henhaobio
Sign in

The Plant CellEvolutionary BiologyPeer-reviewed journal article

Gene expression divergence following gene and genome duplications in spatially resolved plant transcriptomes

Fabricio Almeida‐SilvaiD, Yves Van de PeeriD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Gene and genome duplications are key drivers of plant genome evolution, expanding genetic repertoires and facilitating functional innovation. However, genes originating from different duplication mechanisms undergo different outcomes, especially at the expression level. Previous studies have demonstrated that segmental or whole-genome duplications generate duplicates with similar and somewhat redundant expression profiles across multiple tissues, while other duplicates display increased divergence, ultimately leading to functional innovations. However, little is known about how duplicates diverge in expression across cell types in a single tissue. Here, we used high-resolution spatial transcriptomic data from five species (Arabidopsis thaliana, Glycine max, Phalaenopsis aphrodite, Zea mays, and Hordeum vulgare) to investigate the evolution of gene expression following gene duplications. We found that genes originating from segmental or whole-genome duplications display increased expression levels, expression breadths, spatial variability, and number of coexpression partners. Duplication mechanisms that preserve cis-regulatory landscapes typically generate paralogs with more preserved expression profiles, but such differences by duplication mode disappear over time. Expression divergence also depends on gene functions, with dosage-sensitive gene families displaying highly preserved expression profiles, while families involved in more specialized processes (e.g., flowering and phytohormone biosynthesis) display increased divergence. Paralogs originating from large-scale (including whole-genome) duplications display redundant and/or overlapping expression profiles, indicating functional redundancy and/or subfunctionalization, while small-scale duplicates diverge asymmetrically, indicating neofunctionalization. Collectively, our findings provide new insights into the tempo and mode of gene expression evolution, helping understand how gene and genome duplications shape cell identities.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (27 May 2025) · CC BY 4.0

In spatial transcriptomes from five plant species, segmental (including whole-genome) duplicates show higher, broader, spatially variable expression and redundancy, while small-scale duplicates diverge asymmetrically.

Earlier analyses of paralog expression divergence relied on bulk RNA-seq or microarray compendia that average over heterogeneous cell populations, while single-cell data lose spatial context and under-sample hard-to-protoplast cells. This work uses spatially resolved transcriptomes from five plant species to ask how duplication mode, paralog age and gene function shape expression divergence among compartments of a single tissue.

Key results

  1. ShownGenes from duplication modes that keep the cis-regulatory context (segmental, tandem) tend to be more highly and broadly expressed than dispersed duplicates, and segmental duplicates are enriched among spatially variable genes in every species.

    Kruskal-Wallis P <0.05 for higher expression level and breadth of SD/TD genes, mainly versus dispersed (DD) genes; SD genes overrepresented among DESpace spatially variable genes (FDR <0.05) in all species (Fisher's exact P <0.05). Exceptions: DNA-transposed (dTRD) genes in H. vulgare and P. aphrodite match SD/TD levels and are also SVG-enriched. · Gene duplication mode impacts expression levels…; Inference of spatially variable genes and…

  2. SuggestedParalog expression divergence grows exponentially with synonymous divergence (Ks), and lower divergence for cis-preserving modes (SD, TD, PD) appears only among young pairs (<20 MYA), with Arabidopsis segmental pairs as an exception.

    Log-log fits of divergence vs Ks beat linear fits (smaller BIC and AIC). Soybean SD and PD pairs from ~13 MYA are less diverged, with no mode effect for ~62 MYA pairs; maize SD, TD and PD pairs (~16 MYA) are less diverged, whereas barley pairs (~53 MYA) show no mode effect. A. thaliana SD pairs remain less diverged despite ~51-72 MYA ages. · Gene duplication mode impacts expression…; Calculation of substitution rates and age-based…; Measuring expression divergence between…

  3. ShownClassifying pairs by relative expression breadth (REB) across cell types shows asymmetric divergence, one broad and one restricted copy, is the most common outcome; in most species small-scale duplicates are overrepresented in this class, and segmental pairs typically among redundant pairs.

    Classes: redundancy (both REB >0.7), asymmetric (one >0.7, other <0.3), symmetric (both 0.3 to 0.7). Tandem, proximal, transposed and dispersed pairs overrepresented in the asymmetric class in most species; SD pairs typically overrepresented among redundant pairs, with a large redundant SD fraction in Z. mays and P. aphrodite. · Pervasive asymmetric divergence among…; Classification of paralogous gene pairs into…

  4. ShownCorrelation spread differs widely among PLAZA gene families: families of highly similar paralogs are enriched for core, dosage-sensitive functions, whereas families of highly dissimilar paralogs are enriched for specialized processes.

    Gene Ontology enrichment on top 10% families by highest and lowest median correlation. Similar-paralog families: chromatin assembly, rRNA processing, transcriptional regulation, splicing and photosystem assembly. Dissimilar-paralog families: flower development, detoxification, auxin and abscisic acid biosynthesis, sucrose metabolism/transport, cell wall remodeling. · Expression divergence varies across gene…

  5. SuggestedMost paralog pairs sit in different coexpression modules, but less often than under randomized networks, hinting at constraint on divergence; segmental duplicates are enriched among network hubs in three of five data sets.

    Diverged-pair frequencies >90% in nearly all species and modes yet mostly below nulls from 10,000 degree-preserving permuted networks; eigengene dissimilarities of diverged pairs mostly small; no consistent mode effect. Hub enrichment for SD genes in G. max, P. aphrodite and Z. mays (dTRD also in P. aphrodite), none in A. thaliana or H. vulgare. · Network-based expression divergence reveals…; Gene duplication mode impacts gene…; Inference and analysis of gene coexpression…

Methods, data and reproducibility

Paralogs were classified by duplication mode with doubletrouble/doubletroubledb and binned into age groups from Gaussian mixture peaks in Ks; divergence was scored by Spearman correlations on 30-spot metaspots, co-membership in BioNERO coexpression modules, and relative-expression-breadth classes, alongside DESpace spatially variable gene tests.

  • Spatial transcriptomes of transversal sections of Arabidopsis cauline leaves, obtained from STOmicsDB · Arabidopsis thaliana · n = median 631 spots per sample; median 6 spatial domains
  • Spatial transcriptomes of soybean nodules infected with rhizobia, obtained from the original publication · Glycine max · n = median 6 spatial domains per sample
  • Spatial transcriptomes of developing maize ears (ear primordia), obtained from STOmicsDB · Zea mays · n = median 8062 spots per sample
  • Spatial transcriptomes of Phalaenopsis Big Chili inflorescences (buds and shoot apical meristem) across developmental stages, from STOmicsDB · Phalaenopsis Big Chili (P. aphrodite genome used as reference)
  • Spatial transcriptomes of germinating barley seeds, obtained from the original publication · Hordeum vulgare · n = median 1545 spots per sample; median 14 spatial domains
  • Paralog pairs classified by duplication mode from doubletroubledb; P. aphrodite pairs classified de novo with Amborella trichopoda as outgroup · five study plant species (plus Amborella trichopoda outgroup)
  • Longest-isoform coding sequences (Ensembl Plants release 59, Orchidstra 2.0) plus gene family and functional annotation (PLAZA 5.0) · five study plant species
Methods & tools
zellkonverter (H5AD import), SpatialExperiment, scater (library-size normalization, log counts), scuttle (pseudobulk aggregation), doubletrouble (duplicate classification, γ-MYN Ks/Ka, GMM Ks peaks), doubletroubledb, DESpace (spatially variable genes), HybridExpress (overrepresentation analysis), exdiva (paralog expression divergence metrics), BioNERO (signed hybrid coexpression networks, hub detection), PLAZA 5.0, STOmicsDB
Software & environment
R workflow (zellkonverter, SpatialExperiment, scater, scuttle, doubletrouble, DESpace, BioNERO, exdiva, HybridExpress); reference versions Ensembl Plants release 59, Orchidstra 2.0, PLAZA 5.0; R package versions not stated.
Reproducibility gaps
Zenodo archive mentioned without a DOI or record link; source spatial data come from STOmicsDB and original publications with no accession IDs given in text; R package versions not reported.
Limitations (6)
  • Spatial capture areas are small (for example 6.5 mm x 6.5 mm slides on 10x Visium), so the authors acknowledge it is uncertain whether the patterns replicate in other tissues and conditions. · stated by the authors
  • Few plant spatial transcriptomes are available (five data sets analyzed); the authors note an expanded analysis is needed to test whether conclusions hold broadly or whether tissue- or species-specific exceptions exist. · stated by the authors
  • Whether paralogs diverge first across cell types or across tissues (or independently) is unresolved, because it requires bulk and spatial data from the same species under matched conditions. · stated by the authors
  • Some links between duplication mode and expression features are data set or species specific; for example, SD genes were not enriched among hubs in A. thaliana or H. vulgare, which the authors suggest may reflect species- or tissue-specific effects. · stated by the authors
  • Each species contributes a different tissue (leaf, nodule, ear, inflorescence, seed), so species, tissue and duplicate-age effects are partly confounded; apart from the soybean within-species contrast, age-dependence of mode effects leans on cross-species comparisons such as maize versus barley. · noted in this summary
  • Divergence is measured on 30-spot metaspots and domain-level pseudobulks rather than true single cells, and age-group mode comparisons are described without test statistics or effect sizes in the text. · noted in this summary

Read this if you study duplicate gene fate or regulatory evolution in plants and want a cross-species analysis of how duplication mode, age and gene family shape paralog expression divergence in spatial transcriptomes.

Citations
13
FWCI
6.99
Field percentile
top 2%

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
THISRanaweera '26Coate '20Panchy '17Sandve '17Cai '26Assis '14
81% semantic matchPlant Biology
Thilanka Ranaweera, Rachel E. Kerwin, Huan Chen+4 more·bioRxiv·Oct 30, 2026