Satellite emergence at simulans clade centromeres

To identify the detailed organization of centromeres in the simulans clade, we performed CUT&Tag [32] on embryos from each species (D. simulans, D. sechellia, and D. mauritiana) using a CENP-A antibody. The resulting reads were mapped to versions of each species’ genome assembly with improved representation of heterochromatic regions from previous work [33]. Because centromeres sit in highly repetitive genome regions, we analyzed unique and all reads (including multi-mappers) independently (Figs 1 and S1–S3). We identified centromere candidates as the top reproducible CENP-A-enriched contigs (between-replicate irreproducible discovery rate [IDR] < 0.05; S1 Table and S4 Fig). We also used an assembly-free analysis to detect the enrichment of complex repeats in the CENP-A CUT&Tag reads (see Methods). We validated our approach with CUT&Tag in D. melanogaster, which recovered the same centromere islands as in Chang and colleagues [29] (S5 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Centromeres of chromosomes 2 and 3 in D. simulans, D. sechellia, and D. mauritiana are predominantly 500-bp satellite.

(A, D, G) CENP-A CUT&Tag enrichment on the centromere candidates for the major autosomes (2 and 3) of D. simulans (A), D. mauritiana (D), and D. sechellia (G). The label “Autosome 2/3” indicates that we cannot distinguish between the second and third chromosome centromeres. The y-axis represents normalized CENP-A enrichment in RPM. Black and gray plotted lines represent the enrichment based on uniquely mapping and all reads (including multi-mappers), respectively. The black and gray tracks below each plot correspond to MACS2 peaks showing significantly enriched regions based on the uniquely mapping and all reads (including multi-mappers), respectively. The precise locations of all peaks are listed in S1 Table. The colored cytoband track at the bottom of the plot shows the repeat organization. The color code is shown in the legend at the bottom of the figure. (B, E, H) Assembly-free analysis showing the normalized enrichment score (in RPM) of CENP-A for complex repeats, including TEs and complex satellites across all centromeres. The Top 20 most enriched repeats are represented for D. simulans (B), D. mauritiana (E), and D. sechellia (H). (C, F, I) IF-FISH on mitotic chromosomes from larval brains with CENP-C antibody and 500-bp and dodeca probes, for D. simulans (C) and D. mauritiana (F) or 500-bp and Rsp-like probes for D. sechellia (I). The insets represent a zoom on each major autosome centromere. Bars represent 5 μm. The data underlying this figure can be found at https://doi.org/10.5061/dryad.1zcrjdg2g [40]. RPM, reads per million; TE, transposable element.

https://doi.org/10.1371/journal.pbio.3002911.g001

Like D. melanogaster, all 3 simulans clade species have a pair of large metacentric “major” autosomes (chromosomes 2 and 3), a pair of small autosomes (chromosome 4; referred to as the “dot” chromosome), and a pair of sex chromosomes (X and Y). For each species, there were 5 contigs that were consistently among the most CENP-A-enriched contigs (S4 Fig), which we considered to be the centromere candidates for each chromosome (S2 Table). We found almost no consistent CENP-A signal outside of these centromere candidates (S4 Fig and S1 Table). None of the simulans clade centromere candidates we identified were like D. melanogaster centromeres, suggesting a turnover in centromere identity in the ~2.4 My since these species diverged. Instead, both our assembly-based (Figs 1A, 1D, 1G and S1–S3) and assembly-free (Fig 1B, 1E, and 1H) approaches identify the 500-bp complex satellite among the most CENP-A enriched sequences (Fig 1). The centromere candidate contigs for the major autosomes in D. simulans, D. mauritiana, and D. sechellia (Figs 1A, 1D, 1G, and S1–S3) and the X chromosome in D. simulans and D. mauritiana (Figs 2 and S1–S3) are mainly comprised of the 500-bp satellite repeat. This complex satellite was previously identified as being associated with the centromeres in D. simulans [8]. While the 500-bp satellite is the primary repeat type in these simulans clade centromeres, they also contain transposable element insertions, including G2/Jockey-3 (Figs 1A, 1D, 1G, 2, and S1–S3).

The 500-bp satellite is enriched in, but not specific to, simulans clade centromeres, as we also find it in the proximal pericentromeric regions. In D. melanogaster, the heterochromatin domain makes up approximately 60 Mb of the genome [34], of which centromeres only represent a small fraction (1/200th [29]). In the simulans clade centromeres, the CENP-A domain appears restricted to a 50-kb to 200-kb subset of the 500-bp satellite array (Figs 1A, 1D, 1G, and 2). This is similar to human centromeres, where the CENP-A domain sits on a subset of α-satellite repeats within an array [35]. We also identified a second complex satellite associated with centromere candidates, which we named the 136-bp satellite. While less abundant, 136-bp is interleaved with the 500-bp satellite and is associated with the same centromeres (Figs 1, 2, and S6A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. X chromosome centromeres in D. simulans and D. mauritiana are enriched in 500-bp satellite.

The left panel shows the CENP-A CUT&Tag enrichment on the X centromere candidate in D. simulans (A) and D. mauritiana (B). The y-axis represents the normalized CENP-A enrichment in RPM. Black and gray plotted lines represent the enrichment based on uniquely mapping and all reads (including multi-mappers), respectively. The black and gray tracks below each plot correspond to MACS2 peaks showing significantly enriched regions based on the uniquely mapping and all reads (including multi-mappers), respectively. The precise locations of all peaks are listed in S1 Table. The colored cytoband at the bottom of the plot shows the repeat organization. The color code is shown in the legend at the bottom of the figure. The right panel shows IF-FISH on mitotic chromosomes from larval brains with CENP-C antibody and 500-bp and Rsp-like probes. The inset represents a zoom on each X chromosome centromere. Bars represent 5 μm. The data underlying this figure can be found at https://doi.org/10.5061/dryad.1zcrjdg2g [40]. RPM, reads per million.

https://doi.org/10.1371/journal.pbio.3002911.g002

To validate that the 500-bp and 136-bp satellites are associated with the centromeres, we used a cytogenetic approach with IF-FISH on mitotic chromosome spreads from larval brains using Oligopaints targeting each complex satellite [36]. We confirmed the localization of centromeric protein CENP-C, a kinetochore protein that marks the centromeres and has documented overlap with CENP-A [37], on the 500-bp (Fig 1C, 1F, and 1I) and 136-bp (S6A Fig) satellites. Because mitotic spreads offer limited resolution, it is challenging to distinguish between the centromeric and proximal pericentromeric domains. However, the 500-bp signal extends beyond the CENP-C domain, indicating its presence in both the centromeric and proximal pericentromeric regions, consistent with our genome assemblies and CUT&Tag data. While the major autosomal centromeres primarily consist of the same complex satellites in the 3 species, the distal pericentromere appears more divergent. In D. simulans and D. mauritiana, the major autosomal pericentromeres contain the dodeca satellite (Fig 1C and 1F), while in D. sechellia they contain the Rsp-like satellite (Fig 1I). We also found the Rsp-like satellite on the X pericentromere of D. simulans (Fig 2A), which was absent in D. mauritiana (Fig 2B) [38,39]. The combination of satellites flanking the CENP-A domain (Figs 1C, 1F, 1I, and 2) allows us to assign the 500-bp enriched contigs to either the major autosomes (Fig 1A, 1D, and 1G) or the X chromosome (Fig 2). Unfortunately, we cannot morphologically distinguish between the chromosomes 2 and 3 because of their similarity.

We used a BLAST approach to explore origins of the 500-bp and 136-bp centromeric complex satellites and did not find any evidence of their presence outside of the D. simulans clade, even as single copy sequences (S3 and S4 Tables). For example, in D. melanogaster, the best hit had 85% identity with the 500-bp consensus sequence but only covered 106 bp of the query (S3 Table). This suggests that these satellites emerged after the divergence between D. melanogaster and the D. simulans clade 2.4 Mya [30,31], although it is possible that the primary sequence emerged earlier but was lost in D. melanogaster. In either case, these satellites recently expanded in the D. simulans clade centromeres (S7 Fig).

Dot chromosome centromeres are enriched with a chromosome-specific complex satellite

In D. simulans and D. mauritiana, the centromere of the small autosomal dot chromosome (i.e., Chromosome 4) contains a different complex satellite: the 365-bp satellite (Fig 3). The 365-bp satellite shares no homology with the 500-bp satellite, suggesting an independent origin. This repeat is consistently enriched in CENP-A chromatin in both our assembly-based (Fig 3) and assembly-free (Fig 1B and 1E) approaches. The CENP-A domain is restricted to the 365-bp satellite and flanked by the AATAT satellite on at least 1 side (Fig 3), which is confirmed by our FISH with CENP-C IF on chromosome spreads (Fig 3 insets). Unlike the 500-bp satellite, 365-bp is specific to the dot chromosome centromere. We do not find evidence of the 365-bp satellite outside of one CENP-A enriched contig in each assembly (Fig 3), consistent with the FISH signals (Fig 3 insets).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Dot chromosome centromeres in D. simulans and D. mauritiana are enriched in 365-bp satellite.

The left panel represents the CENP-A CUT&Tag enrichment in D. simulans (A) and D. mauritiana (B). The y-axis represents the normalized CENP-A enrichment in RPM. Black and gray plotted lines represent the enrichment based on uniquely and multi-mapping reads, respectively. Black and gray plotted lines represent the enrichment based on uniquely mapping and all reads (including multi-mappers), respectively. The black and gray tracks below each plot correspond to MACS2 peaks showing significantly enriched regions based on the uniquely mapping and all reads (including multi-mappers), respectively. The precise locations of all peaks are listed in S1 Table. The colored cytoband track at the bottom of the plot shows the repeat organization. The color code is shown in the legend at the bottom of the Fig. The right panel represents the IF-FISH on mitotic chromosomes from the larval brain with CENP-C antibody and 365-bp and AATAT probes. The insets represent a zoom on each dot chromosome centromere. Bars represent 5 μm. The data underlying this figure can be found at https://doi.org/10.5061/dryad.1zcrjdg2g [40]. HTT, Het-A, TART, and TAHRE; RPM, reads per million.

https://doi.org/10.1371/journal.pbio.3002911.g003

We used a BLAST-based approach to explore the origin of the 365-bp satellite and did not find evidence of this satellite outside of the D. simulans clade species (S5 Table). For example, in D. melanogaster, the best hit had 82% identity with the 365-bp consensus sequence but was only 57 bp long (S5 Table) suggesting that, like the 500-bp satellite, the 365-bp satellite emerged after the split with D. melanogaster and likely emerged at the dot centromeres in the ancestor of the D. simulans clade (S7 Fig). One intriguing possibility is that 365-bp may share origins with (or be derived from) a sequence similar to those currently at D. melanogaster centromeres, as some short sequence fragments with similarity to a subset of the 365-bp satellite are on D. melanogaster X and dot centromeres (S5 Table).

Interestingly, 365-bp was lost from D. sechellia: we did not find cytological (S6B Fig) or genomic evidence of this satellite, even as a single copy sequence in the genome assembly, the genomic Illumina reads (S5 Table), or the CENP-A CUT&Tag reads (Fig 1H). However, the pericentromeric AATAT satellite appears to be conserved (S6B Fig).

Centromere shifts to telomere-specialized retroelements: Telocentric chromosomes in D. sechellia

In D. sechellia, the dot and X chromosome are distinct from those of D. simulans and D. mauritiana. We did not identify any 500-bp-enriched contig that might correspond to the X chromosome centromere and 365-bp is completely missing from the D. sechellia genome.

Instead, we identified 2 D. sechellia contigs that are significantly enriched for CENP-A containing an array of non-LTR retroelements well known for their role at telomeres: Het-A, TART, and TAHRE (also known as the HTT elements) [41]. The HTT elements are also among the most CENP-A-enriched elements in our assembly-free approach (Fig 1H). Drosophila species lack telomerases, instead, telomere size and integrity are maintained by the transposition activity of HTT retroelements [41]. HTT elements have specialized functions at telomeres of many Drosophila species, including D. melanogaster and the D. simulans clade [41].

On one HTT-CENP-A enriched contig, the HTT domain is adjacent to the 500-bp satellite, suggesting that it corresponds to the X chromosome centromere (Fig 4C). However, in D. sechellia, CENP-A is enriched on the HTT domain instead of the 500-bp satellite (Fig 4A), suggesting a repositioning of the centromere to the retroelements that normally occupy the telomere. Similarly, on the second HTT-CENP-A enriched contig, the CENP-A domain is flanked by a simple ATAG satellite only found on X and dot chromosomes [42] (Fig 4B). Thus, we infer that this second contig corresponds to the dot chromosome centromere.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The Dot and X chromosome centromere in D. sechellia are telocentric.

CENP-A CUT&Tag enrichment along the X (A) and dot (B) chromosome centromeres. The y-axis represents the normalized CENP-A enrichment in RPM. Black and gray plotted lines represent the enrichment based on uniquely and multi-mapping reads, respectively. Black and gray plotted lines represent the enrichment based on uniquely mapping and all reads (including multi-mappers), respectively. The black and gray tracks below each plot correspond to MACS2 peaks showing significantly enriched regions based on the uniquely mapping and all reads (including multi-mappers), respectively. The precise locations of all peaks are listed in S1 Table. The colored cytoband track at the bottom of the plot shows the repeat organization. The color code is shown in the legend at the bottom of the figure. (C) IF-FISH on mitotic chromosomes from the larval brain with CENP-C antibody and 500-bp and HTT probes. The inset represents a zoom on the X and dot chromosome centromeres. Bar represents 5 μm. (D) IF-FISH on chromatin fibers from the larval brain with CENP-A antibody and 500-bp and HTT probes, representing the telocentric X chromosome of D. sechellia. Bar represents 5 μm. The data underlying this figure can be found at https://doi.org/10.5061/dryad.1zcrjdg2g [40]. HTT, Het-A, TART, and TAHRE; RPM, reads per million.

https://doi.org/10.1371/journal.pbio.3002911.g004

To validate our observations, we designed Oligopaints targeting the HTT array on the X and dot chromosome centromere candidates in D. sechellia. The IF-FISH on mitotic chromosomes from larval brains confirmed that the centromeric protein CENP-C is indeed associated with the HTT domain on both the X and dot chromosomes. The 500-bp satellite appears adjacent to the HTT on the X chromosome (Fig 4C).

To visualize these regions at higher resolution, we performed IF-FISH on stretched chromatin fibers using a CENP-A antibody and Oligopaints targeting the 500-bp satellite and the HTT elements. These fibers confirm that CENP-A nucleosomes are seated on the HTT domain, and are flanked by, but do not overlap, the 500-bp satellite (Fig 4D). On average, 89.82 ± 19.4% of the CENP-A signal overlaps with the HTT signal, while only 6.2 ± 13.6% overlaps with the 500-bp signal (S6 Table). The chromatin fibers appear to end shortly after the CENP-A/HTT signal, strongly suggesting that the centromere is on a telomeric HTT array, making these chromosomes telocentric (Fig 4D). In some fibers, we observed a lack of CENP-A/HTT signal at the very ends, similar to what we show in Fig 4D. It is possible that there is a small amount of non-HTT sequence distal to the HTT signal on these chromatin fibers. However, we believe that the absence of HTT signal at the fiber ends is likely a technical artifact due to the loss of the FISH signal, as this observation was variable across fibers (see S8 Fig). Regardless, the overlap between CENP-A and HTT signal confirms that these centromeres are telocentric.

We also observed patterns from stretched chromatin fibers consistent with our predictions for the other chromosome centromeres (S8 Fig). On the dot chromosome 73.02 ± 32.76% of the CENP-A signal overlaps with the HTT signal, with no 500-bp signal nearby (S8 Fig and S6 Table). On the autosomes, 100% of the CENP-A signal overlaps with the 500-bp signal (S8 Fig and S6 Table).

Interestingly, the dot chromosome centromere of D. mauritiana is flanked by the AATAT satellite on one side and by the HTT on the other side (Fig 3B). Unfortunately, the contig is not long enough to establish how long the HTT domain continues after the centromere, but it suggests that in D. mauritiana, and possibly D. simulans, both centromeric and telomeric domains are very close to each other.

It was very surprising to find the centromeric protein associated with telomeric sequences, as centromeres and telomeres are chromosome domains with distinct functions. Although both the X and the dot chromosomes were considered to be acrocentric chromosomes based on the similarity in karyotype with D. melanogaster [43,44], our high-resolution approach allowed us to reveal that these chromosomes are actually telocentric. We demonstrate here that centromeres can share sequence components with telomeres [45]. Currently, we lack the ability to ascertain whether the centromere and telomere share a common domain or exist as separate domains within the HTT array.

The Y chromosome centromeres are unusual

In all 3 species analyzed, the Y chromosome centromeres are unique in their composition and organization compared to the rest of the centromeres in the genome. Unlike the other chromosomes, we did not identify any complex satellites associated with the Y chromosome centromere. Instead, CENP-A is enriched in a region with high density of transposable elements (Fig 5). Despite being mainly enriched in retroelements, the Y chromosomes from each species have a unique composition (Fig 5 and S7 Table). For example, the most abundant elements associated with the Y centromere are HMSBEAGLE and Jockey-1 in D. simulans, mdg4 in D. mauritiana, and R1 and G2/Jockey-3 in D. sechellia (S7 Table). Interestingly, centromeric sequences form higher order repeats in both D. simulans and D. sechellia, but not in D. mauritiana (S9 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The Y chromosome centromeres of D. simulans, D. mauritiana, and D. sechellia are rich in transposable elements.

The left panel shows the CENP-A CUT&Tag enrichment for the Y centromere of D. simulans (A), D.mauritiana (B), and D. sechellia (C). The y-axis represents the normalized CENP-A enrichment in RPM. Black and gray plotted lines represent the enrichment based on uniquely mapping and all reads (including multi-mappers), respectively. The black and gray tracks below each plot correspond to MACS2 peaks showing significantly enriched regions based on the uniquely mapping and all reads (including multi-mappers), respectively. The precise locations of all peaks are listed in S1 Table. The colored cytoband track at the bottom of the plot shows the repeat organization. The pie chart on the top represents the repeat composition of the CENP-A domain. The color code of the cytoband and pie chart is shown in the legend at the bottom of the figure. The right panel shows the IF-FISH on mitotic chromosomes from the larval brain with CENP-C antibody and cenY Oligopaints specific to each species’ centromere. The insets represent a zoom on each Y chromosome centromere. Bar represents 5 μm. The data underlying this figure can be found at https://doi.org/10.5061/dryad.1zcrjdg2g [40]. HTT, Het-A, TART, and TAHRE; RPM, reads per million.

https://doi.org/10.1371/journal.pbio.3002911.g005

To validate our candidate Y centromeres, we designed Oligopaints specific to the Y contig of each species (cenY). We performed IF-FISH on mitotic chromosomes with a CENP-C antibody and the Oligopaint targeting the putative Y centromeres. Our Oligopaints give a signal specific to a unique region of the Y chromosome which consistently co-localizes with the CENP-C signal (Fig 5), confirming the Y chromosome centromeres.

While simple satellites are present within the pericentromeric region of all the other chromosomes, we do not find any simple satellites in the flanking region of the Y centromere (Fig 5). This is surprising, especially given that these Y chromosomes in these species are highly enriched in simple satellites in general [46,47].

G2/Jockey-3 is associated with centromeres within the D. simulans clade

In D. melanogaster, the only common sequence among all centromeres is G2/Jockey-3 [29]. We asked if this element was also found within the simulans clade centromeres. In D. simulans, G2/Jockey-3 is the most enriched repeat among the CENP-A reads (Fig 1B). We identified G2/Jockey-3 insertions in each centromere except for the X chromosome, where it directly flanks the centromere (Fig 2A). We confirmed the presence of G2/Jockey-3 at each centromere by IF-FISH on mitotic chromosomes (Fig 6C). In D. sechellia, G2/Jockey-3 is also the most enriched repeat in CENP-A chromatin (Fig 1H); however, we only detect it on the Y chromosome and one of the autosomal centromeres (Figs 1G, 5C, and 6C). Similarly, in D. mauritiana, G2/Jockey-3 is associated with only one of the autosomal centromeres (Figs 1D and 6C) and is less enriched than in the two other species (Fig 1E). This suggests that the association of G2/Jockey-3 with the centromere was lost.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. G2/Jockey-3 is associated with the centromeres within the D. simulans clade.

(A) Maximum likelihood phylogenetic tree of G2/Jockey-3 ORF2 from D. melanogaster, D. simulans, D. sechellia, D. mauritiana, D. yakuba, and D. erecta. G2/Jockey-3 within the simulans clade species diverged into 2 different clades, one that is more closely related to the D. melanogaster elements (clade “A”) and one that is more divergent (clade “B”). Centromeric insertions are indicated by a pink * at the tip of the branch. We do not know centromere identity in D. yakuba and D. erecta. (B) ORF2 conservation analyses of the clade “A” G2/Jockey-3 centromere-associated clade. The circles below the species name represent each centromere. Centromeres containing G2/Jockey-3 insertions (based on CUT&Tag and FISH) are shown in black. The pie chart represents the proportion of centromeric (black) and non-centromeric (white) insertions among the clade “A” G2/Jockey-3 within each species’ genome, where we indicate the number of insertions within the pie charts. The consensus sequence of G2/Jockey-3 ORFs is schematized below the pie chart, indicating that only D. melanogaster and D. simulans consensus sequences have an intact ORF2. (C) IF-FISH on mitotic chromosomes from the larval brain with CENP-C antibody and G2/Jockey-3 probes showing consistent centromere-association in D. simulans, but not in D. mauritiana and D. sechellia. In D. simulans, the G2/Jockey-3 insertions on the X chromosome are adjacent to the CENP-A domain, rather than within. The inset represents a zoom on each centromere. Bars represent 5μm. The data underlying this figure can be found at https://doi.org/10.5061/dryad.1zcrjdg2g [40]. ORF, open reading frame.

https://doi.org/10.1371/journal.pbio.3002911.g006

To better understand the evolutionary history of this specific retroelement, we inferred the phylogeny for all G2/Jockey-3 ORFs in the D. melanogaster clade assemblies. G2/Jockey-3 has 2 open reading frames (ORFs), but we only used ORF2 for inferring phylogenies, as ORF1 is more evolutionarily labile across species [48]. While all D. melanogaster G2/Jockey-3 insertions cluster together in a unique clade, the D. simulans clade insertions separate into 2 different clades, which we designate as clade “A”—with sequences more closely related to D. melanogaster G2/Jockey-3—and clade “B” (Figs 6A and S10). Within each clade, insertions largely form species-specific clusters. All centromeric insertions are part of the clade “A” and retain a conserved ORF2. Like D. melanogaster, clade “A” G2/Jockey-3 insertions are enriched at centromeres (Fig 6B). That is, 53% of clade “A” G2/Jockey-3 insertions are centromeric in D. simulans and D. sechellia, which is more than expected if these TEs were randomly distributed in the genome (Fisher’s exact tests: P_sim < 10⁻¹⁶; P_sec < 10⁻¹⁶). The enrichment is less pronounced in D. mauritiana (17%; P_mau = 0.0567). However, the consensus ORF is incomplete in D. sechellia and D. mauritiana, implying that most clade “A” G2/Jockey-3 copies are degenerated in these species, in line with their inconsistent association with centromeres. These findings suggest that a subset of G2/Jockey-3 elements likely had centromere-biased insertion activity in the D. melanogaster clade ancestor. This activity may have continued after the speciation event between D. melanogaster and the D. simulans clade but was lost in D. sechellia and D. mauritiana lineages, explaining the inability of G2/Jockey-3 to jump into centromeres. While the clade “B” appears to have been recently active in the simulans clade, none of the insertions are centromeric. This clade was either lost from D. melanogaster or may have been introduced into the D. simulans ancestor through a horizontal transfer event. The latter appears to be more likely as we find fragmented copies of G2/Jockey-3 from D. yakuba that cluster with clade “B.” However, we do not have sufficient node support to draw strong conclusions about the origins of this clade. Taken together, our data suggest that the clade “A” G2/Jockey-3 targeted the centromeres for insertion in both D. melanogaster and the D. simulans clade specie despite having distinct centromeric sequences, suggesting that this element may preferentially target centromeric chromatin rather than particular DNA sequences.

Fly strains

For D. sechellia and D. mauritiana, we used the same sequenced strains used to build the heterochromatin enriched genome assemblies [30]: Rob12 (Cornell SKU: 14021–0248.25) and w12 (Cornell SKU: 14021–0241.151), respectively. For D. simulans, we used the wXD1 strain that is maintained in the Larracuente lab. While it is the same strain as the one used to build the heterochromatin enriched assembly, our isolate appears to have a structural polymorphism on the X chromosome pericentromeric compared to the assembly [33]. All the experiments conducted in this study were performed using the same isolate. For D. melanogaster, we used an inbred strain from the Netherlands (N25) [79].

Dryad DOI

https://doi.org/10.5061/dryad.1zcrjdg2g [40].

Antibodies used

The list of primary and secondary antibodies that we used for this study is details below:

1. Anti-CENP-A antibody (α-CID20): polyclonal rabbit antibody synthesized for this study (by Covance). The CENP-A antibody was raised against the MPRHSRAKRAPRPSAC peptide [8]. The final serum was proteinA purified. We used this antibody at 1:50 dilution for the CUT&Tag. We validated the specificity of the antibody by western blot (S11 Fig).
2. Anti-CENP-C antibody (α-CENP-C12): polyclonal rabbit antibody synthesis for this study (by Genscript). The CENP-C antibody was raised against the NNRRSMRRSGNVPGC peptide. The final serum was affinity purified. We used this antibody at 1:100 dilution for the Immunostaining on mitotic chromosomes.
3. Anti-CENP-A antibody (α-CIDH32): polyclonal chicken antibody, gift from the Mellone lab. We used the antibody at 1:100 dilution for the Immunostaining on chromatin fibers.
4. Anti-Mouse IgG H&L antibody (abcam, ab46540): rabbit antibody that we used as a negative control for the CUT&Tag at 1:100 dilution.
5. Anti-H3K9me3 antibody (abcam, ab176916): rabbit monoclonal antibody. We used this antibody as a positive control for the CUT&Tag at 1:100 dilution.
6. Anti CENP-C primary antibody: Guinea Pig antibody from [80]. We used this antibody for larval brain squashes for G2/Jockey-3 IF-FISH at 1:500 dilution.
7. Guinea Pig anti-rabbit unconjugated (Novus Biologicals, NBP1-72763). We used this secondary antibody for the CUT&Tag at 1:100 dilution.
8. Goat anti-rabbit IgI H&L conjugate with Alexa Fluor 488 (abcam, ab150077). We used this secondary antibody for the Immunostaining on mitotic chromosomes spread at 1:500 dilution.
9. Goat anti-Chicken IgY (H+L) Secondary Antibody, Alexa Fluor 488 (Invitrogen, A-11039).
10. Goat anti Guinea Pig conjugate with AlexaFlour 546 (Thermo Catalog # A-11074). We used this secondary antibody for the Immunostaining on mitotic chromosomes spread at 1:500, for G2/Jockey-3 IF-FISH.

Western blot

Twenty flies from each species were homogenized in 200 μl 1× Laemmli buffer (diluted from BioRad 4× Laemmli Sample Buffer [1610747] with 2-mercaptoethanol [Sigma] and 1× Pierce EDTA-free Protease inhibitors [Thermo Fisher A32965]), denatured by incubation at 95°C for 10 min, centrifuged at 15,000 reads per million (RPM) for 5 min at 4°C, and 20 μl of each the supernatant and PageRuler Prestained Protein Ladder (Thermo Fisher [26616]) was run 4% to 15% Mini-Protean TGX gel. The protein was transferred to PVDF membrane (Novex Invitrolon [LC2005]), blocked (Li-Cor Intercept Blocking buffer [927–60001]), incubated with 1:1,000 Rabbit anti-CENP-A(lab stock), washed 3 times with TBS/0.1% Tween-20, incubated with 1:20,000 Goat Anti-Rabbit IgG (H+L) DyLight800 (Invitrogen SA5-10036), washed 3 times with TBS/0.1% Tween-20, and imaged with Li-Cor Odyssey CLx imaging system.

CUT&Tag

Centromere identification

We trimmed paired-end reads using trimgalore (v0.4.4) [83] (trim_galore—paired—nextera—length 75—phred33—no_report_file–fastqc) and assessed read quality with FASTQC.

We mapped reads against the reference genome with bwa (v7.4) using the BWA-MEM algorithm (default parameters). We used the heterochromatin-enriched assemblies of D. melanogaster [40], D. simulans, D. sechellia, and D. mauritiana [33]. We converted the resulting sam alignment files into bam files and sorted using respectively samtools (v1.11) view and sort command. We removed PCR duplicates using Markduplicates from Picardtools (v2.12.0) (https://broadinstitute.github.io/picard/). Because we are working with highly repetitive sequences, we analyzed both the unique and multi-mapping reads. We thus performed 2 different filtering based on mapping quality using samtools view [84]. To include multi-mapping reads, we use the following parameters: -b -h -f 3 -F 4 -F 8 -F 256 -F 2048. To keep only the uniquely mapping reads, we use the following parameters: -b -h -f 3 -F 4 -F 8 -F 256 -F 2048 -q30.

We estimated read coverage using the bamCoverage command from deeptools (v3.5.1) using the option—scaleFactor -bs 1—extendReads and normalized the read coverage to RPM.

We called peaks based on fragment size using MACS2 callpeak [85] (v2017-10-26) (option -f BAMPE -g dm -q 0.01 -B—call-summits) and performed an IDR analysis (https://github.com/nboley/idr) to identify high confidence peaks that overlapped between replicates (IDR <0.05, S1 Table). The localization of these high confident peaks allowed us to identify the candidate centromere contigs (S1 Fig).

We calculated mappability along each centromere candidate contig using GenMap (https://github.com/cpockrandt/genmap) with 150-mers to mimic read length.

Repeat enrichment analyses

For this analysis, we used the multi-mapping bam file. We annotated the reference genome (S1–S4 Files) using a custom repeat library specific to each species (S5–S8 Files) with Repeatmasker [86] (options -no_is -a -inv -pa 20 -div 20). Using htseq-count [87], we counted the number of reads that map to each repeat and calculated RPM. To determine the enrichment, we normalized the RPM counts for CENP-A by RPM counts for IgG (negative control). The 25% most enriched repeats are presented in S10 Table, and the top 20 most enriched repeats among all replicates are presented in Fig 1B, 1E, and 1H.

To explore origins of the centromeric complex satellites we blasted (blastn with default parameter) the consensus sequences of 500-bp, 136-bp, and 365-bp satellites against the genome of D. melanogaster [47], the simulans clade [33], and more distant species, D. yakuba, D. annanassae, D. pseudoobscura, D. erecta, and D. virilis [88]. All hits are reported in S3–S5 Tables.

The dotplots of the Y chromosome centromeres cenY (S9 Fig) were generated using re-DOT-able v1.1 (https://www.bioinformatics.babraham.ac.uk/projects/redotable/).

G2/Jockey-3 evolutionary analyses

We surveyed G2/Jockey-3 evolution in additional species with improved genome assemblies of D. simulans, D. sechellia, and D. mauritania [89] and publicly available Nanopore assemblies of D. yakuba, D. erecta, and D. ananassae [90]. We identified G2/Jockey-3 sequences with 2 complementary methods. First, we annotated each genome assembly with our custom Drosophila TE library including the D. melanogaster G2/Jockey-3 consensus sequence [71] using Repeatmasker v4.1.0. The annotations and 500-bp flanking regions were extracted with BEDTools v2.29.0 [81] and aligned with MAFFT [91] to generate a species-specific consensus sequence with Geneious v.8.1.6 [92]. Each assembly was annotated again using Repeatmasker with the appropriate species-specific G2/Jockey-3 consensus sequence. Second, we constructed de novo repeat libraries for each species with RepeatModeler2 v.2.0.1 [93] and identified candidate G2/Jockey-3 sequences which shared high similarity with G2/Jockey-3 in D. melanogaster identified with BLAST v.2.10.0. We did the same with Jockey-1 (LINEJ1_DM) as confirmation of our methods and to use it as an outgroup for the TE fragment alignment. We removed candidates shorter than 100 bp from the analysis. We identified ORFs within consensus TE sequences with NCBI ORFfinder. We used Repeatmasker to annotate the genome assemblies with the de novo Jockey-3 consensus sequences. To infer a phylogenetic tree of TEs, we aligned G2/Jockey-3 fragments identified in each species with MAFFT and retained sequences corresponding to the ORF bounds of the consensus sequences. We removed ORF fragments <400 bp. We inferred the tree with RAxML v.8.2.11 [94] using the command “raxmlHPC-PTHREADS -s alignment_Jockey-3_melsimyak_400_ORF2_mafft.fasta -m GTRGAMMA -T 24 -d -p 12345 -# autoMRE -k -x 12345 -f a.”

Oligopaint design and synthesis

We designed Oligopaint probes targeting 500-bp, 136-bp, 365-bp, Rsp-like, HTTs, and the Y centromere islands of each species using ProbeDealer [95] with some modifications. We extracted the fasta sequences containing the target repeat from the reference genomes and used it as the input for ProbeDealer. After designing all the possible oligo probes, ProbeDealer usually maps them back against the reference genome to eliminate multi-mapping oligos. Because we are working with highly repetitive sequences, we skipped this step. We mapped the oligos to the reference genome to manually inspect for potential off targets. The final oligo list is in S11 Table. Oligopaints libraries were synthesized by Genscript. We then synthesized and purified each Oligopaint sublibrary as described in [29].

IF-FISH on mitotic chromosome

We dissected brains from third instar larvae (both sexes) in PBS, incubated 8 min in 0.5% sodium citrate. We fixed for 6 min in 4% formaldehyde, 45% acetic acid before squashing. We squashed the brains between a poly-lysine slide and coverslip and before immersing in liquid nitrogen. After 5 min in PBS and 10 min in PBS, we blocked slides for at least 30 min in blocking buffer (3%BSA, 1% goat serum in PBST). For immunofluorescence (IF), we incubated slides in primary antibody (α-CENP-C12 1:100) overnight at 4°C. We washed slides 3 times for 5 min in PBST. We incubated slides in secondary antibody (anti-rabbit 1:500) for 1 to 3 h at room temperature and washed 3 times for 5 min in PBST. We post-fixed the slides using 10% formaldehyde diluted in 4XSSC, incubating 20 min at room temperature and washed 3 times for 3 min with 4XSSC and one time for 5 min with 2XSSC. For the hybridization, we used 20 pmol of primary probes (S11 Table) and 80 pmol of the secondary probes (S12 Table) in 50 μl of hybridization buffer (50% formamide, 10% dextran sulfate, 2XSSC). We heated slides for 5 min at 95°C to denature and incubated them overnight at 37°C in a humid chamber. We then washed the slides 3 times for 5 min with 4XSSCT and 3 times for 5 min with 0.1SSC before mounting in slowfade DAPI.

We use acetic acid to obtain high-quality chromosome spreads; however, this also removes histones. Thus, it is not feasible to perform IF on mitotic spread using anti-histone antibodies, such as CENP-A. We therefore use CENP-C—a kinetochore protein that marks centromeres and overlaps with CENP-A [37].

IF-FISH on chromatin fibers

We dissected third instar larval brains in 1XPBS (3 to 4 brains per slide) and incubated in 250 μl of 0.5% sodium citrate with 40 μg of dispase-collagenase, for 12 min at 37°C. The tissue was transferred to a poly-lysine slide using Shandon Cytospin 4 at 1,200 RPM for 5 min. We positioned slides vertically in a tube containing the Lysis buffer (500 nM NaCl, 25 mM Tris-HCL (pH 7.5), 250 nM Urea, 1% Triton X-100) and incubated for 16 min. For the fiber stretching, we allow the buffer to slowly drain from the tube with the hole at the bottom (by removing the tape). A steady flow rate will generate a hydrodynamic drag force which generates longer and straighter fibers. We incubated slides in a fixative buffer (4% formaldehyde) for 10 min and then 10 min in 1XPBST (0.1% Triton). For the IF, we first blocked the slides for 30 min in blocking buffer (1.5% BSA in 1XPBS). We incubated slides overnight at 4°C with the primary antibody (α-CIDH32, 1:100) and washed 3 times for 5 min in 1xPBST. We incubated slides with the secondary antibody (anti-chicken, 1:500) for 1 to 3 h at room temperature and washed 3 times for 5 min with 1XPBST. We post-fixed the slide with 10% formaldehyde for 20 min and washed 3 times for 5 min in 1XPBST. We then incubated slides for 10 min in 2XSSCT at room temperature and 10 min in 2XSSCT—50% formamide at 60°C. For the hybridization, we used 40 pmol of primary probes (S11 Table) and 160 pmol of the secondary probes (S12 Table) in 100 μl of hybridization buffer (50% formamide, 10% dextran sulfate, 2XSSC). We heated slides for 5 min at 95°C to denature and incubated them overnight at 37°C in a humid chamber. We then washed the slides 15 min with 2XSSCT at 60°C, 15 min with 2XSSCT at room temperature, and 10 min with 0.1XSSC at room temperature. We incubated slides for 5 min in DAPI (1 mg/ml) before mounting in SlowFade Gold (Invitrogen S36936).

G2/Jockey-3 IF-FISH

D. simulans, D. sechellia, and D. mauritiana third instar larval brains were dissected in 1× PBS and all attached tissue or mouth parts were removed with forceps. Brains were immersed in 0.5% sodium citrate solution for 8 min in a spot well dish. The tissue was placed in a 6 μl drop of 45% acetic acid, 2% formaldehyde on a siliconized (Rain X) coverslip for 6 min. A poly-lysine–coated slide was inverted and placed on the brains to make a sandwich. After flipping the slide and gently removing excess fixative between a bibulous paper, the brain was squashed using the thumb by firmly pressing down. Slides were then immersed in liquid nitrogen and the coverslip flipped off using a razor blade and transferred to 1× PBS for 5 min to rehydrate before proceeding with IF-FISH. Slides were then washed with 1× PBST (0.1% Triton X-100) for 5 min on a rotator, repeated 3 times. Slides were then transferred to a coplin jar containing blocking solution (1% BSA in 1× PBST) for 30 min while rocking. Diluted antibodies were applied to the slides coating the brains with 50 μl of primary antibodies, covered with parafilm and stored in a dark chamber at 4°C overnight. The following day, slides were washed 4 times with 1× PBST for 5 min while rocking. Secondary antibodies diluted with block were applied to the brains and covered with parafilm, then incubated at room temperature for 1 h. After the 1 h incubation, slides were washed 4 times in 1× PBST for 5 min while rotating. Slides were then post-fixed with 3.7% formaldehyde diluted with 1× PBS for 10 min in the dark. Slides were washed for 5 min in 1× PBS while rotating before proceeding to FISH. The following FISH protocol for G2/Jockey-3 labeling and the synthesis of the G2/Jockey-3 probe was performed as described in the methods of Chang and colleagues [29]. Slides were dehydrated in an ethanol row (3 min washes in 70%, 90%, and 100% ethanol) and allowed to air-dry completely for a few minutes. Probe mix (20 μl) containing 2xSSC, 50% formamide (Sigma-Aldrich), 10% dextran sulfate (Merck), 1 μl RNase cocktail (Thermo Fisher), and 100 ng of DIG-labeled G2/Jockey-3 probe was boiled at 80°C for 8 min, incubated on ice for 5 min, and then applied to slides, covered with a glass coverslip, and sealed with paper cement. Sealed slides were denatured on a slide thermocycler for 5 min at 95°C and incubated at 37°C overnight to hybridize. Slides were then washed 3 times in a coplin jar for 5 min in 2xSSC, 50% formamide at 42°C. Slides were then washed 3 times for 5 min in 0.1xSSC at 60°C, and then blocked in block buffer 1% BSA, 4xSSC, 0.1% Tween-20 at 37°C for 45 min. Slides were incubated with 50 μl of block buffer containing a fluorescein-labeled anti-DIG antibody (sheep, 1:100, Roche) for 60 min at 37°C. Slides were then washed 3 times for 5 min in 4xSSC, 0.1% Tween-20 at 42°C. Slides were washed with 1× PBS briefly in a coplin jar and finally mounted on a coverslip with Slowfade and DAPI, then sealed with nail polish.

Image acquisition

We imaged using a LEICA DM5500 microscope with a 100×/oil immersion objective or Delta vision using an Olympus UPLansApo 100×/1.40 oil immersion objective, maintaining all exposures consistent across each experiment. Images obtained with the Deltavision microscope were deconvolved with Softoworks using 5 iterations with the “conservative” setting. Images were edited, cropped, and pseudocolored using Fiji.