Transgenerational Epigenetic Inheritance During Plant Domestication and Breeding

Abstract

Plants can program and reprogram their genomes to create genetic variation and epigenetic modifications, leading to phenotypic plasticity. While consequences of genetic changes are comprehensible, the basis for transgenerational inheritance of epigenetic variation remains elusive. This review addresses contributions of external (environmental) and internal (genomic) factors to initiation and establishment of epigenetic memory during plant evolution, crop domestication, and modern breeding. Dynamic and pervasive changes in DNA methylation and chromatin modifications provide diverse repertoires of epigenetic variation potentially for transgenerational inheritance. Elucidating and harnessing epigenetic inheritance will help us develop effective breeding and biotechnological strategies and tools to improve crop yield and resilience. Beyond plants, epigenetic principles are shared across sexually reproducing organisms with relevance to medicine and human health.

Epigenetics and epigenetic memory

The term epigenetics was initially used to describe all developmental processes from a fertilized zygote to the mature organism [1], such as canalization of development [2], which has more broad but less precise meaning compared to the contemporary definition [3]. Epigenetics now commonly refers to the study of heritable changes in gene function that cannot be explained by changes in primary DNA sequence [4]. An epigenetic phenomenon has three features, namely, non-Mendelian or above genetics, heritable, and reversible. Examples include genomic imprinting [5–7], X chromosome inactivation [8,9], position effect variegation [10], paramutation [11,12], nucleolar dominance [13,14], and RNA interference [15] or RNA silencing [16]. The heritable nature of these phenomena also refers to epigenetic memory, which defines the set of DNA and chromatin modifications that are inherited through mitosis and meiosis to different cell types and/or subsequent generations. Such modifications can alter gene expression and therefore the physiological traits.

Epigenetic modifications and/or epimutations can be established and altered in response to external (environmental) and internal (genomic) stresses and/or signals in plants and animals [17–22]. In animals, toxins and/or nutritional changes can alter epigenetic states for gene expression, leading to intergenerational variation (not necessary heritable) or transgenerational epigenetic inheritance [18]. Studies in Caenorhabditis elegans have shown that small interfering RNAs (siRNAs) are involved in neural gene expression and chemotaxis behavior in three generations [23] and in a long-term memory of avoidance to pathogens [24]. Although the mechanism for siRNA-induced inheritance in worms remains unclear, in plants siRNAs can induce gene silencing [25] and RNA-directed DNA methylation (RdDM) [26,27] to enforce epigenetic states [28]. The RdDM pathway is responsible for activation of transposable elements (TEs) associated with stress responsive genes in Arabidopsis [29,30] and rice [31,32] and during cotton ovule and fiber development [33]. In rice, stress-induced gene expression is the cause but not the consequence of RdDM [31,32].

Plants are essential in the ecosystem and provide food, feed, fuel, and materials for human civilization accompanied with extensive domestication of hundreds of plants for agricultural, horticultural, and medicinal purposes. 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. Hybridization within and between species occurs naturally in most plants, and plant breeders often use the gene pools in wild relatives to improve agronomic traits in elite crops. In plant interspecific hybrids, B. McClintock described the remarkable consequences of “genome shock,” leading to activation of TEs and other genomic features [34]. A series of studies have suggested these rapid and dynamic changes have an epigenetic basis [35–38]. In Arabidopsis interspecific hybrids or Arabidopsis suecica allotetraploids formed between Arabidopsis thaliana and Arabidopsis arenosa [39,40] (Figure 1), epigenetic modifications including DNA methylation and histone modifications are associated with silencing of selective sets of parental rRNA genes in nucleolar dominance [41,42], silencing of homoeologous protein-coding genes [43], rapid and stochastic changes in homoeolog-biased expression [44], genome-wide nonadditive gene expression, including sunbgenomic transcriptome dominance [39], and epistatic interactions between photoperiodic flowering pathway genes [45]. These changes in the hybrids and allopolyploids result in altered circadian rhythms and their output biological and metabolic pathways [46,47], leading to increased levels of photosynthesis, starch biosynthesis, and energy metabolism [46], a trade-off between growth and stress response [48], and reduced ethylene biosynthesis and increased biomass [49], all giving rise to a widespread phenomenon known as hybrid vigor or heterosis [37,50].

Figure 1. Epigenetic variation induced by hybridization and polyploidization during formation and evolution of Arabidopsis polyploids.

A. thaliana (At) and A. arenosa (Aa) diverged approximately 5–6 million years ago; the hybridization between At and Aa at 14,000 – 300,000 years ago generated natural A. suecica. In the laboratory, a tetraploid A. thaliana (At4) is pollinated with pollen of A. arenosa, an obligated outcrossing tetraploid, to generate F₁ allotetraploid. These newly formed allotetraploids were self-pollinated for 8 or more generations to produce A. suecica-like allotetraploids such as Allo733 and Allo738 lineages. The processes of interspecific hybridization and polyploidization triggered genome-wide epigenetic changes, some of which were maintained through transgenerational inheritance, as observed in both resynthesized and natural A. suecica.

In addition, small RNAs serve as a genetic buffer for TE-induced genome instability and nonadditive gene expression in A. suecica [51,52]. Stable inheritance of TE-associated siRNAs maintains genome stability [51], whereas expression variation of miRNAs leads to changes in gene expression, secondary metabolites, and growth vigor [53]. In the interspecific hybrid between A. thaliana and Arabidopsis lyrata, methylation levels are increased along with histone H3K27me3 modification and nonadditive gene expression [54]. Remarkably, many methylation changes induced in the F₁ resynthesized Arabidopsis suecica-like allotetraploids are shared with those in natural allotetraploid A. suecica [19]. Hybridization between A. thaliana ecotypes also leads to DNA methylation changes [55,56]. The hybridization-induced DNA methylation variation can be transmitted to the selfing progeny, leading to hybrid mimics [57]. Recent studies document that hybridization-induced DNA methylation changes are widespread in hybrid and polyploid crops, including rice [58,59], maize [21,60–62], tomato [63], cotton [20], and wheat [64,65]. Here we review establishment and maintenance mechanisms of epigenetic modifications and epimutations during plant evolution, crop domestication, and modern breeding. Understanding the basis of epigenetic memory and transgenerational inheritance will help us explore and utilize a repertoire of hidden genomic variation for improving crop yield and resilience using molecular breeding, genome editing, and epigenetic engineering.

DNA methylation and epigenetic memory

In plants and animals, DNA methylation is a common epigenetic modification that is stable and reversible and can be transmitted through meiosis [66–68]. DNA methylation is associated with many epigenetic phenomena including imprinting [6,7], X chromosome inactivation [8], and cancer [69]. In mammals most, if not all, DNA methylation occurs exclusively at the CpG islands. In contrast, DNA methylation in plants 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 [68,70]. In Arabidopsis, METHYLTRANSFERASE1 (MET1) is responsible for the maintenance of CG methylation [71], and asymmetrical CHG methylation is dependent on the plant-specific CHROMOMETHYLASE3 (CMT3) along with H3K9 histone methyltransferases [72]. CHH methylation is established de novo through the RdDM pathway [26,27], which targets the DOMAINS REARRANGED METHYLTRANSFERASE1 and 2 (DRM1 and DRM2) for asymmetrical cytosine methylation [73]. Alternatively and independently of RdDM, most asymmetric methylation is established by DECREASE IN DNA METHYLATION1 (DDM1), a chromatin remodeler responsible for the maintenance methylation of TEs and repetitive DNA [74,75], and mediated by the methyltransferase CHROMOMETHYLASE2 (CMT2) [76]. This pathway of CMT2-dependent and RdDM-independent CHH methylation is associated with expression of fiber-related genes during cotton fiber cell development [33].

DNA demethylation is not as well understood as methylation. Directly demethylating 5-methylcytosine is chemically challenging, requiring the disruption of the stable bond C-C [77]. However, active demethylation can occur via a DNA repair pathway. REPRESSOR OF SILENCING1 (ROS1) encodes a member of DNA glycosylase family and initiates a base excision repair pathway through the glycosylase activity to remove 5-methylcytosine in series of enzymatic reactions and replace it with unmethylated cytosine [78]. This process may result in activation of gene silencing and possibly global demethylation. Arabidopsis has three additional members of DNA glycosylase family, DEMETER (DME) [79] and DEMETER-LIKE2 and 3 (DML2 and DML3) [80]. These members exhibit the parent-of origin effect on demethylation of the maternal genome in the central cell prior to fertilization [79,80], which is maintained in the endosperm through a process known as imprinting [81,82]. DMEs can also target heterochromatin, like small TEs and TE fragments that flank coding genes, to induce RdDM and initiate imprinting [68].

Passive DNA demethylation occurs often by gradual dilution (50% per division) through progressive cell divisions [71] and possibly by dilution of trans-acting factors, such as siRNAs [56]. When trans-acting siRNAs in the hybrids are below the threshold for RdDM or absent in another parent, DNA methylation of the locus or allele will be decreased or removed, known as the trans chromosomal demethylation (TCdM) [83,84]. On the other hand, these siRNAs can increase DNA methylation levels in the opposite parental locus or allele via trans chromosomal methylation (TCM) [83]. The siRNA-induced CHH methylation exhibit a parent-of-origin effect [47]. Both TCdM and TCM epialleles are inheritable through 4–5 generations in Arabidopsis hybrids [83,84] and 9 generations in maize backcross and selfing populations [21]. In a previous report [21] and this review, we used trans-acting methylation (TAM) and demethylation (TAdM) interchangeable with TCM and TCdM, respectively, since trans-acting loci depend on siRNAs, which may or may not occur on a chromosomal scale or in a different parental chromosome.

During plant growth and development, the methylation and demethylation homoeostasis can be mediated through RdDM on methylation of a TE in the ROS1 gene promoter, which in turn negatively regulates ROS1 expression to prevent excessive demethylation [85]. Consequently, global methylation levels are dynamically balanced between methylation gain and loss in response to internal (genomic) and external (environmental) signals in plants and animals.

In mammals, DNA methylation is essential for reproductive and embryonic development, and even partial erasure of genomic methylation patterns is lethal to mouse embryos [86]. Plants can tolerate loss of DNA methylation and appear to grow and develop normally during early generations in the ddm1 mutant [74]. However, many epigenetic lesions or epimutations can spread in subsequent generations of plants [87], including the “ball” phenotype [88]; they are transmitted independently of the DDM1 mutation and involve siRNAs [89]. These findings suggest DNA methylation can induce additional epigenetic changes including epigenes or epialleles, which led to the investigation of epigenetic recombinant inbred lines (epiRILs) from the progeny derived from ddm1 mutants and wildtype [90]. Several traits are associated with epigenetic quantitative trait loci (epiQTL), which account for 60 to 90% of the heritability for two complex traits of flowering time and primary root length. These epimutations and traits are subject to selection and may consist of a large number of heritable variation for ecological and physiological traits, including root allocation, drought tolerance, and nutrient assimilation [91], which could increase a potential for evolution of plant phenotypic plasticity [92].

Epigenetic Changes Induced by Environmental and Genomic Stresses

Epigenetic memory can be programmed and reprogramed in response to genomic perturbation or “genome shock” [34] as well as biotic and abiotic stresses. Genome shock such as interspecific hybridization in wheat allotetraploids can induce activation of these antisense or sense transcripts of retrotransposons such as Wis 2–1A, which affects expression of the corresponding genes [93]. Interploidy crosses also induce genomic imbalance in the endosperm of rice [94] and A. thaliana, which depends maternal-derived siRNAs [95] and maternal genome dosage [96] to mediate spatiotemporal regulation of gene expression, imprinting, and seed development [97].

Some retrotransposons such as ONSEN are activated by heat stress in Arabidopsis [30]; new ONSEN insertions are observed in the progeny of stressed plants deficient in biogenesis of siRNAs, and transgenerational inheritance of ONSEN occurs during flower development and before gametogenesis. These observations suggest that genomic and heat stresses can prime or induce and reprogram the epigenetic memory.

The priming may involve DNA methylation [98,99]. DNA methylation variation is often enriched in TE-related genes and regulates their expression. After exposure of A. thaliana plants to a bacterial pathogen or salicylic acid (SA), numerous stress-induced differentially methylated regions (DMRs), many of which were associated with differentially expressed genes [98]. TE-associated DMRs coincide with 21-nt siRNA abundance and correspond to expression changes of the TEs and/or proximal genes. DNA methylation changes via RdDM in plant stress responses act as “priming,” which help plants remember what may occur under the similar stress at transcriptional, metabolic, and physiological levels and subsequently increase survival capacity in response to prolonged stress conditions [100] or after stress removal [32].

Other modifications such as H3K4me3 histone marks induced by the jasmonic acid treatment can also potentiate transcription of RESPONSE TO DESICCATION 29B (RD29B), an abscisic acid (ABA)-dependent gene, for priming dehydration stress in Arabidopsis [101]. Similarly, hybridization between A. thaliana ecotypes can increase histone acetylation in the promoters of salicylic acid pathway genes, priming more rapid and strong induction of defense-related genes than the parents upon pathogen attack [102]. In the DNA methylation triple mutant rdd (ros1/dml2/dml3) of A. thaliana, hypermethylated TEs compromise the transcription of the adjacent stress-related genes against Fusarium oxysporum, resulting in plants susceptible to the pathogen [103]. Herbivory induces methylation changes in targeted wild radish plants and their offspring [100]. Both chemical (glucosinolates) and physical (trichomes) defenses were highly inducible at the seedling stage, but only chemical defenses were inducible through reproduction, indicating transgenerational plasticity in plant defenses against herbivore attack. Notably, epigenetic changes induced by seed priming (preexposure of seed to mild stress) can be maintained, which may influence transgenerational inheritance of stress memory [104], to increase crop resiliency and yield under extreme environments.

Polyploidy or whole-genome duplication (WGD) occurs in all flowering plants and often enhances plants fitness in diverse and extreme environments [105]. For example, tetraploidy in Arabidopsis increases drought tolerance through altering expression of genes involved in ABA signaling and reactive oxygen species (ROS) homeostasis [106]. At the physiological level, several A. thaliana autotetraploids show increased salinity tolerance by absorbing potassium and reducing sodium uptake to elevate the K+/Na+ ratio [107]. In tetraploid rice, salt tolerance is enhanced through lower sodium uptake and correlates with epigenetic regulation of jasmonic acid (JA)-related genes [32]. Genome doubling or interspecific hybridization can be viewed as genomic stress or genome shock [34] that induces epigenetic changes including DNA methylation and histone modifications [35,108].

TEs and TE fragments in the rice genome tend to be enriched in the 5’ flanking sequences of the genes, adjacent to the stress-inducible genes [109]. Tetraploidy in rice induces genome-wide DNA hypomethylation and potentiates genomic loci coexistent with many stress-responsive genes, which are generally associated with proximal TEs [32]. As a result, the stress-responsive genes including those in the JA pathway are more rapidly induced to confer stress tolerance. After stress, elevated expression of stress-responsive genes in tetraploid rice induces hypermethylation and suppression of the TEs adjacent to stress-responsive genes. These induced responses are reproducible in a recurring round of salt stress and shared between two japonica tetraploid rice strains. The priming-resetting of DNA methylation in tetraploid rice may serve as a long-term epigenetic memory, which can facilitate offspring to enhance adaptation to environmental changes during evolution and domestication.

Transgenerational inheritance of epigenetic variation during plant evolution and breeding

Epigenetic changes may have both short- and long-term consequences for plant evolution, through affecting gene expression and morphology [35,108,110–112]. In Linaria vulgaris dorsoventral symmetry of flower development is controlled by Lcyc, a homologue of the cycloidea gene. The lcyc in the mutant strain (aka. peloric Linaria by C. Linnaeus) leads to asymmetrical flower development [113]. Further analysis of the Lcyc gene reveals no mutation in the protein-coding sequences, and the difference lies in the methylation of Lcyc promoter regions between the wild-type and mutant, which is methylated and not expressed in the latter. Thus, this epimutation is associated with the peloric inflorescence mutant phenotype with radially symmetrical flower, which is distinct from the bilaterally symmetrical flower in the wild-type.

Like genetic variation, epigenetic modifications and/or epimutations can occur spontaneously, widespread in natural and experimental populations [19,20,35,91,114]. The rate of spontaneous epiallele formation is reported to be similar to that of mutation rate in a previous study using two mutation accumulation lines of A. thaliana over 30 generations of self-pollination from a common ancestor [115]. This notion has been revised by a couple of recent studies [20,116], indicating that DNA methylation change rates are much higher than the DNA sequence mutation rate in cotton [20] and Arabidopsis [116]. Thus, the methylation change rate can be used as a “molecular clock” to measure population divergence in a relatively short evolutionary timescale such as domestication and breeding [116].

Consistent with heritable DNA methylation patterns during evolution, embryonic epigenetic reprogramming modulate chromatin status of FLOWERING LOCUS C (FLC) in response to vernalization [117]. Many other genes controlling flowering time are found to be epialleles such as FLOWERING WAGENINGEN (fwa), which is hypomethylated alleles found in the progeny of the ddm1 mutant [118] or through mutant screens [119]. Moreover, thousands of methylation quantitative trait loci (mQTLs) have been identified in Arabidopsis, 35% of which are associated with RdDM loci [120], probably through cis- and trans-acting effects of genetic variation. Estimates indicate that half of epimutations is associated with largely cis-acting effects, while the other half is related to trans-acting effects on methylation or demethylation [114]. Many of these epialleles are present in pollen and seeds, suggesting a developmental role in the inheritance of epialleles [121], as also observed in nucleolar dominance [122], an old epigenetic phenomenon with the silencing of rDNA loci from selective parental origins at a scale next to the epigenetic phenomenon X-chromosome inactivation [8,9].

Many “conventional” QTLs in crop plants are found to be associated with epialleles. These examples include DNA methylation on a SBP-box gene promoter as an epi-QTL of fruit ripening in tomato [123], DNA methylation on the CmWIP1 promoter as an epiallele for sex determination in melon [124], histone modification and DNA methylation on DWARF1 (D1) promoter as an epiallele for plant stature in rice [125], DNA methylation on a LINE retrotransposon (aka Karma) as an epiallele for mantled somaclonal variation in oil palm [126], and methylation QTLs associated with vitamin E accumulation in tomato [127]. These epialleles that affect transcription and function of the associated genes have played general roles in plant evolution, crop domestication, and molecular breeding.

Mechanisms for transgenerational inheritance of hybridization-induced epialleles

Allelic interactions in a hybrid (heterozygous state) can induce allelic expression changes in the offspring, as reported as paramutation in plants [11,12,128,129] and later in mice [18], and some paramutation-like events are related to parent-of-origin effects as observed in mice [130], flies [131], and worms [132]. Hybridization between parents can induce one allele to cause heritable changes in the expression of a homologous allele, such as r1, b1, pl1 alleles in maize [11,12]. Map-based cloning in the b1-derived population has identified the Mop1 (mediator of paramutation1) locus required for paramutation; it encodes a RNA-dependent RNA Polymerase2 (RDR2) [133] and is involved in the biogenesis of siRNAs [134]. The paramutation on the b1 locus is related to a hepta-repeat region at 100-kb upstream of the transcription start site (TSS), which produces siRNAs from the B’ allele presumably to trans-act and induce RdDM in the B-I allele (B’*), converting alleles to a paramutated state (B’/B*) [135]. There is evidence that this type of trans-acting siRNAs and/or DNA methylation can occur widely in Arabidopsis intraspecific hybrids [54,55], interspecific hybrids and allotetraploids [51], and rice [59], tomato [63], and maize [21,136] hybrids.

The paramutation induced by hybridization via RdDM represents a common epigenetic phenomenon that can be subject to transgenerational inheritance over nine generations of outcrossing and self-pollination [21] (Figure 2). These epigenetic changes are associated with introgression of genetic materials from one species or strain into another through hybridization and backcross breeding to improve yield and resilience of farm animals and agricultural crops [137,138]. Plant breeders focus on the events and materials that show the transfer of genes and genomic sequences into improved germplasm and breeding stocks. For example, using the maize relative teosinte and backcross breeding, many important traits or genes have been introgressed into different maize elite lines to improve agronomic traits [139], including yield [140], seed protein content and nitrogen use efficiency [141], nutrition and flowering time [142], and salt tolerance [143].

Figure 2. Hybridization-induced trans-acting epialleles in intraspecific or interspecific maize hybrids.

Elite lines A and B and their wild relative (left 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 within (intraspecific) and between (interspecific) species triggers siRNAs that induce trans-acting (chromosome) methylation (TAM or TCM) and demethylation (TAdM or TCdM) in the F₁ hybrids (middle panel). Note siRNAs and gene expression are depicted only in one strand. Some of these F₁ TAM and TAdM loci that acquired from the donor parent (elite line B or wild relative) are heritable like paramutation, through multiple backcrossing (six) to the recurrent parent and self-pollination (three, not shown). As a result, 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 Genome Biology [21, 226]

However, some hidden genetic and epigenetic variation may not be easily observed or traced. In addition to genetic recombination, hybridization can induce abundant epigenetic changes, including small RNAs, DNA methylation, and chromatin modifications [35,37,108,111,144]. In A. thaliana intraspecific hybrids, hybridization can induce both TCM and TCdM [83,84]. Like paramutation, these TCM and TCdM loci are associated with 24-nt siRNAs [83], some of which can be stably inherited in the selfing progeny, resulting in hybrid mimics [57].

The extent of hybridization-induced paramutation-like phenomenon or epigenetic memory has been recently examined in tomato and maize hybrids. 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 [63]. Moreover, thousands of candidate regions for paramutation-like loci have been identified, and the methylation patterns for a subset of loci segregate with non-Mendelian ratios, consistent with paramutation-like interactions. In maize, using single-base resolution DNA methylomes, Cao et al. [21] has addressed the questions how and when the methylation changes are induced and established during hybridization in reciprocal hybrids between two elite maize inbred lines (B73 and Mo17) and to what extent these changes are heritable during six generations of backcrossing and three generations of inbreeding (Figure 2). The hybridization between the inbred parents B73 and Mo17 induces thousands of loci through trans-acting hypermethylation (TAM) and hypomethylation (TAdM) [21], also known as TCM and TCdM [83], respectively. Among these loci, several hundreds (~3%) are transmitted through six generations of backcrossing followed by three generations of selfing [21]. Notably, these transgenerational methylation patterns resemble the epialleles of the nonrecurrent parent, despite over 99% of genomic loci are converted to the recurrent parent from a total of nine generations of backcrossing and self-pollination. In another study, the DNA methylation and associated gene expression changes during maize inbreeding can be partially restored by random mating, suggesting a reversible nature of epigenetic memory [61].

This transgenerational inheritance may require trans-acting factor(s) to initiate and maintain these epigenetic states, which can involve siRNAs and RdDM [21]. Indeed, the initiation of these epialleles depends on 24-nt siRNAs (Figure 2), which are eliminated in the isogenic hybrid Mo17xB73:mop1–1, defective in siRNA biogenesis [133,134]. Interestingly, these siRNAs are present in the transgenerational differentially methylated regions (tgDMRs) between the inbred parents in backcrossing-selfing lines, which resemble their respective nonrecurrent parents, suggests that 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 [28,68,70]. Inheritance of TAdM loci may be counterintuitive and result from hindrance of transcription by PolIV to generate 24-nt siRNAs. Alternatively, 24-nt siRNAs may dilute in tgDMRs during meiosis through backcrossing and selfing generations. Together, TAM and TAdM loci in the F₁ hybrids are established by the canonical RdDM pathway and can be maintained after excessive backcrossing and selfing.

This trans-acting methylation noted above occurs not only in the intraspecific hybrids but also in the interspecific maize hybrids that were formed between the modern maize W22 and teosinte (Z. may L. ssp. parviglumis) accessions Bravo (BR) or Blanco (BL) [62] and transmit to their backcross lines [21]. Both maize and its wild relatives (teosintes) are outcrossing under the natural environment, which are predicted to create modern maize through introgression of one [145] or two [146] teosintes with several regulatory genes [147], including teosinte branched1 (tb1) [148,149] and teosinte glume architecture1 (tga1) [150]. In these interspecific hybrids, methylation levels in most TAM and TAdM loci in the backcrossing and selfing progeny resemble the teosinte parents, while the genome is largely converted to W22, which is also correlated with 24-nt siRNA levels [21] (Figure 2). The trans-acting methylation 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, and at least some of the induced epigenetic patterns are maintained for transgenerational inheritance during backcrossing and hybrid breeding.

Because many of these transgenerational DMRs are associated with the genes involved in stress response [21] and domestication related traits [20], the hybridization-induced trans-acting epialleles may serve as a long-term epigenetic memory for adaptation, evolution, and domestication [129], as shown in rice [151]. Notably, many paramutation genes are related to TE regulation [133,152] of the genes such as r1, bronze 2 (bz2), and b1 loci [12,128] involved in anthocyanin regulatory and biosynthetic pathways. These genes often activate 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 [29,30] and rice [32]. These examples show a general mechanism for siRNAs to initiate and maintain trans-acting epigenetic patterns on a TE-associated DNA methylation in the promoter to regulate expression of stress-responsive genes. These stress-responsive genes may contribute to phenotypic variation and adaptation to environmental changes during breeding.

Transgenerational inheritance of epigenetic variation during polyploid evolution and crop domestication

Interspecific hybridization followed by chromosome doubling leads to formation of allopolyploids, with fixed heterozygosity and heterosis [35–37]. For example, A. suecica was formed from an polyploid event following hybridization between A. thaliana and A. arenosa [153] at 14,000–300,000 years ago [19,154]. This evolutionary type can be replayed in the laboratory to produce A. suecica-like allotetraploids through pollinating the autotetraploid A. thaliana with pollen of the tetraploid A. arenosa (Figure 1). The allotetraploid F₁ (AlloF₁) are subsequently self-pollinated for eight or more generations to generate genetically stable and tractable Arabidopsis allotetraploid (Allo733 and Allo738). A. suecica [39,155] and allotetraploid cotton [156,157] represent one group of allopolyploids that are genetically stable, whereas some other allopolyploids such as in Brassica napus [158,159] and Tragopogon miscellus [160,161] can undergo rapid genomic reshuffling. This paradox can be explained in part by epigenetic remodeling of the subgenomes in stable allopolyploids [19,35].

The overall methylation levels are higher in A. arenosa than in A. thaliana, which is related to natural epigenetic variation. In the resynthesized allotetraploids, high methylation levels of the A subgenome from A. arenosa are reduced immediately in the F₁ and continuously during self-pollination (F₁₀), while methylation levels in the T subgenome from A. thaliana are increased convergently to a similar level in natural A. suecica [19]. This changes in DNA methylation may be associated with expression changes of Repressor of Silencing1 (ROS1). ROS1 encodes a DNA glycosylase/AP lyase [78] and is responsible for demethylation and maintains methylation homeostasis through RdDM [85]. Both A. thalian and A. arenosa ROS1 genes are highly expressed in A. suecica, which may reduce overall methylation levels of A subgenome and maintain methylation homeostasis in natural A. suecica [19].

The result of convergent changes in the DNA methylation levels of the two subgenomes in A. suecica is consistent with finding in cotton interspecific hybrids between Gossypium arboreum and G. raimondii, where the low methylated D subgenome tends to increase methylation levels and reach equilibrium levels in cotton allotetraploids [20], which were formed 1–1.6 million years ago [156,162].

The dual processes of convergent and conserved epigenomic modifications may provide a basis for allotetraploids to stabilize the two subgenomes derived from divergent hybridizing species [44]. Newly formed A. suecica-like allotetraploids often lead to infertility or meiotic instability [40,163], which are related to two processes. One is silencing of self-incompatibility (S) locus from outcrossing A. arenosa in neo-allotetraploids and natural A. suecica [35,164,165]. S locus system comprises a combination of S-locus cysteine-rich (SCR) protein in pollen coat and S-locus receptor kinase (SRK) expressed on stigma surface [166]. Two SRK genes in A subgenome are weak alleles in the S locus dominance hierarchy than in T subgenome [167]. These weak alleles are immediately silenced by miRNA, leading to a loss of self-incompatibility in neo-allotetraploids and become nonfunctional in natural A. suecica. The other process may involve the genes related to mitosis and meiosis [168]. Coincidently, these convergent and conserved DNA methylation changes are related to the genes overrepresented in reproductive process, pollen development, seed production [19]. As examples, STRUCTURAL MAINTENANCE OF CHROMOSOMES3 (SMC3) is an essential gene for sister chromatid alignment and plant viability [169,170]. PHYTOENE DESATURASE5A (PDS5A) regulates mitotic sister chromatid cohesion [171], and AUXIN SIGNALING F-BOX3 (AFB3) is associated with pollen maturation and stamen development [172]. CG methylation levels of these genes were reduced from newly formed allotetraploids to A. suecica, and their expression levels are upregulated in natural A. suecica [19]. Meiotic instability is often associated with newly formed allotetraploids (F₁) and is gradually improved in resynthesized allotetraploids by self-pollination [44]. The dynamic DNA methylation changes in association of reproduction-related genes may serve as an epigenetic process to facilitate survival and adaptation of A. suecica allotetraploids in response to ever changing environments (Figure 1).

This hybridization-induced epigenetic remodeling is a general phenomenon among many allopolyploids including most important crops such as cotton, canola, and wheat. Compared to A. suecica (~20,000 years) [173], hexaploid wheat (~8,000 years) [174], tetraploid canola (~7,500 years) [175], and tetraploid Tragopogon (~150 years) [176], the evolutionary history of the polyploid cotton clade is ancient [156,162]. Polyploidization between an A-genome African species (Gossypium arboreum-like) and a D-genome American species (G. raimondii-like) in the New World, which diverged 4.7–5.2 Mya, created a new allotetraploid or amphidiploid (AD-genome) cotton clade [177], approximately 1–1.6 Mya [156,162], which has diversified into five polyploid lineages, Gossypium hirsutum (Gh) (AD)₁, G. barbadense (Gb) (AD)₂, G. tomentosum (Gt) (AD)₃, G. mustelinum (Gm) (AD)₄, and G. darwinii (Gd) (AD)₅ over 300,000 to 600,000 years. Gh and Gb were separately domesticated ~8,000 years ago from perennial shrubs to become annualized Upland and Pima cottons [178]. Moreover, an interspecific hybrid was formed between extant G. arboreum and G. raimondi in 1940’s and clonally propagated [179]. These materials are suitable to determine genetic and epigenetic variation accompanied by interspecific hybridization, evolution, domestication, and modern breeding [20] (Figure 3).

Figure 3. Polyploidy-induced changes during polyploid plant evolution and crop domestication.

(a) Allotetraploid cotton (AADD) was formed between A-genome species like G. 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 Genome Biology (Song et al. 2017). Divergence estimates were reported in Nature Genetics (Chen et al. 2020). Ma: million years ago. (b) Diagram of polyploid formation and evolution (for simplicity using one pair of chromosomes to depict each species). Diploid progenitors can evolve and adapt to their original environments, while newly formed polyploids can diversify into different species and some of these species can be selected and domesticated into crops. Different colors of chromosomes represent origin from different species. Over time sequence changes (bars/boxes) can occur, and epigenetic modifications (puffed chromosomal regions) can induce epigenes that can be inherited during selection and domestication. For epigene, grey dots, short lines, and green arrow indicate DNA methylation and small interfering RNAs in the promoter and gene expression, respectively. Panels (a), except for flowers (curtsey of Atsumi Ando and David M. Stelly), and (b) are adopted from Advances in Agronomy [111].

At the species level, phylogenetic trees constructed using CG and non-CG methylation sites recapitulated the known evolutionary relationships of cotton species [20,180], including sister taxa relationships between G. hirsutum and G. tomentosum and between G. barbadense and G. darwinii. 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 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 [20]. This notion has been elaborately demonstrated as an epigenetic clock using self-pollinating A. thaliana and clonal-propagating seagrass [116].

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 [20]. This result indicates that a large portion of hybridization-induced DNA methylation changes is conserved over one million years during polyploid evolution, diversification, and domestication. Although the exact progenitors of ancient allotetraploids are dispute, 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 expression of methylation-associated genes, which are called epigenes or epialleles.

In two cultivated allotetraploid cotton G. hirsutum and G. barbadense, nearly a thousand DMRs are conserved between and associated with 519 epigenes or epialleles [20]. These genes are enriched in several important biological processes, including metabolic process, stress response, and domestication traits such as regulation of seed dormancy. These epialleles could contribute to common morphological and physiological changes, including fiber length, reduction in seed dormancy, and photoperiod sensitivity, during domestication of G. hirsutum (Upland) and G. barbadense (Pima) cotton. Loss of photoperiod sensitivity is a major “domestication syndrome” trait [181] is controlled by an epigene in allotetraploid cotton [20]. In Arabidopsis, CONSTAINS (CO) can control photoperiodic flowering through the induction of FLOWERING LOCUS T (FT) and regulates FT expression via diurnal rhythms [182]. The cotton CONSTAINS-LIKE2 homoeolog (COL2) is an epigene [20]. COL2A is hyper-methylated and silenced in all cotton allotetraploids, while COL2D 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 COL2 by virus-induced gene silencing in Upland cotton results in delayed flowering and square formation for two weeks. These data indicate domestication-induced loss of DNA methylation in the epiallele COL2D, which changes the photoperiod sensitive wild G. hirsutum and G. barbadense species to photoperiod insensitive cultivated Upland and Pima cottons, which could contribute to worldwide cotton cultivation.

Bread wheat (Triticum aestivum, 2n = 6x = 42, AABBDD) is complex hexaploid, resulting from two rounds of interspecific hybridization and polyploidization, which is believed to have occurred naturally ~ 8000–10,000 years ago by crossing between tetraploid wheat (Triticum turgidum, AABB) and a goatgrass species (Aegilops tauschii, DD), accompanied with genome doubling [183]. Interestingly, the tetraploid wheat (ETW, AABB) could be extracted by hybridization between a natural hexaploid wheat (NHW, AABBDD) and a natural tetraploid wheat (NTW, AABB), and subsequently nine generations of backcrossing using NHW as the recurrent parent, followed by three generations of self-pollination to exclude D chromosome and genotype stabilization [184,185]. Although chromosomes pair and segregate normally in the extracted tetraploid wheat, it displays severe developmental abnormalities. These abnormal phenotypes such as dwarfed plants, compacted spikes, shriveled seeds, decreased starch content, smaller starch granule, and reduced fertility are restored in the resynthesized allohexaploid wheat (RHW) formed by hybridization between ETW and a D-genome diploid (Ae. tauschii) [64,185].

The phenotypic restoration is at least in part associated with concerted changes in DNA methylation changes[64]. Overall levels of DNA methylation, especially in CHG (H=A, T, or C) context, are dramatically decreased in the ETW relative to natural hexaploid wheat and lead to many hypo differentially methylated regions (DMRs). These hypo DMRs in ETW were remethylated in the resynthesized hexaploid wheat (RHW) after the addition of the D genome, and highly correlated with expression levels of DMR-associated genes. For example, both A and B homoeologs of TaDOS, a rice homolog OsDOS gene regulating spike development and fertility [186], are demethylated and highly expressed in the ETW, suggesting a potential role for DNA methylation changes in TaDOS expression and seed phenotypes during subgenome separation and merger. Hypo DMRs in ETW are also associated with reduced levels of H3K9me2 marks and increased levels of histone variant gene expression, suggesting concerted epigenetic changes after separation from the hexaploid. These results indicate epigenetic memory induced by hybridization and polyploidization can be established, erased or reversed, and re-established during evolution and domestication.

Transgenerational inheritance of epigenetic variation during rice domestication and de-domestication

Rice (Oryza sativa L.) is an important cereal crop that feeds nearly one-third of the global population. The domestication process of rice is debatable. It is believed that domestication occurred from the wild rice (O. rufipogon L.) approximately 9,000 years ago in Southern China [187]. One genomic study suggests that an ancient japonica rice was domesticated and then crossed to local landraces in Southeast and South Asia to generate indica rice [188]. However, the genetic diversity study indicates indica and japonica subspecies evolve independently [189] or from indica rice with diversification into japonica rice [187]. Rice domestication has led to dramatic changes in morphological and physiological traits, including loss of awn, pericarp and hull color, seed shattering, and increased grain size [189–192]. The japonica or “sticky rice” is largely distributed in high-latitude Northeast Asia and upland areas of South Asia, while the indica or “non-sticky rice” is cultivated in lowlands of tropical Asia. Aus, Boro and Rayada, as well as Basmati and Sadri aromatic (Aro) ecotypes are more geographically restricted from Bangladesh, Pakistan, and India [189,193,194].

However, weedy plants can occur spontaneously in rice cultivation fields and revert back into self-sustainable “wild-like” plants under natural selection, a process called de-domestication [191,195]. The de-domesticated rice (Oryza sativa L. f. spontanea) is a pernicious weed in rice fields and resembles cultivated rice morphologically but lacks its agronomic traits and values, such as seed shattering and strong seed dormancy [192,196]. The weedy rice can cross-fertilize with cultivated rice and cause severe yield and economic loss. There is growing evidence to suggest that de-domestication is not a simple reversal of domestication [195]; All three weed strains including two from the US and one from China examined had evolved after rice domestication and after differentiation within the cultivated rice crop [191,195].

The paradox is that de-domesticated and cultivated rice have shared genomic features, morphological and adaptative traits, in spite of convergent evolution among critical genes of different weedy types [191,195]. This suggests an epigenetic basis for de-domestication. A recent study has analyzed single-base resolution DNA methylomes from 20 wild rice (Oryza rufipogon), 18 cultivated japonica (CJ, Oryza sativa), 20 cultivated indica (CI, Oryza sativa), 9 aromatic (Aro), 10 Aus (Aus), 10 weedy japonica-like (WJ, Oryza sativa f. spontanea), and 8 weedy indica-like (WI, Oryza sativa f. spontanea) rice [197]. Compared to wild rice, DNA methylation levels have decreased in all lines of cultivated indica and japonica rice examined during domestication (Figure 4). Most DMRs induced by domestication in the cultivated CJ and CI rice groups are associated with stress-responsive genes, which may serve as an epigenetic memory to remember the domestication syndrome [191,198,199] and/or adapt to changing environments [32,200]. Unmethylated or low methylated regions with the length of commonly ~1 kb or longer are known as DNA methylation valleys (DMVs). There are 29,585 CG DMVs identified in the wild and four cultivated rice, including CJ, CI, Aro, and Aus, which span over 13% of the rice genome [197]. Interestingly, 171 of conserved DMVs in four cultivated rice groups showed extensive DNA methylation (mCG > 0.4) in the wild rice, which may have contributed to gene expression changes during rice domestication (see below). Consistently, DMVs are involved in cell differentiation and cancer pathways in mammals [201] and associated with transcriptional regulation during seed development in soybean (Glycine max L.) [202].

Figure 4. Epigenetic variation involved during rice domestication and de-domestication.

Cultivated rice was domesticated from wild rice ~6,000 years with changes in grain color from dark to light, seed shattering from strong to weak, and fertilization from outcrossing to self-pollination (left panel). A stretch of DNA sequence showing cytosine methylation (mC) and unmethylated (C) (2^nd from the left). Genome-wide DNA methylation levels were decreased during domestication and increased during de-domestication(3^rd from the left). Consequently, some epialleles such as the promoter of ORF2 were methylated in wild rice and demethylated in cultivated rice (right panel, from top to middle), excluding and recruiting, respectively, binding of transcription factor(s) (TFs). As a result, the gene (ORF2) is upregulated in the cultivated rice (middle) to breakdown self-incompatibility from wild rice (top). During de-domestication, weedy rice regained some traits of wild rice (bottom), which was accompanied by increased methylation levels compared to cultivated rice (middle). However, most methylation changes are not shared by domestication and de-domestication processes, despite they act on a common set of stress-responsive genes, suggesting that de-domestication is not a simple reversal of domestication. An example is hypomethylation and increased chromatin accessibility of OsPR5 in the wild japonica rice, leading to increased binding affinity of ERF3 to the promoter region of rap2.6 gene. Two panels (from left) were modified from a thumb icon published in [197].

A key trait for rice domestication is conversion from outcrossing in wild rice to self-pollination in the cultivated rice [203,204]. This reproductive isolation in domesticated rice is related to a transposable element associated with OPEN READING FRAME2 (ORF2) and ORF3 [205]. ORF2 encodes a toxin inhibiting pollen development, whereas ORF3 encodes the antidote for pollen viability. The ORF2 is a gain-of-function protein, probably by artificial selection of a nonfunctional allele from wild rice during domestication. This domestication-induced gain-of-function is associated with a DMV spanning the promoter of ORF2 and the neighboring gene ORF1 in all cultivated rice tested, whereas this DMV is heavily methylated in wild rice [197] (Figure 4). As a result, ORF1 and ORF2 are expressed at high levels in cultivated accessions, including weedy rice, but at very low levels among wild rice accessions. This result suggests natural and artificial selection during domestication can induce heritable epigenetic changes, leading to gain- or loss-of-function of the genes related to rice growth and development.

In addition to shared DMRs between domesticated japonica and indica rice, CJ and CI groups possess abundant group-specific DMRs, supporting the model that the CJ and CI rice evolved independently after a shared domestication event [189,194]. These specific DMRs contribute to morphological diversity among rice groups. For example, CJ rice has wider leaves and shorter grains than CI rice, which is controlled by several genes including SHORT GRAIN1 (SG1) [206]. A DMR is located in the promoter region of SG1 in both CJ and Aro groups, which often have shorter grains and wider leaves [197]. By contract, this region is heavily methylated in both CI and Aus groups with longer grains and narrower leaves. The hypermethylation and hypomethylation patterns correlate with silencing of SG1 in most CI and Aus accessions and upregulation of SG1 in most CJ and Aro accessions, respectively. These results indicate epigenetic regulation of a heritable morphological trait that has diverged between japonica and indica rice subspecies.

Compared to overall reduced DNA methylation during domestication, DNA methylation levels are dramatically increased in the weedy rice during rice de-domestication [197] (Figure 4). This may result in the loss of physiological and agronomic traits, which is often accompanied with de-domestication [191,195,207]. Most DMR changes from domestication to de-domestication are associated with banding site motifs of transcription factors such as ethylene responsive factors (ERFs), affecting expression of stress-related genes such as OsPR5 in response to natural and rice cultivation environment [197].

DNA methylation is by far the most well-documented epigenetic variation, and most changes in DNA methylation go hand-in-hand with changes in chromatin accessibility (ATAC-seq), chromatin loops (Chromatin Interaction Analysis with Paired-End-Tag or ChIA-PET and Chromatin Immunoprecipitation-sequencing approaches) [208], histone modifications (enhanced ChIP-seq) [209], and 3D chromosomal conformation (Hi-C seq) [210]. Most variation in DNA methylation together with changes in chromatin structures and histone modifications regulate expression of both adjacent genes locally and distant genes via long-range chromatin loops, such as hypo-DMR on chromosome 5 in association of elevated expression levels of the neighboring gene OsNDPK1 and the distal gene OsCAX1b, located ~70 kb downstream of the DMR [197]. Transcriptional regulation by chromatin loops has been well documented in mammals [211], including spatial regulation of CTCF factors for the beta-globin locus [212]. New sequencing technologies have enabled findings of long-range chromatin interactions in rice [208] and maize [213,214]. Specifically, two DMRs are associated with the domestication-related loci vegetative to generative transition1 (vgt1) [215] and teosinte-branched1 (tb1) [149], respectively [213]. They regulate expression of downstream genes controlling flowering time [215] and apical dominance [149] through interactive chromatin loops in modern maize lines [148,213], but these loops are absent in the wild maize ancestor teosinte. These studies suggest that DNA methylation can establish epigenetic memory and inheritance through formation of complex chromatin loops during plant evolution and crop domestication.

Concluding remarks and future perspectives

Epigenetic memory and transgenerational inheritance of epigenetic variation are different from Lamarckian “inheritance of acquired characters” [216] and Lysenko’s pseudo-inheritance [217], despite they are frequently refuted and occasionally confused with the emerging field of epigenetics [17,218]. Compared to genetic variation, epigenetic variation can be transitory and reversible, and not all epigenetic variation can be transmitted through mitosis and meiosis [219]. 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. It is critical to understand why some epialleles are stably transmitted through generations and during evolution, while others are metastable and often reversible.

While most current studies are focused on mechanisms for DNA methylation and demethylation, chromatin modifications, programming, and reprogramming, the bases for the stability of epigenetic variation and durability of epigenetic traits remain 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, epimutations produced during hybrid and polyploid formation 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 (see also outstanding questions).

Outstanding questions.

Future research in transgenerational inheritance of epigenetic variation should elucidate:

1. why and 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

2. what kind of DNA methylation and/or chromatin modifications can generate stable information that is transmissible through meiosis

3. why some epigenetic changes are transitory while others are stably inherited through generations

4. what kind of specificity or quantitative threshold levels are required to establish stable inheritance patterns of an epigenetic variation, epiallele, or epigene

5. how can we develop genome-editing methods for effective and targeted modifications of the epigenetic variation to improve agronomic traits and crop resilience

With a better understanding of stable epigenetic modifications such as DNA methylation and chromatin modification, one may use the genome-editing technology with neotype CRISPR/dCas9 to methylate or demethylate [220–222], acetylate or deacetylate [223,224] specific genomic features to alter chromatin states of epialleles in animals, which will open new venues for epigenetic engineering. In plants, removing DNA methylation in a specific locus by gene-editing [220] is more challenging than increasing DNA methylation via RdDM by gene-editing [225] or virus-induced gene silencing [61]. Many genes in response to domestication, selection, and adaptation to environmental cues in plants are subject to epigenetic modifications, therefore further understanding the mechanisms of epigenetic inheritance and exploring utilization of epialleles will help us improve crop yield and resilience in agriculture. Finally, plants provide a powerful model for the study of genetics and epigenetics and leads the field in many areas. There is growing evidence that the principles of DNA methylation and epigenetic inheritance like the Mendelian laws of genetics and fundamentals of transposons [34] and RNA interference [15,25] 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 apply broadly across sexually reproducing organisms including humans to improve public health and medicine.

Highlights.

An epigenetic phenomenon has three features: not a DNA mutation, heritable, and reversible.

Epigenetic memory is related to DNA methylation, chromatin modification, and/or RNA-mediated mechanisms.

Both external (environmental) and internal (genomic) stresses and signals can induce heritable epigenetic variation.

Most internally-induced epigenetic variation are relatively stable and can be explored to improve crop yield and resilience.

Beyond plants, epigenetic principles are shared across sexually reproducing organisms with relevance to medicine and human health.

Glossary

Epiallele or Epigene

A gene or allele whose expression is controlled by DNA methylation or chromatin modification, without genetic mutation.

Epigenetics

The study of traits and gene expression changes that are independent of primary DNA sequences but dependent on DNA and chromatin modifications. Epigenetic traits are heritable and reversible.

Epigenetic memory

The set of DNA and chromatin modifications that are inherited through mitosis and meiosis to different cell types and/or subsequent generations in plants and animals.

Genome shock

The term first coined by Barbara McClintock to describe rapid genomic changes largely associated with activation of transposable elements in interspecific hybrids. The phenomenon is commonly defined as epigenetic modifications and (nonadditive) gene expression changes due to intergenomic interactions in interspecific hybrids and allopolyploids.

Heterosis or hybrid vigour

Growth and fitness levels are superior in the offspring compared to those in either or both parents.

Homoeolog

Genes or genomic loci with an origin from different species are present in the same allopolyploid species.

Inheritance of Acquired Characters

The theory proposed by Lamarck that if an organism changes during life in order to adapt to its environment, those changes are passed on to its offspring. The theory does not provide any basis or experimental evidence for inheritance.

Nonadditive gene expression

The expression level of a gene in a hybrid or polyploid is unequal to the sum of two parental alleles (aka, 1 + 1 ≠ 2), suggesting activation (> 2) and gene repression (< 2). There is an epigenetic basis for nonadditive gene expression.

Nucleolar dominance

An old epigenetic phenomenon that describes silencing or activation of one parental set of rRNA genes in an interspecific hybrid or allopolyploid of plants and animals.

Paramutation

An epigenetic phenomenon that was discovered in maize and sorghum where one allele influences expression of the other allele at the same locus when two alleles are combined in a hybrid (heterozygote). The first allele is defined as paramutagenic, and the second allele as paramutable. Paramutated alleles display transgenerational inheritance and can be metastable.

Polyploidy

A situation where the number of chromosome sets is greater than two. A polyploid refers to a cell or organism having more than two sets of chromosomes. Polyploids can be classified into autopolyploids, allopolyploids, and paleopolyploids. An autopolyploid originates by the multiplication of one basic set of chromosomes, while an allopolyploid results from combination of distinct sets of chromosomes often through interspecific hybridization followed by chromosome doubling or fusion of unreduced gametes.

Pseudo-inheritance

A pseudo-science and political ideology proposed by Lysenko that rejects natural selection and Mendelian genetics; it was proposed to increase wheat crop yield by cold and humidity, a process known as vernalization, and further to change durum (tetraploid) wheat to bread (hexaploid) wheat. The political movement of Lysenkoism was a tragedy for science and society in former Soviet Union countries and their affiliated institutions.

Transgenerational epigenetic inheritance

A phenomenon of non-genetic patterns that are transmitted from one generation to another and in some instances from grandparents to grandchildren.

Vernalization

The process that plants require a period of cold temperature for seed germination and flowering.

X chromosome inactivation

One of the two X chromosomes in mammalian females is repressed in somatic cells as a mechanism of dosage compensation (compared to a single X chromosome in males). The inactivated X chromosome is reversible during reproductive stage in the female eggs. This epigenetic phenomenon is stable and heritable from mothers to daughters.