Plant environmental memory: implications, mechanisms and opportunities for plant scientists and beyond

Plants live in fluctuating environments and, being sessile, must continuously integrate environmental information to optimize growth, development, survival, and reproduction. Environmental cues perceived in one developmental stage can affect later stages or even subsequent generations—an environmental memory. When such memory improves future performance under predictable conditions, it has adaptive significance and can shape evolutionary trajectories; however, it may also constrain future development. Plant memory varies among populations across environments and contributes to local adaptation. Molecularly, plant memory arises from multilayered genetic, epigenetic, transcriptional, and metabolic mechanisms enabling transient to stable (mitotic/meiotic) information storage, with costs balanced by benefits and with mechanisms for dissipation. Many abiotic and biotic stressors (temperature, drought, salinity, pathogens) induce tissue-specific changes in gene expression and chromatin states (histone modifications, DNA methylation), making epigenetics a strong candidate in memory regulation. Mathematical modelling complements ecological and molecular insights, offering predictive frameworks for plant responses within and across generations, with applications to conservation, agriculture, and weed management. This review synthesizes ecological/evolutionary evidence for environmental memory, details molecular and metabolic mechanisms, and highlights modelling opportunities to understand and predict plant memory.

This review synthesizes empirical and theoretical work on plant environmental memory across four major themes: (1) Ecological and evolutionary contexts: Memory evolves when environmental cues are temporally (or spatially) autocorrelated with later selective environments, or when abiotic cues predict biotic selection (e.g., density dependence, herbivory). Studies across Arabidopsis, Plantago, Mimulus, Biscutella, Taraxacum, Populus, Quercus and others illustrate within- and transgenerational effects shaped by cue predictability, life history, and dispersal, with genotype-by-environment dependence. (2) Molecular mechanisms: Detailed within-generation memory is exemplified by vernalization (FLC repression via PRC2/H3K27me3 and lncRNAs COOLAIR/COLDAIR/COLDWRAP/ASL), and thermomemory (HSFA2/HSFA3 complexes, H3K4me3 marks). Transgenerational phenomena involve DNA methylation (CMT2 variation, RdDM, ROS1 demethylation effects), small RNAs, histone modifiers (REF6/BRM), and stress-activated transposons (ONSEN). (3) Metabolic interfaces: One-carbon/folate and methionine/SAM cycles (MTHFD1, METS1, SAMS, SAH) link metabolism to epigenetic states; ROS/proline metabolism influence chromatin regulators and flowering pathways; metabolic priming may underlie persistent stress imprints. (4) Modelling: Evolutionary theory delineates conditions favouring within- vs transgenerational memory; mechanistic and machine-learning models (e.g., for FLC dynamics, APSIM) demonstrate predictive utility and highlight data and transferability challenges. The cited literature spans lab, field, and meta-analyses, emphasizing the need for multi-factor and long-term studies, especially in long-lived species.

• Environmental prerequisites: Adaptive plant memory evolves when environmental cues correlate with later selective conditions (temporal/spatial autocorrelation), including cases where abiotic cues predict biotic selection (e.g., rainfall predicting next-year density). Life history matters: transgenerational memory is more common in short-lived species but can benefit long-lived species (e.g., Pinus sylvestris). Meta-analysis indicates transgenerational effects are more frequently detected in annuals than perennials (Yin et al. 2019).
• Within-generation molecular memory: Vernalization in Arabidopsis involves PRC2-mediated H3K27me3 at FLC with a digital ON–OFF cellular switch, multi-phase memory states (establishment, consolidation, perpetuation), and lncRNA regulation; FLC resetting occurs in embryogenesis via ABI3-dependent reactivation. Thermomemory relies on HSFA2/HSFA3 complexes and sustained H3K4me3 at target genes, enabling type I/II transcriptional memory and enhanced re-activation upon recurring heat.
• Transgenerational molecular memory: Heat can induce a REF6/BRM–HSFA2 positive feedback that reduces H3K27me3 and is inherited by progeny, impacting PTGS via SGS3/SGIP1 and tasiRNA pathways. DNA methylation dynamics (e.g., ROS1 demethylation at TE-rich promoters like RMG1) critically tune immunity; RdDM modulates maternal control over seed traits (size, dormancy) and responses to maternal environments. Stress-evoked DNA methylation changes can persist across generations but may reset after a stress-free generation (e.g., hyperosmotic stress in Arabidopsis).
• Genetic–epigenetic–environment interplay: Natural variation at methylation machinery (CMT2) associates with climate variables and CHH methylation patterns; epigenomic diversity correlates with growth temperatures, supporting selection on methylation determinants.
• Transposable elements and memory: Stress-activated TEs (e.g., ONSEN) introduce regulatory sequences and variation affecting gene responsiveness, potentially facilitating rapid adaptation; their activation is governed by epigenetic control.
• Metabolism–epigenetics crosstalk: One-carbon/folate and methionine pathways supply SAM and regulate SAH, constraining methyltransferase activity; mutations (mthfd1-1) alter genome-wide DNA methylation and histone marks. Proline and ROS influence developmental transitions and possibly FLC expression. Parental water-stress experiences can enhance offspring antioxidative capacity (e.g., Plantago), suggesting metabolic priming in transgenerational memory.
• Modelling insights and applications: Theory predicts when within- vs transgenerational memory is favoured (costs, predictability, cue reliability), and how multiple cues are weighted. Mechanistic models of FLC and field-informed phenology models forecast flowering under climate change; crop system models (APSIM) implicitly embed within-generation memory and aid yield predictions. Machine learning can augment predictions but faces transferability/data limitations; hybrid mech-ML approaches show promise.

The review integrates ecological theory, empirical studies, and molecular mechanisms to demonstrate that plant environmental memory is an adaptive strategy contingent on cue predictability and life history. Molecularly, robust, multilayered networks (chromatin modifications, small RNAs, transcription factors, and metabolic state) enable plants to encode, maintain, and dissipate information from past environments. These mechanisms explain how plants balance responsiveness with stability—e.g., FLC’s polycomb-based memory and reset, thermomemory’s histone marks—allowing context-appropriate phenotypic adjustments within and across generations. The findings underscore that metabolism supplies substrates and regulatory feedback to epigenetic enzymes, coupling environmental status to chromatin and gene expression. Incorporating these insights into mathematical models enhances predictive power for phenology, stress tolerance, and yield under future climates, guiding breeding, conservation, and management. However, genotype-by-environment specificity, multi-cue interactions, and temporal scales necessitate field validation and multi-factor experiments to translate memory mechanisms into reliable predictions.

Plant environmental memory is governed by multilayered, partly redundant and fail-safe molecular systems integrating genetic, epigenetic, and metabolic components. These systems allow adaptive within- and transgenerational responses that can shape eco-evolutionary trajectories. The review highlights (1) ecological conditions favouring memory evolution; (2) detailed mechanistic paradigms (vernalization, thermomemory, immunity priming); (3) a critical role for metabolism in modulating epigenetic states; and (4) the promise of modelling (mechanistic and data-driven) to predict responses and identify targets for agriculture and conservation. Future work should prioritize multi-cue, field-based, and long-term studies (especially in long-lived species), deeper dissection of metabolic–epigenetic feedbacks, and development of computational frameworks incorporating cross-generational environments to improve predictive capacity and practical deployment.

• Many empirical studies occur under controlled, single-cue conditions, potentially missing interactions among simultaneous abiotic and biotic cues observed in nature; responses can switch from adaptive parental effects to within-generation plasticity under combined treatments.
• Genotype and population differences are common; broader sampling is needed to capture (epi)G × E effects and spatial variation in cue predictability (e.g., herbivory regimes).
• Long-lived species are underrepresented; sustained funding and long-term experiments are required to assess memory benefits across extended lifespans.
• Mechanistic uncertainties persist: duration and stability of epigenetic marks across generations, rules of dissipation/resetting, and causal links between TE dynamics, methylation changes, and adaptive outcomes.
• Modelling constraints include limited mechanistic knowledge for many species/traits, data scarcity/format issues for parameterization, and challenges in model transferability to novel climates; integrating cross-generational environmental information adds computational and data demands.