Empowering plant epigenetics to breed resilience of crops: From nucleolar dominance to transgenerational epigenetic inheritance

Abstract

Advancements in genomic and epigenetic research in both plants and animals have transformed breeding methods and biotechnological strategies for crop improvement, particularly in the face of extreme weather challenges. These breakthroughs in plant biology and agriculture have laid a strong foundation for ensuring food security, promoting environmental sustainability, enhancing nutritional health, and driving basic science advances, as exemplified by Mendel's discovery of genetic principles and McClintock's discovery of transposable elements. Plant epigenetics has held a transformative potential for developing high‐yielding and resilient crops. In this review, I will examine various relevant epigenetic phenomena, including nucleolar dominance, paramutation, imprinting, somaclonal variation, and transgenerational epigenetic inheritance, to explore strategies for overcoming yield limitations in an increasingly volatile climate. This perspective aligns with the vision for plant breeding and sustainable agriculture championed by the late Professor Ronald L. Phillips.

Core Ideas

• Hybridization and polyploidy induce genetic and epigenetic variation.

• Many epigenetic variations are reversible and heritable.

• Epigenetic variations are associated with agronomic and domestication traits.

• Plant epigenetics provides new strategies for molecular breeding, genome editing, and epigenetic engineering.

Abbreviations

AGL

AGAMOUS‐LIKE

aza‐dC

5′‐aza‐2′‐deoxycitidine

CmWIP

Cucumis sativus Trp‐Ile‐Pro domain transcription factor

CNS

cytoplasmic‐nuclear substitution

CO

CONSTAINS

COL

CONSTAINS‐LIKE

DME

DEMETER

easiRNAs

epigenetically activated small interfering RNAs

EBN

endosperm balance number

EIN2

ETHYLENE INSENSITIVE 2

FT

FLOWERING LOCUS T

Ga

G. arboreum

Gb

G. barbadense (AD)₂

Gd

G. darwinii (AD)₅

Gh

Gossypium hirsutum (AD)₁

Gr

G. raimondi

Gt

G. tomentosum (AD)₃

Igf2

Insulin‐like growth factor 2

Ks

synonymous substitutions

LCM

laser‐captured microdissection

Mop1

mediator of paramutation 1

NOR

nucleolar organizer region

NRPD1

nuclear RNA polymerase subunit D1

PolIV or p4

RNA polymerase IV

QTLs

quantitative trait loci

RdDM

RNA‐directed DNA methylation

TAdM

trans‐acting demethylation

TAM

trans‐acting methylation

Vgt1

vegetative to generative transition 1

1. INTRODUCTION

Plant biologists have two major missions in response to the challenges and opportunities in the world. The first is to create new knowledge and technology that will directly or indirectly improve agricultural productivity, enhance human health, and protect sustainable environment and ecosystems. The second is to make groundbreaking advancements using plants as experimental systems, as garden peas were used for the discovery of genetic principles (Feng et al., 2025) and maize was used for the discovery of transposable elements (TEs) (McClintock, 1941). Plants provide food, feed, fuel, and renewable materials for human civilization through extensive domestication of hundreds of plants for agricultural, horticultural, and medicinal purposes (Huang et al., 2022). Increasing crop resilience and yield by 50% or more will be needed to meet a growing population that is estimated to reach ∼10 billion by 2050 (Bailey‐Serres et al., 2019; Eshed & Lippman, 2019). Moreover, urbanization has dramatically reduced arable land for agriculture, and deforestation has accelerated deterioration of agricultural land and ecosystems in the face of rapidly changing climate. The atmospheric CO₂ level is increasing annually at ∼2 ppm(by volume) from 280 ppm at preindustrial levels to ∼412 ppm in 2020, which contributes to the predicted increase in global temperature. To mitigate this problem, we should not only lower the carbon emissions but also preferably find solutions to CO₂ sequestration and storage. Approximately ∼25% of global carbon emissions are captured by plant ecosystems as aboveground biomass and belowground soil‐root organic matter (Kell, 2012; Mayer et al., 2018). Plants provide an alternative and clean solution to combating climate change while increasing crop resilience and ensuring food security (Eckardt et al., 2023; Mayer et al., 2018).

As sessile organisms, plants cannot move and must incorporate ever‐changing environmental (external) and genomic (internal) signals to optimize their growth and development to complete life cycles, leading to phenotypic plasticity. This unique life cycle may explain why many epigenetic phenomena were first discovered in plants, from nucleolar dominance in Crepis (Navashin, 1934), wheat (Flavell, 1986), and Arabidopsis and Brassica (Chen et al., 1998a; Chen & Pikaard, 1997a, 1997b), paramutation in maize and sorghum (Brink, 1973; Chandler, 2007; Deans et al., 2024; Hollick, 2017), transposon activation in maize (McClintock, 1941; McClintock, 1984), to cosuppression in Petunia (Napoli et al., 1990) and RNA silencing in Nicotiana benthamiana, an Australian weed and relative of cultivated tobacco (Baulcombe, 2023; Hamilton & Baulcombe, 1999). Despite that these phenomena have been well documented with experimental evidence, the emerging field of epigenetics has been skeptical about inheritance and stability, which is complicated by a series of debates between the popular Lamarckian “inheritance of acquired characters” and infamous Lysenko's “pseudo‐inheritance” (Cao & Chen, 2024; Grossniklaus et al., 2013; Heard & Martienssen, 2014).

Core Ideas

• Hybridization and polyploidy induce genetic and epigenetic variation.

• Many epigenetic variations are reversible and heritable.

• Epigenetic variations are associated with agronomic and domestication traits.

• Plant epigenetics provides new strategies for molecular breeding, genome editing, and epigenetic engineering.

The skepticism has abated after a plethora of findings to support that many epialleles or epi‐mutations are evolutionary stable and linked to quantitative trait loci (QTLs) in crops. For example, asymmetrical flower development in toadflax (Linaria vulgaris) is associated with an epiallele of cycloidea gene, a mutation that was originally described by Linnaeus more than 250 years ago (Cubas et al., 1999). FLOWERING WAGENINGEN (fwa) is an epi‐mutations for flowering time found in the progeny of the demethylation mutant (Kakutani, 1997) or through mutant screens (Soppe et al., 2000). Many “traditional” QTLs in crop plants are found to be associated with epialleles, including SBP‐box epi‐QTL for fruit ripening in tomato (Manning et al., 2006), CmWIP1 (where CmWIP is Cucumis sativus Trp‐Ile‐Pro domain transcription factor) epiallele for sex determination in melon (Martin et al., 2009), DWARF1 (D1) epiallele for plant architecture in rice (Miura et al., 2009), LINE retrotransposon (aka Karma) epiallele for mantled somaclonal variation in oil palm (Ong‐Abdullah et al., 2015), and epi‐QTLs for vitamin E accumulation in tomato (Quadrana et al., 2014). Moreover, many epigenetic changes are heritable in newly formed and natural Arabidopsis suecica (X. Jiang et al., 2021) and in allotetraploid cotton species that were formed 1–1.5 million years ago (Mya) (Chen et al., 2017, 2020), including domestication and agronomic traits such as flowering time and seed dormancy (Song et al., 2017). These epialleles that affect transcription and function of the associated genes have played critical roles in plant evolution, crop domestication, and molecular breeding.

In this review, I focus on establishment and maintenance of a few epigenetic phenomena, including nucleolar dominance, endosperm development in Arabidopsis and Brassica, and transgenerational inheritance in maize and cotton, which are directly or indirectly influenced by some original work from the late Ronald L. Phillips and his former postdocs and students (Kaeppler et al., 2009; Kaeppler & Springer, 2024). For example, Dr. Phillips and his former associates and colleagues pioneered the work on plant tissue culture (Green & Phillips, 1975) and tissue culture‐induced epigenetic variation (Kaeppler & Phillips, 1993). They identified noncoding genomic sequences associated with flowering‐time QTL (vegetative to generative transition 1 [Vgt1]) in maize (Salvi et al., 2007). Vgt1 is a 2‐kb noncoding region located ~70 kb upstream of an AP2‐like transcription factor to be involved in flowering‐time control. Further analysis has found a miniature transposon insertion into the Vgt1 region, which is associated with DNA methylation that regulates expression of the gene to modulate early or late flowering (Castelletti et al., 2014). This type of gene regulation via cis‐regulatory elements as well as trans‐acting factors (Ng et al., 2011; Shi et al., 2012) has emerged to become a general mechanism for natural variation and tissue‐specific gene expression (Marand et al., 2025; Marand & Schmitz, 2022). Readers should refer to other excellent reviews in these related topics, such as nucleolar dominance (Pikaard, 1999; Pikaard et al., 2023; Reeder, 1985), imprinting in seeds (H. Jiang & Kohler, 2012; Khouider & Gehring, 2024), paramutation (Chandler, 2007; Hollick, 2017; Kulikova et al., 2024; Stam & Mittelsten Scheid, 2005), RNA silencing (Baulcombe, 2004; Baulcombe, 2023; Jorgensen et al., 1998), and epigenetic regulation in plants and crops (Cao & Chen, 2024; Springer & Schmitz, 2017; H. Zhang & Zhu, 2025). Understanding the basis of epigenetic and transgenerational inheritance will help us explore and utilize a repertoire of hidden epigenomic variation for improving crop yield and resilience using molecular breeding, genome editing, and epigenetic engineering.

1.1. Nucleolar dominance—New insights into an old epigenetic phenomenon

Nucleolar dominance refers to selective transcriptional activity of ribosomal RNA genes in an interspecific hybrid or allopolyploid in plants and animals (Pikaard, 1999; Pikaard et al., 2023; Reeder, 1985). Each plant or animal species has at least one nucleolar organizer region (NOR), a cytological marker that is separated by a secondary constriction; the primary constriction is the centromere (Figure 1). Each NOR consists of hundreds and thousands of copies of rDNA repeat (∼350 copies per NOR in Arabidopsis thaliana) (Pikaard et al., 2023), which are transcribed by RNA polymerase I into 35–48S pre‐rRNA transcripts and processed into 18S, 5.8S, and 25–28S rRNAs, which together with 5S rRNA (transcribed by RNA polymerase III), form the catalytic core of the protein‐synthesizing machine ribosomes (Grummt & Pikaard, 2003).

FIGURE 1.

Transcriptional nucleolar dominance in Arabidopsis suecica. (A) Diagram of A. thaliana (At) haploid chromosomes with NOR2 and NOR4 on chromosomes 2 and 4, respectively. An autotetraploid At Ler plant is shown. (B) Diagram of A. arenosa (Aa) haploid chromosomes with NOR3, NOR4, and NOR7 locations. Arabidopsis arenosa is a natural outcrossing tetraploid. (C) Diagram of haploid A. suecica‐like allotetraploid chromosomes with active Aa NOR3, NPR4, and NOR7. At NOR2 and At NOR4 are silenced in natural A. suecica strains, and At NOR2 is lost in some strains. (D) An autoradiogram showing transcription of rRNA transcripts in At, Aa, and A. suecica. Note that At rRNA transcripts (*) are detected in A. suecica when the seedlings are treated with 10 and 25 (mg/L) of 5′‐aza‐2′‐deoxycitidine (aza‐dC), respectively. Gel image was copied from Chen et al. (1998b, Fig. 7A). Dashed lines were added for visual separation of At and Aa subgenomic RNA in A. suecica.

B. McClintock first described NOR in maize (McClintock, 1934) after chromosome breakage‐fusion‐bridge cycle by joining two NORs in one chromosome (McClintock, 1941). In 1971, Prof. Phillips identified chromosomal location of ribosomal RNA genes in the maize NORs using DNA/RNA hybridization (Phillips et al., 1971). This was the first localization of ribosomal RNA genes to a chromosome in plants. The copy number of rDNA repeat units ranges from 5000 to 12,000 per 2C nucleus, and chromosome 6 contained most rRNA genes in maize (Phillips et al., 1974). Decades later, Dr. Phillips and his long‐term collaborator, Dr. Howard W. Rines, had made a series of oat‐maize additional lines (Kynast et al., 2001; Riera‐Lizarazu et al., 1996), when I worked on cloning maize sequences from one of the additional line (Chen et al., 1998b). Interestingly, cytological observation found maize NOR6 was inactive in the oat‐maize chromosome 6 additional line. This observation has inspired me to study the molecular mechanism of nucleolar dominance in Arabidopsis and Brassica allopolyploids (Chen et al., 1998a; Chen & Pikaard, 1997a, 1997b).

The phenomenon of nuclear dominance was first described in Crepis, which is commonly known as hawksbeard or hawk's‐beard in the family Asteraceae (including sunflower). The number of NORs in the interspecific hybrid (2n = 2x = 7) between Crepis capillaris (2n = 2x = 6) and Crepis neglecta (2n = 2x = 8) is not additive (Navashin, 1934), that is, only the C. capillaris NOR appears, suggesting its dominance over the C. neglecta NOR. This dominance phenomenon has been reported in many other plants, including wheat (Crosby, 1957; Flavell & Odell, 1979), barley interspecific hybrids (Kasha & Sadasivaiah, 1971), maize hybrids (Jupe & Zimmer, 1993), triticale (a man‐made species between wheat and rye) (Houchins et al., 1997), and Solanum species hybrids (Yeh & Peloquin, 1965), as well as animals such as Drosophila interspecific hybrids (Durica & Krider, 1977) and frog interspecies hybrids (Cassidy & Blackler, 1974).

The link between active NOR and rRNA transcription has been demonstrated in the study of Xenopus interspecific hybrids (Honjo & Reeder, 1973; Macleod & Bird, 1982). The intergenic spacers of rDNA loci in the species Xenopus laevis and Xenopus borealis that are defined by different numbers of multiple 42‐bp repeats known as enhancer elements with more numbers in the former than in the latter. These enhancer elements tend to recruit nucleolar transcription factors such as upstream binding factor (Bell et al., 1988) to initiate rRNA transcription. When X. laevis and X. borealis rRNA minigenes are co‐transfected into X. laevis oocytes, the former is four times more actively transcribed than the latter, suggesting transcriptional dominance of X. laevis rRNA genes, which has been a predominant model to explain the nucleolar dominance (Reeder, 1985).

However, the transcriptional dominance effect cannot be recapitulated in protoplasts of Brassica napus allotetraploids, which consist of Brassica oleracea and Brassica rapa subgenomes. Both B. rapa and B. oleracea rRNA genes are transcribed at similar levels from the mini‐genes that are co‐transfected into protoplasts of B. napus allotetraploids (Frieman et al., 1999), whereas the endogenous B. oleracea rRNA genes are silenced in B. napus (Chen & Pikaard, 1997b). This result argues against transcriptional dominance but supports a silencing mechanism of B. oleracea rRNA genes by chromatin modifications in the allotetraploid (Figure 1). Indeed, the silencing occurs at the transcriptional level by nuclear run‐on assays, and the silenced transcription is derepressed by treatment of 5′‐aza‐2′‐deoxycitidine (aza‐dC), an analog inhibitor of (DNA) cytosine methylation (Chen & Pikaard, 1997a). Silencing can also be relieved (Chen & Pikaard, 1997a) by the treatment of butyrate and trichostatin A, general and potent chemical inhibitors, respectively, for histone deacetylases (Kruh, 1982; Yoshida et al., 1990).

Arabidopsis suecica is formed by hybridizing A. thaliana (Figure 1A) and Arabidopsis arenosa (Figure 1B), which occurred spontaneously in nature approximately 150,000–300,000 years ago (Novikova et al., 2017) or in laboratory conditions by manual pollination of A. thaliana with A. arenosa pollen. In A. suecica (Comai et al., 2000; J. Wang, Tian, Lee, Wei, et al., 2006) (Figure 1C), three NORs from A. arenosa are active, whereas A. thaliana NOR2 and NOR4 are inactive in natural and newly formed A. suecica or in some A. suecica strains (Burns et al., 2021; X. Jiang et al., 2021). Arabidopsis thaliana NOR2 is lost in a few accessions of natural A. suecica (Burns et al., 2021; Pontes et al., 2003). At the transcription level, A. thaliana rRNA genes are silenced, whereas A. arenosa rRNA genes are transcribed (Chen et al., 1998a) (Figure 1D). Similarly, the silenced A. thaliana rRNA genes are derepressed by aza‐dC treatment in a concentration‐dependent manner. Moreover, A. arenosa rRNA transcript levels are also upregulated, suggesting not all A. arenosa rRNA genes are transcribed in a given time, which is subject to a similar silencing mechanism involving DNA methylation and chromatin modification. Indeed, recent studies have shown differential activities of NOR2 and NOR4 within A. thaliana and that different switching‐off mechanisms may operate to control rRNA transcript levels during growth and development (see review Pikaard et al., 2023).

Studies of nucleolar dominance have provided several new insights. This epigenetic phenomenon involves NORs (parts of chromosomes) comprising thousands of copies of rDNA gene repeats. Once it is established, the silencing status is stable and can be transmitted through 15,000–300,000 years of evolution in A. suecica. However, the silencing is reversible by increasing dosage of the under‐dominant genome (i.e., A. thaliana) and during reproductive development (Chen et al., 1998a), which is probably due to rapid cell division during flowering. This dosage effect on nucleolar dominance reversal is different from the dosage compensation or counting mechanism on X‐chromosome inactivation, whereas additional X chromosomes or autosomes carrying X‐inactivation center become inactive (Avner & Heard, 2001). The chromosomal boundary of silencing NORs has been actively pursued but has yet to be clearly defined (Pikaard et al., 2023).

The initiation mechanism of nucleolar dominance remains unknown and may resemble the imprinting mechanism for X‐chromosome inactivation (Lee, 2000). During the early stage (implantation, 3–5 days) of embryo development, paternal X chromosome is selected for silencing, but choice of silencing becomes random after the blastocyst stage. In nucleolar dominance, the paternal (A. arenosa) rDNA genes are active, although it is difficult to examine this during fertilization and in early stages of plant embryos. It is also difficult to make the reciprocal hybrid using A. thaliana as a pollen donor, presumably due to self‐incompatibility in outcrossing A. arenosa (Chen, 2007; Jiang et al., 2021).

A recent study reported that nucleolar dominance could be variable among different accessions of A. suecica, including co‐expression of both A. thaliana and A. arenosa rRNA genes in many accessions, expression of only A. arenosa rRNA genes in several accessions, and only A. thaliana rRNA genes in one accession (Burns et al., 2021). Further analysis found at least three active NOR loci, Aar3, Aar4, and Aar7, in A. suecica (Jiang et al., 2021) (Figure 1D). This complexity of stochastic may reflect rapid and dynamic changes between A. thaliana and A. arenosa subgenomes in A. suecica. Alternatively, it may take several generations of epigenetic modifications to cement silencing chromatin in the underdominant set of rDNA genes, as observed in the protein‐coding genes (Wang, 2006; Wang et al., 2004). The establishment mechanism of rDNA silencing may also require a cis‐acting factor such as Xist RNA in X‐chromosome inactivation (Lee, 2000) or unknown trans‐acting factors from the other species. Identification of these factors will advance our understanding of selective silencing of parental rDNA subject to nucleolar dominance.

Going back to the early study by Phillips et al., 1974 about amplification of rRNAs in the endosperm, the number of rDNA repeats appears to be correlated with genotypes during maize domestication and improvement (Liu et al., 2017). The basis for this remains unknown. It is likely that ribosome activities are upregulated when the cell cycle is arrested, accompanied by transcriptional attenuation, as observed during cotton fiber cell development (Ando et al., 2021). Endoreduplication in the endosperm may result from coupling of DNA replication and transcription of rDNA with suppression of other activities such as cell cycle and division. As a result, translational activities dominate during endosperm and seed development, which remains to be investigated.

1.2. Genomic imprinting and endosperm development in seeds

Crop seeds provide nearly 70%–80% of calories and 60%–70% of all proteins consumed by the human population (Borlaug, 1973; Li & Berger, 2012). Endosperm is the direct or indirect source for most of the nutritional content of the seed, and it resembles the placenta in mammals (Moore & Haig, 1991), which is the source of nutrition for embryo development (Stebbins, 1976). In maize, mitotic activity sharply decreases in endosperm cells 10–12 days after pollination, a phenomenon called endoreduplication (Kowles & Phillips, 1985). Nuclear size and DNA content per nucleus could increase to about 90–200 times per haploid genome or even up to 690. This dramatic process of endoreplication is associated with rapid cellularization and enlargement of endosperm, leading to seed maturation.

Endosperm formation is part of double fertilization in flowering plants. Each male gamete (pollen) contains two sperm nuclei after miotic division. One sperm fertilizes the egg to form a zygote with a 1:1 maternal to paternal genome ratio (1 m:1p), whereas the other fertilizes two central cell nuclei to form an endosperm cell with a 2:1 maternal to paternal genome ratio (2 m:1p). In A. thaliana, increasing the paternal genome ratio (2 m:2p) in the endosperm by pollinating a diploid “mother” with a tetraploid “father” (2 × 4) delays endosperm cellularization (EC) and produces larger seeds (Lu et al., 2012) (Figure 2A). In contrast, increasing the maternal genome ratio (4 m:1p) in endosperm by pollinating a tetraploid mother with a diploid father (4 × 2) leads to precocious EC and smaller seeds (Scott et al., 1998).

FIGURE 2.

Seed size variation is dependent on maternal genome dosage and small RNAs. (A) Size variation of developing seeds (from 3 to 7 days after pollination, DAP) in diploids, tetraploids, and triploids (2 × 4 or 4 × 2) in A. thaliana C24. (B) Seed size correlates negatively with maternal genome dosage and siRNA abundance during endosperm and seed development. A group of maternally expressed p4‐siRNAs (where p4 is RNA polymerase IV), whose biogenesis is dependent on maternal genome dosage, negatively regulates expression of siRNA‐target genes including AGAMOUS‐LIKE (AGLs) in the endosperm, which affect endosperm cellularization and size. Multiple dots and an elongated black rod in each diagram represent the endosperm and embryo cells, respectively. m:p = maternal:paternal genome number. (C) RNA‐directed DNA methylation (RdDM) regulates endosperm development. A model for siRNAs to control expression of genes (e.g., AGL) in the endosperm. Early in development, RdDM activity is low, spatially restricting AGL expression to the CZE. Later in development, siRNAs and the RdDM pathway are activated to silence AGL expression. Loss of the RdDM pathway results in the failure to restrict spatial expression of AGL to the CZE and temporal expression of AGL up to the heart stage. CZE, chalazal endosperm; CZSC, chalazal seed coat; EP, embryo proper; MCE, micropylar endsperm; PEN, peripheral endosperm; SC, distal seed coat. (A) and (B) are adopted and modified from Lu et al. (2012) and (C) is adopted and modifiedfrom Kirkbride et al. (2019).

These findings are consistent with the concept of endosperm balance number (EBN) (Carputo et al., 1997; Johnston & Hanneman, 1982), which refers to the optimal maternal‐to‐paternal genome (2:1) ratio that permits proper endosperm development. Interploidy or intraspecific hybrids where the EBN is disrupted can lead to seed size morphology and postzygotic hybridization barriers (Comai, 2005; Scott et al., 1998).

At the molecular level, the EBN is directly associated with maternally expressed small interfering RNAs (siRNAs) (Lu et al., 2012). In plants, RNA polymerase IV (PolIV or p4) encoded by NRPD1a (where NRDPI is nuclear RNA polymerase subunit D1) is required for biogenesis of a major class of 24‐nt small interfering RNAs (also known as p4‐siRNAs) (Mosher et al., 2009), which are predominately expressed in developing endosperm (Lu et al., 2012). The accumulation of p4‐siRNAs depends on the maternal genome dosage (Figure 2B), and maternal p4‐siRNAs target TEs and TE‐associated genes (TAGs) in seeds (Lu et al., 2012). The p4‐siRNAs correlate negatively with expression levels of TAGs such as AGAMOUS‐LIKE (AGL) genes in endosperm of interploidy crosses (tetraploid × diploid). In the reciprocal crosses (diploid × tetraploid), NRPD1a is associated with p4‐siRNA reduction is associated with AGL upregulation in the endosperm. These results provide strong genetic evidence for maternal siRNAs in the control of parental genome imbalance and endosperm‐expressed genes during seed development.

Using the samples from laser‐captured microdissection (LCM) in the reciprocal crosses between the wild‐type and nrpd1 plants, Kirkbride et al., 2019 identified four distinct groups of NRPD1‐siRNA loci, which reglates expression of their target genes such as AGL91 and AGL40. The first group of siRNAs is dependent on maternal NRPD1 and present in both endosperm and seedcoat and is associated with endosperm‐expressed genes such as AGL91 and AGL40 (Figure 2C). These p4‐siRNAs regulate AGL expression and hence endosperm and seed development. The second group is the most abundant and dependent on maternal NRPD1; these p4‐siRNAs are exclusively expressed in the seedcoat. The third group is present in both endosperm and seedcoat with the seedcoat fraction is dependent on maternal NRPD1 while the endosperm fraction dependent on either maternal or paternal NRPD1. The last group is largely expressed in the endosperm but dependent on paternal NRPD1. In a separate study, this group of paternal‐NRPD1 dependent p4‐siRNAs is also known as the paternally epigenetically activated small interfering RNAs (easiRNAs) (Martinez et al., 2018). These easiRNAs affect paternal gene expression in postzygotic seed abortion in interploidy crosses (Kirkbride et al., 2015) and interspecific hybrids (Dziasek et al., 2024).

Consistent with siRNA distribution patterns, pAGO4:GFP::AGO4 (promoter:GFP::protein) is maternally expressed in the endosperm and seed coat but biparentally expressed in the embryo. This result supports spatiotemporal expression of AGO4 and other RNA‐directed DNA methylation (RdDM) pathway genes such as NRPD1 during seed development (Kirkbride et al., 2019).

The role of p4‐sRNAs in regulating endosperm‐expressed genes has been tested using a sensor of GUS (encoding β‐glucuronidase) fused with the AGL91 and AGL40 promoters that are targeted by p4‐siRNAs (Kirkbride et al., 2019). The results indicate the maternal p4‐siRNAs control both the spatial expression boundary in the chalazal endosperm and the temporal expression intensity of AGL91 and AGL40 from pre‐globular and globular to heart and linear cotyledon stages. This spatial‐temporal endosperm‐expression pattern is disrupted and spreads to all other endosperm domains when the maternal nrpd1 mutant was used in the cross (Figure 2C, middle). Moreover, the loss of late‐stage silencing implies that the RdDM in the seed is established after fertilization, when maternal NRPD1 is required for RdDM. These data indicate that p4‐siRNAs control the RdDM pathway and hence spatial and temporal expression domains of endosperm‐expressed genes during seed development.

Interestingly, compared with AGL91, the p4‐siRNAs silence maternal AGL40 less completely in the chalazal endosperm at early stages. This can result from the different siRNA targets present in AGL91 and AGL40 promoter regions. Alternatively, maternal AGL40 and AGL91 alleles are subjected to different modifications in the central cell through such mechanisms as demethylation by DEMETER (DME) (Gehring et al., 2006). Overexpressing AGL91 or AGL40 increases seed size, while seed size is reduced in the agl91 and agl40 mutants. These results indicate maternally expressed p4‐siRNAs can serve as suppressors of gene expression in developing seeds, as well as mediate spatial‐temporal expression of the endosperm‐expressed genes, which regulate seed size. These p4‐siRNAs may also exert parent‐of‐origin effects on EBN in interploidy and hybrid crosses (Dziasek et al., 2024; Kirkbride et al., 2015).

Alternatively, alterations to EBN are correlated with changes in easiRNAs that regulate expression of imprinted and non‐imprinted genes (Martinez et al., 2018). Additional information about the effect of imprinting on endosperm and seed development have been extensively reviewed (Gutierrez‐Marcos et al., 2003; Khouider & Gehring, 2024; Kohler et al., 2012; Li & Berger, 2012; Wilkins & Haig, 2003).

Imprinting is defined as an equal expression of paternal and maternal alleles in the offspring (Moore & Haig, 1991; Tilghman, 1999), which occurs in sexually reproducing organisms including humans and flowering plants (Feil & Berger, 2007; Ferguson‐Smith, 2011). There are a couple of hundred imprinted genes in mammals and plants. In flowering plants, most imprinted genes are found in the endosperm, whereas in mammals, imprinting occurs in the embryo (Feil & Berger, 2007). Endosperm resembles the function of placenta in mammals and provides nutrients for embryo development.

A prevailing hypothesis for imprinting is parental conflict model to balance the resource allocation between the mother and offspring (Moore & Haig, 1991; Tilghman, 1999). This is because many imprinted genes are related to growth factors and regulate growth and development in mammals (Ferguson‐Smith, 2011; Sasaki et al., 1992). For example, maternally expressed factors inhibit growth, while paternally expressed factors promote growth. As a result, the mutant of paternally expressed Insulin‐like growth factor 2 (Igf2) leads to a 40% reduction in growth (DeChiara et al., 1990), whereas the mutation of maternally expressed Igf2 receptor (Igf2r) results in overgrowth and death (Lau et al., 1994). Intriguingly, the double mutant is viable and develops normally (Filson et al., 1993), suggesting that imprinting is dispensable. In plants, both maternally and paternally expressed imprinted genes are required for maintaining optimal development and fitness of offspring (seeds), as observed in rice (Yuan et al., 2017). These data are consistent with the parental‐offspring coadaptation model (Bateson, 1994; Wolf & Hager, 2006), suggesting that selection favors coadaptation of maternal‐ and paternal‐offspring traits to achieve the optimal offspring seed development.

Imprinted gene expression is largely controlled by CG methylation in mammals (Ferguson‐Smith, 2011; Reik, 2007). In flowering plants, DNA methylation is found in both CG and non‐CG sites; the latter is known as CHG (H = A, T, or C) and CHH methylation through distinct pathways (Law & Jacobsen, 2010; Zhang et al., 2018). CG methylation largely depends on METHYLTRANSFERASE1 (MET1) (Kankel et al., 2003), while CHROMOMETHYLASE3 is responsible for CHG methylation (Lindroth et al., 2001). CHH methylation is established de novo through the RdDM pathway (Pelissier & Wassenegger, 2000; Wassenegger et al., 1994). Demethylation is less well understood and can occur through a DNA repair pathway. REPRESSOR OF SILENCING1 (ROS1) encodes a member of DNA glycosylase family to remove 5‐methylcytosine and replace it with unmethylated cytosine (Gong et al., 2002). Arabidopsis has three additional family members, DME (Choi et al., 2002) and DEMETER‐LIKE2 and 3 (DML2 and DML3) (Penterman et al., 2007). These members exhibit the parent‐of‐origin effect on demethylation of the maternal genome in the central cell prior to fertilization (Choi et al., 2002; Penterman et al., 2007).

Despite many imprinted genes being found in the endosperm of Arabidopsis (Gehring et al., 2011; Hsieh et al., 2011) and maize (Waters et al., 2011; Zhang et al., 2014), the function of these genes remains poorly understood. ETHYLENE INSENSITIVE 2 (EIN2) encodes a membrane protein that links the ethylene perception to transcriptional regulation. During seed development, EIN2 is maternally expressed in the endosperm of Arabidopsis and maize, which have been separated by over 150 million years of evolution (Chaw et al., 2004). In Arabidopsis, EIN2 is expressed specifically in the endosperm, and the maternal‐specific EIN2 expression regulates temporal patterns of EC (Ando et al., 2023). The maternal‐specific expression of EIN2 in the endosperm is controlled by DNA methylation and DME but not by H3K27me3 or by ethylene pathway genes. As a result, the size increases in the genetic cross using the ein2 mutant as the maternal parent or in the ein2 mutant. This EIN2 regulation involves cellularization‐related AGL genes such as AGL23, AGL28, AGL40, and AGL62, indicating a mechanism for the imprinted genes like EIN2 to control expression of endosperm‐expressed genes during seed development.

EIN2/3‐dependent ethylene signaling pathway facilitates pollen tube growth and attraction during fertilization (Volz et al., 2013; Zhang et al., 2018). After fertilization, EIN3 is involved in blocking other pollen tubes and programmed cell death in synergid cells (Heydlauff et al., 2022). Thus, in the embryo, expression of ethylene signaling pathway genes, including EIN2, is turned off to prevent cell degeneration and ensure proper embryo development (Volz et al., 2013). In the endosperm, the maternal allele of EIN2 is expressed due to demethylation in the central cell to regulate expression of cellularization‐related genes during early stages of endosperm development. EIN2 expression is uncoupled from ethylene signaling in sporophytic tissues and plays an unique role in EC and consequently regulates seed size. Similarly, understanding of EIN2 imprinting in maize may shed light on improving crop seed yield.

In seeds, the parent‐of‐origin effect can result from imprinting or combinational effects between cytoplasmic and nuclear genomes. For example, cytoplasm‐nuclear interactions contribute to changes in the metabolome and growth (Douglas et al., 2015). Several cybrids (same cytoplasm with different nuclei) have a strong effect on photosynthetic phenotypes (Flood et al., 2020; Joseph et al., 2015), consistent with the role of cytoplasmic genomes in photosynthesis and growth.

These confounding factors have been investigated in a recent study using cytoplasmic‐nuclear substitution (CNS) lines in A. thaliana ecotypes (June et al., 2024). After six generations of backcrossing, these CNS lines differ only in the nuclear genome (imprinting) or cytoplasm (maternal effect). The seedling phenotypes of CNS lines resembled the paternal parent. Notably, the parent‐of‐origin effect on seed size was found only in the reciprocal CNS crosses where the cytoplasm is the same, suggesting that seed size is largely controlled by imprinting of nuclear genes. This notion is supported by a large number of imprinted genes identified in the endosperm using LCM samples; only a few imprinted genes, including a long non‐coding RNA, are found in the embryo. Moreover, the parent‐of‐origin effect on seed and embryo size is largely controlled by NRPD1 (Kirkbride et al., 2019; Lu et al., 2012), supporting a role of the maternal NRPD1 allele in seed development. Together, these results suggest that imprinting and the maternal NRPD1‐mediated small RNA pathway play roles in seed size heterosis in plant hybrids.

1.3. Transgenerational epigenetic inheritance in corn hybrids and cotton tetraploids

Hybridization is the basis for genetics and breeding of plants and animals. McClintock (1984) predicted environmental stresses and genomic interactions between hybridizing species could induce “genome shock,” which referred largely to the activation of transposons and other genomic features. A contemporary view is that genomic interactions in hybrids lead to gene expression changes through epigenetic modifications, such as DNA methylation and chromatin remodeling (Cao & Chen, 2024; H. Zhang & Zhu, 2025). Most important agricultural crops, including wheat, cotton, and canola, are allopolyploids that are derived from interspecific hybridization, and many others, such as corn and rice, are grown as hybrids. In addition to genetic variation caused by hybridization, many epigenetic changes are heritable through evolution and modern breeding (Chen, 2007; Ding & Chen, 2018; Li & Chen, 2022; Spadafora, 2023; Weigel & Colot, 2012).

Like genetic variation, epigenetic modifications and/or epi‐mutations can occur spontaneously and widespread in natural and experimental populations (Chen, 2007; Hofmeister et al., 2017; Jiang et al., 2021; Johannes et al., 2008; Kakoulidou et al., 2021; Song et al., 2017). Recent studies indicate that DNA methylation change rates are much higher than the DNA sequence mutation rate in cotton (Song et al., 2017) and Arabidopsis (Yao et al., 2023). Thus, the methylation change rate can be used as a “molecular clock” to estimate population divergence in a relatively short evolutionary timescale such as domestication and breeding (Yao et al., 2023).

As the genetic principles that were first discovered in plants, many epigenetic phenomena and mechanisms such as paramutation (Brink, 1973; Chandler, 2007) and RNA silencing (Hamilton & Baulcombe, 1999) or co‐suppression (Napoli et al., 1990) were reported in plants. Allelic interactions in a hybrid (heterozygous state) can induce heritable changes of one allele over the other in the offspring, which is known as paramutation in plants (Brink, 1973; Chandler, 2007; Hollick, 2017; Stam & Mittelsten Scheid, 2005) and later in mice (Perez & Lehner, 2019; Rassoulzadegan et al., 2006), flies (de Vanssay et al., 2012), and worms (Sapetschnig et al., 2015). In maize, paramutation of the b1 locus is dependent on Mop1 (mediator of paramutation 1) locus, which encodes an RNA‐dependent RNA polymerase 2 (RDR2) (Alleman et al., 2006) and is involved in the biogenesis of siRNAs (Nobuta et al., 2008). A hepta‐repeat region at 100 kb upstream of the transcription start site produces siRNAs from the paramutagenic allele B’ to trans‐act and induce RdDM in the homologous paramutable allele (B‐I), converting both alleles to a paramutated state (Hollick, 2017).

Notably, this type of trans‐acting siRNAs and/or DNA methylation is widespread in Arabidopsis intraspecific hybrids (Shen et al., 2012; Zhu et al., 2017), interspecific hybrids and allotetraploid Arabidopsis (Ha et al., 2009), and hybrid rice (Chen et al., 2010), tomato (Gouil & Baulcombe, 2018), and maize (Barber et al., 2012; Cao et al., 2022; Deans et al., 2024).

The hybridization‐induced paramutation via RdDM in maize represents a common epigenetic phenomenon subject to transgenerational inheritance (Cao & Chen, 2024). In tomato, paramutation‐like changes in the H06 locus in hybrids of Solanum lycopersicum and its relatives and cultivars depend on the timing of siRNA production and an RNA‐directed mechanism (Gouil & Baulcombe, 2018). In maize, the hybridization between the inbred parents B73 and Mo17 induces thousands of loci through trans‐acting methylation (TAM) and demethylation (TAdM) (Cao et al., 2022) (Figure 3A), also known as TCM and TCdM (Greaves et al., 2014), respectively. Among these loci, several hundreds (∼3%) are transmitted through six generations of backcrossing followed by three generations of selfing (Cao et al., 2022). Notably, these transgenerational methylation patterns resemble the epialleles of the nonrecurrent parent, despite over 99.9% of genomic loci being converted to the recurrent parent from a total of nine generations of backcrossing and self‐pollination.

FIGURE 3.

Hybridization‐induced trans‐acting epialleles in interspecific maize hybrids and cotton allotetraploids. (A) Elite line and wild relative (right panel) have different methylation patterns on the promoter region of a gene. These loci are initiated and maintained by siRNAs via RNA‐directed DNA methylation (RdDM) pathway. Hybridization between species triggers siRNAs that induce trans‐acting (chromosome) methylation (TAM or TCM) and demethylation (TAdM or TCdM) in the F₁ hybrids. Note that siRNAs and gene expression are depicted only in one strand. Some of these F₁ TAM and TAdM loci that acquired from the donor parent (wild relative) are heritable like paramutation, through multiple generations of backcrossing to the recurrent parent and self‐pollination. These transgenerational DMRs affect expression of adjacent genes (epigenes or epialleles) involved in different biological pathways including stress responses to regulate plant growth, development, and adaptation. Adopted and modified from Cao and Chen (2024) (B) Allotetraploid cotton (AADD) was formed between A‐genome species like Gossypium arboreum (Ga) and D‐genome species like G. raimondii (Gr), giving rise to five allotetraploid species, including wild G. hirsutum (wGh), wild G. barbadense (wGb), G. tomentosum (Gt), G. darwinii (Gd), and G. mustelinum (Gm). Wild Gh and Gb are domesticated into cultivated G. hirsutum (cGh) and G. barbadense (cGb), respectively. Images are adopted from Cao and Chen (2024). Divergence estimates were reported in Chen et al. (2020). Mya, million years ago.

This transgenerational inheritance requires trans‐acting factor(s) to initiate and maintain these epigenetic states, which can involve siRNAs and RdDM (Cao et al., 2022), as observed in paramutation (Alleman et al., 2006). The initiation of these epialleles depends on 24‐nt siRNAs, which are eliminated in the isogenic hybrid Mo17xB73:mop1‐1, defective in siRNA biogenesis (Alleman et al., 2006; Nobuta et al., 2008). Moreover, these siRNAs are involved in maintaining these loci during excessive backcrossing and selfing. The inheritance of TAM loci can be explained by the canonical RdDM pathway (Law & Jacobsen, 2010; Matzke & Mosher, 2014; H. Zhang et al., 2018). Inheritance of TAdM loci may be counterintuitive, probably through dilution or gradual loss of 24‐nt siRNAs during meiosis through backcrossing and selfing generations.

The hybridization‐induced paramutation also occurs in the interspecific maize hybrids between the modern maize W22 and teosinte (Z. may L. ssp. parviglumis) accessions Bravo (BR) or Blanco (BL) (Xue et al., 2019). In these interspecific hybrids, methylation levels in the backcrossing and selfing progeny resemble the teosinte parents, while the genome is largely converted to W22, which is correlated with 24‐nt siRNA levels (Cao et al., 2022). The TAM induced by the nonrecurrent parent teosinte can be maintained for at least six backcross generations. Thus, divergent siRNAs between the hybridizing parents can induce trans‐acting epialleles in hybrids, some of which are subject to transgenerational inheritance during backcrossing and hybrid breeding.

Interestingly, many paramutation genes are related to TE regulation of the genes, such as r1, bronze 2 (bz2), and b1 loci (Chandler, 2007; Hollick, 2017). These genes often activate anthocyanin biochemical pathways under the stress, leading to visible color phenotypes. Stress‐responsive genes are often associated with TEs and tend to show epigenetic regulation and inheritance in Arabidopsis (Ito et al., 2011; Sanchez & Paszkowski, 2014) and rice (Wang et al., 2021). These examples represent a general mechanism for siRNA‐induced methylation to initiate and maintain trans‐acting epigenetic patterns to regulate expression of stress‐responsive genes, which may contribute to phenotypic variation and adaptation in response to the changing environment and breeding manipulation.

Polyploidy, or whole genome duplication, is a unique evolutionary feature of many animals (Otto, 2007) and all flowering plants (Jiao et al., 2011). Autopolyploids result from duplication of own genomes, while allopolyploids are formed by interspecific hybridization following chromosome doubling and, alternatively, by fusion of unreduced gametes from distantly related species (Chen, 2007). Interspecific hybridization in the allopolyploids leads to fixation of heterozygosity and heterosis, which provides genetic bases for selection and adaptation in response to climate change and modern breeding. Most important crops, such as wheat, cotton, and canola, are polyploids.

Cotton is a crop model for the study of genome evolution and function in polyploid plants (Chen et al., 2007; Wendel & Grover, 2015). Polyploidization between an A‐genome African species (Gossypium arboreum (Ga)‐like) and a D‐genome American species (Gossypium raimondii (Gr)‐like) in the New World, which diverged 4.7–5.2 Mya, created a new allotetraploid or amphidiploid (AD‐genome) cotton clade (Wendel & Grover, 2015), approximately 1–1.6 Mya (Chen et al., 2020; Zhang et al., 2015), which has diversified into five polyploid lineages, Gossypium hirsutum (Gh) (AD)₁, Gossypium barbadense (Gb) (AD)₂, Gossypium tomentosum (Gt) (AD)₃, Gossypium mustelinum (Gm) (AD)₄, and Gossypium darwinii (Gd) (AD)₅ over 300,000–600,000 years (Chen et al., 2020). Gh and Gb were separately domesticated ∼8000 years ago from perennial shrubs to become annualized Upland and Pima cottons (Splitstoser et al., 2016). These materials are suitable to determine genetic and epigenetic variation accompanied by interspecific hybridization, evolution, domestication, and modern breeding (Cao & Chen, 2024; Song et al., 2017) (Figure 3B).

At the species level, phylogenetic trees constructed using the rate of CG and non‐CG methylation sites (Song et al., 2017) recapitulated the known evolutionary relationships of cotton species (Wendel & Cronn, 2003), including sister taxa relationships between Gh and Gt and between Gb and Gd. Among CG body‐methylated genes, the percentage of CG methylation changes (peaks at 0.18∼0.24) was substantially higher than the substitution rate of coding sequence (Ks (synonymous substitutions) value peaks at 0.007∼0.034), suggesting that the methylation change rate is (1–2 orders of magnitude) faster than the neutral sequence substitution rate and can be used to study genetic relationships within a recent and short evolutionary history (Song et al., 2017). This notion has been elaborately demonstrated as an epigenetic clock using the self‐pollinating A. thaliana and clonally propagating seagrass (Yao et al., 2023).

Compared to A. suecica (∼20,000 years) (Sall et al., 2003), hexaploid wheat (∼8000 years) (Feldman et al., 1997), tetraploid canola (∼7500 years) (Lu et al., 2019), and tetraploid Tragopogon (∼150 years) (Chester et al., 2012), the evolutionary history of the polyploid cotton clade is ancient (Chen et al., 2020; Zhang et al., 2015). These polyploid species can be compared with an interspecific hybrid that was formed between extant Ga and G. raimondii in 1940s and clonally propagated (Beasley, 1942). Remarkably, the differentially methylated regions (DMRs) between the interspecific hybrid and the extant progenitors Ga and Gr are shared 66%–96% and 17%–51% in at least one and all allotetraploid species tested, respectively (Song et al., 2017). This result indicates that a large portion of hybridization‐induced DNA methylation changes is conserved over 1 million years during polyploid evolution, diversification, and domestication. Although the exact progenitors of ancient allotetraploids are disputed, the data suggest that a wide range of hybridization‐induced DNA methylation changes can serve as epigenetic memory along with polyploid evolution. If methylation changes are stably inherited, the epigenetic modifications can program and reprogram expression of methylation‐associated genes, which are called epigenes or epialleles.

In two cultivated allotetraploid cottons, Gh and Gb, nearly a thousand DMRs are conserved between and associated with 519 epigenes or epialleles (Song et al., 2017). These genes are enriched in several important biological processes, including metabolic process, stress response, and domestication traits. These epialleles could contribute to a variety of morphological and physiological traits, including fiber length, reduction in seed dormancy, and photoperiod sensitivity, during domestication of Gh (Upland) and Gb (Pima) cotton. Loss of photoperiod sensitivity is a major “domestication syndrome” trait (Olsen & Wendel, 2013) controlled by an epigene in the allotetraploid cotton (Song et al., 2017). In Arabidopsis, CONSTAINS (CO) and CONSTAINS‐LIKE (COL) control photoperiodic flowering through regulation of FLOWERING LOCUS T (FT) expression via diurnal rhythms (Samach et al., 2000). There are eight groups of cotton (Gh) COLs (GhCOLs) (Zhang, et al., 2015). Group I GhCOLs are in the same clade with Arabidopsis CO, and only GhCOL2 is expressed rhythmically with GhFT, indicating that GhCOL2 is a major regulator of GhFT. Interestingly, the homoeolog (GhCOL2) is an epigene (Song et al., 2017). GhCOL2A is hypermethylated and silenced in all cotton allotetraploids, while GhCOL2D is also hypermethylated and silenced in wild cotton allotetraploids, but expressed in cultivated Upland and Pima cottons, coincident with loss of DNA methylation in all cultivated cottons tested. Reducing expression of GhCOL2 by virus‐induced gene silencing in Upland cotton results in a delay of flowering and square formation for 2 weeks. These data provide unique evidence for domestication‐induced loss of DNA methylation in the epiallele GhCOL2D, which could have reduced the photoperiod sensitivity from wild Gh and Gb species to cultivated Upland and Pima cottons. This change in photoperiodic sensitivity could contribute to worldwide cotton cultivation.

1.4. Future perspectives

Transgenerational inheritance of epigenetic variation provides the molecular basis for the rapidly evolving field of epigenetics, which is different from Lamarckian “inheritance of acquired characters” (Burkhardt, 2013) and Lysenko's pseudo‐inheritance (Kolchinsky et al., 2017); the latter has been frequently refuted and occasionally confused with epigenetics (Becker & Weigel, 2012; Heard & Martienssen, 2014). Compared to genetic variation, epigenetic variation can be transitory and reversible, and not all epigenetic variation can be transmitted through mitosis and meiosis (Becker et al., 2011). This is partly because epigenetic inheritance is based on DNA sequence modifications such as DNA methylation, RNA‐mediated processes, and/or chromatin modifications, which are largely metastable and reversible. Methylcytosine is considered to be the fifth base, which is common and largely heritable compared with other chromatin modifications (Lister & Ecker, 2009). It is critical to understand why some epialleles are stably transmitted through generations and during evolution, while others are metastable and often reversible.

The basis for the stability of epigenetic variation and durability of epigenetic traits remains unknown. For example, epialleles of disease‐ and stress‐tolerant traits may switch from laboratory to field conditions, which has hindered the progress of utilizing epigenetic variation for crop improvement. Notably, most internally induced epigenetic variation, such as paramutation, nucleolar dominance, and epi‐mutations produced during hybridization and polyploidization, are heritable, while stress‐ or externally induced epialleles are metastable and reversible. This suggests that internal and external factors may provoke different mechanisms of DNA methylation and/or other modifications. Future epigenetic research should elucidate (1) what kind of DNA methylation and/or chromatin modifications can generate stable information that is transmissible through meiosis; (2) what kind of specificity or quantitative threshold levels are required to establish stable inheritance patterns of an epigenetic variation, allele, and/or gene; (3) how internal genetic perturbations such as hybridization and polyploidy and external factors such as pests and environmental stresses can induce specific sets of epigenetic variation; and (4) why some epigenetic changes are transitory while others are stably inherited through generations.

Since many genes in response to domestication, selection, and adaptation to environmental cues in plants are subject to epigenetic modifications, further understanding the mechanisms of epigenetic inheritance and exploring utilization of epialleles will help us improve crop yield and resilience in agriculture. For example, one may use the genome‐editing technology with neotype CRISPR/dCas9 to methylate or demethylate (X. S. Liu et al., 2016; Vojta et al., 2016) specific genomic features to alter chromatin states of epialleles in animal cells for epigenetic engineering. In plants, removing DNA methylation in a specific locus by gene editing (Papikian et al., 2019) is more challenging than increasing DNA methylation via RdDM by gene editing (Veley et al., 2023) or virus‐induced gene silencing (Han et al., 2021; Song et al., 2017). Finally, there is growing evidence that the principles of DNA methylation and epigenetic inheritance, like the Mendelian laws of genetics and fundamentals of transposons (McClintock, 1984) and RNA interference (Fire et al., 1998) or RNA silencing (Hamilton & Baulcombe, 1999), are shared among plants and animals. Indeed, many human diseases, including cancers, have an epigenetic cause. The mechanisms learned from plant epigenetics and epigenetic inheritance should not only provide new breeding and biotechnological applications for improving crop resilience but also apply more broadly across sexually reproducing organisms, including humans, to improve public health and medicine.

AUTHOR CONTRIBUTIONS

Zengjian Jeffrey Chen: Conceptualization; formal analysis; funding acquisition; investigation; project administration; writing—review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Chen, Z. J. (2025). Empowering plant epigenetics to breed resilience of crops: From nucleolar dominance to transgenerational epigenetic inheritance. The Plant Genome, 18, e70064. 10.1002/tpg2.70064

DATA AVAILABILITY STATEMENT

No original datasets are associated with this review article.