The significance of a telomere-to-telomere, or T2T, assembly is best understood by considering what earlier genome projects routinely missed. Standard short-read sequencing struggles with the highly repetitive DNA that forms centromeres, the structures that anchor chromosomes during cell division, and telomeres, the protective caps at chromosome ends. These regions are not genetic dead weight; as Yi Wang, first author of the study, explained, they often hold key information about genome stability and phenotypic variation. The previous assembly of C. gladiata left centromeric and telomeric regions unresolved, which meant that any attempt to link DNA sequence to traits of agronomic importance was working from an incomplete blueprint. A T2T assembly was therefore essential, in the authors’ view, for functional genomics and breeding. With the complete sequence in hand, researchers can now ask precise questions about how genes are organized, how repeats shape chromosome behavior, and which genomic features underlie the seed chemistry that makes sword bean nutritionally interesting.

Building the assembly required a combination of technologies, each compensating for the weaknesses of the others. The team used Oxford Nanopore ultra-long reads, which can span repetitive DNA tracts that defeat shorter reads, together with Hi-C scaffolding, a method that uses the physical proximity of chromosomes in the nucleus to order and orient sequence contigs correctly. Multiple assembly strategies were applied and reconciled. The result is an eleven-chromosome genome containing all 22 telomeres and 11 centromeres, a complete chromosomal set for the species. Within the centromeres, the researchers found the characteristic architecture of tandemly repeated DNA, but with an instructive twist: a single repeat family, a 504-base-pair sequence designated TR504, dominates the centromeres of six chromosomes, while the remaining chromosomes carry distinct repeat families of their own. This mosaic of centromeric repeats offers a natural experiment in how centromere DNA evolves, and it provides markers that will help chromosome biologists track centromere identity across legume species.

Placing the new genome in an evolutionary framework revealed where sword bean sits on the legume family tree. Comparative genomics identified C. gladiata as a basal member of the tribe Phaseoleae, the group that also includes soybean and several other major crops. The analysis estimated that sword bean diverged from soybean approximately 45 million years ago, a deep split that predates much of the diversification within the tribe. The distribution of synonymous substitution rates, a standard measure known as Ks analysis, showed that C. gladiata shares an ancient whole-genome duplication with soybean, a doubling event inherited from a common ancestor. Crucially, however, sword bean did not undergo the additional, more recent genome duplication that occurred specifically in the soybean lineage. This makes the sword bean genome a valuable reference for reconstructing ancestral legume karyotypes, because it preserves a simpler genomic structure that has been scrambled and reduplicated in soybean. Researchers seeking to understand what the genome of the last common ancestor of the Phaseoleae looked like now have a cleaner template to work from.

The genome alone would have been a substantial contribution, but the team went further by asking how the genetic blueprint plays out during seed development. They sampled seeds at three stages: 40 days after flowering, designated S1 and representing early development; 60 days after flowering, designated S2 and representing the middle phase; and 80 days after flowering, designated S3 and representing maturity. Widely targeted metabolomics, a technique that quantifies hundreds to thousands of small molecules simultaneously, detected 2,723 metabolites across the samples, with flavonoids emerging as the most abundant chemical class. In parallel, transcriptomic profiling identified 14,016 differentially expressed genes across the three developmental stages, a strikingly large fraction of the genome’s coding capacity that shifts in activity as the seed matures. Together, the two datasets allowed the researchers to connect metabolic changes to the genes that drive them.

The integrated analysis revealed a clear metabolic phase transition during seed development, a kind of molecular switch that redirects the plant’s carbon economy. In the early stage, S1, the seed is a site of active starch and sucrose metabolism, branched-chain amino acid degradation, and flavonoid biosynthesis. At the center of this early program sits glucose-6-phosphate, or G6P, a hub metabolite that links sugar breakdown to downstream biosynthetic pathways. G6P showed a remarkably strong correlation, with a coefficient of 0.99, with the transcription factor CgKNAT7, suggesting that this regulatory protein plays a role in carbon allocation, deciding how the seed partitions its sugar resources among storage, structure, and defense chemistry. Twelve flavonoid metabolites, including the well-known compounds catechin and quercetin, accumulated at the early stage alongside the coordinated expression of structural genes in the flavonoid pathway, including CHS, F3H, DFR, LAR, and ANR. This tight coupling of metabolite accumulation and gene expression gives breeders a concrete set of targets if they wish to modulate flavonoid content in the seed.

By the middle and late stages, S2 and S3, the picture changed decisively. Carbon flux that had earlier fed starch and flavonoid production was redirected toward fatty acid accumulation, consistent with the seed’s shift toward building its oil reserves. Free fatty acids, including the oxygenated lipid species 13(S)-HODE and 9(S)-HODE, peaked at the mature stage, and their accumulation was consistent with the expression of genes involved in linoleic acid metabolism, notably CgLOX5 and CgMMS19. The transition from a sugar-and-flavonoid program to a lipid-accumulation program mirrors patterns documented in other oilseed crops, but seeing it mapped gene by gene in sword bean converts a general principle into an actionable roadmap. Genes whose expression tracks the lipid shift become candidates for manipulation, whether the goal is increasing oil yield, altering the fatty acid profile toward more unsaturated species, or stabilizing seed composition across environments.

Why does this matter beyond the laboratory? Underutilized legumes are increasingly discussed as assets in a food system facing climate stress and protein demand. Crops that fix their own nitrogen, tolerate marginal conditions, and deliver high-quality seed protein can diversify agriculture away from a narrow set of staple commodities. Sword bean fits that profile on paper, with its 26 to 30 percent seed protein and unsaturated-fatty-acid-rich oil, but paper is where it stayed as long as breeders lacked the genomic tools to work with it efficiently. A complete reference genome changes the calculus. It enables molecular marker development, accelerates the identification of genes controlling nutritional quality, and allows comparative genomics to transfer knowledge from well-studied legumes like soybean to this orphan crop. The candidate genes surfaced by the multi-omics analysis, from CgKNAT7 and the flavonoid pathway genes to the lipid metabolism genes CgLOX5 and CgMMS19, provide a starting catalog for that effort.

The study also signals a broader shift in plant genomics. T2T assemblies, once a tour de force reserved for a handful of model organisms, are becoming the expected standard for new crop genomes, and this project demonstrates why. Repetitive centromeric and telomeric regions, far from being ignorable filler, carry information about genome stability, chromosome evolution, and potentially phenotypic variation. Combined with time-resolved transcriptomics and metabolomics, a complete genome becomes a dynamic portrait of a developing organ rather than a static parts list. The authors have deposited the T2T assembly, transcriptome data, and metabolome data in public databases, making the resource available to the global research community. For C. gladiata, the release of a gap-free genome marks the moment a neglected legume joins the ranks of genetically tractable crops, and for legume biology more broadly, it supplies a basal Phaseoleae reference that will sharpen evolutionary reconstructions across the entire tribe.

Subject of Research: Telomere-to-telomere genome assembly and multi-omics analysis of the legume Canavalia gladiata

Article Title: First telomere-to-telomere genome of Canavalia gladiata released: Multi-omics analysis uncovers candidate genes for nutritional composition

Article References: First telomere-to-telomere genome of Canavalia gladiata released: Multi-omics analysis uncovers candidate genes for nutritional composition. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: Canavalia gladiata, sword bean, T2T genome, genomics, transcriptomics, metabolomics, seed development, legumes, flavonoids, fatty acids, molecular breeding, Phaseoleae

The endosperm begins its life in an unusual way. Rather than immediately forming cells separated by membranes and walls, it undergoes coenocytic growth, a phase in which nuclei divide repeatedly without accompanying cytokinesis, producing a single multinucleate cytoplasm that expands to fill the growing seed. During this period, in Arabidopsis, the expansion of the maternal seed coat together with the proliferation of the endosperm largely determines the final size of the seed. This continues until the embryo reaches the early heart stage, a morphological landmark at which the first obvious organogenesis of the plant body becomes visible. At that point, the endosperm changes course dramatically and begins to cellularize, partitioning its many nuclei into individual cells that will later serve as storage and transport tissue.

What makes this transition remarkable is that it is not uniform across the seed. Cellularization starts at the micropylar pole, the region of the seed where the embryo resides, and it stops before reaching the opposite, chalazal pole. The result is a polarized tissue: cellularized endosperm near the embryo and a persistent non-cellularized, multinucleate zone at the chalazal end that remains specialized for nutrient uptake from maternal tissues. This spatial asymmetry is not a curiosity; it is functionally central to how the seed feeds itself. Yet despite decades of work on endosperm development, the molecular mechanism that imposed this polarity remained unclear. Two genes were known to be essential players: HAIKU2, abbreviated IKU2, which encodes a leucine-rich repeat receptor kinase, and MINI3, which encodes a transcription factor. Loss-of-function mutations in either gene produce small seeds because endosperm cellularization is disrupted, but how these components generated a polarized pattern was unknown.

The new study resolves this puzzle by showing that the transcription factor MINI3 is expressed in two distinct phases. Early in seed development, MINI3 is active in the central cell before fertilization and in the proliferating endosperm afterward. Later, its expression becomes restricted to the micropylar endosperm, precisely the region where cellularization begins. This biphasic pattern immediately suggested a regulatory logic: an early, broadly distributed phase followed by a later, spatially restricted phase that must depend on some positional cue. The question was where that cue comes from, and the answer turned out to lie in a dialogue between the embryo and its surrounding endosperm.

Genetic evidence pointed the way. In iku2 mutants, the late micropylar expression of MINI3 is reduced, indicating that IKU2 is required for the polar phase of MINI3 transcription. Even more striking, the polar expression of MINI3 is also diminished in seeds in which only the central cell has been fertilized and no zygote or embryo is present. This observation implies that a signal derived from the zygote or the young embryo is needed to establish the micropylar pattern of MINI3 expression. Consistent with this interpretation, IKU2 protein is localized to the membrane of the central cell in the immediate vicinity of the egg cell at the micropylar pole, placing the receptor exactly where an embryo-derived signal would be expected to arrive.

Receptor kinases of the leucine-rich repeat family typically detect small extracellular peptides, and IKU2 belongs to subclade XI of this receptor family, a group for which members of the PIPL peptide family have previously been reported as ligands. This gave the researchers a defined search space. They screened for PIPL-family peptides on the basis of their expression patterns, looking specifically for genes active in zygotes or early embryos, the tissues positioned to send a signal toward the neighboring endosperm. This strategy yielded PROPIPL7 as the leading candidate, a gene whose expression profile matched the predicted source of the polarizing signal.

Functional tests then confirmed the hypothesis. When PIPL7 was applied to a chimeric receptor consisting of the extracellular domain of IKU2 fused to the intracellular signaling domain of FLAGELLIN SENSING 2, a well-characterized immune receptor commonly used as a reporter for ligand-receptor interactions, the construct was activated in Nicotiana benthamiana. This demonstrated biochemically that PIPL7 can be recognized by IKU2. The genetic evidence was equally compelling: propipl7 mutants recapitulate the phenotype of iku2 mutants, and double mutants between the two show no additive effect, the classic signature of two genes acting in the same pathway. Together, these results indicate that PIPL7 is the in vivo ligand of IKU2, completing a signaling axis that runs from the embryo, through a peptide ligand, to an endosperm-expressed receptor kinase, and finally to the transcription factor MINI3 that drives cellularization at the micropylar pole.

One additional finding adds an intriguing layer of redundancy and regulation to the system. PIP2, a peptide from the same family whose expression is induced by pathogens in other tissues, can rescue the propipl7 mutant phenotype when supplied in the relevant context. This suggests that PIPL-family peptides are functionally interchangeable to a considerable degree, and that the specificity of the developmental pathway is achieved not by unique ligand-receptor chemistry but by tissue-specific transcriptional regulation. In other words, many peptides could in principle activate the receptor, but only the embryo-expressed PROPIPL7 gene is transcribed at the right time and place to do so during normal seed development. This is an elegant example of how plants can build developmental specificity on top of a promiscuous signaling system, reserving a general-purpose signaling module for a highly specialized patterning role.

The broader significance of this work extends beyond Arabidopsis seed biology. Seed size is an agronomically critical trait, and the IKU2-MINI3 pathway has long been recognized as a key regulator of the maternal and filial determinants that set final seed dimensions. By identifying the embryo-derived peptide signal that polarizes endosperm cellularization, the study closes a mechanistic gap in our understanding of how the two products of double fertilization coordinate their development. It establishes, at the level of a defined ligand-receptor pair, that the embryo actively instructs the endosperm rather than merely responding to it, reframing the seed as a system of intercompartmental communication. The micropylar-to-chalazal polarity of the endosperm, which underpins the seed’s nutrient uptake architecture, can now be traced to a peptide signal originating in the zygote or early embryo and read out by a receptor positioned at the central cell membrane adjacent to the egg. For researchers interested in plant polarity, peptide signaling, and seed development, the finding provides both a satisfying answer to a long-standing question and a template for dissecting similar intertissue dialogues in other plant organs, where short-range peptide signals are increasingly recognized as the lingua franca of coordinated development.

Subject of Research: Embryo-to-endosperm peptide signaling that polarizes endosperm cellularization in Arabidopsis seeds

Article Title: Embryo talks to endosperm

Article References: Trösch, R. (2026). Embryo talks to endosperm. Nature Plants. https://doi.org/10.1038/s41477-026-02443-0

Image Credits: AI Generated

DOI: 10.1038/s41477-026-02443-0

Keywords: Arabidopsis thaliana, double fertilization, endosperm, seed development, IKU2, MINI3, PIPL7, receptor kinase, peptide signaling, cellularization, plant polarity, seed size

The core problem the researchers confronted is a familiar one in genomics: experimental annotation is slow and expensive, while genome sequencing is fast and cheap. When a new plant genome is sequenced, thousands of proteins are catalogued with little more than a name and a sequence. Computational tools can transfer functional labels from well-studied relatives, but this homology-based approach breaks down for proteins that have no close characterized counterparts, which is precisely the situation in many orphan crops. General-purpose deep learning function predictors, trained on multi-taxon datasets, tend to underrepresent experimentally annotated plant proteins, so their celebrated performance on animals and microbes does not necessarily transfer to the green lineage. DeepGreenGO was built to close that transferability gap by learning the patterns of plant protein function in a dataset where plants are not a footnote but the entire curriculum.

Technically, the model combines two complementary views of a protein. The first comes from ProtBERT-BFD, a protein language model trained on the enormous Big Fantastic Database, which converts each amino acid residue into a rich numerical embedding that encodes evolutionary and biochemical context learned from billions of sequences. The second view is structural: the researchers derived contact maps from protein structures, essentially matrices recording which residues lie close together in three-dimensional space. These contact maps define a graph in which residues are nodes and physical proximity defines the edges. This graph representation is then processed by two types of graph neural network layers in sequence: graph convolutional network layers, which aggregate information from neighboring residues, and GATv2 layers, a more expressive form of graph attention that learns to weight which neighbors matter most for each residue. An attention pooling step then compresses the residue-level representations into a single protein-level vector that feeds a multilabel classifier.

The output of that classifier is a set of Gene Ontology terms, the standardized vocabulary that biologists use to describe protein functions across three branches: molecular function, biological process, and cellular component. Because Gene Ontology is hierarchical, predicting a function is inherently a multilabel problem, and DeepGreenGO was trained to predict terms specific to the ontology it was trained on. The training corpus was a carefully curated Viridiplantae dataset of 7,534 experimentally annotated protein structures drawn from the Protein Data Bank, cross-referenced through the SIFTS resource to link structures to sequences and taxonomy. Crucially, the team split these proteins into training, validation, and test sets of 6,026, 754, and 754 PDB chains respectively using sequence-similarity-aware clustering, a methodological safeguard that prevents near-identical proteins from appearing in both training and test data and inflating performance estimates.

The benchmarking results reveal where the plant-specific approach pays off. DeepGreenGO was evaluated against sequence homology-based tools and against other recent sequence- and structure-based deep learning methods, and it showed its strongest performance in the biological process ontology, arguably the hardest and most informative of the three GO branches. There it achieved the highest protein-centric maximum F1 score, known as Fmax, of 0.227, and the lowest Smin score of 19.73, a metric that penalizes both missed and spurious predictions in a semantically aware way. Perhaps most striking was its showing on the information-content weighted area under the precision-recall curve, where it scored highest among all compared methods for biological process. That metric rewards models for correctly predicting rare, informative annotations rather than playing it safe with broad, shallow terms, suggesting that DeepGreenGO is genuinely learning specific plant biology rather than recycling generic predictions.

Ablation experiments, in which components of the model are systematically removed to measure their contribution, added an honest and instructive note to the study. The pretrained ProtBERT-BFD sequence embeddings turned out to provide most of the predictive signal, confirming the now well-established principle that self-supervised protein language models capture a remarkable amount of functional information on their own. The structural contact maps, processed through the graph neural network stack, contributed more modestly, and the authors identify improving the integration of structural information as a clear opportunity for future work. This kind of transparency matters in a field where architectural novelty is often claimed as the source of gains; here the data show that the language model backbone is central, while the graph machinery offers a framework that can be sharpened as structural prediction quality improves across the plant kingdom.

To demonstrate that the model is more than a benchmark winner, the researchers deployed DeepGreenGO across the entire proteome of rice, Oryza sativa, screening all 43,649 rice proteins for functions related to seed development. The scan flagged 372 candidate proteins associated with this agriculturally critical process. The team then validated the biological plausibility of these predictions through two independent lines of evidence. Functional enrichment analysis showed that the predicted proteins were collectively linked to biological processes known to operate during seed formation, and transcriptomic analysis revealed that many of them display elevated expression in ovary, embryo, and endosperm tissues, the anatomical arenas where seed development unfolds. For a purely computational prediction to align with tissue-level expression patterns is a meaningful consistency check, and it suggests the candidate list is a credible starting point for experimental follow-up.

The implications reach well beyond rice. Seed development genes influence yield, grain quality, and stress resilience, traits at the heart of breeding programs aimed at sustainable agriculture in a changing climate. A tool that can prioritize which of tens of thousands of uncharacterized proteins deserve scarce laboratory resources is a practical accelerator for crop science, particularly for orphan crops and underexplored species where homology-based annotation fails. Because DeepGreenGO is taxonomically focused, its predictions are also more trustworthy for proteins with limited detectable similarity to the training set, exactly the proteins that dominate non-model plant genomes. The framework could in principle be retrained or extended as new plant structures are experimentally solved and as AlphaFold-style predicted structures expand the structural universe available to graph-based methods.

There are also broader lessons here for the design of biological AI systems. The study is a case study in domain-specific curation: rather than scraping the largest possible dataset, the team built a plant-only corpus, controlled for sequence similarity during data splitting, and evaluated with metrics that distinguish informative predictions from easy ones. The modest absolute Fmax values in the biological process ontology are a candid reminder that protein function prediction remains genuinely hard, and that headline numbers from multi-taxon benchmarks should not be assumed to hold in specialized domains. The work also highlights the growing role of research groups outside the traditional centers of AI and genomics; the Sri Lankan team, working with donated workstations acknowledged from the Colombo University Faculty of Science Alumni Association in North America and a faculty high-performance computing facility, produced a competitive model without industrial-scale resources.

DeepGreenGO arrives at a moment when the intersection of protein language models, graph learning, and structural biology is reshaping what is computable in the life sciences. Its strongest advantages, in informative biological process annotations and in proteins far from any characterized relative, point toward a future where plant genomes are no longer annotated by proxy from animals and fungi but on their own terms. The rice seed development results offer a template for how such models should be deployed: predict broadly, then triangulate with enrichment and expression data before anyone commits a greenhouse season to a candidate gene. As plant structural datasets grow and structural integration improves, models of this kind could become standard instruments in the crop biologist’s toolkit, quietly converting sequence catalogs into testable biology, one green protein at a time.

Subject of Research: Graph neural network-based deep learning for plant-specific protein function prediction

Article Title: DeepGreenGO: a graph neural network-based deep learning model for plant-specific protein function prediction

Article References: Sridharan, G., Weththasinghe, S. A., Sridharan, A., Upeka, W. M. M., Abeywardhana, D. L., & Fernando, P. C. (2026). DeepGreenGO: a graph neural network-based deep learning model for plant-specific protein function prediction. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06677-9

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06677-9

Keywords: DeepGreenGO, protein function prediction, graph neural networks, plant biology, Gene Ontology, ProtBERT-BFD, rice, seed development, bioinformatics, deep learning, Viridiplantae, sustainable agriculture

The heart of the discovery concerns a chemical tag on histones, the protein spools around which DNA is wound. Specifically, the team focused on trimethylation of lysine 27 on histone H3, universally abbreviated H3K27me3. This mark is the calling card of Polycomb repressive complexes, evolutionarily ancient machines that silence genes by altering chromatin structure without changing the underlying DNA sequence. In animals, H3K27me3 is central to maintaining cell identity and to the epigenetic crosstalk between mother and offspring during pregnancy. In flowering plants, Polycomb proteins were already known to coordinate seed coat growth with fertilization, but whether the maternal tissues’ epigenetic state could reach into the embryo and steer its patterning had never been demonstrated.

Using CUT&Tag, a sensitive chromatin-profiling technique that maps histone modifications in tiny amounts of tissue, the researchers charted the H3K27me3 landscape in the integuments, the maternal cell layers that develop into the seed coat, of the reference plant Arabidopsis thaliana. They compared the landscape before fertilization and at 24 hours after pollination, a window during which the fertilized egg, the zygote, elongates and divides into the first embryonic lineages. What emerged was a strikingly dynamic picture: thousands of genomic regions gained or lost the repressive mark after fertilization, and the genes associated with those regions shifted in expression accordingly. The maternal seed coat, in other words, undergoes a coordinated epigenetic reprogramming precisely when the embryo is making its earliest patterning decisions.

To test whether this reprogramming matters functionally, the team turned to mutants lacking the three Arabidopsis H3K27me3 demethylases, the enzymes ELF6, REF6 and JMJ13, which erase the repressive mark. Plants carrying mutations in all three genes, a triple mutant the authors call erj-1, produced embryos with pronounced defects in their apical region, the part that gives rise to the shoot and the cotyledons. Crucially, the defect was not a failure of the embryo’s basic cell fate decisions. Single-cell transcriptome analysis of the apical and basal cell lineages showed that the early lineage-specification program ran normally in erj-1 embryos. Instead, the embryos failed at a later step: organizing the apical structures that define a mature body plan. This distinction, between specifying cell identities and arranging them into a pattern, proved to be the key to the whole mechanism.

Genetic detective work then localized the action to the mother. In reciprocal crosses between wild-type and mutant plants, the embryonic defects followed the maternal genotype, not the embryo’s own. When the demethylases were expressed only in maternal tissues, driven by the integument-specific SEEDSTICK promoter, the mutant phenotype was rescued. Allele-specific expression analysis confirmed that ELF6, REF6 and JMJ13 transcripts in early embryos are overwhelmingly of maternal origin, and fluorescently tagged versions of the proteins were detected in the maternal integuments rather than in the embryo itself. The epigenetic erasers, in short, work in the seed coat, and their influence travels across the maternal-embryonic boundary.

How does a chemical change in the seed coat reach the embryo? The answer is the plant hormone auxin, the great morphogen of plant development, which forms gradients that pattern organs from roots to leaves. In the erj-1 integuments, where H3K27me3 accumulates abnormally because it can no longer be erased, the researchers found that auxin levels dropped specifically in the micropylar integuments, the region where the embryo attaches to maternal tissue. Auxin signaling reporters such as R2D2 and DR5::GFP lit up weakly in these tissues, and the embryos themselves showed reduced auxin signaling. The maternal tissue, it seems, normally builds an auxin reservoir at the embryo’s doorstep and hands the hormone over to the developing offspring.

The molecular link between the histone mark and the hormone turned out to be two auxin transport genes, AUX1 and LAX1, which encode influx carriers that move auxin across cell membranes. In the erj-1 mutant, both genes carry elevated H3K27me3 and are transcriptionally downregulated in the integuments. The team showed that REF6, one of the demethylases, binds directly to specific sequence motifs in the promoters of AUX1 and LAX1, using yeast one-hybrid assays and published chromatin immunoprecipitation data as evidence. When AUX1 was expressed specifically in the integuments of erj-1 or aux1 lax1 mutant plants, using the TT10 seed coat promoter, auxin accumulated again in the micropylar region and embryonic patterning was substantially restored. The causal chain is therefore remarkably clean: maternal demethylases remove H3K27me3 from auxin transporter genes, the transporters load auxin into the micropylar integuments, and that auxin flows into the embryo where it instructs apical patterning.

Notably, the study found that the opposite side of the Polycomb system, the methyltransferases that write H3K27me3, played no detectable role in early embryonic patterning. Mutants in components of Polycomb repressive complex 2, including CURLY LEAF and SWINGER, produced embryos at normal frequencies. This asymmetry suggests that the dynamic removal of the mark, rather than its deposition, is the regulatory lever that maternal tissues pull during the fertilization window. It also echoes a theme familiar from animal biology: in mammals, oocytes actively remodel the epigenetic state of the genome they contribute, and placental epigenetics shapes pregnancy outcomes. The plant seed coat, functionally analogous to a placenta, now appears to run an epigenetic script of its own with direct consequences for the offspring’s form.

The implications extend well beyond basic developmental biology. Seed formation is the foundation of agriculture, and embryonic patterning defects translate directly into poor seed set and abnormal seedlings. Understanding that the maternal seed coat exerts epigenetic control over embryo architecture opens a potential avenue for crop improvement: modulating the expression of histone demethylases or auxin transporters in maternal tissues could strengthen embryo development under stress conditions or in hybrid breeding programs, where maternal effects are often decisive. The datasets generated in the study, including RNA sequencing of eight-cell embryos and seed coats and CUT&Tag maps deposited in the National Genomics Data Center, provide a public resource for exploring these possibilities.

Conceptually, the work reframes the maternal microenvironment from a passive incubator into an active epigenetic instructor. A mother plant, through the controlled erasure of a repressive histone mark in her seed coats, decides how much auxin her embryo will receive and, by extension, how that embryo will be shaped. The finding that this instruction crosses the generational boundary via a hormone rather than via inherited chromatin states adds a new mechanism to the growing catalog of maternal effects in biology. As Li, Zhao and their colleagues continue to dissect the pathway, one thing is already clear: in the quiet world of seeds, the mother’s epigenome speaks, and the embryo listens.

Subject of Research: Maternal epigenetic regulation of embryonic patterning via H3K27me3 demethylation and auxin transport in plant seed coats

Article Title: Maternal epigenetic regulation of embryonic patterning in plants

Article References: Maternal epigenetic regulation of embryonic patterning in plants. (n.d.). https://doi.org/10.1038/s41477-026-02421-6

Image Credits: AI Generated

DOI: 10.1038/s41477-026-02421-6

Keywords: plant embryogenesis, epigenetics, H3K27me3, histone demethylases, auxin, seed coat, Arabidopsis, maternal effects, Polycomb, AUX1, LAX1, seed development

The core obstacle is a trade-off that has haunted the field of synthetic apomixis since its inception. Apomixis is the natural ability of some plants, mostly wild relatives of crops, to produce seeds asexually: the embryo inside the seed is a genetic clone of the mother plant, no meiosis and no fertilization required. If engineers could install this trait into hybrid crops, each seed would carry the exact hybrid genotype, locking in the vigour of the original cross. Researchers have made real progress by hijacking the plant’s own developmental machinery. By mutating three genes involved in meiosis, they can replace the reductive cell division with a mitosis-like division, producing gametes that retain the full parental genome, a process called clonal gametogenesis or mitosis instead of meiosis, known as MiMe. By expressing the BABY BOOM transcription factor in egg cells, they can coax those unreduced egg cells into embryos without sperm delivery, generating clonal seeds.

The problem is that these two interventions fight each other. Inducing fertilization-independent embryogenesis in the egg cell is inefficient, and the resulting seeds are often fewer, smaller or less viable than normal. The egg cell is not naturally poised to launch an embryo; it is a terminally differentiated cell waiting for a sperm signal. Forcing a master embryonic regulator such as BABY BOOM to act there produces only a fraction of clonal seeds, and the plants that do form frequently show reduced fertility, undermining the very yield advantage that apomixis is supposed to preserve. This is the apomixis trade-off: the more aggressively you push asexual embryo initiation, the more you compromise the reproductive machinery that makes the seed viable in the first place. Breaking that trade-off has been the central engineering challenge of the field.

The new study, highlighted in the commentary, takes a different route by shifting the trigger from the maternal side to the paternal side. Instead of reprogramming the egg cell, the researchers identified HUAXU, a transcription factor that is normally expressed specifically in sperm cells. In ordinary sexual reproduction, the sperm delivers its paternal genome to the egg and, crucially, contributes signals that help activate the zygotic programme. HUAXU appears to be part of that paternal activation machinery: when deployed appropriately, it can trigger embryogenesis independently of the normal fertilization event. By coupling this sperm-derived embryogenesis cue with clonal gametogenesis, the team achieved near-complete clonal seed production in hybrid rice, meaning almost every seed carried the intact hybrid genome without genetic segregation.

The mechanistic insight matters as much as the agronomic result. Paternal activation of embryogenesis has long been recognized as a critical step in flowering plants, where the sperm cell does far more than deliver DNA. It contributes cues that reconfigure the egg cell from a quiescent state into an actively dividing zygote. The identification of HUAXU as a sperm-specific transcription factor capable of initiating this programme independently gives researchers a handle on a process that was previously accessible only through maternal reprogramming. It suggests that the egg cell may be more receptive to paternal signals than to forced maternal expression of embryonic regulators, which would explain why earlier strategies that placed BABY BOOM or similar factors under egg cell promoters produced low efficiencies and fertility costs.

Efficiency and fertility are not academic details; they determine whether synthetic apomixis can ever leave the laboratory. Earlier landmark work, published in Nature in 2018 by Khanday, Skinner, Yang, Mercier and Sundaresan, demonstrated that synthetic clonal seeds could be made in rice, but at rates far too low for practical breeding, and subsequent refinements published in Nature Biotechnology in 2019 improved the system without eliminating the penalty. The commentary authors, who include researchers affiliated with the Laboratory of Advanced Breeding Technologies at the Institute of Genetics and Developmental Biology in Beijing, frame the new result as a turning point precisely because it resolves the efficiency-fertility conflict: near-complete clonality combined with minimal yield penalty means the hybrid’s agronomic performance is essentially preserved through the seed generation. A scalable strategy for fixing heterosis requires exactly this combination.

The implications for global agriculture are substantial. Hybrid varieties underpin a large share of rice production in Asia, and hybrid seed production depends on manual or chemical emasculation, carefully timed planting of male-sterile and restorer lines, and enormous logistical expense. If clonal hybrid seed became routine, breeders could propagate elite hybrids indefinitely, smallholder farmers could save their own seed without losing vigour, and the genetic gains locked into each cross would no longer evaporate after a single generation. The commentary also points toward broader applicability: because the strategy relies on conserved reproductive processes, sperm-derived embryogenesis cues and clonal gametogenesis, the same logic could in principle be transferred to other cereals and broadleaf crops where hybrid breeding is central but hybrid seed is costly.

Caution remains warranted, as the commentary itself makes clear. The reported work is a mechanistic and proof-of-concept advance in rice; translating it into varieties grown across millions of hectares will require testing stability across environments, confirming that clonal fidelity holds over multiple generations, and navigating regulatory frameworks for genome-edited crops, which differ sharply between jurisdictions. The commentary’s corresponding author, Yazhong Wang, is listed as an inventor on a European patent application related to earlier work in this area, a reminder that intellectual property will shape how quickly these tools reach breeders. Still, the trajectory of the field is unmistakable: from the discovery that meiosis can be replaced by mitosis, through the first synthetic clonal seeds, to a system that now approaches the completeness and fitness that practical agriculture demands.

What makes the HUAXU result resonate beyond the rice paddy is what it reveals about the logic of plant reproduction itself. The egg cell, it turns out, may not need to be forcibly reprogrammed from scratch; it may simply need the right paternal instruction at the right moment. By finding that instruction and delivering it in an unreduced gamete, the researchers effectively persuaded the plant to treat an asexual embryo as if fertilization had occurred, without the genetic consequences of fertilization. Breaking the apomixis trade-off, as the commentary’s title puts it, is therefore both an engineering milestone and a biological insight, one that converts a long-standing theoretical promise into a strategy that is, for the first time, plausibly scalable for fixing hybrid vigour in one of the world’s most important staple crops.

Subject of Research: Synthetic apomixis and clonal seed production in hybrid rice via the sperm-specific transcription factor HUAXU

Article Title: Breaking the apomixis trade-off

Article References: Li, T., Wang, B., & Wang, Y. (2026). Breaking the apomixis trade-off. Nature Plants. https://doi.org/10.1038/s41477-026-02404-7

Image Credits: AI Generated

DOI: 10.1038/s41477-026-02404-7

Keywords: apomixis, hybrid rice, HUAXU, clonal seeds, heterosis, sperm cell, embryogenesis, plant biotechnology, clonal gametogenesis, seed development, plant breeding, transcription factor

The study, published in the journal Genome Biology, was led by Junpeng Shi of Sun Yat-sen University together with colleagues at Shandong Agricultural University, the Chinese Academy of Sciences and other institutions. The endosperm, the triploid tissue that nourishes the developing embryo and ultimately makes up most of what we eat when we consume corn, has long been recognized as the primary arena for genomic imprinting in plants. Because it contains two maternal genome copies and one paternal copy, the endosperm provides a natural system in which researchers can distinguish whether a gene’s transcripts came from the mother or the father, simply by counting the relative abundance of parental alleles.

Previous efforts to catalog imprinted genes in maize relied on bulk RNA sequencing of endosperm tissue that had been painstakingly dissected by hand. Those studies identified hundreds of candidate imprinted genes, but they averaged expression signals across an entire tissue that is anything but uniform. The endosperm is a mosaic of specialized cell types, each with distinct developmental roles, and averaging across them risks blurring the very signals that matter. The new work set out to resolve that heterogeneity directly, asking whether imprinting looks the same in every endosperm cell or whether it, too, is compartmentalized.

To answer that question, the researchers deployed two complementary technologies. The first was 10x Visium spatial transcriptomics, which captures gene expression while preserving information about where in the tissue each transcript was detected. The second was single-nucleus RNA sequencing, or snRNA-seq, which profiles the transcriptomes of individual nuclei, allowing each cell’s identity to be inferred from its expression signature. Crucially, the team applied both methods to reciprocal maize kernels, meaning kernels produced by crossing two inbred lines in both directions, so that the maternal and paternal genomes could be swapped and parental origin of each allele tracked unambiguously.

The combined datasets revealed six major endosperm cell types, providing the most granular cellular atlas of this tissue to date. When the researchers examined allele-specific expression, the pattern of transcripts attributable to one parent versus the other, they found something striking: this expression was strongly cell-type specific. Roughly two-thirds of the genes showing allele-specific expression were detected in only a single endosperm cell type. In other words, the parental bias governing a gene’s activity is not a fixed property of the gene itself but depends heavily on which cellular neighborhood that gene happens to be operating in.

The two technologies proved highly consistent with one another. Nearly 70 percent of the imprinted genes identified through spatial profiling could be validated at the cell-type level using single-nucleus sequencing, a concordance rate that strengthens confidence in both platforms and in the biological conclusions drawn from them. Yet when the team compared imprinted gene sets across the different endosperm cell types, they found that fewer than half of the imprinted genes were shared among multiple cell types. The implication is profound: the imprinting landscape of the endosperm is substantially shaped by spatial position and cellular identity, not merely by the epigenetic state inherited from the gametes.

The study also addressed a long-standing question about the balance of parental influence. In maize endosperm, paternally expressed genes, those that preferentially silence the maternal copy and express the paternal one, outnumbered maternally expressed genes by approximately twofold. This asymmetry adds to evidence that the paternal genome exerts a stronger transcriptional imprint on endosperm development than the maternal genome, at least in terms of the number of genes affected, though the functional consequences of that imbalance remain an open question for future work.

Among the most intriguing discoveries was a maternally expressed long non-coding RNA located within a region of the genome where the maternal copy is hypomethylated, meaning it carries fewer DNA methylation marks than the paternal copy. This lncRNA sits alongside Mez1, one of the canonical, best-characterized imprinted genes in maize, which is known to reside in a differentially methylated region. The co-localization of a novel imprinted non-coding RNA with a classic imprinted protein-coding gene in a shared hypomethylated landscape suggests that these loci may be subject to coordinated epigenetic regulation, hinting at regulatory architectures that earlier bulk approaches simply could not see.

Technically, the work demonstrates how modern genomics can dissect a phenomenon that has resisted fine-scale analysis for decades. Allele-specific expression analysis requires distinguishing transcripts from two highly similar genome copies, and doing so in spatially resolved or single-cell data adds layers of computational complexity, from assigning nuclei to cell types to correcting for the uneven coverage typical of single-cell libraries. By validating spatial results against single-nucleus data, and by using reciprocal crosses to control for the effects of the specific inbred lines, the team built a methodologically robust framework that other plant researchers can now adapt to their own systems.

The broader significance of the study lies in what it says about the developmental logic of imprinting. The authors conclude that post-fertilization endosperm differentiation, the process by which the initially uniform triploid tissue diversifies into specialized cell types, actively shapes the spatial and cellular landscape of genomic imprinting in maize. Rather than being a static epigenetic legacy of fertilization, imprinting emerges as a dynamic, context-dependent program that unfolds as the kernel develops. The comprehensive resource generated by the team, including cell-type atlases, allele-specific expression catalogs, DNA methylation data and lncRNA annotations, is expected to serve as a foundation for investigating the functional roles of imprinting during kernel development, work that could ultimately inform efforts to improve seed traits in one of the world’s most important crops.

Subject of Research: Cell-type-resolved mapping of genomic imprinting and allele-specific expression in developing maize endosperm using spatial transcriptomics and single-nucleus RNA sequencing

Article Title: Spatial and single-nucleus transcriptomics reveal the complexity of genomic imprinting in maize

Article References: Li, T., Jiang, Y., Zhang, M., Zhu, K., Jiang, S., Yang, X., Liu, H., Wang, J., Dong, X., & Shi, J. (2026). Spatial and single-nucleus transcriptomics reveal the complexity of genomic imprinting in maize. Genome Biology. https://doi.org/10.1186/s13059-026-04291-9

Image Credits: AI Generated

DOI: 10.1186/s13059-026-04291-9

Keywords: maize, genomic imprinting, spatial transcriptomics, single-nucleus RNA sequencing, allele-specific expression, endosperm, epigenetics, DNA methylation, long non-coding RNA, Mez1, plant genetics, seed development