A deep-time landscape of plant cis-regulatory sequence evolution
Kirk AmundsoniD, Anat HendelmaniD, Danielle CireniD, Hailong YangiD, Amber E de NeveiD, Shai Tal, Adar Sulema, David P. JacksoniD, Madelaine Elisabeth BartlettiD, Zachary B. LippmaniD, Idan EfroniiD
Abstract
Developmental gene function is often conserved over deep time, but cis-regulatory sequence conservation is difficult to identify. Rapid sequence turnover, paleopolyploidy, structural variation, and limited phylogenomic sampling have impeded conserved non-coding sequence (CNS) discovery. Using Conservatory, an algorithm that leverages microsynteny and iterative alignments to map CNS-gene associations over evolution, we uncovered [~]2.3 million CNSs, including over 3,000 predating angiosperms, from 284 plant species spanning 300 million years of diversification. Ancient CNSs were enriched near developmental regulators, and mutating CNSs near HOMEOBOX genes produced strong phenotypes. Tracing CNS evolution uncovered key principles: CNS spacing varies, but order is conserved; genomic rearrangements form new CNS-gene associations; and ancient CNSs are preferentially retained among paralogs, but are often lost as cohorts or evolve into lineage-specific CNSs. One Sentence SummaryConservatory maps ancient cis-regulatory elements and uncovers regulatory evolution dynamics.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 11
- FWCI
- 33.13
- Field percentile
- top 1%
- Organism
- Assay
- Research Focus
- Approach
In-depth summary
Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 3 (18 Feb 2026) · CC BY 4.0
Conservatory, an orthogroup-centric conserved noncoding sequence finder, catalogs 2.3 million unique CNSs across 314 plant genomes and reveals regulatory elements older than angiosperms and seed plants.
Whole-genome alignment cannot recover conserved noncoding sequences (CNSs) between distant plants, because regulatory DNA diverges quickly and repeated polyploidy, fractionation and rearrangement hide orthology, so most known plant CNSs stay within single families. Conservatory resolves CNS orthology across green plants, which allows a deep-time account of how cis-regulatory sequences arise, persist, move and disappear.
Key results
ShownDeeply conserved noncoding sequences are widespread in plants. Most CNSs are restricted to one family, but tens of thousands predate the eudicots and hundreds predate seed plants. Genes near the ancient CNSs are enriched for developmental and transcriptional functions.
32.8 million CNSs found across 314 genomes (284 species) using 10 reference genomes, collapsed into 2.3 million unique CNSs; median length 12-40 bp; 2.15 million (92%) restricted to single families; 31,314 (1.34%) predate eudicot diversification, 3,321 (0.14%) predate angiosperms, 633 (0.02%) predate seed plants. · Ancient cis-regulatory elements are common in…
ShownCNSs carry the hallmarks of functional cis-regulatory elements. They are enriched for transcription factor binding, activating histone marks and accessible chromatin, depleted for DNA methylation and repressive marks, and sit at the centers of open chromatin regions.
CNSs with at least one TF binding site: 96% (Arabidopsis), 80.1% (maize); lacking functional genomics support: 0.67% (Arabidopsis) vs 22.7% (maize); significant enrichment in accessible chromatin in Arabidopsis, tomato and maize (p < 1e-4, permutation test); 0.05%-1.8% of CNSs matched known peptides and were removed. · Ancient cis-regulatory elements are common in…; Functional genomic marks
ShownIn tomato, CRISPR-Cas9 alleles that disrupt ancient promoter CNSs of SlWOX9 and SlWOX2 cause embryonic lethality and vegetative defects. Previously characterized CREs of eight developmental transcription factors also coincide with ancient CNSs.
Two new SlWOX9 alleles disrupting seed-plant CNS S217 reproduced partially penetrant embryonic lethality and vegetative defects; five of six SlWOX2 promoter deletions made across about 3 kb, each removing all or part of the angiosperm-level CNSs, gave complete or near complete lethality, while Pro-6, leaving them largely intact, did not; lethality screens used F2 populations of at least 125 seedlings. · Mutations in deeply conserved CNSs disrupt…; Plant phenotyping; CRISPR-Cas9 mutagenesis, plant transformation…
ShownAncient CNSs sit closer to start codons and change position less than family-level CNSs. Even so, many CNS-gene links are long-range or skip intervening genes, and chromatin loops in Arabidopsis, tomato and maize support these links.
73% of CNSs lie upstream; median distance 2,252 bp for angiosperm/seed-plant vs 11,865 bp for family-level CNSs (p < 0.001, Wilcoxon); 25% sit >25 kb away, 47.5% are nearer a different gene, 25% to 50% skip genes; intergenic loop edges with CNSs more often link to genes (p < 0.001, chi-square); ancient CNS mobility lower (p<2.2e-16). · Genome rearrangements alter CNS position and…; Chromatin interactions
SuggestedAncient CNSs are lost or diverge gradually overall, with punctuated large-scale losses in specific families. After whole-genome duplication, paralogs diverge asymmetrically in CNS content while ancient CNSs are preferentially kept, and new paralog-specific CNSs appear to derive from ancestral sequence.
Parsimony mapping found punctuated angiosperm-CNS losses or divergence in Brassicaceae, Asteraceae, Poaceae and Lemnaceae; across WGD paralog pairs in the 10 reference genomes, shared CNSs fell with age as paralog-specific CNSs accumulated, asymmetrically favoring one dominant paralog; syntenic replaced-new CNS pairs had higher identity than nonsyntenic controls (p < 0.001, Dunn test). · Evolution of new cis-regulatory sequences from…
Methods, data and reproducibility
Conservatory aligns up to 120 kb of flanking intergenic DNA for each orthogroup member (up to 16 co-orthologs per genome) to a family reference and calls conserved blocks with phyloP. It reconstructs family-ancestral CNSs with FastML, searches them in other families, and links distant homologs through bridge genomes. Tomato CRISPR, ATAC-seq and Hi-C test function.
- Public plant genome assemblies and annotations (NCBI, Ensembl, CoGe, Phytozome) from 68 families, including 10 family reference genomes · green plants (284 species) · n = 314 genomes; 284 species; 68 families
- New tomato leaf ATAC-seq on young M82 terminal leaflets (100,000 nuclei per reaction), peaks called with Genrich · Solanum lycopersicum cv. M82 (tomato) · n = six biological replicates · GSE307325
- New Hi-C (Arima Hi-C 2.0) on inflorescence tissue of M82 anantha mutants; 9,573 FitHiC2 loops at 5 kb resolution · Solanum lycopersicum cv. M82 (tomato) · n = two biological replicates · GSE307325
- Published Arabidopsis ATAC-seq (Plant Chromatin State Database), TF and histone ChIP-seq (ReMap 2022), DNA methylation (GEO) and chromatin loop data · Arabidopsis thaliana
- Published maize B73 ChIP-seq and DAP-seq (GEO), MaizeGDB histone/ATAC/bisulfite tracks, JASPAR 2024 motif scans, Hi-C and HiChIP loops · Zea mays (B73)
- Published tomato TF ChIP-seq reads, meristem ATAC-seq and histone ChIP-seq peaks, and bisulfite sequencing data · Solanum lycopersicum (tomato) · PRJNA545571 (TF ChIP-seq); GSE285260 (bisulfite)
- Tomato CRISPR-Cas9 alleles in M82: SlWOX9 promoter alleles hitting CNS S217, SlWOX2 coding nulls and six SlWOX2 promoter deletions · Solanum lycopersicum cv. M82 (tomato) · n = F2 populations of at least 125 seedlings for lethality screens
- 1KP low-copy orthogroups (386 of 410 recovered) used to build the dated all-species tree · green plants · n = 100 orthogroups
- Methods & tools
- Conservatory (Perl/R CNS discovery algorithm), LastZ, phyloP, FastML ancestral sequence reconstruction, MAFFT, OrthoFinder, RAxML-NG and IQTree2, ape (penalized likelihood node dating), CRISPR-Cas9 mutagenesis (tomato), ATAC-seq with Genrich peak calling, Arima Hi-C 2.0, HiC-Pro and FitHiC2, DupGenFinder
- Data & accessions
- https://www.conservatorycns.com/
- Software & environment
- Conservatory written in Perl and R; BUSCO v6.0.0, RAxML-NG 1.1.0, IQTree2 2.0.3, mafft v7.313, nf-core/chipseq v2.0.0, HiC-Pro v3.1.0, CrossMap v0.7.3, bedtools v2.31.1; tuned parameters listed in table S12
- Assay conditions & compute
- Run on the Hebrew University HPC cluster using up to 1,000 nodes at a time; producing and merging the full dataset took approximately 7 days.
- Reproducibility gaps
- The Zenodo data record is cited without an identifier in the text; the empirically tuned parameters appear only in supplementary table S12; biological materials are available only on reasonable request to the corresponding authors.
Limitations (6)
- Sparse phylogenomic sampling, especially of monocots outside grasses and duckweeds, constrains CNS orthology calls; apparent losses of ancient CNSs may partly be lineage-specific CNSs that remain orthologous to ancestral angiosperm elements. · stated by the authors
- Limited guide RNA availability and efficiency prevented finer dissection of individual CNSs in the SlWOX2 promoter. The deletions therefore implicate one or more ancient CNSs together rather than a single defined element. · stated by the authors
- Maize CNS support is weaker: 22.7% lacked biochemical evidence, and 10.2% of Poaceae CNSs called on the maize reference overlapped transposons annotated after the analysis. The authors attribute this to sparser data and changing annotation. · stated by the authors
- How common ancestrally derived paralog-specific CNSs are remains unassessed, and separating de novo origin from preferential retention would need denser phylogenomic coverage. · stated by the authors
- Alignment and filtering parameters were tuned to recover a curated set of known CREs. The overlap between characterized developmental CREs and ancient CNSs (fig. S3) may therefore be partly circular as validation. · noted in this summary
- Functional tests cover only embryonic WOX genes in tomato, with no reporter or expression assays. Regulatory function for most of the 2.3 million CNSs rests on correlative enrichment with chromatin and binding data. · noted in this summary
Who should read:Read this if you study plant cis-regulatory evolution, want deeply conserved noncoding targets for promoter editing, or need comparative-genomics methods that cope with polyploidy and genome rearrangement.
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
- Extensive modulation of a conserved cis-regulatory code across 589 grass species
Charles O. Hale, Sheng‐Kai Hsu, … Edward S. Buckler·Molecular Biology and EvolutionPeer-reviewed·Dec 9, 2025
- STAG-CNS: An Order-Aware Conserved Non-coding Sequences Discovery Tool For Arbitrary Numbers of Species
Xianjun Lai, Sairam Behera, … James C. Schnable·Molecular PlantPeer-reviewed·Jun 6, 2017
- Evolution of conserved noncoding sequences in Arabidopsis thaliana
Alan E. Yocca, Zefu Lu, … Patrick P. Edger·Molecular Biology and EvolutionPeer-reviewed·Feb 8, 2021
- Extreme restructuring of cis-regulatory regions controlling a deeply conserved plant stem cell regulator
Danielle Ciren, Sophia G. Zebell, Zachary B. Lippman·PLoS GeneticsPeer-reviewed·Mar 4, 2024
- Conserved statistical organization of cis-regulatory landscapes defines a fluctuation-structured regulatory phase
Yoshihiro Ohmori·bioRxiv·Feb 6, 2026
- Extensive Loss and Gain of Conserved Non-Coding Elements during Early Teleost Evolution
Elisavet Iliopoulou, Vasileios Papadogiannis, … Tereza Manousaki·Genome Biology and EvolutionPeer-reviewed·Apr 1, 2024