Meiotic drive supergenes are complexes of alleles at linked loci that together subvert Mendelian segregation resulting in preferential transmission. In males, the most common mechanism of drive involves the disruption of sperm bearing one of a pair of alternative alleles. While at least two loci are important for male drive- the driver and the target- linked modifiers can enhance drive, creating selection pressure to suppress recombination. In this work, we investigate the evolution and genomic consequences of an autosomal, multilocus, male meiotic drive system, Segregation Distorter (SD) in the fruit fly, Drosophila melanogaster. In African populations, the predominant SD chromosome variant, SD-Mal, is characterized by two overlapping, paracentric inversions on chromosome arm 2R and nearly perfect (~100%) transmission. We study the SD-Mal system in detail, exploring its components, chromosomal structure, and evolutionary history. Our findings reveal a recent chromosome-scale selective sweep mediated by strong epistatic selection for haplotypes carrying Sd, the main driving allele, and one or more factors within the double inversion. While most SD-Mal chromosomes are homozygous lethal, SD-Mal haplotypes can recombine with other, complementing haplotypes via crossing over, and with wildtype chromosomes via gene conversion. SD-Mal chromosomes have nevertheless accumulated lethal mutations, excess non-synonymous mutations, and excess transposable element insertions. Therefore, SD-Mal haplotypes evolve as a small, semi-isolated subpopulation with a history of strong selection. These results may explain the evolutionary turnover of SD haplotypes in different populations around the world, and have implications for supergene evolution broadly.

We sequenced haploid embryos using the following scheme: we crossed SD-Mal/CyO stocks to homozygous ms(3)K81 males and allowed them to lay eggs overnight. We inspected individual embryos under a dissecting scope for evidence of development and then isolated them for whole genome amplification using the REPLI-g Midi kit from Qiagen (catalog number 150043) and tested for the presence of Sd-RanGAP using PCR. We prepared sequencing libraries for Illumina sequencing with TruSeq PCR free 350bp.  To trim reads, we used Trimgalore v0.3.7 and the parameters: q 28 --length 20 --paired -a GATCGGAAGAGCACACGTCTGAACTCCAGTCAC -a2 GATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT --phred33 --fastqc --retain_unpaired -r1 21 -r2 21 --dont_gzip --length 20. Trimmed reads are available in SRA (Bioproject PRJNA649752, SRA accession numbers in Table S1). We sequenced a total of 10 SD-Mal genomes. 

For the Nanopore library, we extracted High-Molecular-Weight DNA from ~200 frozen female SD-ZI125/SD-ZI125 virgins, extracted DNA using a standard phenol-chloroform method and spooled DNA using capillary tubes. We constructed a library with ~1 ug DNA using RAD004 kit and the ultra-long read sequencing protocol. We sequenced the library using R9.4 flow cells and called bases with the ONT Albacore Sequencing Pipeline Software version v2.2.10.

For SNP calling, we mapped the Illumina reads from our SD-Mal libraries and the 20 SD⁺ libraries from the DPGP3 dataset to D. melanogaster (BDGP6) genome (ftp://ftp.ensembl.org/pub/release-88/fasta/drosophila_melanogaster/dna/; last accessed 6/25/20) using BWA mem (v0.7.9a). We removed duplicated reads with Picard (2.0.1) and applied the GATK (3.5) “best practices” pipeline for SNP calling. We did local realignment and base score recalibration using SNPs data from DPGP1 ensembl release 88 (ftp://ftp.ensembl.org/pub/release-88/variation/vcf/drosophila_melanogaster/; last accessed 6/25/20). To call SNPs and indels, we used HaplotypeCaller and performed joint genotyping for each of the five genotypes using GenotypeGVCFs. SNPs filtered with following parameters: 'QD < 2.0 || FS > 60.0 || MQ < 40.0 || MQRankSum < -12.5 || ReadPosRankSum < -8.0'. We annotated SNPs as synonymous or nonsynonymous using SNPeff  with the integrated D. melanogaster database (dmel_r6.12) database and parsed these annotations with SNPsift . To classify the SNPs as ‘shared’ between SD-Mal, SD⁺In(2L)t and SD⁺In(2R)NS, or ‘private’ to each one of them, we used BCFtools intersect (v1.6).

Please see the manuscript for complete methods.

Please see README.md for instructions for how to reproduce all figures and analyses.