Non-B-Form DNA Is Enriched at Centromeres

Abstract

Animal and plant centromeres are embedded in repetitive "satellite" DNA, but are thought to be epigenetically specified. To define genetic characteristics of centromeres, we surveyed satellite DNA from diverse eukaryotes and identified variation in <10-bp dyad symmetries predicted to adopt non-B-form conformations. Organisms lacking centromeric dyad symmetries had binding sites for sequence-specific DNA-binding proteins with DNA-bending activity. For example, human and mouse centromeres are depleted for dyad symmetries, but are enriched for non-B-form DNA and are associated with binding sites for the conserved DNA-binding protein CENP-B, which is required for artificial centromere function but is paradoxically nonessential. We also detected dyad symmetries and predicted non-B-form DNA structures at neocentromeres, which form at ectopic loci. We propose that centromeres form at non-B-form DNA because of dyad symmetries or are strengthened by sequence-specific DNA binding proteins. This may resolve the CENP-B paradox and provide a general basis for centromere specification.

Fig. 1.

Patterns of DNA dyad symmetry at eukaryotic centromeres. (A) Examples of dyad symmetries in centromeric DNA sampled randomly from human, African green monkey, mouse, chicken, and fission yeast WGS data sets. (B) Dyad density, which is defined for a given sequence as the total number of palindromic positions with palindrome length > 4 and spacer length < 20 normalized by sequence length, at centromeres relative to composition-matched background genomic regions. Asterisks indicate two-sample Kolmogorov-Smirnov P < 0.05. (C) Enrichment (relative to permuted sequence) of dyad symmetries over varying palindrome lengths in read ends mapping to centromeres or from sequences sampled from genome assemblies for a variety of organisms. The displayed phylogeny is based on NCBI Taxonomy annotations.

Fig. 2.

Centromeric dyad symmetries are predicted to adopt non-B-form structures. (A) Scores from SIST model predictions of DNA melting (left) and cruciform extrusion (right) for centromeric sequences and composition-matched background genomic regions from human, African green monkey, mouse, chicken, and fission yeast genomes. Asterisks indicate two-sample Kolmogorov-Smirnov P < 0.05. (B) Examples of minimum free energy secondary structure predictions for randomly selected α-satellite monomers from human and African green monkey. (C) DNA secondary structure folding free energy predictions for read ends mapping to centromeres or from sequences sampled from genome assemblies from the indicated species. The displayed phylogeny is based on NCBI Taxonomy annotations.

Fig. 3.

Non-B-form DNA detected experimentally at dyad-depleted functional human and mouse centromeres. (A) Abundance of heterochromatic major (MaSat) and centromeric minor (MiSat) satellite fragments (left) and the minimal CENP-B box (right) in mouse WGS reads. (B) DNA secondary structure free energy distributions from computational prediction for MaSat and MiSat-containing Sanger reads. (C) Fraction of permanganate-seq reads from resting (R) and LPS-activated (A) cells mapping to the masked mouse genome (Masked mm10) or Sanger reads containing MiSat or MaSat monomers. (D) Examples of Sanger reads harboring MiSat and CENP-B box sequences (left) or devoid of predicted centromeric features (right) and signal from CENP-A ChIP-seq, permanganate-seq, and dyad symmetry analysis. (E) Correlation between total permanganate-seq signal from activated B cells (normalized to control) and input-normalized CENP-A occupancy for MiSat-containing Sanger reads. Scatter plot points are colored based on CENP-B box score tertile, where scores are defined as the sum of scores for all FIMO-defined CENP-B boxes occurring on a read. (F) Fraction of permanganate-seq reads aligning to the repeat-masked hg38 assembly and HuRef Sanger alphoid reads normalized to number of reads from WGS of HuRef mapping to the respective assemblies. (G) Examples of permanganate-seq, CENP-A ChIP, dyad symmetry, and predicted DNA melting and cruciform transition probabilities for functionally active (D5Z2) and inactive (D5Z1) α-satellite repeat arrays. Note that the CENP-A ChIP tracks are on different scales. (H) Correlation between total permanganate-seq signal and CENP-A occupancy for alphoid Sanger reads. Scatter plot points are colored based on CENP-B box score tertile.

Fig. 4.

Primate CENP-B binding is restricted to dyad-depleted great ape centromeres. (A) Estimated abundance of α-satellite sequences in a sampling of simian primates. (B) Enrichment of minimal CENP-B box sequences in raw reads from WGS. The displayed phylogeny is a chronogram based on mitochondrial genomes and is adapted from the 10kTrees Project (Arnold et al. 2010). Abundance of α-satellite sequence in aligned reads (C) and abundance of matches to the minimal CENP-B box sequence in raw reads (D) in CUT&RUN experiments performed in human (K562) and African green monkey (Cos-7) cell lines normalized to α-satellite abundance in WGS reads.

Fig. 5.

The human Y centromere and vertebrate neocentromeres are associated with dyad symmetries and non-B-form DNA. (A) Predicted ensemble free energies for DYZ3 and non-DYZ3 alphoid satellites classified based on CENP-A ChIP enrichment and centromere activity in artificial chromosome assays (Hayden et al. 2013; Henikoff et al. 2015). (B) Examples of minimum free energy structures for DYZ3 and D5Z2 alphoid fragments. A human chromosome 13 neocentromere (C) and a chicken chrZ neocentromere (D) with profiles from CENP-A ChIP-seq and SIST-predicted DNA melting and cruciform extrusion probabilities (left panels). Dyad symmetry and SIST DNA melting and cruciform extrusion scores for neocentromeres (“neo”) and composition-matched noncentromeric background genomic intervals (right panels). Data from genome-wide analysis of palindrome formation with sequencing (GAP-seq), which was performed in human cell lines, is also included in (A). Asterisks indicate two-sample Kolmogorov–Smirnov P < 0.05.

Fig. 6.

Centromeric dyad symmetries are features of yeast centromeres. Average CenH3 signal, dyad symmetry, and SIST melt and cruciform profiles (left panels) and comparison of dyad densities and SIST melt and cruciform scores for centromeres versus nucleotide composition-matched, randomly selected genomic intervals (right panels) in S. cerevisiae (A). (B) Predicted ensemble free energy distributions for centromeric sequences from sensu strictu and sensu lato saccharomycetes with well-annotated genomes. (C) Enriched CDEI motifs for saccharomycetes and average estimated dyad densities over CDEII.

Fig. 7.

Models for genetic centromere specification. (A) Summary of centromeric DNA sequence type, association with helix-deforming DNA binding protein, dyad symmetry, and predicted secondary structure forming tendency for various eukaryotes. (B) Repetitive centromeres vary in their predilection for forming cruciform structures exemplified by alphoid sequences of OWMs, which are predicted to form stable non-B-form DNA structures, and great apes, which do not preferentially adopt non-B-form DNA structures. In great apes, CENP-B binding may facilitate formation of non-B-form DNA such as cruciforms. Cruciform structures are recognized by HJURP/Scm3 chaperones, which deposit CENP-A nucleosomes. (C) Alternatively, OWM AS units may be spontaneously transcribed, while CENP-B binding may facilitate transcription of great ape alphoid units, with the RNAs contributing to deposition of CENP-A.