The variation and evolution of complete human centromeres

Key results

1. ShownAll CHM1 centromeres were completed, with ultra-long Oxford Nanopore (ONT) reads bridging fragmented HiFi contigs. Read mapping confirmed integrity for all but two chromosomes, and k-mer checks plus an independent assembler supported accuracy and population relevance.

   ~66-fold HiFi and ~98-fold ONT coverage; hifiasm alone completed only four centromeres; final assemblies estimated >99.9999% accurate (QV>60), >99.99% identical to independent Verkko assemblies; vs 112 HPRC/HGSVC (Human Genome Structural Variation Consortium) haplotypes, 20.9% match ≥99% and 46.9% of these fit CHM1 better than CHM13. · Complete sequence and validation of CHM1…; Targeted sequence assembly and validation of…

2. ShownThe α-satellite higher-order repeat (HOR) arrays are the most divergent centromeric component: much of their sequence cannot be aligned between haplotypes, CHM1 arrays are usually longer, and several chromosomes carry different array structures or new HOR variants.

   63.0-71.5% of HORs align at >90% identity between CHM1 and CHM13, dropping to 53.2-55.3% across 56 HPRC/HGSVC genomes; identity 98.6 ± 1.6% in HOR arrays vs 99.9 ± 0.3% in non-satellite pericentromere; CHM1 arrays ~1.3-fold longer (16 of 23); eight arrays structurally distinct (chromosomes 5, 7, 8, 10-14), four rich in novel HORs. · Genetic variation among human centromeres; DISCUSSION

3. ShownKinetochore sites, marked by hypomethylated DNA and CENP-A chromatin, often sit in different places between the two haplotypes, and CENP-A-bound HOR sequence is not consistently more conserved than the rest of the array.

   18 of 23 CHM1 sites are smaller than in CHM13 (178 vs 214 kbp average); 16 of 23 shifted >100 kbp and six >500 kbp (chromosomes 4, 6, 11, 12, 18, 20); chromosome 6 moved 2.4 Mbp alongside a switch from 16/18- to 15/18-monomer HORs; five CHM1 chromosomes show two kinetochores >150 kbp apart. · Epigenetic differences between CHM1 and CHM13…; CpG methylation analysis; Native CENP-A ChIP-seq and analysis

4. ShownOrthologous centromeres from chimpanzee, orangutan, and macaque show near-complete species-specific α-satellite turnover, lineage-specific array architectures, and suprachromosomal family (SF) changes, while pericentromeric flanks mutate faster than unique DNA.

   31 nonhuman primate centromeres (six chromosomes); chimpanzee HOR arrays average 67.8% of human size; orangutan arrays form three to four blocks at 80-90% identity; macaque dimeric arrays average 12.2 ± 1.6 Mbp (93-97% identity); chimpanzee chromosome 5 has 2.1- or 2.4-Mbp of SF5 α-satellite; flank mutation rate up 1.1-fold (X) to 4.1-fold (chromosome 5). · Different centromere evolutionary trajectories…; Human haplotype phylogenetic reconstruction and…; Human and NHP α-satellite suprachromosomal…; Human and NHP phylogenetic analysis

5. SuggestedTrees built from p- and q-arm flanks agree, consistent with little recombination across the array, and haplotypes carrying new HORs on chromosomes 11 and 12 are monophyletic, pointing to single recent origins followed by saltatory amplification.

   New HORs dated ~13-23 kya (chromosome 12) and ~80-153 kya (chromosome 11), assuming 6 million years of human-chimpanzee divergence; 392-2,490 nucleotide differences per generation, scaling to ~30-60 (chromosome 11) vs ~500-1000 (chromosome 12) per Mbp per generation; arrays grew from 554 kbp to 2 Mbp on average with ~100 structural changes. · Human haplotype phylogenetic reconstruction and…; DISCUSSION; Human-specific phylogenetic analysis