. The analysis was done using PCA-GEN 66 . The significance of each principal component was assessed from 5000 randomisations of genotypes. The allocation of each species across the two principal components provides a quantitative measure of the degree of the genetic dissimilarity among the populations/species 18 .

The Bayesian statistical framework provided by the program STRUCTURE [version 2.2.3, 67] was used to understand the genetic structure among populations and to determine which individuals from the allopatric and sympatric populations of I. elegans and I. graellsii can be classified to a high degree as pure species. STRUCTURE applies a Bayesian Markov chain Monte Carlo (MCMC) approach that uses model-based clustering to partition individuals into groups. The model accounts for the presence of Hardy-Weinberg and linkage disequilibrium by introducing group structure and attempts to find groupings that (as far as possible) are in equilibrium 68 . We applied the 'admixture model' with 'correlated allele frequencies' for more details 69 . For the model, a 'burn-in' period of 20,000 MCMC replicates and a sampling period of 100,000 replicates were used. We performed runs for a number of genetic clusters (K), ranging from one to ten; and for each K, 20 iterations were run. In this way, multiple posterior probability values (log likelihood (lnL) values) for each K were generated, and the most likely K was evaluated by the ΔK-method following Evanno et al. 66 .

Admixture analyses in STRUCTURE were also used to assign all individuals of the Spanish Ischnura populations into each of two genetic clusters, one representing I. graellsii genotypes and one I. elegans. We used the 'prior population information' option in the models to (i) facilitate the clustering process of the reference individuals (i.e. pure I. elegans from central and eastern Europe, and I. graellsii, respectively), and (ii) to calculate the admixture proportions (and ± 90% credible regions) of each individual in the Spanish I. elegans populations. This approach was hence used to measure of the degree of introgression of I. graellsii genetic material into the genome of I. elegans in Spain. The model was run for K = 2, where one cluster corresponded to I. graellsii and the other to I. elegans. We used the 'population flag' option to exclude Spanish I. elegans as reference individuals, which implied that the clustering process was based on only I. graellsii samples and I. elegans samples collected outside of Spain. The model was run for 100,000 MCMC replicates, after an initial burn-in period of 20,000 replicates, using the admixture model and correlated frequencies for five iterations 32 33 34 63 . To generate simulated genotypes of hybrids and backcrosses, we applied the program HYBRID-LAB  70 using the genotypes of 66 individuals of I. graellsii and 240 genotypes of I. elegans collected outside of Spain as initial genotypes. We generated 50 genotypes of each of the following crosses: first-generation hybrid (F₁; i.e. I. graellsii × I. elegans), second-generation hybrid (F₂; i.e. F₁ × F₁), first backcross with I. elegans (1 EB; i.e. F₁ × I. elegans), first backcross with I. graellsii (1 GB; F₁ × I. graellsii), second backcross with I. elegans (2 EB; 1 EB × I. elegans), third backcross with I. elegans (3 EB; 2 EB × I. elegans), and forth backcross with I. elegans (4 EB; 3 EB × I. elegans). We then evaluated the admixture proportions (± 90% credible intervals) of these artificial crosses with STRUCTURE in the same way as done for the I. elegans samples from Spain (above). To determine the level of introgression of I. graellsii into the Spanish I. elegans populations, the individual admixture proportions of the I. elegans samples from Spain were compared to the admixture proportion for the artificial hybrids and backcrosses.

Three to six damselflies from sympatric and allopatric localities were amplified by polymerase chain reaction (PCR) for part of the mitochondrial Cytochrome C Oxidase I and II genes (COI-COII), part of the mitochondrial Cytochrome B (CYTB) gene and part of the mitochondrial 12S rRNA (12S) and NADH Dehydrogenase 1 (ND1). The amplification was done using universal primers: 591 bp of the COI with the primers COI-H (5'-TCAGGGTGACCAAAAAATCA-3') and COI-L (5'-GGTCAACAAATC ATAAAGATATTGG-3') (Marina Magaña Ramos, personal communication), 673 pb of the COII with the primers TL2-J-3037 (5'- ATGGCAGATTAGTGCAATGG-3') and C2-N-3494 (5'-GGTAAAACTACTCGATTATCAAC-3') and C2-J-3400 (5'-ATTGGACATCAATGATATTGA-3') and TK-N-3785 (5'-GTTTAAGAGACCAGTACTTG-3') 34 , 457 pb of the CYTB with the primers CB-J-10933 (5'-TATGTACTACCATGAGGACAAATATC-3') and TS1-N-11683 (5'-TATTTCTTTATTATGTTTTCAAAAC-3') 34 , 370 bp of the 12S with the primers SR-J-14233 (5'-AAGAGCGACGGGCGATGTGT-3') and SR-N-14588 (5'-AAACTAGGATTAGATACCCTATTAT-3') 30 , and 591 bp of the ND1 gene with CB-J-11545 (5'ACATGAATTGGAGCTCGACCAGT-3') and N1-N-12051 (5'-GATTTTGCTGAAGGTGAATCAGA-3') 34 . DNA amplification was done in a total reaction volume of 20 μl. The amplification conditions were as follows: 50 ng of DNA (1 μL), 1 unit (0.2 μL) of Taq DNA polymerase (Ecogen), 2 μL 10x of reaction buffer (Ecogen), 0.5 μL of MgCl₂ (50 mM) (Ecogen), 0.5 μL of dNTPs Mix Sigma (200 μM), and 1 μL of each primer (10 pmol). All PCR reactions were completed in a "GeneAmp PCR system 2700" thermocycler (Applied Biosystems). The PCR program had an initial cycle of 95°C for 3 min, the annealing temperature for 1 min, and an elongation period at 72°C for 45 s, followed by 34 cycles at 95°C for 30 s, with annealing for 45 s, and an elongation phase at 72°C for 45 s, and a final extension phase at 72°C for 10 min. PCR products were sent to an external sequencing service (University of Valencia) where bidirectional sequencing reactions were conducted using Bigdye™terminator cycle sequencing kit (Applied Biosystems) using automatic sequencer ABI3100 (Applied Biosystems). Forward and reverse sequences were edited in Codon Code Aligned (CodonCode, Dedham, MA, USA) and consensus sequences were aligned with Clustal X 71 implemented in MEGA v. 4.0 72 73 . Variable positions were revised by eye, and only high quality sequences were considered.

Samples of both species were pooled in two groups, depending on the sympatric or allopatric origin. For each gene, genetic diversity was assessed in terms of number of haplotypes (S), and haplotype diversity (H) and nucleotide diversity (π) according to Nei 74 with DNASP V4.10 74 . Further analyses could not be done because overall low genetic diversity and shared haplotype frequencies.

We reviewed female morph frequencies from the literature in seven I. graellsii populations from Iberia and northern Africa and seven I. elegans populations from Iberia 41 . In addition, we estimated female colour morph frequencies in four sympatric populations; Xuño and O Vilar from north-western Spain, and Alfaro and Las Cañas from north-central Spain. Populations were visited between June and September during sunny days, in two different years (2006 and 2007) and the sampling was done with entomological nets. Only single, solitary and mature females were used, and colour morph frequencies were estimated from the number of each morph divided by the total number of females.