2.4.1 Assignment to hybrid classes: Hybrid Index (HI) and Heterozygosity (Het)
We used the R package INTROGRESS v1.2.3 (Gompert & Alex Buerkle, 2010) to calculate individual introgression coefficients; hybrid index (HI-values) and individual heterozygosity (HET-values), and used both of them to classify individuals into different hybrid classes (c.f. Jordan et al., 2018). INTROGRESS was used with the subset of 381 species-specific SNPs, as the assignment to hybrid classes can be inexact when using nondiagnostic markers (Buerkle, 2005). INTROGRESS when using species-specific allele SNPs, calculates the hybrid index as the proportion of alleles inherited from one species, and the heterozygosity, as the proportion of alleles that are heterozygous, ranging from 0 (pure species) to 1 (F1 hybrids) because of pure species are 100% homozygous, while F1 hybrids are 100% heterozygous (Gompert & Buerkle, 2010). Thus, the HI-value gives the proportion of alleles inherited from one species, in this caseI. elegans (e.g. 1.00=100% I. elegans , and 0.00 =100%I. graellsii , alleles), whereas HET-values, which range from 0.00 to 1.00 (0.00=all sites are homozygous, 1.00=all sites are heterozygous), indicate the timing of the hybridization event. First generation hybrids (F1 individuals) are expected to be heterozygous at all species-specific SNPs, while later-generation hybrids and backcrosses will have a lower heterozygosity levels, and the HI-values of F1 and F2 individuals will be close to 0.5, while backcrosses will have a HI-value below (or up to) 0.5 (Fitzpatrick, 2012). However, because of the lack of pure I. elegans and pure I. graellsii in the three hybrid regions [only 3 out of the 102 individuals were pure I. elegans (2) or pure I. graellsii (1)], criteria for F1 and F2 hybrid classes were relaxed. Thus, we classified individuals into eight parental and hybrid classes (cf., Milne & Abbott, 2008; Walsh et al., 2015): (i) pure I. elegans (HI=1.000; HET≤0.000), (ii) pure I. graellsii (HI=0.000; HET≤0.000), (iii) introgressed-elegans (HI=0.900-0.999; HET≤0.118), (iv) introgressed-graellsii (HI=0.001-0.100; HET≤0.118), (v) backcross-elegans (HI=0.601-0.899; HET=0.118-0.449), (vi) backcross-graellsii (HI=0.101-0.399; HET=0.118-0.449), (vii) relaxed F1 hybrids (HI=0.400-0.600; HET≥0.600), and (viii) relaxed F2 hybrids (HI=0.400-0.600; HET=0.450-0.599).
To investigate whether the hybrid regions have similar proportions of hybrid classes, the observed hybrid-class distributions of individuals from north-central and north-west (north-east was not included in the statistical analyses due to small sample sizes) were used to estimate the predicted distribution of individuals in admixture-classes using a contingency table (assuming random distribution) and then compared to the observed admixture-class distribution in both hybrid regions. Z-tests with Yates’s corrections for small sample sizes were used to test for differences in the proportions of each hybrid class category between north-west and north-central hybrid regions.
The twelve analyzed populations from the Spanish hybrid zone were assigned to three qualitative measures that reflects their genotypic compositions. This classification depends on the frequency distribution of the different hybrid classes: 1) unimodal hybridization pattern, when the distribution spans a range of admixture and backcrosses towards both parental species; 2) bimodal hybridization pattern, when the distribution is deviated to the two parental or parental-like genotypes, and few hybrids (F1 and F2 hybrids) are present (Jiggins & Mallet, 2000); and 3) introgressed hybridization pattern when the distribution was deviated to one parental-like genotype.