Demographic and genetic structure of a severely fragmented population of the endangered hog deer (Axis porcinus) in the Indo-Burma biodiversity hotspot

estimation of observer-specific detection probabilities, and abundance estimation is carried out under the assumption that no animal goes undetected during the sampling periods. Dur- ing the exercise, details such as number of individuals sighted, time of sighting, sighting dis- tance, sighting angle, age, sex and habitat type were recorded. The animals were classified as adult male, sub-adult male, adult female, sub-adult female, juvenile and fawn [24, 25]. Statistical analyses The mean faecal pellet group density was calculated (number/km2) using MS Excel. The per- centage areas of the faecal pellet densities (high, medium and low) were also calculated. The area was then stratified according to the percentage area and faecal pellet distribution map was generated using the ArcView GIS software (9.1). The habitat data were pooled into a single matrix. Continuous independent variables were first evaluated for independence using Pear- son correlations. Multiple regression models were run using SPSS 16 [26] to find the best com- bination of habitat variables that predicted hog deer occurrence. The best combination was determined from the explained variance, overall F and regression diagnostics. The data were normalised using the Z-transform. The scaling of the habitat use can be represented by a linear model of the form Y = a + bx1 + cx2 + dx3. . ., where Y is the predicted density and xn repre- sents the habitat variables. The Distance 6.0 software package was used to derive the abundance [27]. The hog deer

density was calculated using maximum peak sightings for 30 minutes to avoid duplicate counts. The best-fitting model on the basis of the Akaike information criterion (AIC) [28] was used to derive the density. The exponential rate of increase was derived by regressing the natu- ral log (ln) of the population density (hog deer km-1) against year. The exponential rate of increase (r ˙ ¯) is the slope of the regression line. The percentage change per year was calculated as percent change = (er—1)�100 [29]. The population trend was calculated for the present study based on the data generated during this study (2006–08) and also from the data generated by the Manipur Forest Department for 2003–2018 “S1 Table”. Microsatellites genotyping The total genomic DNA was extracted from tissue hair, antlers and faecal samples (n = 38) using Kit (QIAGEN), QIAamp1 DNA Micro Kit (QIAGEN) and QIAGEN Stool Mini Kit, respectively, following the manufacturer’s protocol [30]. The DNA fragments were separated on an automated DNA sequencer (Applied Biosystems). Different individuals were genotyped as either homozygote or heterozygote on the basis of the band pattern shown for each micro- satellite locus. GeneMapper v.3.7 software was used to analyze the allele frequencies, heterozy- gosity and other parameters of genetic variability. To reduce the occurrence of allelic dropout, a multi-tube approached was employed [31]. Two repeats were performed, and if there was a heterozygote in three of four repeats, the site was considered to be a heterozygous locus. The

same procedure was followed for a homozygote loci [32]. The overall summary statistics were calculated for each locus: observed heterozygosity (Ho), expected heterozygosity (He), number of alleles per locus and effective number of alleles and polymorphic information content (PIC) were calculated using allelic frequencies in CER- VUS 3.0 [33]. The software CERVUS 3.0 was also used to detect the presence of any matching genotype between sampled individuals [33, 34]. The PID and PID-Sibs were also calculated using the same software [35, 36]. GENEPOP 4.7 software was used to derive the deviations from the Hardy-Weinberg equilibrium (HWE) [37]. Demographic and genetic structure of hog deer in the Indo-Burma biodiversity hotspot PLOS ONE | https://doi.org/10.1371/journal.pone.0210382 February 6, 2020 4 / 16