Genomes of Galápagos mockingbirds reveal the impact of island size and past demography on inbreeding and genetic load in contemporary populations

Restricted range size brings about noteworthy genetic consequences that may affect the viability of a population and eventually its extinction. Particularly, the question if an increase in inbreeding can avert the accumulation of genetic load via purging is hotly debated in the conservation genetic field. Insular populations with limited range sizes represent an ideal setup for relating range size to these genetic factors. Leveraging a set of 8 differently sized populations of Galápagos mockingbirds (Mimus) we investigated how island size shaped effective population size (Ne), inbreeding, and genetic load. We assembled a genome of M. melanotis and genotyped 3 individuals per population by whole genome resequencing. Demographic inference showed that the Ne of most populations remained high after the colonisation of the archipelago 1 - 2 Mya. Ne decline in M. parvulus happened only 10 - 20 Kya, while the critically endangered M. trifasciatus showed a longer history of reduced Ne. Despite these historical fluctuations, the current island size determines Ne in a linear fashion. In contrast, significant inbreeding coefficients, derived from runs of homozygosity, were identified only in the 4 smallest populations. Index of additive genetic load suggested purging in M. parvulus, where the smallest populations showed the lowest load. By contrast, M. trifasciatus carried the highest genetic load, possibly due to a recent rapid bottleneck. Overall, our study demonstrates a complex effect of demography on inbreeding and genetic load, providing implications in conservation genetics in general and in a conservation project of M. trifasciatus in particular.

Methods
We extracted genomic DNA using the MasterPure DNA Purification Kit (Lucigen). PCR-free sequencing libraries were prepared by the Norwegian Sequencing Centre and sequenced on Illumina HiSeqX10 in a pair-end mode. We used Trimmomatic (v. 0.36, Bolger) to remove adapters and trim low-quality base pairs from raw reads. Quality-filtered reads of at least 120bp (minimum base quality Phred score of 15) were mapped onto our M. melanotis reference genome by BWA–mem (v. 0.7.15) (Li & Durbin 2009). We removed duplicated reads by Picard (v. 2.8.1). Variant calling was done by GATK (v. 3.7) (McKenna et al. 2010) with a series of stringent filtering steps to minimise the presence of false genotypes. Reliable biallelic SNPs were obtained following the GATK best practices for hard filtering as described in DePristo et al. (2011). We also excluded sites that either exceeded the average coverage depth per individual by more than two standard deviations, showed excess heterozygosity (p <0.05 as determined by VCFTools (v. 0.1.16; Danecek et al. 2011), or had more than 25% missing data. We also excluded sites found within repetitive regions and, finally, invariant sites with low quality (QUAL < 15). Such filtering resulted in a vcf file containing ~8 M bi-allelic SNPs