The formation and propagation of human Robertsonian chromosomes

Abstract

Robertsonian chromosomes are a type of variant chromosome that is commonly found in nature. Present in 1 in 800 humans, these chromosomes can underlie infertility, trisomies and increased cancer incidence^1-5. They have been recognized cytogenetically for more than a century⁶, yet their origins have remained unknown. Here we describe complete assemblies of three human Robertsonian chromosomes. We identified a common breakpoint in SST1, a macrosatellite DNA located on chromosomes 13, 14 and 21, which commonly undergo Robertsonian translocation. SST1 is contained within a larger shared homology domain⁷ that is inverted on chromosome 14, which enables a meiotic crossover event that fuses the long arms of two chromosomes. Robertsonian chromosomes have two centromeric DNA arrays and have lost all ribosomal DNA. In two cases, we find that only one of the two centromeric arrays is active. In the third case, both arrays can be active but owing to their proximity, they are often encompassed by a single outer kinetochore. Thus a combination of array proximity and epigenetic changes in centromeres facilitates the stable propagation of Robertsonian chromosomes. Investigation of the assembled genomes of chimpanzee and bonobo highlights that the inversion on chromosome 14 is unique to the human genome. Resolving the structural and epigenetic features of human Robertsonian chromosomes at a molecular level provides a foundation for a broader understanding of the molecular mechanisms of structural variation and chromosome evolution.

Fig. 1. Complete assembly of ROBs.

a, Working model for the dependence of ROB formation on recombination between SST1 repeats (pink triangles) located in PHRs (coloured bands) on the short arms of human chromosomes 13, 14 and 21. The adjacent 45S rDNA arrays facilitate 3D proximity by co-locating in nucleoli. b, Schematic representation of the main SST1 arrays and flanking sequences in acrocentric chromosomes from the CHM13 genome. This region is similar on chromosomes 13 and 21 and is inverted on chromosome 14. c–e, Representative images of ROBs from GM03786 (c), GM04890 (d) and GM03417 (e) cell lines. Left image, chromosome labelled with an SST1 probe (magenta) and whole chromosome paints as indicated. Centre image, chromosome labelled with an SST1 probe (magenta) and centromeric satellite probes for cen14/22 (orange) and cen13/21 (green). DNA was counterstained with DAPI. Right image, magnified view showing SST1 localization between the two centromere arrays. Scale bar, 1 µm. Right, averaged intensity profiles of lines drawn through the centromeres of multiple ROBs (GM03786: n = 10, GM04890: n = 20, GM03417: n = 11). Intensity profiles were aligned to the peak of the Gaussian of the SST1 signal and normalized to the maximum intensity of each channel. Error bars denote s.d. Bottom, synteny plots comparing the assembled ROB to the CHM13 genome sequence. The structure of each fused region is shown in detail. a.u., arbitrary units.

Fig. 2. Evidence for SST1-mediated interchromosomal exchange in human genomes.

a, All SST1 monomers from previous analysis of CHM13 and from HG002 were collected and phylogenetic analysis was performed using the maximum-likelihood method based on the best-fit substitution model (Kimura two-parameter + G, parameter = 5.5047) inferred by Jmodeltest2 with 1,000 bootstrap replicates. Bootstrap values higher than 75 are indicated at the base of each node. The colour indicates the source chromosome and the shape indicates the source genome. Three major subfamilies were identified: sf1, primarily on the acrocentrics (Acros); sf2, primarily on the remaining autosomes; and sf3, primarily on the Y chromosome. Black arrows indicate the location on the phylogenetic trees of sf2 monomers S and L from the acrocentric chromosomes (Fig. 1b). b, Predicted PRDM9 DNA binding site frequency (mean sites per kb, each dot indicates one haplotype) in SST1 arrays in n haploid genomes are plotted by chromosome. ANOVA analysis with the two-sided Tukey–Kramer test for pairwise mean comparisons. c, Schematic representation of the three subfamilies of SST1. SST1-sf1 has a central gap and a predicted PRDM9 DNA binding site (red box). d, A segmental duplication of 27 kb or larger was identified on several autosomes in CHM13 that includes Y-like α-satellite DNA (α-sat) and SST1-sf3. Phylogenetic analysis was performed using the maximum-likelihood method and general time reversible (GTR)-plus-gamma substitution parameters. Bootstrap values are shown. e, Comparison of overlaps between segmental duplications (SDs; ≥10 kb) and random regions or SST1 monomers across 147 genomes. Distributions show the number of overlaps versus density. A permutation test with 10,000 iterations per genome was used to generate random region overlaps. The significant difference between distributions (Wilcoxon signed-rank test, paired, two-sided) indicates non-random association between segmental duplications and SST1 regions.

Fig. 3. Evidence for exchange of SST1 on rDNA array-bearing chromosomes in chimpanzee and bonobo genomes.

a, Ideograms of all the rDNA array-bearing chromosomes in human, chimpanzee and bonobo, annotated with the human numbering system (indicated by Hsa prefix). The directionality of 45S rRNA gene arrays (grey) and SST1 arrays (coloured bars) are indicated with arrowheads. b, Predicted PRDM9 binding sites were identified in the chimpanzee genome, and the number of sites per kb is plotted for SST1 arrays for the indicated subfamily. Random regions of the genome (randBins and randGC (GC-matched random regions)) were used to determine background. c, All SST1 monomers from the chimpanzee genome were subjected to phylogenetic analysis using the maximum-likelihood method. The colour indicates the source chromosome. The SST1 monomers from Hsa13, Hsa14, Hsa18, Hsa21 and Hsa22 chromosomes form a single branch, indicating a high degree of similarity. d, All SST1 monomers from the bonobo genome were subjected to phylogenetic analysis using the maximum-likelihood method. The SST1 monomers from chromosomes 14 and 22 form a single branch, indicating a high degree of similarity. e, SST1 monomers from human (Hs), chimpanzee (Pt) and bonobo (Pp) were subjected to phylogenetic analysis using the maximum-likelihood method. The three subfamilies are apparent.

Fig. 4. Centromere activity in dicentric ROBs.

a–c, ImmunoFISH, DNA methylation and CENP-A CUT&Tag analyses were performed for GM03786 (a), GM04890 (b) and GM03417 (c). Left, representative SIM images of ROBs labelled by immunoFISH with centromeric satellite probes for cen14/22 (orange), cen13/21 (green) and CENP-C antibody (red). DNA was counterstained with DAPI. Bottom images, magnified views depicting single CENP-C foci on cen14 in GM03786 (a) and GM04890 (b), and double CENP-C foci on cen21 and cen14 in GM03417 (c). Scale bars, 1 µm. Bottom left, averaged intensity profiles of lines drawn through the individual kinetochore regions of sister chromatids of multiple ROBs. GM03786: n = 22 (11 chromosomes) (a). GM04890: n = 26 (13 chromosomes) (b); GM03417: n = 24 (12 chromosomes) (c). Intensity profiles were aligned to the peak of the Gaussian of the cen14 signal and normalized to the maximum intensity of each channel. Error bars denote s.d. Top centre and right, corresponding heat maps of sequence similarity calculated for 5-kb bins for each centromere. Below the heat maps, DNA methylation tracks show methylation calls from ONT (orange) or PacBio HiFi (turquoise) sequencing, with hypomethylated regions suggesting active centromere localization. Active centromere regions are indicated by CENP-A enrichment on CUT&RUN (blue) and CUT&Tag (black) tracks.

Extended Data Fig. 1. Assembly graphs of the t(14;21)-bearing GM03417 cell line.

Assembly graph of GM03417 integrating PacBio and ONT reads visualized using Bandage (https://doi.org/10.1093/bioinformatics/btv383). Each chromosome is mostly resolved as a single connected component, with the exception of the acrocentric chromosomes (13, 14, 15, 21 and 22), which are joined by the highly similar rDNA arrays. The graph is colored by Hi-C information used to phase both haplotypes (haplotype 1, red; haplotype 2, blue). The inset shows a portion of the graph involving the acrocentric chromosomes, highlighting the graph node representing the Robertsonian translocation, which connects chromosomes 14 and 21.

Extended Data Fig. 2. Assembly graphs of the t(13;14)-bearing GM03786 cell line.

Assembly graph of GM03786 integrating PacBio and ONT reads visualized using Bandage. Each chromosome is mostly resolved as a single connected component, with the exception of the acrocentric chromosomes (13, 14, 15, 21 and 22), which are joined by the highly similar rDNA arrays. The graph is colored by Hi-C information used to phase both haplotypes (haplotype 1, red; haplotype 2, blue). The inset shows a portion of the graph involving the acrocentric chromosomes, highlighting the graph node representing the Robertsonian translocation, which connects chromosomes 13 and 14.

Extended Data Fig. 3. Assembly graphs of the t(13;14)-bearing GM04890 cell line.

Assembly graph of GM04890 integrating PacBio and ONT reads visualized using Bandage. Each chromosome is mostly resolved as a single connected component, with the exception of the acrocentric chromosomes (13, 14, 15, 21 and 22), which are joined by the highly similar rDNA arrays. The graph is colored by Hi-C information used to phase both haplotypes (haplotype 1, red; haplotype 2, blue). The inset shows a portion of the graph involving the acrocentric chromosomes, highlighting the graph node representing the Robertsonian translocation, which connects chromosomes 13 and 14.

Extended Data Fig. 4. Cytogenetic evidence for the structural changes in ROBs.

Fluorescence in situ hybridization (FISH) was used to visualize the arrangement of centromeres, SST1 and 45S rDNA arrays on acrocentric chromosomes in the ROB-bearing cell lines. A. Representative nuclei of ROB cell lines labeled by FISH with probes for Cen 14/22 (orange), Cen 13/21 (green) and DAPI. Bar, 10 µm. Magnified insets show centromeres of ROB chromosomes (bar 1 µm). B. Extended karyograms of GM03786, GM04890, and GM03417 cell lines labeled by FISH with SST1 probe (magenta) and whole chromosome paints (14 in orange, 13 and 21 in green). All acrocentric chromosomes and other chromosomes with detectable SST1 signals are shown. Acrocentric chromosomes lacking SST1 signal are denoted with blue numbers. DNA was counter-stained with DAPI. C. Extended karyograms of acrocentric chromosomes only labeled by FISH with rDNA probe (red), whole chromosome paints, and DAPI. Note lack of rDNA signal on ROBs in all cell lines.

Extended Data Fig. 5. SST1 array deletion on chromosome 14.

Alignment of chromosome 14 contigs from 29 haplotypes, including 26 from the Human Pangenome Reference Consortium (HPRC) year 1 dataset and 3 from ROB cell lines (GM04890, GM03417, and GM03786, highlighted in bold). Each row represents a distinct haplotype aligned against the T2T-CHM13 reference chromosome 14 (bottom track). Gold bars indicate aligned regions, while gaps represent absent and unaligned sequences. The vertical black line marks the position of the SST1 array within the Pseudo-Homolog Region (PHR). The bottom track displays the CHM13 centromeric satellite annotation, with different colors representing various satellite families. Approximately 34.5% (10 out of 29) of the analyzed chromosome 14 haplotypes lack the SST1-containing PHR, including the non-ROB chromosome 14 in the GM03786 cell line. Haplotypes are clustered based on sequence similarity to better visualize the deletion pattern. The x-axis shows the genomic position in megabases (Mb).

Extended Data Fig. 6. Schematic representation of common ROB formation.

Top panel: Structure of the short arms of chromosomes 13/21 and chromosome 14 before fusion. Centromeric, rDNA, and SST1 arrays are highlighted. Note the inverted orientation of the SST1 type 1 array on 14 relative to 13/21. Middle panel: Alignment of 13/21 and 14 during meiosis, showing potential recombination within the SST1 type 1 array region. Bottom panel: Resulting ROB structure after fusion. The ROB retains two centromeres and the SST1 arrays from both parent chromosomes. The central multicopy SST1 array is flanked by sequences from 13 and 14 or 14 and 21. The fusion is within the SST1 subfamily 1 array. The size of the array is variable. The SST1 type 2 L monomer is derived from 13 or 21, while the SST1 type 2S monomer is from 14. The rDNA regions are lost. Key features are color-coded: magenta for SST1 type 1 array, pink for SST1 type 2S monomer, blue for SST1 type 2 L monomer, and gray for centromeres and rDNA regions. Distances between features are indicated in kilobases (kb).

Extended Data Fig. 7. Sequence alignment comparison among SST1 monomer consensuses from each chromosome in CHM13 with a major array.

Each row corresponds to the consensus sequence for monomers of SST1 derived from major arrays on chromosomes 13, 14, 21, 17, 19, and 4. The consensuses from chromosomes 13, 14, and 21 (gray labels) have in common a large gap in the middle of the macrosatellite as well as a predicted PRDM9 DNA binding site (red). Darker shades indicate more conservation.

Extended Data Fig. 8. CENP-A and NDC80 localization on the dicentric ROB chromosome in GM03417.

A-B. Representative SIM images of ROB chromosome (A) and normal copies of chromosomes 14 and 21 (B) labeled by immuno-FISH with centromeric satellite probes for Cen 14/22 (orange), Cen 13/21 (green), and anti-CENP-A antibody (white). DNA was counterstained with DAPI. Plots below show averaged line intensity profiles along the individual kinetochore regions of sister chromatids normalized to the maximum intensity of each channel for 20 ROBs (A) and 11 normal copies of chr.14 and chr.21 (B). Error bars denote standard deviations. C. Example SIM images of four individual ROB chromosomes labeled by immuno-FISH with a probe for Cen 13/21 (green), and antibodies against CENP-A (white) and NDC80 (magenta). Magnified insets show centromeric regions labeled with each antibody individually. Line scans (1 and 2) along kinetochores of sister chromatids show normalized fluorescence intensity profiles for CENP-A (blue) and NDC80 (magenta), illustrating variations between kinetochores. D. SIM images and corresponding line scans of representative normal copies of acrocentric chromosomes 13 and 21 labeled and analyzed as in (C). Scale bar in all images is 1 µm.

Extended Data Fig. 9. CENP-A CUT&Run on the ROB chromosomes.

Tracks displaying CENP_A enrichment in three ROBs (A) GM03786, (B) GM04890, and (C) GM03417. From top to bottom, tracks denote the two α-satellite arrays with centromeric regions colored by chromosome, CENP-A CUT&RUN enrichment (blue) (two replicates), IgG negative control, Input coverage, and CpG methylation profile. DNA methylation tracks show methylation calls from ONT (orange) or PacBio HiFi (light blue) sequencing. CENP-A enrichment is associated with a dip in CpG methylation (CDRs).

Extended Data Fig. 10. Imaging and genomic analysis of chromosomes 13, 14, and 21 in assembled genomes.

ImmunoFISH, DNA methylation, and CENP-A CUT&RUN, and CUT&Tag analysis of normal copies of acrocentric chromosomes from (A) GM03786, (B) GM04890 and (C) GM03417 cell lines are shown. The left panels show representative structured illumination super-resolution images of normal acrocentric chromosomes labeled by immuno-FISH with centromeric satellite probes for Cen 14/22 (orange), Cen 13/21 (green), and anti-CENP-C antibody (red). DNA was counterstained with DAPI. Scale bar is 1 µm. Note single CENP-C foci on corresponding centromeres. Plots below show averaged intensity profiles of lines drawn through the individual kinetochore regions of sister chromatids of multiple chromosomes: GM03786 n = 22 for each chromosome (11 chromosomes 13 and 11 chromosomes 14), GM04890 n = 24 for each chromosome (12 chromosomes 13 and 12 chromosomes 14), GM03417 n = 24 for each chromosome (12 chromosomes14 and 12 chromosomes 21). Intensity profiles were aligned to the peak of the Gaussian of the corresponding Cen signals and normalized to the maximum intensity of each channel. Error bars denote standard deviations. The right panels display corresponding heatmaps of sequence similarity calculated for 5 kb bins for each centromere. Below the heatmaps, DNA methylation tracks show methylation calls from ONT (orange) or PacBio HiFi (blue) sequencing. Hypomethylated regions correspond to CENP-A enrichment on CUT&RUN (blue) and CUT&Tag (black) tracks below, indicating active centromere location.