The organization and evolution of the Responder satellite in species of the Drosophila melanogaster group: dynamic evolution of a target of meiotic drive

Background Satellite DNA can make up a substantial fraction of eukaryotic genomes and has roles in genome structure and chromosome segregation. The rapid evolution of satellite DNA can contribute to genomic instability and genetic incompatibilities between species. Despite its ubiquity and its contribution to genome evolution, we currently know little about the dynamics of satellite DNA evolution. The Responder (Rsp) satellite DNA family is found in the pericentric heterochromatin of chromosome 2 of Drosophila melanogaster. Rsp is well-known for being the target of Segregation Distorter (SD)— an autosomal meiotic drive system in D. melanogaster. I present an evolutionary genetic analysis of the Rsp family of repeats in D. melanogaster and its closely-related species in the melanogaster group (D. simulans, D. sechellia, D. mauritiana, D. erecta, and D. yakuba) using a combination of available BAC sequences, whole genome shotgun Sanger reads, Illumina short read deep sequencing, and fluorescence in situ hybridization. Results I show that Rsp repeats have euchromatic locations throughout the D. melanogaster genome, that Rsp arrays show evidence for concerted evolution, and that Rsp repeats exist outside of D. melanogaster, in the melanogaster group. The repeats in these species are considerably diverged at the sequence level compared to D. melanogaster, and have a strikingly different genomic distribution, even between closely-related sister taxa. Conclusions The genomic organization of the Rsp repeat in the D. melanogaster genome is complex—it exists of large blocks of tandem repeats in the heterochromatin and small blocks of tandem repeats in the euchromatin. My discovery of heterochromatic Rsp-like sequences outside of D. melanogaster suggests that SD evolved after its target satellite and that the evolution of the Rsp satellite family is highly dynamic over a short evolutionary time scale (<240,000 years). Electronic supplementary material The online version of this article (doi:10.1186/s12862-014-0233-9) contains supplementary material, which is available to authorized users.


Background
Genomes are frequently in conflict with selfish genetic elements that propagate in genomes or populations despite the harm that they cause to the host [1][2][3]. Genetic elements can range in their degree of selfishness from the expansion of blocks of tandemly repeated satellite DNAs [4] typically found near centromeres and telomeres [5], to the invasive properties of transposable elements or the ultra-selfish behavior of meiotic drivers. Meiotic drivers spread in populations by gaining a transmission advantage through gametogenesis [6]. Segregation Distorter (SD) is an autosomal male meiotic drive system in Drosophila melanogaster that has biased transmission? while heterozygous females transmit SD fairly to half of their progeny, heterozygous males transmit SD to nearly all of their progeny [7]. SD targets Responder (Rsp), a satellite DNA in the pericentric heterochromatin of 2R [8,9]. The sensitivity of the Rsp locus to segregation distortion correlates positively with the number of repeats on 2R? SD targets wild-type chromosomes with many Rsp repeats [8,9] resulting in wild-type sperm dysfunction through a currently unknown mechanism (reviewed in [10]).
The structure of the Rsp locus is complex: the canonical form of the repeat is a dimer of two related 120-bp repeats referred to as ? Left? and ? Right? Rsp (84% identical), but the tandemly arrayed canonical repeats are interspersed with more divergent variants of Rsp [11]. At least two additional locations of Rsp repeats exist outside of the SD target in 2R pericentric heterochromatin: a cluster of repeats occurs on 3L in cytological band 80C [12], and a single fragment of a Rsp repeat occurs on 2R at cytological band 60A [11].
The evolutionary dynamics of the Rsp satellite are currently unknown. SD is a selfish genetic system specific to D. melanogaster. While the divergence between the Right and Left Rsp repeats suggests that the repeat is old, it has never been found outside of D. melanogaster? implying that the repeat arose in an ancestor of the melanogaster group and was subsequently lost outside of D. melanogaster, or it is rapidly evolving in D. melanogaster. These inferences resulted from studies based on DNA-DNA hybridization or poorly assembled and fairly low coverage genomes, however, and currently available genomic resources could provide new insights into the evolutionary history of Rsp repeat family evolution.
We know little about the evolutionary dynamics of satellite DNAs in general. Many repetitive DNAs undergo concerted evolution [13][14][15], whereby unequal recombination and/or gene conversion events cause repeats within a species to be more similar to each other than to their homologous repeats between species [15][16][17][18][19][20]. While concerted evolution is documented among various repetitive DNA sequences, the effect of intragenomic conflict on any particular family of repeats is understudied. While some satellite DNAs are expected to be selfish themselves, the Rsp repeats of D. melanogaster are instead (or perhaps in addition), the target of a selfish meiotic driver, making these especially interesting repeats to study. An evolutionary genetic analysis of the Rsp family of repeats could reveal important details about the evolutionary history of SD, as well as the dynamics of satellite DNAs and the effect of intragenomic conflict on genome evolution.
In this paper, I present an evolutionary genomic analysis of the Rsp satellite in D. melanogaster and the three closely related species of the simulans clade (D. simulans, D. sechellia and D. mauritiana). I combine traditional Sanger sequencing data (from BACs and Whole Genome Shotgun assemblies), Next Generation Sequencing (NGS) data (from genomic Illumina reads), and in situ hybridization to study patterns of Rsp satellite evolution on a short evolutionary time scale. I show that Rsp repeats have euchromatic locations throughout the D. melanogaster genome, that Rsp arrays show evidence for concerted evolution, and that Rsp repeats exist outside of D. melanogaster, in species of the melanogaster group. My analyses suggest that Rsp repeat family evolution is highly dynamic, and are consistent with the rapid evolution of Rsp in D. melanogaster, where it is a target of meiotic drive.

Results and discussion
Rsp in D. melanogaster I used BLAST to identify Rsp repeats in the WGS assembly of D. melanogaster and found hits on nearly every chromosome arm ( Figure 1). As expected, large blocks of canonical Rsp repeats? defined as the sequences similar to those correlated with the sensitivity to segregation distortion in Wu et al. [9]? occur on chromosome 2R (these repeats in versions of the D. melanogaster genome prior to v6.01 were found in ArmU and ArmU extra scaffolds). These repeats correspond to the large block of satellite in the pericentric heterochromatin on chromosome 2R. Consistent with previous findings [12], at least one large block of canonical Rsp repeats occurs on chromosome 3L. However, I found several small blocks of Rsplike repeats in the euchromatic regions of X, 2R, 3L and 3R ( Figure 1B and C). Interestingly, the largest block of euchromatic repeats is found on chromosome 3L, in an intron of the gene Argonaute 3 (Ago3). The Ago3 repeats are canonical Rsp repeats ( Figure 1B). The euchromatic Rsp repeats consist of between 1 and 12 repeats which, because the assembly of repetitive sequences tends to collapse repeats with nearly-identical sequences, may be considered estimates of the minimum repeat number. The second largest euchromatic Rsp blocks occur on the X chromosome. The Rsp-like repeats found on the X chromosome are not canonical Rsp repeats, but divergent copies of a Rsp family repeat (hereon referred to as RlX for Rsp-like on X; P-value from permuted alignments with D. melanogaster canonical Rsp <10 −4 ; Figure 1B). There are three clusters of interspersed RlX repeats in cytological band 4C on the X chromosome (occurring within a 150 kb interval). The three clusters span 715 bp, 120 bp and 1430 bp, respectively, with the largest cluster occurring within 1 kb of the gene CG12688. I compared all individual WGS reads matching canonical Rsp to estimate genome-wide variability in Rsp repeat sequence. Overall, individual reads matching the Left and Right canonical Rsp sequences from across the genome shared 85.8% and 87.6% identity, indicating that there is considerable variability in canonical Rsp repeats in the genome.
Because during genome assembly, reads from identical or nearly identical sequences may be collapsed into a single sequence or mapped to the wrong location, repetitive regions of the genome are often misassembled [21]. This makes it difficult to reliably compare repeats in different regions of the WGS assembly. To compare repeats between regions of the genome, I instead used BLAST to identify Rsp repeats in sequenced BACs mapping to known genomic locations. I obtained hits on BACs that map to euchromatic and heterochromatic locations in the genome, including the 2R heterochromatic division called h39, the target of SD [8]. The BAC categorized as h39 (AC246306.1) contains clusters of Bari-1 repeats that define the distal boundary of the Rsp locus at h39 [22], (Additional file 1:   defined as repeats that fall between 79.5%-90% identical to right or left canonical Rsp repeats ( Figure 1B). A neighbor-joining tree constructed from Rsp repeats found in clusters on different BACs confirms that repeats within an array tend to be more similar to each other than repeats outside of an array ( Figure 2), a pattern consistent with a history of concerted evolution. Surprisingly, for some repeat clusters, it appears that there is even exchange between arrays of repeats in different genomic locations and perhaps even different chromosome arms (between unmapped BACs whose likely location is chromosome 2R and 3L BACs; Figure 2). It is however possible that the unmapped BACs map to pericentric heterochromatin on 3L instead of 2R, however if this were the case, this result still demonstrates exchange between distinct clusters of repeats (no portion of Ago3 occurs on this unmapped BAC). Unfortunately, I was unable to conclusively determine the genomic location of the unmapped BACs [23], (Additional file 1: Figure S1). The Rsp repeats found on BACs mapping to 2R (at h39) and 3L are very similar [23] (Additional file 1: Table S2), consistent with previous observations [24]. While this may be evidence for rare interchromosomal exchange between 2R and 3L, such exchange has not been documented for other repeat families in Drosophila. Alternatively, selective sweeps involving the 2R pericentric heterochromatin could also cause repeats on different chromosomes to be more closely related to each other than nearby repeats, as the different chromosomes would have the same recent common ancestor [24].

Rsp family repeats in simulans clade species
Although the Rsp satellite has not been described outside of D. melanogaster [9,24,25], I found repeats similar to Rsp in each species of the simulans clade, and in D. erecta and D. yakuba (hereon referred to as Rsp-like; [23], Additional file 1: Table S3). I used BLAST to identify sequences related to the D. melanogaster Rsp repeat in the WGS assemblies of D. sechellia, D. simulans, D. erecta and D. yakuba (P-values from permuted alignments with D. melanogaster canonical Rsp are <10 −4 for each species). The repeat unit of Rsp-like is larger than canonical Rsp: Rsp-like is~160 bp in the simulans clade and 173 bp in D. erecta, whereas canonical Rsp in D. melanogaster is~120 bp. Throughout the paper, only the~120 bp homologous to canonical Rsp is analyzed. Interestingly, D. erecta Rsp-like repeats appear to have a dimeric structure analogous to the Left and Right repeats of D. melanogaster. I refer to the D. erecta repeats as Rsp-like-1 and Rsp-like-2 (there is 9.5% divergence between the pairs) to avoid confusion with the Left and Right repeats of D. melanogaster because the dimeric structures appear independently derived.
Because repetitive sequences are often underrepresented and misassembled in traditional WGS assemblies [21,26,27], I also queried the reads of Illumina NGS datasets in D. melanogaster and species of the simulans clade? D. simulans, D. sechellia and D. mauritiana? for Rsp sequences. Using the consensus of Rsp-like sequences as references, I identified Rsp and Rsp-like repeats among the NGS reads of D. melanogaster, D. sechellia, D. simulans, and D. mauritiana using Bowtie2 (KP016744-KP016746). I collected all unique Rsp sequences (individual repeat units) by constructing de novo assemblies of all canonical Rsp and Rsp-like reads for each species (see Methods; Table 2). Species vary widely in their number of unique repeats: D. sechellia, in particular, has an order of magnitude more unique repeats than the other species (Table 2).
Because several indels differentiate repeats within and between species, I constructed a neighbor joining tree based on distance using a model that considers adjacent indels as a 5th nucleotide state. This tree was constructed for unique Rsp and Rsp-like sequences (excluding RlX repeats and partial repeat units) in D. melanogaster and species of the simulans clade with D. yakuba as the outgroup [23]. The tree topology reveals that Rsp-like sequences in D. simulans, D. sechellia and D. mauritiana are very similar (Figure 3). The divergence between the canonical D. melanogaster Left and Right Rsp (79-82% ID) implies that the repeat family originated before the speciation of melanogaster group species, assuming a molecular clock. However, the Rsp repeats in D. melanogaster form a monophyletic group, whereas the Rsp-like repeats of the simulans clade species are intercalated throughout the tree (Figure 3). I also inferred the best maximum likelihood tree for all unique Rsp and Rsp-like sequences and found the same pattern [23], (Additional file 1: Figure S2). The canonical D. melanogaster Rsp and simulans clade species Rsp-like repeats are highly similar in the first 28 bp and the last 66 bp, but difficult to align in the middle. I am therefore uncertain of the length of the internal branch leading to the D. melanogaster repeats (Figure 3). Taken together, the excess divergence and monophyly of Rsp repeats on the D. melanogaster branch implies that there has been accelerated evolution in this lineage, where it is the target of SD.

Relationship between heterochromatic and euchromatic repeats
In addition to the heterochromatic Rsp-like repeats, the WGS assemblies of D. sechellia and D. simulans include the euchromatic RlX repeats. To examine the relationship between the heterochromatic Rsp and Rsp-like repeats and the euchromatic RlX in D. melanogaster, D. simulans, D. sechellia, D. mauritiana, D. erecta and D. yakuba, I constructed phylogenies using MrBayes. To test for concerted evolution of the euchromatic RlX repeats, I selected three RlX repeats occurring in orthologous positions upstream of the X-linked gene CG12688 for D. melanogaster, D. sechellia and D. simulans. The tree suggests that the D melanogaster canonical Rsp repeats are indeed monophyletic and evolving rapidly. The tree topology indicates that the Rsp-like repeats of the simulans clade groups with canonical Rsp repeats of D. melanogaster, but with low posterior probability (60%; Figure 4). Comparing overall percent identity, the RlX repeats are more similar to the Rsp-like repeats of the simulans clade than canonical D. melanogaster Rsp repeats: whereas RlX to Rsp-like percent identify varies between 75.6-80.7%, RlX to Rsp (canonical) only varies from 55.6-61.5% [23], again suggesting that canonical Rsp repeats evolve rapidly in D. melanogaster.
Although it is possible that canonical Rsp is an old repeat that was lost in the simulans clade, the tree topology suggests that this is not the case (Figure 4). The polytomy between the simulans clade RlX repeats, the D. melanogaster RlX repeats and the heterochromatic Rsp and Rsp-like repeats of D. melanogaster and the simulans clade (Figure 4) may be influenced by gene conversion events between RlX repeats and canonical Rsp repeats in D. melanogaster [23]. A comparison of the three tandem RlX repeats in orthologous positions on the X chromosome (in cytological band 4C upstream of CG12688) revealed a pattern of concerted evolution: repeats within a species are more similar than repeats at the orthologous positions between species (Figure 4).

Dynamic genomic distribution of Rsp-like repeats
The WGS assembly and NGS reads cannot give a complete picture of the satellite DNA distribution in these genomes because of assembly difficulties in heterochromatin. To determine the large-scale organization of the Rsp repeats in these species, I used FISH on mitotic chromosomes from larval neuroblasts using probes specific to    The chromosome arms of these species are homologous and have extremely high degrees of synteny [28]? they appear to mostly differ at a gross scale in their distribution of satellite repeats.

Conclusions
We currently know little about the evolution of satDNA, despite that it can make up a significant fraction of eukaryotic genomes [29]: >50% of the genome in kangaroo rats [30] and tenebrionid beetles [31]. Because of its ability to spread in genomes without offering any benefit to the host, satDNA has been considered selfish, junk DNA [1][2][3]. However, subsequent work in evolutionary, molecular, cellular, and cancer biology has converged on the idea that the heterochromatic fraction of the genome, including satDNAs, has important functional consequences [32][33][34][35][36][37][38][39][40][41][42][43][44][45][46]. While some blocks of satellite DNA themselves may exhibit meiotic drive, or biased transmission through gametogenesis [6] in females [47,48], the topic of this study? the Responder (Rsp) satellite? is a instead the target of meiotic drive in male D. melanogaster [7]. My discovery of highly dynamic evolution in the Rsp satellite family offers a model system to study the evolutionary dynamics of satDNA and the genetic conflict surrounding this enigmatic compartment of the genome. Most of what we know about satellite DNA dynamics is at the resolution of large blocks of satellite DNA on chromosome arms. Assembly issues with repetitive DNA have stymied our understanding of satellite DNA evolutionary dynamics at the genomic level [21]. Using information from the physical map (e.g. BACs [27,49]) and deep Illumina sequencing in combination with FISH offers a higher resolution image of satellite DNA evolution within and between species. My analysis of the Rsp satellite DNA family demonstrates that Rsp: 1) exists outside of the large block of pericentromeric satellite in locations across the D. melanogaster genome; 2) exists in species of the melanogaster group; 3) shows evidence for concerted evolution; and 4) is evolutionary dynamic in its abundance and genomic distribution over short evolutionary time scales.

Euchromatic satellite repeat islands
The Rsp satellite is a particularly interesting satellite DNA because it is the target of the SD meiotic drive system in D. melanogaster. Moschetti et al. [12] showed using FISH on polytene chromosomes that a small block of Rsp exists on chromosome 3L, outside the region associated with SD (at heterochromatic division h39 on 2R). I show here that in addition to the block of repeats on 3L, Rsp family repeats exist in ? euchromatic satellite repeat islands? in locations across the D. melanogaster genome, most notably on 3L and the X chromosome (RlX repeats). Variable tandem repeats in several taxa have been considered as possible sources of gene regulatory variation (reviewed in [50]). It is possible that these Rsp repeat islands have some regulatory role in the expression of nearby genes or local chromatin condensation: in some cases they occur in or near genes in D. melanogaster. Similar to the genomic distribution of Rsp, short, euchromatic blocks of up to 5 tandem repeats were recently reported clustering in or near genes for satellites of the 1.688 family [14] in D. melanogaster.
The largest euchromatic Rsp islands are on 3L and the X chromosome. The X chromosome has a special role in the SD system: escapers from SD-mediated meiotic drive have biased sex ratio [51], and X-linked suppressors of SD segregate at high frequencies in natural populations [52,53]. Some X chromosomes therefore seem to offer a protective effect against SD. Factors that suppress SD map to at least three regions of the X chromosome [54], and some of these intervals contain RlX repeats. It will be interesting to determine if the RlX repeats are involved in suppression of the SD system. One recently proposed hypothesis for the molecular mechanism of segregation distortion suggests that repeat-associated small interfering RNAs (rasiRNAs) corresponding to Rsp are necessary for proper packaging of the satellite during spermiogenesis, and that SD interferes with the production or localization of these rasiRNAs [10,55,56]. Many satellite DNAs in D. melanogaster [37] and other taxa [34,57,58] are transcribed and processed into small RNAs, including Rsp ([59]; Larracuente, unpublished). One possibility is that these euchromatic Rsp satellite repeat islands correspond to Rsp rasiRNA-producing clusters.
Dynamic satellite DNA evolution on small time scales SD is a melanogaster-specific drive system [60,61], but the history of the Rsp satellite has been unclear. While divergence between the Left and Right canonical Rsp repeats suggests that the repeat is old, the repeats had not been reported outside of D. melanogaster previously [9], although some studies indicated that divergent copies of the repeat may exist in D. simulans [24,25]. My results demonstrate conclusively that Rsp-like sequences exist in abundance in the closely-related species of the simulans clade, adding important insight into the evolution of the SD system? SD arose in D. melanogaster in a background that already had Rsp repeats to target. The D. melanogaster canonical Rsp is highly divergent compared to the Rsp-like repeats of the simulans clade species (e.g. 56-63% ID) and monophyletic, suggesting that Rsp has accelerated evolution in D. melanogaster. What might drive the rapid evolution of this repeat family in D. melanogaster? One possibility is that the mutation rate is unusually high for D. melanogaster Rsp compared to the Rsp-like repeats in the simulans clade species. An alternative explanation is that in D. melanogaster, segregation distortion against the canonical Rsp repeats by SD creates selection pressure to diverge from the target sequence. At present, we do not know the molecular mechanism of distortion, or what feature of the Rsp locus makes it a target (i.e. if it is sequence-specific). Efforts to compare the detailed evolutionary history of Rsp family repeats in the melanogaster group are underway.
Comparing Rsp repeat sequences within and between species revealed that Rsp and Rsp-like sequences show at least two lines of evidence for concerted evolution: 1) repeats within the D. melanogaster genome are more similar within a genomic cluster than between genomic clusters; and 2) RlX repeats between D. melanogaster and species of the simulans clade are most closely related to nearby repeats within each species than to repeats at orthologous positions between species.
At least in part due to the mutational properties of repetitive DNA [62], the turnover in satellite DNAs between closely-related species can be extreme [5,63] and in some cases may contribute to genetic incompatibilities between species [5,38]. On a genome-wide level, Rsp family evolution is highly dynamic. In the time since D. sechellia, D. simulans and D. mauritiana diverged? approximately 240 Kya [64]? the Rsp-like satellite has dramatically changed its genomic distribution. In D. melanogaster, large blocks of the Rsp satellite occur in the pericentric heterochromatin of 2R, in D. sechellia, the Rsp-like satellite expanded and occurs in the pericentric heterochromatin of 2R, 3L and 3R, in D. simulans, the Rsp-like satellite only occurs at the base of the X chromosome and the Rsp-like satellite is undetectable at the chromosome level in D. mauritiana and D. yakuba. Rsp repeat copy number is highly polymorphic within D. melanogaster [65]? this polymorphism is directly related to SD, as large blocks of satellite confer sensitivity to segregation distortion [8,9]. The Rsp-like repeats in non-melanogaster species are presumably not targets of a meiotic drive system. It will be interesting to compare the population genetics of Rsp between species where it is and is not a target of meiotic drive.

Methods
Querying Whole Genome Shotgun (WGS), BAC assemblies, and reads D. melanogaster reads were downloaded from the NCBI Trace Archive; BACs [27,49] and WGS contigs [66] were downloaded from Genbank. Rsp repeats were found in the D. melanogaster WGS genome assembly (version 6.01), BACs and NCBI Trace Archive using local basic local alignment search tool (BLAST) searches. Individual traces from four BACs (AC246323.1, AC246299.1, AC007548.10, and AC009843.9; [23], Additional file 1: Figure S1) were obtained from Sue Celniker and Kenneth Wan. An iterative BLAST search protocol was used to gather all sequences matching Rsp. Initial BLAST searches were performed using Rsp sequences deposited in Genbank as queries. To capture as much variation in Rsp sequence as possible, and to recover more divergent Rsp sequences, several BLAST iterations were performed in which subsequent BLAST searches used incrementally refined lists of the hits from previous BLAST searches as queries. This iterative process ended when no additional significant hits were obtained. BLAST hits with an e-value >0.1 and a length less than 30 bp were excluded from the analysis. BLAST alignments for hits with an e-value > .001 and length <50 were inspected by eye. Redundancies between sequences were removed using custom Perl scripts. Alignments of Rsp sequences were made using BWA-SW [67]. Pairwise percent identity between Rsp repeats was calculated in Geneious (version 6.1.7, created by Biomatters; http://www.geneious.com; [68] and the mean and 95% confidence intervals of pairwise percent identity were calculated in R. The same procedure was used to query the WGS assemblies of D. sechellia [28], D. simulans [69], D. erecta [28], and D. yakuba [28]. To determine that Rsp and Rsp-like repeats are indeed related sequences, the nucleotides of each repeat sequence were randomly shuffled and percent identity was re-calculated for pairwise alignments using custom perl scripts. To create a distribution of percent identity based on nucleotide composition, 10,000 permutations were completed for each sequence and P-values were obtained using the empirical cumulative distribution function for each set of permuted alignments (in R).

Querying Next Generation Sequencing (NGS) reads
Illumina GAIIX paired end reads from D. melanogaster (SRR060098; [70]), D. simulans (SRR520350; [69]), D. sechellia (SRR869587; [64]) and D. mauritiana (SRR483621; [64]) were downloaded from the NCBI? s SRA (http://www. ncbi.nlm.nih.gov/sra). NGS reads were trimmed of adapters and low quality bases using Trim Galore (version 0.2.8; Babraham Bioinformatics http://www.bioinforma tics.babraham.ac.uk/projects/trim_galore/). Using Bowtie2 [71], the trimmed reads were mapped to consensus Rsp sequences (from the iterative BLAST procedure described above) that capture variation within and between species from WGS and BAC sequences. The alignments were analyzed using Samtools-0.1.18 [72]. To create a list of unique Rsp sequences for each species, de novo assemblies were constructed using the mapped reads extracted from SAM files created from Bowtie2. ABySS (version 1.3.6; [73] was used for the de novo assembly with the following parameters: k =64; se; m = 30; l =64. The unique contigs were parsed for individual Rsp sequences using BLAST to the original query Rsp and custom Perl scripts (e.g. a 400bp contig was split into three individual Rsp repeats as determined by BLAST to a consensus Rsp sequence). Some assembled repeat units were less than the canonical repeat unit length (~120 bp). Because some of these shorter units represent true fragmented repeats and some are expected to be fragments of unique repeats that the de novo technique couldn? t complete (for reasons such as insufficient read depth or sequencing error), repeats with identical sequence but length variants were merged. Specifically, short, overlapping fragments with identical sequence and overhangs (non-overlapping sequence) of <50% of the total fragment length were merged into a single unique repeat. The list of unique Rsp repeats in each species was imported into Geneious (version 6.1.7 [68] to create and edit alignments.

Trees
Neighbor-joining trees (Saitou and Nei 1987) were constructed using bionj with the ape package in R. Blocks of adjacent indels were included in the model using the setting ? model = indelblock? and bootstrapping was done using the boot.phylo function in the ape package (B = 100). Bayesian inference of the phylogenetic relationships between the RlX, Rsp-like and canonical Rsp repeats was performed using MrBayes [74] using a GTR nucleotide substitution model with Gamma rate variation. The Rsp-like sequences of the simulans clade species were consensus sequences (GenBank accession numbers: KP016744, KP016745, KP016746) of the de novo contigs assembled from each species NGS reads. The Rsp-like repeats of D. yakuba and D. erecta are consensus sequences from the BLAST of WGS assemblies. The RlX-1, RlX-2 and RlX-3 are orthologs of three tandemly-repeated RlX sequences from cytological band 4C in the region~1 kb upstream of CG12688.
Maximum likelihood inference of the best tree for the unique Rsp sequences from the NGS reads of D. melanogaster, D. sechellia, D. simulans and D. mauritiana was performed in RAxML v7.4.2 [75] using the CIPRES Gateway (http://www.phylo.org); [76]. Bootstrapping was performed using a GTR plus Gamma nucleotide substitution model (−m GTRGAMMA ? ? 1000). The maximum likelihood tree was drawn using the ape package in R [77]. The full tree including bootstrap confidence at the nodes is deposited in the Dryad Digital Repository (http://dx.doi.org/10.5061/dryad.3sh6d; [23]).

Fluorescence in situ hybridization (FISH)
FISH to mitotic chromosomes in larval neuroblasts was performed as previously described [78,79]. Briefly, brains were dissected from 3rd instar larvae and fixed in 1.8% paraformaldehyde, 45% acetic acid. Fixed brains were denatured at 95? C and hybridized overnight at 30? C. Slides were washed in 4X SSCT three times, and 0.1X SSC three times before mounting in Vectashield with DAPI. The probe was a biotinylated, nick translated PCR product specific to Rsp repeats D. melanogaster (F-5? GG AAAATCACCCATTTTGATCGC and R-5? CCGAATT CAAGTACCAGAC for D. melanogaster FISH) or Rsp-like repeats in D. sechellia (F-5? ACTGATTATCATCGCCTGGT and R-5? TCCAGTTCGCCTGGTAGTTT; for D. sechellia, D. simulans, and D. mauritiana FISH).