Results

DNA extraction and amplification of mitochondrial D-loop region
DNA was extracted from 971 giant panda fecal samples, with a roughly equal sample number from each of the three nature reserves (HZG: 322, MG: 343, MB: 306). We successfully extracted DNA from 731 fecal samples (HZG: 275, MG: 230, MB: 226; Figure S1 for partial electrophoresis).
In this study, the mitochondrial D-loop region was used for genetic diversity assessment, and for fecal DNA quality assessment. The quality of extracted fecal DNA was evaluated by PCR amplification of a 750bp fragment in the mitochondrial D-loop region of the giant panda, to determine the freshness of collected feces. We successfully amplified and sequenced 686 DNA samples equally divided between the three reserves (HZG: 240, MG: 228, MB: 218; Figure S2 for partial electrophoresis).
Individual identification
We found that the minimum number of loci to have greater than 99% confidence of individual identification was six. Analyses determined that six microsatellite combinations were most effective for individual identification for the three populations, calculating PID (sib) 0.00911 and 0.00964 (Table 3; Figure 3). According to Waits et al. (2001), a PID less than 0.01 is required to evaluate population size. PID (sib) avoids errors associated with PID and provides a conservative upper estimate of the number of loci required to identify individuals.
Individual identification using microsatellites determined that the number of giant pandas in the three reserves was 27 (MB), 22 (MG) and 43 (HZG). Two of the reserves’ individual numbers were higher than the fourth survey (18 (MB), 22 (MG) and 29 (HZG)). Identification of larger populations in MB and HZG may have two causes, not necessarily mutually exclusive. The first being that different methods were used for our study and the fourth survey. The fourth survey predominantly used the “Distance-Bamboo Stem Fragments Method” adopted in the third survey (Shi et al., 2016). The fourth survey did employ non-invasive DNA quantity survey technology, but it was only used as an auxiliary survey. Therefore, the limitations of the fourth survey may have underestimated the number of giant pandas. Secondly, the population’s emigration, immigration, births and deaths will may influence differences in population number estimation. For example, we detected two individuals from Meigu Nature Reserve in Heizhugou Nature Reserve.
Genetic diversity based on microsatellite markers
Seven microsatellite markers were successfully amplified from the three populations. Ninety-six alleles were detected at seven loci in 92 individuals from the three populations. There were 13 common alleles in the three populations. We identified five unique alleles in each of Mabian and Meigu populations (Figure 4). Similarly, these two populations had the same number of rare alleles (9). The numbers of unique and rare alleles in Heizhugou population were 15 and 13, respectively (Figure 4). These rare alleles are at risk of being lost due to inbreeding or genetic drift.
The number of alleles at each locus ranged from 1 to 10. The average number of alleles in Heizhugou population was the largest, followed by Meigu population and Mabian population. The average observed heterozygosity (Ho) of the three populations was 0.632 (MB), 0.598 (MG) and 0.466 (HZG), the average expected heterozygosity (He) was 0.577 (MB), 0.502 (MG) and 0.555 (HZG), and the polymorphic information content (PIC) was 0.514 (MB), 0.441 (MG) and 0.508 (HZG), respectively (Table 4). Therefore, the three populations showed a moderate-low level of genetic diversity. The HWE test results showed that four of the seven microsatellite loci in Mabian population deviated from HWE (P < 0.01), while three in Meigu population and two in Heizhugou population (Table 4). A positive mean inbreeding coefficient (Fis) value was found in Heizhugou population (Table 4). High inbreeding coefficient suggests a heterozygote deficiency due to inbreeding. Our results are similar to Guan et al. (2009), who concluded that their observed HWE deviation was due to inbreeding and genetic drift.
Genetic diversity based on mitochondrial control region sequence
We successfully sequenced the mitochondrial D-loops from 85 of the 92 individuals from the three reserves, with sequencing peaks shown in Figure S3. The number of mitochondrial D-loop sequences (n), haplotypes (H), variation sites (s), haplotype diversity (h), and nucleotide diversity (π) of the three populations are summarized alongside other wild and captive populations in Table 5. Compared to other populations, the mitochondrial genetic diversity of giant pandas in these three reserves was significantly lower than in wild populations from Qinling, Minshan and Qionglai Mountains. Mitochondrial genetic diversity of three populations was also lower than captive populations from Wolong, Chengdu and Shaanxi, but higher than Daxiangling and Xiaoxiangling populations.
Geographic isolation and genetic differentiation
According to the distribution map of giant panda fecal samples (Figure 2), feces collected in Heizhugou and Meigu Nature Reserves were often in close proximity to the border between the two reserves. Samples collected in Mabian Nature Reserve were far away from collection sites in Meigu and Heizhugou Nature Reserves.
The software STRUCTURE (Pritchard, Stephens & Donnelly, 2000) was used to analyze the population genetic structure. Our results showed that when K=2, the value of △K peaked and decreased with increasing values of K. As shown in Figure 5, the giant pandas of three reserves were clearly divided into two genetic structural units, Heizhugou and Meigu populations formed a genetic structural unit, while the Mabian population formed a relatively independent genetic structural unit. Thus, the two genetic structural units indicated that the gene exchange between Heizhugou and Meigu populations was more frequent than with the Mabian population.
The Fst of giant panda population pair-wise comparisons from the three reserves was calculated and measured by GenALEx 6.5 (Peakall and Smouse, 2012). We used the Fst to represent interpopulation differentiation (Zeng, 2014). Studies have shown that if the range of Fst is 0.00-0.05, the genetic differentiation between populations is small and can be ignored. If Fst is between 0.05 and 0.15, there is a moderate degree of differentiation and between 0.15 and 0.25 indicates a high degree of differentiation (Wright, 1972; Liu, 2012). The results showed that there was a significant genetic differentiation between the three giant panda populations, with Fst ranging from 0.0756 to 0.1588 (Table 6). The Mabian population had a significantly higher degree of genetic differentiation with Meigu and Heizhuguo population, while there is a moderate degree of differentiation between Heizhuguo and Meigu population.