Population genomics of the critically endangered kākāpō

Abstract

The kākāpō is a flightless parrot endemic to New Zealand. Once common in the archipelago, only 201 individuals remain today, most of them descending from an isolated island population. We report the first genome-wide analyses of the species, including a high-quality genome assembly for kākāpō, one of the first chromosome-level reference genomes sequenced by the Vertebrate Genomes Project (VGP). We also sequenced and analyzed 35 modern genomes from the sole surviving island population and 14 genomes from the extinct mainland population. While theory suggests that such a small population is likely to have accumulated deleterious mutations through genetic drift, our analyses on the impact of the long-term small population size in kākāpō indicate that present-day island kākāpō have a reduced number of harmful mutations compared to mainland individuals. We hypothesize that this reduced mutational load is due to the island population having been subjected to a combination of genetic drift and purging of deleterious mutations, through increased inbreeding and purifying selection, since its isolation from the mainland ∼10,000 years ago. Our results provide evidence that small populations can survive even when isolated for hundreds of generations. This work provides key insights into kākāpō breeding and recovery and more generally into the application of genetic tools in conservation efforts for endangered species.

Keywords: bottleneck; conservation; inbreeding; kākāpō; mutational load; purging.

Graphical abstract

Figure 1

Sampling locations, population structure, and past demography of kākāpō (A) Sampling locations for historical and modern specimens on a map showing the vegetation cover circa 1840. (B) Principal-component analysis (PCA) for 14 mainland and 35 Stewart Island kākāpō. Asterisks indicate museum samples likely to have been mislabeled (STAR Methods). (C) Demographic history and divergence time between the mainland and Stewart Island population inferred using the PSMC method. Each colored curve represents an individual bird. The black dashed curve represents the sex chromosome comparison (i.e., Z chromosome), with population size reaching infinity at the time of divergence between the two populations. (D) Parameter estimates for a scenario of post-glacial population divergence and expansion using ABC. See also Figures S5–S10 and Tables S2 and S3.

Figure 2

Heterozygosity and inbreeding estimates for kākāpō (A) Individual genome-wide heterozygosity. Horizontal lines within boxplots and bounds of boxes represent means and standard deviations, respectively. Vertical lines represent minima and maxima. (B) Individual inbreeding coefficients estimated using ROH (F_ROH). Open bars show the total proportion of the genome in ROH ≥ 100 kb and solid bars show proportions in ROH ≥ 2 Mb. Bars extending from the mean values represent the standard deviation (Welch’s two-sample t test; ∗∗∗p < 0.001). Richard Henry is shown with a purple triangle. See also Figures S11–S15.

Figure 3

Mutational load estimates for kākāpō (A) Individual relative mutational load measured as the sum of all homozygous and heterozygous derived alleles multiplied by their conservation score (GERP score > 2) over the total number of derived alleles. (B) Number of loss of function (LoF) variants per individual. Horizontal lines within boxplots and bounds of boxes represent means and standard deviations, respectively (Welch’s two-sample t test; ∗∗∗p < 0.001). Vertical lines represent minima and maxima. (C) R_xy ratio of derived alleles for synonymous, missense, and LoF variants. R_xy < 1 indicates a relative frequency deficit of the corresponding category in Stewart Island compared to mainland kākāpō. Whiskers represent 95% confidence interval (CI). See also Figures S16–S22.

Figure 4

Forward simulations of demographic scenarios and impact on deleterious mutations (A) Simulated demographic scenarios representing a Stable scenario as control (N_e = 10,000), a Mainland scenario (LGM bottleneck and long-term N_e = 6,000), a Stewart Island scenario (LGM decline and long-term N_e = 1,000), and an Extreme decline scenario (LGM decline and long-term N_e = 100). (B) Additive genetic load calculated as the sum of selection coefficients for homozygous mutations plus the sum of selection coefficients multiplied by the dominance coefficients for heterozygous mutations. Black dots and whiskers show the means and 95% CIs for each demographic scenario. See also Figures S23–S27.