Twelve high-quality loci were tested in 30 individuals from the three Abies pinsapo populations studied (Table 1). Three of them (Pin14, Pin22 and Pin25) were monomorphic. The number of alleles in the polymorphic loci ranged from two to five (mean ± SE: 2.074 ± 0.150). Their observed and expected heterozygosities (H_o and H_e) ranged both from 0.100 to 0.600. Null allele frequency (N_a) ranged from 0.000 to 0.907. After Bonferroni correction for multiple comparisons (adjusted α = 0.002), no locus showed a significant deviation from expectations under Hardy-Weinberg equilibrium (HWE) and no linkage disequilibrium (LD) was detected for the 36 paired loci comparisons (Table 2).

Primer amplifications in other Mediterranean Abies species were successful in most cases. Seven primers amplified in all species, Pin8 did not amplify in any of them and the remaining four amplified at least in three species. For several loci, there was a lower number of alleles in A. pinsapo than in the other Mediterranean species, in spite of using fewer individuals for the latter (three individuals in A. nebrodensis and five individuals in the rest of them); see Table 3. This low number of alleles in Abies pinsapo could be due to a low effective population size [10].

Genomic DNA was extracted from leaf tissue from a single Abies pinsapo individual using a QIAGEN DNeasy Plant Mini Kit (QIAGEN, Valencia, CA, USA). A size selected DNA library was built and enriched for microsatellites using Dynabeads, as by Glenn and Schable [11] and sequenced on a 454 Genome Sequencer FLX System (454 Life Sciences, a Roche company, Branford, CT, USA) at the Savannah River Ecology Laboratory (Aiken, SC, USA). The 454 sequencing technique is described in detail in Abdelkrim et al.[12] and Lance et al.[13]. Development of microsatellites via “next generation sequencing” can be quicker, cheaper and it recovers more loci than enrichment methods. Moreover, this technique is more suitable for this purpose because of the large average size of the fragments obtained, which facilitates primer design [14]. CAP3 [15] was used to assemble sequences at 98% sequence identity using a minimal overlap of 75 bp. Sequence data were screened for di-, tri- and tetra-nucleotide repeats using the program MSATCOMMANDER version 0.8.1 [16] with minimum repeats set to eight, seven and six respectively, and primers were designed with Primer3 [17]. Two default 5′-tails (CAG: 5′-CAGTCGGGCGTCATCA-3′ and M13R: 5′-GGAAA CAGCTATGACCAT-3′) options were considered for designed primers, as with Faircloth [16], to allow a cheaper labeling technique than the direct labeling of the primers [18], as the inclusion of the 5′-tail allows the use of a third primer in the polymerase chain reaction (PCR) (M13R or CAG) that is fluorescently labeled for detection in the DNA Analyzer sequencer [19]. Additionally, a not-tagged primer from a primer pair was designed with a 5′-GTTT tail to promote adenylation and thus facilitate genotyping [20]. Primers could be designed for 497 of the 3617 sequences obtained. Out of these 497 primer pairs, we removed 196 with low quality (i.e., redundant, and with high likelihood of forming secondary structures such as hairpins and self or primer dimers). Therefore 301 primers with high quality were selected.

Preliminary screenings from 50 loci (out of the 301 selected) were done to test the amplification quality using two previously obtained high quality DNA samples. Of these, 31 loci amplified well, but 19 primer pairs produced unclear patterns in the electropherograms. Conditions for PCR amplification were an initial denaturation step of 5 min at 94 °C, followed by five cycles of 94 °C for 30 s, 60 °C for 30 s, and 72 °C for 30 s; 21 cycles of 94 °C for 30 s, 60 °C for 30 s (decreased by 0.5 °C per cycle), and 72 °C for 30 s; 15 cycles of 94 °C for 30 s, 50 °C for 30 s and 72 °C for 30 s; and a final step of 8 min at 72 °C. PCR reactions were carried out using approximately 50 ng of genomic DNA in a total volume of 20 μL, containing 2 μL of PCR Buffer 1× (0.2 mM MgCl₂), 2 μL of dNTP 0.1 mM (0.025 mM each), 0.5 μL of BSA (0.025 mg/mL), 0.2 μL of i-Start Taq DNA polymerase (1 U/μL) (iNtRON Biotecnology Inc., Sungnam, Korea), 1.25 μL of primer with 5′-GTTT tail (0.25 μM), 0.3 μL of primer with 5′-CAG or M13R tail (0.06 μM) and 0.5 μL of universal primer with 6-carboxyfluorescein (FAM), nitrobenzoxadiazolyl (NED), or VIC fluorescent label (0.25 μM). PCR products were run on a 3730 DNA Analyzer sequencer (Applied Biosystem, Foster City, CA, USA) and sized with LIZ 500 standard (Applied Biosystem, Foster City, CA, USA). Fragments were analyzed using GENEMARKER version 1.8 (SoftGenetics, State College, PA, USA).

Twelve high-quality loci were tested in 30 individuals, 10 from each of the three unique A. pinsapo localities (Sierra de Grazalema Natural Park, Sierra de las Nieves Natural Park, and Sierra Bermeja Natural Area). Cross-amplification tests for all loci were performed with the 11 remaining Abies species from the Mediterranean Basin (see Supplementary Information), using five samples per species across their distribution ranges (three individuals for the micro-endemic A. nebrodensis).

Observed heterozygosity (H_o), expected heterozygosity (H_e), null allele frequency (A_n) and Hardy-Weinberg equilibrium test (HWE) were obtained for each locus and population in CERVUS 3.0.3 software [21]. A test for genotypic linkage disequilibrium was conducted with GENEPOP 4.0.10 software [22].