henhao
Sign in
GenomicsbioRxiv
Preprint

An atlas of eukaryotic centromere architecture reveals recurrent evolutionary dynamics

Piotr WłodzimierziD, Estela Pérez-RomániD, Jacob Gonzalez IsaiD, Michael Hong, Meng ZhangiD, Ludmila Oliveira, Katharine M. JenikeiD, Robin BurnsiD, Pío SierraiD, Matthias HeubergeriD, Chenxi ZhouiD, Sagarika Koppera, Mia Becker, Nicola GorringeiD, Petr Novak, Yennifer Mata‐SucreiD, Giulio FormentiiD, Felipe Karam TeixeiraiD, Ines Anna DrinnenbergiD, Ioannis KontoyiannisiD, Kamil S. JaroňiD, Marcela Uliano Da Silva, André MarquesiD, Jir̆ı́ MacasiD, Mark BlaxteriD, , Ian R. HendersoniD

PDFDOICodeData
Post to X
Threads on XPost to BlueskyThreads on Bluesky
In-depth full-text digest available · Jump to digest ↓

Abstract

Centromeres evolved at the root of eukaryotes to segregate chromosomes during cell division. Despite their ancient origin, centromeric DNA sequences evolve rapidly and adopt diverse architectures, including point centromeres, satellite arrays, transposon clusters, and holocentrics. To analyse centromere evolution at a broad scale, we characterised architectures across 325 diverse Darwin Tree of Life genome assemblies. Centromere architecture is evolutionarily labile, and similar configurations arise independently across divergent lineages. In plants and animals, we modelled centromere evolution as a recurrent cycle, in which satellite- and transposon-based architectures interconvert, with independent origins of holocentricity. We curated >23 million satellite repeats comprising 263 families from 165 species. Despite sequence divergence between satellite families, higher order repeats are prevalent, indicating constraint on repeat architecture rather than primary sequence. Satellite arrays are heavily invaded by diverse transposon families, consistent with convergent adaptation to the centromeric niche. In 89 species, transposons themselves constitute the primary centromere structure. We observed centrophilic transposons forming tandem arrays, suggesting mechanisms for satellite regeneration. Our sample includes five independent origins of holocentricity in plants and animals, which vary in association with periodic satellite arrays. We propose that genetic instability, centrophilic transposition, and transmission distortion promote recurrent centromere architectural interconversions during evolution.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
1
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 1 (23 Sept 2026) · CC BY 4.0

Across 325 Darwin Tree of Life genomes, satellite-array and transposon-cluster centromeres are phylogenetically scattered, fitting a model of repeated interconversion; satellites share higher-order repeat structure but not sequence.

Centromeric DNA ranges from point centromeres to satellite arrays, transposon clusters and holocentric layouts, and the paper notes that what drives switches between these types, and whether types are confined to lineages, has been unclear. Using standardised chromosome-scale Darwin Tree of Life assemblies, this work classifies centromere architecture across 325 eukaryotes with one validated pipeline and models transitions on a dated species tree.

Key results

  1. SuggestedCentromere architecture is labile: close relatives often differ while distant lineages converge. Ancestral state reconstruction favours satellite and transposon states converting into each other, with holocentricity arising independently as an evolutionary sink.

    CAP assigned 288 of 325 genomes (89%): 147 (45%) satellite α-state, 89 (27%) transposon Ω-state, 11 (3%) mixed, 41 (13%) in holocentric clades; 31 (10%) were cryptic. Of seven Mk models compared by AIC on the full, Metazoa and Viridiplantae trees, all-rates-different with irreversible holocentricity (ARD_irrevH) fit best. CAP agreed with CENP-A/CENH3 ChIP-seq in 8 genomes and plant FISH/KNL1 staining. · An atlas of centromeric architectures in 325…; A recurrent model for centromere evolution…; Ancestral state reconstruction modelling of…

  2. ShownCentromeric satellites share almost no sequence or base-composition signature across families, yet higher-order repeats (HORs) are widespread in plants and animals. This implies that repeat organisation, not primary sequence, is what recurs.

    23,577,177 repeats in 263 families from 165 species; mean monomer 140 bp, 128/263 in 100-200 bp. Mean 17-mer sharing between families 0.03%; similarity half-life ~12 My. 135/263 families more and 128/263 less AT-rich than their genome. Rank-frequency fits Zipf's law (log-log r²>0.9). Within-chromosome similarity exceeded between-chromosome; HOR scores peaked in array centres. · Abstract; Satellite repeat diversity and recurrent higher…; Satellite repeat analysis; Satellite higher order repeat and regimentation…

  3. ShownSatellite arrays are colonised by transposons spanning the full known range of integration mechanisms. LTR elements inside arrays are younger and more often intact than those outside, consistent with convergent, ongoing insertion that targets centromeres.

    Non-satellite interruptions averaged 17% of array space, 70% of it transposon; array transposon content 18% in plants, 10% invertebrates, 7% chordates. LTR retrotransposons made up 11% of centromeric sequence. In-array LTR elements were more often intact (Welch t-test P<9.85×10^-7) and had higher 5′-3′ LTR identity (P<1.55×10^-12). Athila entered centromeres independently in at least 13 angiosperms. · Convergent centrophilic transposon adaptation…; Annotation of transposons

  4. SuggestedGeum records a recent switch: sister species G. urbanum and G. rivale have transposon-based and satellite-based centromeres at syntenic positions. In other species, transposons are recombined into tandem, satellite-like arrays, including CENH3-bound arrays in Rosa canina.

    G. urbanum centromeres are 87.5% Gypsy-Athila (~18 Mb); G. rivale's are 65.7% drGeuRiva.159 satellite (~57 Mb); split ~2 My ago; satellite copies 3,187 vs 236,781. Excision showed Athila nested in up to three strata (106 intact elements, median LTR identity >99.8%). Rosa canina has CENH3-enriched tandem arrays (up to 238.7 kb) of 2.2 kb and 1.6 kb Athila-derived monomers. · Transposon-dominated centromeres in plants…; Formation of tandem repeats by centrophilic…; Analysis of transposon-state centromeres

  5. ShownCENH3/CENP-A evolves faster than canonical H3, yet no residues separate species with satellite centromeres from those with transposon centromeres. The sample holds five independent holocentric origins, with periodic satellites in Poales and Limnephilus and no CENP-A orthologs in holocentric insects.

    From 422 CENH3/CENP-A and 897 H3-like proteins: longer CENH3/CENP-A terminal branches (Wilcoxon P<0.001); high entropy in N-tail, histone-fold loop 1 and C-terminus. GroupSim recovered known CENP-A vs H3 differences but no position at z≥2 for α vs Ω genomes. Periodic satellite arrays in all but one of 11 Poales holocentrics and 3 Limnephilus species (0.07-0.20 vs 0.89-2.11 arrays/Mb in plants). · A recurrent model for centromere evolution…; Independent origins of holocentricity with…; CENH3/CENP-A sequence curation and analysis

Methods, data and reproducibility

Built the Centromere Annotation Pipeline (CAP), which scores gene-poor, repeat-rich, low-entropy regions from Helixer, EDTA/TEsorter, TRASH and context-tree-weighting annotations (plus Hi-C in fungi). CAP was run on 325 Darwin Tree of Life (DToL) assemblies and checked by FISH/KNL1 staining and CENP-A/CENH3 ChIP-seq; state transitions were modelled on a dated BUSCO species tree.

  • Darwin Tree of Life chromosome-scale assemblies (PacBio HiFi + Hi-C): 89 Archaeplastida, 49 fungi, 184 metazoa, 3 TSAR/Discoba · multiple eukaryotes · n = 325 genomes · ENA PRJEB40665
  • CENP-A/CENH3 ChIP-seq and input for CAP benchmarking: human (CHM13), mouse, Timema douglasi, rice, maize, Triticum monococcum, pea, Cryptococcus deuterogattii R265 · multiple species · n = 8 genomes · ChIP: SRX255043, SRX20577194, SRX28824591, SRX27798377, SRX17512265, DRX799804, ERR9981080, SRX3049403
  • Centromeric satellite repeats curated with CAP and TRASH · multiple species · n = 23,577,177 repeats; 263 families; 165 species
  • CENH3/CENP-A and canonical H3-like protein sequences from Helixer proteomes, with an archaeal histone outgroup · multiple species · n = 422 CENH3/CENP-A; 897 H3-like; 10 archaeal
  • Intact Athila LTR retrotransposons used for phylogeny · angiosperms · n = 48,339 elements from 90 species
  • Plant root-tip metaphase spreads for FISH and KNL1 immunostaining, mostly field-collected in the Czech Republic (e.g. Geum urbanum, Mercurialis annua, Solanum dulcamara) · plants
  • TimeTree 5 divergence-time intervals used to date the species tree · multiple species · n = 62 calibration nodes
Methods & tools
CAP, TRASH (TRASH_2), EDTA v2.0.1, TEsorter v1.4.6, Helixer, BCT R package (context-tree weighting), FastSpeciesTree v1.0, IQ-TREE v2.3.4, phytools, DIAMOND v2.1.11, GroupSim, soloLTRseeker
Code
https://github.com/vlothec/eukaryotic_centromere_architecturehttps://github.com/vlothec/CAPhttps://github.com/vlothec/TRASH_2https://github.com/estpr/soloLTRseekerhttps://github.com/LSLeClercq/PAReTThttps://github.com/kavonrtep/TideCluster
Data & accessions
https://portal.darwintreeoflife.org/https://www.girinst.org/server/archive/RepBase28.06/https://www.dfam.org/classification/tree
Software & environment
Tool versions given (EDTA v2.0.1, TEsorter v1.4.6, FastSpeciesTree v1.0, IQ-TREE v2.3.4, DIAMOND v2.1.11, HMMER v3.4, miniprot v0.15, MAFFT v7.490/v7.526, BLASTn v2.12.0/v2.16.0+); no container or dependency file described.
Reproducibility gaps
Transposon-centromere boundaries were set by visual dot-plot inspection rather than a scripted rule; A. nigricornis transposons were called with a development version of pantera; compute resources are not reported.
Limitations (5)
  • Microbial eukaryotes are thinly sampled (two alveolates, one Discoba species). Amoebozoa, Haptista, Cryptista and Metamonada are absent because DToL lacks chromosome-scale assemblies for them, though such lineages are known for centromeric novelty. · stated by the authors
  • Mixed and cryptic states were too rare for reliable rate estimates, so they were pruned before ancestral state reconstruction, leaving three states and 277 tips. · stated by the authors
  • 31 genomes (10%) outside known holocentric taxa showed no centromeric repeats. The authors suggest these may be unrecognised repeatless holocentrics or non-repetitive centromeres. · stated by the authors
  • For most genomes the centromere call rests on sequence signatures alone. Direct CENP-A/CENH3 or kinetochore evidence covers 8 ChIP-seq benchmark genomes and a small set of FISH-validated plants. · noted in this summary
  • Final architecture classes combined CAP output with Hi-C, literature surveys and manual review, and transposon-centromere boundaries were drawn by eye from dot plots, so some calls involve judgement. · noted in this summary

Who should read:Read this if you study centromere, satellite or transposon evolution, need to locate centromeres in new chromosome-scale assemblies, or want a cross-eukaryote catalogue of centromeric repeats.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISWłodzimierz '26Glushkevich '26Naish '21McCulloch '26Courret '23Samano '26
76% semantic matchGenomics
Piotr Włodzimierz, Estela Pérez-Román, Amanda Souza Câmara+11 more·bioRxiv·Jan 18, 2026· 3 citations