Rapid evolutionary turnover underlies conserved lncRNA-genome interactions

Abstract

Many long noncoding RNAs (lncRNAs) can regulate chromatin states, but the evolutionary origin and dynamics driving lncRNA-genome interactions are unclear. We adapted an integrative strategy that identifies lncRNA orthologs in different species despite limited sequence similarity, which is applicable to mammalian and insect lncRNAs. Analysis of the roX lncRNAs, which are essential for dosage compensation of the single X chromosome in Drosophila males, revealed 47 new roX orthologs in diverse Drosophilid species across ∼40 million years of evolution. Genetic rescue by roX orthologs and engineered synthetic lncRNAs showed that altering the number of focal, repetitive RNA structures determines roX ortholog function. Genomic occupancy maps of roX RNAs in four species revealed conserved targeting of X chromosome neighborhoods but rapid turnover of individual binding sites. Many new roX-binding sites evolved from DNA encoding a pre-existing RNA splicing signal, effectively linking dosage compensation to transcribed genes. Thus, dynamic change in lncRNAs and their genomic targets underlies conserved and essential lncRNA-genome interactions.

Keywords: ChIRP; Drosophila; RNA structure; dosage compensation; lncRNAs; roX.

Figure 1.

Summary of lncRNA ortholog search strategy and queried species. (A) The search strategy found lncRNA orthologs in query species by integrating synteny, microhomology, and secondary structure features of a known lncRNA. The search features were iteratively refined by bootstrapping new ortholog candidates and the phylogenetic relationships between queried species. To initiate the search, a priori knowledge of the lncRNA in only a single species is needed. (B) Phylogenetic tree of the 35 Drosophilid species queried in this study. Whole-genome sequencing (WGS) assemblies were available for 27 species. Nine roX1 and 10 roX2 orthologs have previously been described ([K] known roX ortholog from Amrein and Axel 1997 [i]; Meller et al. 1997 [ii]; Park et al. 2007 [iii]; Byron et al. 2010 [iv]; Alekseyenko et al. 2013 [v]); our search identified 47 new roX orthologs. (Y) New ortholog; (N) no ortholog found. X chromosome karyotypes are indicated by Müller elements. (n.d.) No data.

Figure 2.

Identified roX candidates are bona fide orthologs despite sequence divergence. (A,B) Heat map showing the sequence conservation between the identified roX1 (A) and roX2 (B) ortholog candidates relative to the lower limit of homology (scrambled sequences, 36%). Phylogenetic trees as in Figure 1B. Red dashed boxes highlight exceptionally poor conservation between distantly related species. (C) RT–PCR of roX1, roX2, and GPDH RNA in male and female flies. roX1 and roX2 orthologs exhibit strong male-biased expression; GPDH mRNA is a sex-independent control. (n/a) No ortholog found. (D) RNA FISH of roX1 and roX2 in polytene chromosomes from male and female Drosophila busckii larvae. roX2 paints the male X chromosome (white arrowhead) but not the female X; roX1 was not detected. (E) Rescue of male lethality in roX-null D. melanogaster (D.mel) males by transgenic D. busckii (D.bus) roX2 or chimeric busckii–melanogaster roX2. RNA cartoons depict secondary structures, with roXboxes (red) and inverted roXboxes (blue) indicated. Error bars show standard deviation. Expression was calculated relative to wild-type roX2 transgene ± standard deviation.

Figure 3.

Genomic occupancy maps of roX orthologs highlight the loss of roX1–roX2 functional redundancy in other species. (A) ChIRP-seq identifies the genome-wide binding sites of an RNA target, performed directly from chromatin prepared from Drosophila larvae. (B) roX1 and roX2 signal enrichment (ChIRP/input) in 1-kb windows of MEs A–F in four Drosophila species. Signal is enriched on the X chromosome. roX1 enrichment is lower than roX2 in D. willistoni (D.wil), D. virilis (D.vir), and D. busckii (D.bus). (N.D.) No data, as no genome scaffolds aligned to ME-F. (C, left) The log ratio of roX1 to roX2 ChIRP signal at binding sites shows that roX2 is the dominant roX RNA in D. willistoni, D. virilis, and D. busckii. Average roX1/roX2 bias is shown as fraction. (Right) Known functional domains (red outlines), secondary structures, and roXboxes (filled red or blue rectangles) of roX1 are absent in D. willistoni, D. virilis, and D. busckii. Only D. melanogaster roX1 has a full complement of these repetitive elements. See also Supplemental Figure S5. (D) Rescue of male lethality in roX-null D. melanogaster males improves with the number of repetitive roXbox stem–loops. LacZ with D. melanogaster stem–loop (SL) rescues poorly, D. virilis roX1-D3 rescues modestly, and addition of the D. melanogaster stem–loop to D. virilis roX1-D3 further improves rescue, approaching the wild-type D. melanogaster roX1 rescue efficiency. Error bars and relative expression are as in Figure 2E.

Figure 4.

Evolutionary dynamics of roX-bound HASs. (A,B) roX2 ChIRP-seq tracks in representative windows on the X chromosome. (A) Some HASs are evolutionarily dynamic. The strong HAS in the 3′ untranslated region (UTR) of D. melanogaster HDAC4 is absent or present elsewhere in other species. (B) Other HASs are evolutionarily conserved, such as the HAS in the last intron of CG14806 (gray highlight). (C) roX2 ChIRP-seq identified hundreds of HASs on the X chromosome of each species. (n) Total number of HASs. (D) HASs contain GA dinucleotide repeats, characteristic of the MRE motif. (E) Gene-level conservation of HASs between four species. Forty-five genes are roX-bound (overlapping or neighboring a HAS) in all four species. D. willistoni has the most species-specific roX-bound genes because of its larger X chromosome. (Top) Number of shared roX-bound genes. (Bottom) Number of HASs within shared genes in each species. (F) Pairwise proximity of orthologous HASs between D. melanogaster and D. willistoni. About 60 HASs directly overlap in genomic lift-over (distance = 0); however, if an exact HAS homolog is lost (distance >0), another HAS is likely nearby. There are more overlapping HASs or nearby HASs (<30 kb) than expected by random permutation of HASs within each chromosome or over the whole genome. See also Supplemental Figure S12.

Figure 5.

Transplanted roX RNAs from other species bind to D. melanogaster-specific binding sites. (A) ChIRP-seq of roX RNAs transplanted into roX-null D. melanogaster. (Left) Transgenic roX constructs showing structures and roXboxes. Dotted lines indicate mutated structures. (Right) A representative ∼150-kb window on the X chromosome showing ChIRP-seq tracks for each roX transgene relative to wild-type D. melanogaster roX1 and roX2 (black tracks) and transgenic D. melanogaster roX1 and roX2 (green tracks). Indicated above is the HDAC4 locus, as in Figure 4A. D. virilis roX1 and D. busckii roX2 transgenes exhibit the same binding pattern as D. melanogaster roX RNAs, even at species-specific HASs. The mutated roX2 does not bind the X. (B) Pairwise comparisons of ChIRP-seq signal enrichment for each 1-kb window on the X chromosome. roX1 and roX2 loci are indicated as outliers due to genetic deletion of each locus. Correlation scores and overall enrichment are lower for D. virilis roX1 and D. busckii roX2 relative to D. melanogaster roX RNAs; log-scaled axes.

Figure 6.

HASs exapt pre-existing regulatory signals and are selected for even spacing on the X chromosome. (A) HASs are enriched in genic, noncoding regions of the genome, primarily within introns and 3′ UTRs. HASs are subcategorized by strong roX occupancy and/or high evolutionary conservation. Fold enrichment over the genomic distribution is shown. (B) Intronic HASs are proximal to PPTs. See also Supplemental Figure S14A. (C) The MRE motifs within HAS classes exhibit significant and distinct orientation biases. Intronic HASs are biased in the reverse complement orientation (CT repeat), whereas exonic HASs are biased in the forward orientation (GA repeat); intergenic HASs have no bias. (D) Alternative HAS spacing models on the X chromosome. HASs may be clustered together, randomly spaced, or evenly spaced. The observed distribution is more even and less clustered than random. (E) The difference between the observed and random HAS (conserved, strong only) distributions on the X chromosome. The positive value near the theoretically perfect spacing distance indicates a more even spacing model relative to random spacing; conversely, the negative value at short distances indicates a less clustered model relative to random spacing. See also Supplemental Figure S13.

Figure 7.

Models of roX and roX-binding site evolution. (Left) lncRNAs like roX can evolve function through the accumulation and maintenance of repetitive structures or sequence elements and gene duplication. (Right) lncRNA-binding sites are evolutionarily dynamic, losing function at one genetic element while gaining function nearby. The DCC and roX RNAs can coopt existing PPTs within gene introns, which are refined into the MRE sequence.