Testing of microsatellite multiplexes for individual identification of Cape Parrots (Poicephalus robustus): paternity testing and monitoring trade

Ethics

Ethical clearance for this study was received from the University of KwaZulu-Natal Animal Ethics sub-committee (Ref numbers: 074/13/Animal, 017/14/Animal, and 042/15/Animal). All sampling procedures followed the criteria laid out by this committee.

Sampling and DNA extraction

Samples were collected from 76 captive Cape Parrots (Table S1). This includes samples taken from one international and five South African breeding facilities. The captive specimens included in this study comprise 22.3% of the Cape Parrot regional studbook (Wilkinson, 2015). The majority of these samples were sourced from one breeding facility, which holds the largest captive Cape Parrot-breeding populations in the world. Five of the specimens included in this study were wild-caught birds that were recently introduced into the captive breeding program. These five birds originated from the KwaZulu-Natal (KZN) Province. To test the utility of the molecular markers, captive birds with known pedigree were included. Both parents of 31 specimens were sampled, with only the sire sampled for seven of the captive-bred birds (Table S1).

Whole blood was collected from 45 captive Cape Parrots using Whatman™ FTA™ Elute cards (Buckinghamshire, UK) and was stored at room temperature in a dark cool storage area. Seventeen samples were whole blood stored in absolute ethanol at −20 °C. Samples were also collected from deceased birds provided by two breeding facilities (n = 14). Biopsies of 5 mm × 5 mm were collected from each carcass and stored in absolute ethanol at −20 °C.

DNA extraction from the Whatman™ FTA™ Elute cards (Buckinghamshire, UK) was performed following the manufacturer’s protocols. The DNA extraction from the tissue and whole blood stored in absolute ethanol was performed with the NucleoSpin® Tissue kit (Macherey-Nagel, Düren, Germany), following the manufacturer’s protocols. All DNA extracts were stored at −20 °C. DNA quantity was established via NanoDrop® ND-1000 Spectrophotometer (Thermo Fisher Scientific, DE, USA) analysis.

To assess any genetic differences between captive-bred and wild Cape Parrots, 85 genotypes from wild Cape Parrot populations were taken from Coetzer (2015). This study assessed the phylogeographic relationships between wild Cape Parrot populations. Strong genetic sub-structuring was observed, with three distinct genetic clusters linked to three geographical regions within the Cape Parrot distribution range (Coetzer, 2015). The wild data set consisted of 52 genotypes from the south genetic cluster (Eastern Cape region), 19 from the central cluster (KZN region), and five genotypes from the north cluster (Limpopo region). Details on these specimens are provided in Table S2.

Microsatellite amplification

The 16 microsatellite loci selected for this study were specifically developed for use in Cape Parrots (Pillay et al., 2010). The markers were divided into six multiplex sets (multiplex 1: Prob06, Prob15, and Prob26; multiplex 2: Prob30 and Prob36; multiplex 3: Prob18, Prob25, and Prob31; multiplex 4: Prob01, Prob09, and Prob17; multiplex 5: Prob23 and Prob28; multiplex 6: Prob29, Prob34, and Prob35). In-silico testing of all multiplexes was done prior to PCR amplification to test for the presence of primer dimers and hairpin structures using the program AutoDimer v1 (Vallone & Butler, 2004). The selected multiplex combinations did not show any signs of primer dimers or hairpin structures among the primer pairs. The forward primer in each microsatellite pair was fluorescently labeled on the 5′ end. The KAPA2G Fast Multiplex mix (KAPA Biosystems, Wilmington, MA, USA) was used for all amplifications, with each PCR reaction mixture consisting of: ∼2–60 ng template DNA, 5 μl KAPA2G Fast Multiplex mix (KAPA Biosystems, Wilmington, MA, USA), 0.2 μM of each primer and dH₂O to give a final reaction volume of 10 μl. The annealing temperature prescribed by the KAPA2G Fast Multiplex kit’s manufacturers was initially tested and provided positive results for all multiplex sets. The PCR cycling conditions consisted of an initial denaturation step for 3 min at 94 °C followed by 30 cycles at 94 °C for 30 s, 60 °C for 30 s, 72 °C for 30 s, and a final extension step for 5 min at 72 °C. The whole PCR setup, excluding the DNA addition step, was performed in a DNA-free area.

The amplified PCR products were sent to the Central Analytical Facility at Stellenbosch University, South Africa, for fragment analysis, using a standard ROX 500 internal size standard. The software package GeneMarker® v2.4.0 (Soft Genetics, State College, PA, USA) was used for all genotype scoring. We reamplified 20% of the data set to check for genotype consistency.

Data analysis

Parentage testing of captive population

The eight selected microsatellite panels were evaluated by testing the accuracy of each panel for parentage assignments using captive specimens with both known and unknown pedigrees. For this analysis, the full-pedigree maximum likelihood method implemented in COLONY v2.0.4.6 (Jones & Wang, 2010) was used. This program compensates for genotyping errors and null alleles (Wang, 2004) and has been used previously to identify parentage in captive (Ferrie et al., 2013; Loughnan et al., 2015) and wild vertebrate populations (Masello et al., 2002; Riehl, 2012; Bergner, Jamieson & Robertson, 2014). The offspring data set consisted of 38 individuals. For seven of these, only the paternal parent was known. A monogamous mating system with no inbreeding was selected, using the full-likelihood method. A medium run length with no prior sibship selected. The marker type and null allele frequencies for each locus were uploaded with an error rate of 0.02 as suggested by Wang (2004). We uploaded the genotypes of 38 offspring, 30 paternal candidates, and 21 maternal candidates, with the probability of the sire or dam included in the data set at 0.75 and no paternal or maternal exclusion information.

Power of microsatellite panel to detect origin of illegally traded birds

A partial Bayesian exclusion approach (Rannala & Mountain, 1997) as implemented in GenClass2 (Piry et al., 2004) was used to further assess the eight microsatellite panels. This method estimates the likelihood that a test sample belongs to one of the reference populations provided for analysis and calculates a sample exclusion probability for each of the reference populations (Ogden & Linacre, 2015). The use of assignment methods to identify the population or area of origin of confiscated wildlife is a well-known tool in wildlife forensics (Alacs et al., 2010; Ogden & Linacre, 2015). The partial Bayesian assignment analysis implemented in GeneClass2 is a well-established method for assignment of specimens to their population of origin (Primmer, Koskinen & Piironen, 2000; Grobler et al., 2005; Pruett et al., 2010). All wild born individuals were excluded from the captive population reference data set. To simulate an assignment study, six captive-bred and six wild-caught individuals were selected at random for the “samples to be assigned” data set. The captive population from the current study and the three wild populations from Coetzer (2015) were used as reference populations. The 12 individuals selected for the “samples to be assigned” data set were excluded from these data sets. The Bayesian method from Rannala & Mountain (1997) was followed, with probability computation using Monte-Carlo resampling and the simulation algorithm from Paetkau et al. (2004). The number of simulated individuals was set at 100,000, with the Type I error set at 0.01 and the assignment threshold at 0.05. These parameters were used for each of the eight microsatellite panels.

Captive vs wild Cape Parrots

The genetic diversity of the captive bred sample group was compared to the three wild Cape Parrot populations identified in the recent phylogeographic study (Coetzer, 2015). Values compared included the average number of alleles (N_A), number of private alleles (P_A), observed heterozygosity (H_O) and unbiased expected heterozygosity (uH_E) estimated in GenAlEx, inbreeding coefficient (F_IS) using Genepop v4.2 (Rousset, 2008) and Ar using rarefaction in FSTAT (Goudet, 2001). A Wilcoxon signed-ranked test was performed to assess the level of genetic differences between the captive population and the three wild populations using the per locus estimates for each of the N_A, H_O, uH_E, F_IS, and Ar estimates. Pairwise F_ST values and analysis of molecular variance (AMOVA) were estimated to assess if the captive population constitutes a separate genetic unit using GenAlEx. Bonferroni correction was implemented to all p-values to correct for problems with multiple testing (Rice, 1989). In this analysis, the five wild born individuals were removed from the captive data set.

Marker analysis

For this study, 76 captive Cape Parrots were successfully genotyped across 16 microsatellite loci. Loci were amplified across a range of DNA template concentrations, with low template concentration (2–5 ng/μl) successfully amplifying with minimal signs of allelic dropout (<3% over all loci). The replicate genotypes did not show any signs of discrepancies. Co-amplification of each multiplex was also highly successful, with the most amplification failures (4 of 76) observed for the locus with the largest bp size (Prob17). The data set contained less than 1 missing data, with a mean null allele frequency over all loci and samples of 0.039. The per locus null allele frequencies ranged from 0 to 0.186 (Table 1). Only two loci showed null allele frequencies above 0.1 (Prob15, Na = 0.186; Prob36, Na = 0.125). The mean number of alleles per locus varied greatly among loci, ranging from 1.75 (Prob36) to 17 (Prob17) alleles. A large difference in Ar values was observed across the loci, with values ranging from 2 (Prob36) to 22 (Prob17). Seven loci showed high levels of heterozygosity (Table 1), with negative F_IS values. Only two loci showed signs of heterozygote deficiency (Prob09, F_IS = 0.439; Prob36, F_IS = 0.471). Fourteen of the 16 loci were moderately to highly informative, with values ranging from 0.415 (Prob29) to 0.888 (Prob17). Only two loci (Prob35 and Prob36) were identified as uninformative (PIC < 0.3; Table 1). The P_ID values ranged from 0.019 (Prob17) to 0.591 (Prob36). A combined P_ID over all 16 loci was calculated as 1.831E−13 following the product rule. The P_E2 ranged from 0.658 (Prob17) to 0.032 (Prob36), with the combined P_E2 at 0.995. It was observed that the P_ID and P_E2 values improved as the number of loci analyzed increase (Fig. 1). Five of the 16 loci significantly deviated from HWE (Prob09, Prob15, Prob28, Prob30, and Prob36), following Bonferroni correction (p < 0.003). Of the 120 per locus comparisons made during the linkage disequilibrium (LD) analysis, more than half of the locus pairs showed signs of LD (52.5%).

Parentage analyses

All eight microsatellite panels showed very low combined P_ID values, with moderate to high informativeness levels (PIC range: 0.581–0.703; Table S4). The P_ID values for the eight panels ranged from 1.8E−13 for the 16 locus panel to 5.7E−10 for the nine locus panel (Table S3). These values suggest that 1 in 5.5E+12 (16 loci) to 1.8E+9 (nine loci) randomly chosen individuals will share the same genotype. The assessment from this parameter alone suggests that any of these panels could be suitable for forensic use, as the total number of wild Cape Parrots does not exceed 1,600 individuals. The ability of these eight panels to successfully identify known parents, however, differed. The seven larger panels were generally equally successful in identifying parent pairs and individual parents, with only slight differences in the mean probability values and a slightly higher sire identification success rate for the 10 locus panel (Fig. 2; Table S4). The nine-locus panel was less successful in correctly identifying parent pairs, with only 71% of parent pairs correctly identified with high probability (p > 0.75). The nine-locus panel also showed a lower success rate at identifying the correct sires and dams, with 73.7% of sires and 96.8% of dams correctly identified with high probability (p > 0.75). All known dams were correctly identified using the seven larger panels. Although the seven larger panels had similar assignment success rates (parent pair assignment success = 83.9%), the full 16 locus panel had overall higher mean probability rates making this panel most suited for use in future parentage analyses.

The majority of the test individuals were assigned to the correct population of origin following the partial Bayesian exclusion analyses using the eight microsatellite panels (83.33%–66.67%). The highest assignment success was achieved with the six larger microsatellite panels (16 loci to 11 loci), with 83.33% of the specimens correctly assigned (Fig. 3; Table S5). The 12 locus panel had the best average assignment probability value out of the eight tested panels (average assignment probability = 0.565, SE = 0.087), with 5 out of the 10 individuals correctly assigned with assignment probabilities above 0.6. The remaining five individuals were assigned to the correct populations with assignment probabilities lower than 0.6 (assignment probability = 0.170–0.591; Table S5). The two individuals (FH12 and FH32) that were incorrectly assigned, were sampled from the Eastern Cape but assigned to the captive (FH12) and KZN (FH32) populations. The assignment probabilities of these individuals did, however, differ only slightly between the actual population of origin and the assigned population (Table S5).

Genetic diversity: captive vs wild populations

The genetic diversity estimates for the captive data set, using 16 microsatellite loci, were largely similar to that observed for the wild Cape Parrot populations (Table 2). Significant differences were observed between the captive/central and captive/north per locus Na estimates (p < 0.003). The average number of alleles observed for the captive data set was higher than that observed in the wild population (captive, N_A = 6.813; southern, N_A = 6.563; central, N_A = 5.313; northern, N_A = 2.875; Table 2). The significant differences observed for the Na estimates could, however, be influenced by differences in sample size (captive born samples, n = 71; south, n = 60; central, n = 20; north, n = 5). The Ar estimates provide a more accurate estimation, with no significant difference observed among the captive and wild per locus Ar estimates (p < 0.003). The captive data set did, however, have the highest number of private alleles (P_A = 21), which was almost double that of the highest value observed among the wild populations (southern, P_A = 13). All private alleles occurred at low frequencies, with private allele frequencies for the captive data set ranging from 0.007 to 0.100 and frequencies for the southern wild population ranging from 0.008 to 0.133. No significant differences were observed between the captive vs wild per locus observed H_O, unbiased expected uH_E and F_IS comparisons.

The H_O for the captive data set was only slightly lower than that observed for the three wild populations (captive, H_O = 0.591; southern, H_O = 0.605; central, H_O= 0.647; northern, H_O = 0.6), with an uH_E comparable to that observed for the south and central wild populations (captive, uH_E = 0.625; southern, uH_E = 0.632; central, uH_E = 0.635). A low positive F_IS value was observed for the captive data set indicating only slight inbreeding (F_IS = 0.054), with low heterozygote deficiency. Low genetic differentiation was found only between the captive data set and the south population (F_ST = 0.017; p = 0.001). No significant genetic differentiation was observed between the captive and north populations (F_ST = 0.104; p = 0.004) or the captive and central populations (F_ST = 0.01; p = 0.024), following a Bonferroni correction (p = 0.003). The global F_ST value calculated for the captive and three wild populations did not significantly differ from zero (F_ST = 0.008; p = 0.008). AMOVA indicated that 92% of the observed genetic variance occurred within individuals, with 5% of the genetic variance between individuals and only 3% among the populations.

Microsatellite evaluation

The high LD levels observed in the current study was, however, not observed during the phylogeographic analysis of Cape Parrots by Coetzer (2015). This could be explained by the population history of the captive population, as admixture, rapid population expansion, bottleneck events, and genetic drift can affect LD (Rogers, 2014). The selection of a suitable locus combination was therefore not based on these LD estimates. The null allele frequencies and expected heterozygosity values observed for the loci from the current study is generally comparable to that reported by Pillay et al. (2010). The majority of these loci were highly to moderately informative, following the current study. The Ar values from the current study only marginally deviated from the allele numbers reported by Pillay et al. (2010). Locus Prob15 was previously reported as Z-linked in Cape Parrots (Pillay et al., 2010). Two of the known females for the current study were, however, heterozygous at this locus. The same trend was observed by Taylor (2011) who found no evidence of sex linkage of Prob15 in other Poicephalus species, including P. fuscicollis fuscicollis and P. f. suahelicus. The level of variation and informativeness observed in the current study is comparable to that observed in other studies. The PIC values observed in the current study fall within the same range as the values observed by Klauke, Segelbacher & Schaefer (2013) during a study investigating the breeding system of the endangered El Oro parakeet from southwest Ecuador. The combined P_ID observed by Russello et al. (2007) for 14 loci used in the South American Monk parakeet was lower than that observed for Cape Parrots in the current study.

Microsatellite loci for parentage analysis

The locus informativeness analyses performed on the 16 microsatellite loci allowed for the ranking of the 16 loci according to their level of informativeness. The full locus set of 16 markers showed to be the best combination of markers for parentage analysis of the eight panels tested. This panel had the highest average assignment probabilities for parent pair, sire, and dam assignment tests (Fig. 2). This panel is highly suited for individual-level identification, with a P_ID value suggesting that 1 in 5.5E+12 individuals will share the same genotype. All the known dams were positively identified with this panel, but only 31 of the 38 offspring’s known sires were identified with high probability. It was observed that the 10 locus panel did have a better assignment rate for the known sires, with 32 of the 38 sires identified, but the average assignment probability value was much lower than that observed for the 16 locus panel (Fig. 2). Similar success rates were observed by Labuschagne et al. (2015) when assessing the utility of 12 microsatellite loci for parentage assignment in the African Penguin (Spheniscus demersus). The failure to assign a parent, or assignment of a parent with low probability, can be linked to the occurrence of amplification errors during PCR causing allelic dropout, null alleles, stuttering, or polymerase slippage (Buckleton, Triggs & Walsh, 2005; Dewoody, Nason & Hipkins, 2006; Ferrie et al., 2013). It is possible to compensate for such errors in the parentage analysis program COLONY by importing the expected error rates of each locus, including null allele frequencies, prior to analysis. When this was done in the current study, failed or incorrect assignments were still observed and it is advisable to not only rely on genetic data, but also make use of a complete studbook of legally registered captive bred birds, as suggested by Ferrie et al. (2013). It is therefore important to compile a complete studbook of all captive bred Cape Parrots, complemented by a complete DNA database using the full 16 locus microsatellite panel described in this study. The inclusion of additional markers such as single nucleotide polymorphisms (SNPs) can improve the success rate of the parentage analysis, as demonstrated in Labuschagne et al. (2015).

Population of origin analysis

The assignment analysis performed with six of the eight microsatellite panels (16 loci–11 loci) all had a success rate of 83.33%. The average assignment probabilities of the correct assignments did however differ, with the 12 locus panel outperforming the rest (Fig. 3). The two Eastern Cape individuals were not assigned to their population of origin, with sample FH12 assigned to the captive population and FH32 assigned to KZN (Table S5). The probabilities that these two samples should be assigned to the captive and KZN populations were only marginally higher than the probabilities observed for assignment to the Eastern Cape population. This could be due to the occurrence of ancestral admixture, as it is shown that the southern (Eastern Cape) populations form a source to the central (KZN) populations (Coetzer, 2015), and the captive-bred populations in turn is largely sourced from the KZN populations (C. T. Downs, 2015, unpublished data). These individuals could therefore have ancestral links to individuals in the central (KZN) and the captive populations (via the KZN populations).

Captive vs wild Cape Parrots

The majority of the genetic differentiation estimates showed little to no genetic difference between the captive data set and the three wild Cape Parrot populations. Similar AMOVA results were observed in a study focusing purely on the genetic variation among the wild Cape Parrot populations (Coetzer, 2015). A clear difference was, however, observed for the private allele estimate. The captive data set contained almost double the number of private alleles observed in the southern wild population, which the recent phylogeographic study suggests is the most genetically diverse wild population (Coetzer, 2015). In theory, the higher number of private alleles in the captive population could be due to rare alleles, which are generally not often seen in the wild, being present in the founders of the captive populations. Gautschi et al. (2003) observed a similar trend in a captive bearded vulture (Gypaetus barbatus) population, with a higher level of genetic diversity in the captive population when compared to that observed in the wild population. This was linked to founder individuals, who are still present in the breeding population, carrying rare alleles and thereby passing these alleles down to their offspring (Gautschi et al., 2003). It is therefore possible that the captive Cape Parrots have not been in captivity for an appropriate amount of time to lose the observed rare alleles, and that these alleles are still being passed down to the new generations. New wild birds are also regularly introduced to the captive stock, through the addition of injured or confiscated wild birds (Wilkinson, 2015). These introductions could then also supplement the genetic diversity of the captive population, especially if the birds originate from different regions of the Cape Parrot’s natural distribution range. Many of the birds in the captive data set used in the current study are F1–F3 descendants from wild birds, and could therefore still carry these rare alleles.

The captive and wild birds are also subjected to different environmental forces, which can lead to genetic adaptation to captivity (Frankham, 2008). The difference in selective pressures such as a lack of predators, food availability, and pathogenic exposure could promote the selection of certain traits in captive animals, which would normally be detrimental in the wild (Lynch & O’Hely, 2001). It is possible for selection of certain rare, fitness-linked, loci to influence the genetic diversity of neutral loci like microsatellites, although it was observed to mostly decrease the genetic diversity of neutral loci (Montgomery et al., 2010). It could, therefore, also be argued that the large number of private alleles observed in the captive sample set could in some way be linked to the selection of rare alleles, due to human mediated mate selection of breeding pairs. Further analyses using fitness-linked loci, like the major histocompatibility complex (MHC) genes or toll-like receptor (TLR) genes, should be performed to better understand the effects captive breeding has on the genetic health of the Cape Parrot population.