Sex chromosomes have evolved repeatedly across the tree of life and often exhibit extreme size dimorphism due to genetic degeneration of the sex-limited chromosome (e.g. the W chromosome of some birds and Y chromosome of mammals). However, in some lineages, ancient sex-limited chromosomes have escaped degeneration. Here, we study the evolutionary maintenance of sex chromosomes in the ostrich (Struthio camelus), where the W remains 65% the size of the Z chromosome, despite being more than 100 million years old. Using genome-wide resequencing data, we show that the population-scaled recombination rate of the pseudoautosomal region (PAR) is higher than similar-sized autosomes and is correlated with pedigree-based recombination rate in the heterogametic females, but not homogametic males. Genetic variation within the sex-linked region (SLR) (π = 0.001) was significantly lower than in the PAR, consistent with recombination cessation. Conversely, genetic variation across the PAR (π = 0.0016) was similar to that of autosomes and dependent on local recombination rates, GC content, and to a lesser extent, gene density. In particular, the region close to the SLR was as genetically diverse as autosomes, likely due to high recombination rates around the PAR boundary restricting genetic linkage with the SLR to only ~50Kb. The potential for alleles with antagonistic fitness effects in males and females to drive chromosome degeneration is therefore limited. While some regions of the PAR had divergent male-female allele frequencies, suggestive of sexually antagonistic alleles, coalescent simulations showed this was broadly consistent with neutral genetic processes. Our results indicate that the degeneration of the large and ancient sex chromosomes of the ostrich may have been slowed by high recombination in the female PAR, reducing the scope for the accumulation of sexually antagonistic variation to generate selection for recombination cessation.

Blood samples of Struthio camelus were obtained from Western Cape Department of Agriculture’s ostrich research facility in Oudtshoorn, South Africa. Since 1995, individuals have been bred in pairs at the research facility to create pedigrees. At the time of sampling, the pedigrees contained 1531 males and 2067 females. We selected 5 males and 5 females for sequencing using the program PedMine. PedMine identifies individuals with most distant links within pedigrees allowing the maximum amount of genetic diversity in populations to be sampled. All procedures were approved by the Departmental Ethics Committee for Research on Animals (DECRA) of the Western Cape Department of Agriculture, reference no. AP/BR/O/SC14. Samples were sequenced at Science for Life Laboratory, the National Genomics Infrastructure, using paired end with 126 base pairs on Illumina HiSeq 2500, following manufacturer's protocol. Reads were trimmed with cutadapt version 2.10 and then mapped to the optical map improved reference genome (Struthio_camelus.20130116.OM.fa.masked) with bwa version 0.7.17.r1188. Duplicates were marked with Picard MarkDuplicate. Variant calling was performed with GATK version 4.1.4.1 following best practice procedures developed at the Broad Institute. The GATK HaplotypeCaller was run individually on each sample to generate GVCF output. GVCF files for all samples were imported to a GenomicsDB datastore, followed by genotyping with GATK GenotypeGVCFs to produce a final raw variant call set. Several filtering steps were performed on the raw call set to obtain the final call set of high quality. Biallelic SNPs were selected with GAKT SelectVariants and filtered with GATK VariantFilteration using best practice options QUAL < 30, QualByDepth (QD) < 2.0, RMSMappingQuality (MQ) < 40.0, MappingQualityRankSumTest (MQRankSum) < -12.5, FisherStrand (FS) > 60.0, ReadPosRankSumTest < -8.0 and StrandOddsRatio (SOR) > 3.0. We removed variants overlapping with repeats annotated by the aves repeat library using BEDTools intersect. We filtered SNPs with more than twice the average coverage (>70 reads) and less than 5 reads per site. SNPs in the SLR in females are expected to occur only as haploid. However, heterozygous SNPs in the SLR in females can occur either due to genotyping error or due to the divergence of the Z and W sequences in the gametologous region. We therefore filtered the heterozygous SNPs in females in the SLR.