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

Bottleneck detection To test the evidence of recent population bottleneck, a comparison of the He of a sample with the heterozygosity that would be expected of a sample from a population at mutation drift equilibrium with the same size and allele number (Heq), was carried out to investigate the sig- natures of bottleneck events [38]. As the allele number decreases faster than the heterozygosity, a bottleneck is indicated by He being greater than Heq in subsequent generations [39]. Tran- sient heterozygosity excess and mode-shift tests were performed through 1000 simulations using the BOTTLENECK 1.2.02 software package [40], assuming that mutations of the micro- satellite loci followed an infinite allele model (IAM), a stepwise mutation model (SMM) or a two-phase model (TPM). In TPM, it was assumed that 70% of the mutations consist of subsist of one step and 30% consist of multistep changes with a variance of 30. To determine if the het- erozygosity excess was significant, the sign test, standardized differences test and Wilcoxon test were performed. The allelic distribution was determined to see if it was approximately L- shaped curve (as expected under a mutation-drift equilibrium) or not (recent bottlenecks pro- voke a mode shift) [39, 41]. Results Distribution pattern of hog deer in the Park Faecal pellets were recorded in 22.34 ± 0.20 km2 area in the park. The high-density area (9.31 ± 0.30 km2) had 15.9 ± 1.0 pellets km-2, the medium-density area (8.10 ± 0.40 km2) had 6.66 ±1.30 pellets km-2 and the low-density area (4.8 ± 0.6 km2) had 2.61 ±0.15 pellets km-2.

The overall pellet density was 0.34 ± 0.02 pellets km-2. The hog deer habitat usage was best explained by the combination of phumdi thickness, vegetation cover and abundance of short grass (R2 = 0.423; F = 11.475, df = 50, p < 0.001). The estimated coefficient of habitat use showed that the preference for a specific habitat by hog deer increased with increasing phumdi thickness (β = 0.087 ± 0.05), decreasing vegetation cover (β = -0.142 ± 0.044) and increasing abundance of short grass (β = 0.124 ± 0.048). In other words, it showed that hog deer favoured areas with thick phumdi, less vegetation cover and dominated by short grasses. Population trend Out of the eight models tested during the analysis using DISTANCE 6.0 software [42], based on the lowest AIC value, we used uniform cosine model to derive density of hog deer in the park “S2 Table”. The encounter rates were 0.36 ± 0.33, 0.22 ± 0.20 and 0.24 ± 0.20 individuals km-2 during 2006–2008 respectively and the probabilities of detection were 0.38 ± 0.58, 0.31 ± 0.29 and 0.30 ± 0.25, and the effective detection radius (EDR) values were 1.92 ± 0.35, 1.19 ± 0.17 and 1.25 ± 0.17, respectively, during the study period. The mean cluster sizes were 1.53 ± 0.10, 2.29 ± 0.18 and 2.02 ± 0.18 respectively, during the study period (Table 1). The estimated densities were 2.94 ± 0.57 (CV 19.5%), 2.75 ± 0.44 (CV 16.3%) and 2.51 ± 0.40 (CV 16.2%) individuals km-2 respectively, with a minimum of 1.82 individuals km-2 and a maxi- mum of 4.32 individuals km-2 at 95% confidence level during the study period. The population

size was 65 ± 12.6, 61 ± 9.9 and 57 ± 9.2 individuals, with a minimum of 44, 44 and 41 and a maximum of 96, 84 and 79 individuals at the 95% confidence level during 2006–2008 respec- tively (Table 1). The maximum number of sightings was recorded between 0600 and 0630 hours at distances between 100 m and 200 m (Fig 1A and Fig 1B). Results of the regression analysis for the period of 2003–08 (p < 0.05, R2 = 0.939) showed a 17% decline in population (Fig 2A). For the period 2006–08 (p < 0.05, R2 = 0.992) the decline was 8% annum-1 (Fig 2B). 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 5 / 16