Supplementary Methods
We used dnaPipeTE (Goubert et al. 2015) to quantify the extent that repetitive element content differed across the populations. Our approach assumed a genome size of 168Mbp, with the number of randomly sampled reads equal to 1-fold coverage of the genome, resampling each strain 2 times to get the average estimate of each strains TE content. Following TE identification, we grouped sequences by known super-families and compared the proportion of the genome composed of each superfamily across strains in the populations. We also used a reciprocal blast (e-value < 0.00000001) (Altschulet al. 1990) to identify TE families present in each strain.
We confirmed TE families shared between the previous RepeatModeler (Smit and Hubley 2008) annotation of the D. innubilareference genome and the dnaPipeTE annotation using blast (e-value < 10e-08) (Altschul et al. 1990). After confirming that they did not differ in content, we called TE insertions in each strain across the genome using PopoolationTE2 (Kofleret al. 2016), then merged the output and calculated the frequency of insertions, grouping by TE order, population and if the insertion was exonic, intronic, non-coding or flanking a gene (500bp up or downstream of start or end). When considering individual TE families in D. innubila , we used the RepBase TE names and identifications (Bao et al. 2015).