A total of 384 clones were isolated and sequenced bidirectionally. The Codoncode Aligner 3.7.1 software (CodonCode Corporation: Centerville, MA, USA) revealed a redundancy of 17% in the library. Of the unique clones selected for analysis in Microsatellite Repeats Finder [16], 176 (46%) had at least one microsatellite. A predominance of dinucleotide repeats (56%) was found in the motifs that make up the library. The CA_N/GT_N repeats (130 motifs identified) were the most numerous, followed by CT_N/GA_N (48 motifs). This predominance of CA_N/GT_N repeats is typical of the eukaryote genome [17]. Considerable numbers of other types of motifs were also recorded, in particular the dinucleotide AT_N/TA_N (36 motifs), the trinucleotides CAT_N/GTA_N (18 motifs), CTT_N/GAA_N (17), AAT_N/TTA_N and CTC_N/GAG_N (6 motifs each), and the tetranucleotides CTAT_N/GATA_N (13 motifs) and CATT_N/GTAA_N (6 motifs). These data provide a baseline that will support the development of additional probes for the isolation of new microsatellites in P. moratoi.

A total of 29 pairs of primers were designed and optimized successfully for the PCR amplification of microsatellite loci (Table 1). Of the loci analyzed, Pmoratoiμ1 presented several nonspecific amplifications even after optimization (with varying concentrations of magnesium chloride and different annealing temperatures) and was excluded. Six loci—Pmoratoiμ2, Pmoratoiμ3, Pmoratoiμ4, Pmoratoiμ9, Pmoratoiμ20, and Pmoratoiμ22—were monomorphic. With the exception of Pmoratoiμ22, all these monomorphic loci represent interrupted or interrupted compound microsatellites characterized by a small number of repetitions, with predictably low polymorphism [18]. Twenty-two microsatellites were polymorphic (Table 2) in at least some populations (Pmoratoiμ7, Pmoratoiμ8, Pmoratoiμ10, Pmoratoiμ11, Pmoratoiμ14, Pmoratoiμ17, Pmoratoiμ18, and Pmoratoiμ21). The Pmoratoiμ5 locus was not amplified in some populations, possibly due to local mutations in the primer annealing site. The identity analysis calculated using Cervus 3.0.3 [19] detected four pairs of specimens with identical genotypes (exact match), suggesting the recapture of the same animal during fieldwork. In these cases, the duplicate genotype was excluded from the analyses.

The total number of alleles per locus (N_A) varied between 2 and 18 (Table 2). Observed heterozygosity (H_O) ranged from 0.02 to 0.96, expected heterozygosity (H_E) from 0.02 to 0.87, and polymorphism information content (PIC) from 0.02 to 0.87. As might be expected from the relatively large number of markers developed for this study, evidence of linkage disequilibrium was found in three pairs of loci (Pmoratoiμ10-Pmoratoiμ25, Pmoratoiμ12-Pmoratoiμ15, and Pmoratoiμ15-Pmoratoiμ25) following Bonferroni correction (p < 0.002). Significant deviations from Hardy–Weinberg Equilibrium (HWE) were found in Pmoratoiμ24 and Pmoratoiμ27 from the São Carlos population and in Pmoratoiμ29 from Brotas, due to a deficit of heterozygotes. These deviations can be attributed to the presence of null alleles in these populations. The estimated null allele frequency for the Pmoratoiμ24 locus from São Carlos was 0.23, and that for Pmoratoiμ27 from this same population was 0.18. The estimated frequency for Pmoratoiμ29 from Brotas was 0.11.

In addition to the loci with deviations from HWE, null alleles were detected in the Pmoratoiμ13 (0.10), Pmoratoiμ23 (0.19), and Pmoratoiμ24 (0.10) loci from Bauru, and in Pmoratoiμ27 from Brotas (0.07). In all these cases, the evidence of the presence of null alleles was relatively weak and thus insufficient to confirm a significant departure from HWE following the Bonferroni correction. Micro-Checker 2.2.3 [20] did not detect small allele dominance, but found evidence of the presence of stutter bands in one locus, Pmoratoiμ13 from Bauru. The identity analysis indicated that the combination of all the loci would permit the individual identification of each of the specimens.

We constructed an enriched partial microsatellite genomic library using an approach based on the selective hybridization method of Kijas [21]. The library was constructed using DNA extracted from the muscle tissue of one specimen of P. moratoi using the procedure of Sambrook et al. [22] with modifications. Six micrograms of genomic DNA were digested with 50 units of Afa I (Invitrogen) and the fragments were then ligated to Rsa I linkers (Rsa21: 5′-CTCTTGCTTACGCGTGGACTA-3′/Rsa25: 5′-TAGTCCACGCGTAAGCAAGAGCACA-3′) using 2 units of T4 DNA ligase (Promega). The fragments were then amplified by polymerase chain reaction (PCR) with a reduced number of cycles (20 cycles) using the primer Rsa21. The PCR products were purified, denatured and hybridized with biotinylated microsatellite probes (GT₈ and CT₈) at room temperature for 20 min. The hybrid mixtures containing microsatellites were then collected by streptavidin-coated magnetic beads (Promega). The selected fragments were amplified via PCR and the products were ligated into a pGEM-T easy cloning vector (Promega). Escherichia coli XL1-Blue cells (Stratagene) were transformed with recombinant plasmids by electroporation and grown overnight in solid Luria-Bertani agar medium containing ampicillin, IPTG and X-Gal. The positive colonies were selected and grown in liquid medium with 2YT HMFM containing ampicillin. After growing for 16 h, they were stored at −80 °C.

Of the total of 596 clones obtained, 384 were sequenced bidirectionally in an ABI Prism 3100 automatic sequencer (Applied Biosystems: Foster City, CA, USA). The DNA sequences were exported into Codoncode Aligner 3.7.1 (CodonCode Corporation) which assembled the contigs and verified the redundancy of the library. The Bioedit program was used to check the quality of the sequences by chromatogram and to align them to form a consensus sequence. The repetitive elements were located using the Microsatellite Repeats Finder program [16]. After removal of the vector sequences, adapters, and restriction endonuclease sites by the Microsat software (version 1.0; CIRAD: Montpellier, France, 2005), the primers were designed using Primer 3 [23].

The polymorphic microsatellite markers were characterized by the amplification of the genomic DNA obtained from buccal epithelial cells (non-destructive method) following a modified version of the procedure described by Pidancier et al. [24]. Samples were obtained from five remnant P. moratoi populations in the Brazilian state of São Paulo: 41 samples were collected in the municipality of São Carlos (22°01′00.5″ S, 47°56′21.0″ W), 41 in Brotas (22°12′53″ S, 47°54′41″ W), 27 in Bauru (22°20′48.46″ S, 49°0′56″ W), 3 in Avaré (22°53.227′ S, 48°56.803′ W), and 1 in Lençóis Paulista (22°49′13.17″ S, 48°53′0.28″ W). The PCRs were prepared in a final volume of 15 μL containing 10 ng of the DNA template, 1× reaction buffer, 0.3 mM dNTP, 0.6–4.0 mM MgCl₂ (Table 1), 0.6 μM of each primer, and 1 unit of Taq polymerase (Invitrogen). The reactions were conducted following the same cycling conditions: 5 min at 94 °C followed by 41 cycles of 30 s at 94 °C, 1 min at the locus-specific annealing temperature (Table 1), and 1 min at 72 °C, followed by a final extension of 30 min at 72 °C to minimize stutter bands. The PCR products were analyzed in a Dual Dedicated Height Sequencing Kit (CBS Scientific) vertical electrophoresis system in 6% denaturing polyacrylamide gel and stained with silver nitrate [25]. Allele size was estimated by comparison with a 10 bp DNA ladder (Invitrogen) and using the GelAnalyzer 2010a software [26].

The levels of polymorphism of the microsatellites were evaluated as the number of alleles per locus (N_A), observed (H_O) and expected (H_E) heterozygosity calculated by Popgene 1.32 [27]. The polymorphism information content (PIC) was calculated with Cervus 3.0.3 [19], which was also used to conduct a test of individual discrimination (identity analysis). The Genepop 4.0.9 software [28] was used to detect an excess or deficiency of heterozygotes, linkage disequilibrium between pairs of loci, and deviations from the Hardy–Weinberg Equilibrium (HWE), for which significance levels were determined using the Markov chain algorithm [29], with 10,000 dememorization steps, 100 batches and 5000 iterations per batch. All significance levels were adjusted by the sequential Bonferroni correction for multiple tests [30]. Micro-Checker 2.2.3 software [20] was used to identify genotyping errors and verify the presence of null alleles using the Brookfield method [31].