Synopsis
Epigenetic mechanisms such as DNA methylation modulate gene expression in a complex fashion are consequently recognized as among the most important contributors to phenotypic variation in natural populations of plants, animals, and microorganisms. Interactions between genetics and epigenetics are multifaceted and epigenetic variation stands at the crossroad between genetic and environmental variance, which make these mechanisms prominent in the processes of adaptive evolution. DNA methylation patterns depend on the genotype and can be reshaped by environmental conditions, while transgenerational epigenetic inheritance has been reported in various species. On the other hand, DNA methylation can influence the genetic mutation rate and directly affect the evolutionary potential of a population. The origin of epigenetic variance can be attributed to genetic, environmental, or stochastic factors. Generally less investigated than the first two components, variation lacking any predictable order is nevertheless present in natural populations and stochastic epigenetic variation, also referred to spontaneous epimutations, can sustain phenotypic diversity. Here, potential sources of such stochastic epigenetic variability in animals are explored, with a focus on DNA methylation. To this day, quantifying the importance of stochasticity in epigenetic variability remains a challenge. However, comparisons between the mutation and the epimutation rates showed a high level of the latter, suggesting a significant role of spontaneous epimutations in adaptation. The implications of stochastic epigenetic variability are multifold: by affecting development and subsequently phenotype, random changes in epigenetic marks may provide additional phenotypic diversity, which can help natural populations when facing fluctuating environments. In isogenic lineages and asexually reproducing organisms, poor or absent genetic diversity can hence be tolerated. Further implication of stochastic epigenetic variability in adaptation is found in bottlenecked invasive species populations and populations using a bet-hedging strategy.
Introduction
The issue of chance and randomness is one that transcends disciplines and raises many questions: to what extent can one claim an event is due to stochastic processes? In this context, stochasticity is used to describe undetermined processes that “cannot be predicted a priori from readily measurable variables” (Honegger and de Bivort 2018). As such, one can argue that this is merely a term we use to describe processes that we do not yet truly understand, or variability that our current models have yet to explain. In laboratory experiments, despite efforts to maintain animals in homogeneous and stable conditions, individuals may experience different microenvironments such as slight differences in temperature, light, olfactory signals, or distribution of food (Bierbach et al. 2017), which might induce unexplained variability in the responses. In evolutionary biology, it is broadly admitted that genetic mutations are mostly random. They occur independently of their phenotypic consequences, and do not occur more frequently when they are advantageous (Lenski and Mittler 1993). However, if evolution is conveniently qualified as a stochastic process, the question of chance is still under debate. The reason mainly lies on whether evolution is predictable or repeatable, which can be studied by replicated evolution in similar environments (Lenormand et al. 2009). Consequently, Lenormand et al. (2009) grouped stochastic effects on evolution into three main categories: stochasticity of genes’ mutation rate (genetic level), of life histories of individuals (individual level), and of environmental fluctuations (population level). Stochasticity is receiving increasing attention in a wide range of domains, including developmental biology, evolutionary ecology, and viral infection, because of its potential selective advantages (Feinberg and Irizarry 2010; Branciamore et al. 2014; Vogt 2015; Zhao et al. 2019).
Besides the classical view of random genetic mutation, stochasticity may find a form in epigenetics, more precisely in the emergence of random epigenetic variability in marks and mechanisms (Jeltsch and Jurkowska 2014). As described by Jablonka (2017), epigenetics initially referred to “the branch of biology that studies the causal interactions between genes and their products which bring the phenotype into being.” Today, the generally accepted definition of epigenetics is the study of changes in gene function that are mitotically and/or meiotically heritable and that do not entail a change in DNA sequence (Dupont et al. 2009). It includes three main types of mechanisms: DNA methylation, histone modifications, and noncoding RNA (ncRNA). DNA methylation is found across all taxa of life (Lee et al. 2010; Zemach et al. 2010) and primarily occurs at 5-methylcytosine (5mC) bases in eukaryotes and prokaryotes. In vertebrates and plants, methylation is mainly restricted to symmetrical CG dinucleotides (CpG sites), although there is evidence that asymmetrical non-CpG methylation is also correlated to gene expression (Monti et al. 2020). CpG islands (CGIs) are specific regions in the genome containing high densities of mostly unmethylated CpG sequences, predominantly associated with promoter regions (Saxonov et al. 2006). Different methylation patterns are described across taxa, from plants and animals showing global methylation to some species displaying little to none at all (Suzuki and Bird 2008; Lee et al. 2010; Zemach et al. 2010). Within the animal kingdom, 5mC shows more diversity and complexity than previously anticipated, and the idea that genome hypermethylation is exclusive to vertebrates has recently been challenged (de Mendoza et al. 2020; Schmitz et al. 2019). It is broadly admitted that DNA methylation of regulatory regions is associated with gene down-regulation or silencing. However, we now recognize that this relation is far from being systematic and that gene body methylation is the most conserved genomic region for 5mC in animals, positively correlated with transcriptional activity in most species (de Mendoza et al. 2020). The goal of the present review is to discuss the mechanisms sustaining the stochastic component of epigenetic variability in animal populations, and their possible implications for adaptive evolution. We will focus on DNA methylation, as most of the current knowledge about stochastic epigenetic variation relates to this mechanism.
Epigenetics and phenotypic variation
Epigenetic variance can interact with the terms of this equation to significantly affect the phenotype. As its definition states, epigenetic mechanisms regulate gene expression in a complex fashion. The phenotype being the results of gene expression, epigenetic mechanisms can affect the phenotype via the fine-tune regulation of gene expression (Burggren 2014). Nowadays, epigenetic variance is even recognized as one of the most important contributors to phenotypic variation within a population (Peaston and Whitelaw 2006; Allis and Jenuwein 2016) (Fig. 1).
Origin of epigenetic variation and its interaction with genetic, environmental, and phenotypic variation. Epigenetic variation can be generated by environmental triggers (environmental epimutation), by the genotype (obligatory or facilitated epimutation), or can be the result of stochasticity (spontaneous epimutation). This epigenetic variation can then induce genetic variation by modifying the mutation rate, which can then affect the phenotype. It can also change gene expression patterns which directly results in a modified phenotype. Note that epimutations can be inherited through generations or not.
Epigenetic variation can affect VG by creating DNA instability and promoting genetic mutations: the mutation rate of DNA sequences can be modified by epimutations (i.e., epigenetic modifications), and by doing so, alter the dynamics of genome evolution (Vogt 2015; Skinner 2015; O’Dea et al. 2016). For example, the methylation of cytosines in CGIs strongly facilitates their transition to a thymine (Danchin et al. 2019). As a result, CGIs show particularly high mutation rates (Hodgkinson and Eyre-Walker 2011; Makova and Hardison 2015). Epigenetic modifications, and notably if they are environmentally induced, can therefore increase the mutation rate, enhancing the evolutionary potential of a population (O’Dea et al. 2016; Verhoeven et al. 2016).
It is important to keep in mind that epigenetic modifications themselves can depend on nucleotide sequences. Three classes of epimutations can be considered according to their dependency on DNA sequences (Richards 2006): first, obligatory epigenetic variation concerns epigenetic modifications that are completely associated with nucleotide sequence. For example, Janowitz Koch et al. (2016) showed co-occurrence of methylation patterns with transposable elements. Second, facilitated epigenetic variation refers to cases in which the genotype potentiates the epigenetic pattern in a probabilistic manner, that is, mutations in an epigenetic regulator that would facilitate the formation and maintenance of an epigenetic variant in certain conditions (Baluska et al. 2018). Third, pure epimutations are independent of the DNA sequences. These mechanisms include stochastic errors in replicating methylation patterns as described below (Richards 2006). Summarized in other words, the genome influences the epigenome, and the epigenome influences the genome.
What is really distinctive about epigenetic regulation, however, is that it is at the interface between genetic control and the environment, as epigenetic marks can be modified by environmental conditions (referred to as environmental epimutations). The environment can, therefore, induce changes in gene expression as well as genetic variance through the modification of the epigenome, which reflects the importance of epigenetic variation in determining VE and VGxE.
Regarding Ve, several molecular mechanisms can be involved in stochastic variance, such as a few copies of a transcription factor present in the nucleus, a positive feedback in gene network leading to bistability, and epigenetic variance (see details below) (Honegger and de Bivort 2018). In addition to its possible contribution to VG, VE, and VGxE, epigenetics represent an important set of mechanisms that can promote stochastic variability and brings additional potential for diversifying phenotypes (Oey and Whitelaw 2014; Vogt 2015). Stochastic changes of DNA methylation, also called spontaneous epimutations, have been studied for years in plant species (recently reviewed in Johannes and Schmitz 2019), but this source of epigenetic variation and its consequences in terms of adaptation and evolutionary potential in animal populations is less described.
In summary, the source of epigenetic variability can be attributed to genetic, environmental, or stochastic factors and it can affect the phenotypic diversity by interacting with the different components of Equation (1). Moreover, epimutations can also be inherited or not. Inherited epigenetic variation has been reported in various species and consequently blurs the line between genetic and environmental variation by introducing mechanisms other than genetic variability to explain the heritability of phenotypic traits (Richards 2006; Richards et al. 2010; Heard and Martienssen 2014; Boskovic and Rando 2018; Perez and Lehner 2019). This transgenerational epigenetic inheritance has been widely debated due to its implication in the so-called neo-lamarckian theories which exceed the scope of the present review (Burggren 2014; Skinner 2015). The molecular mechanisms of epigenetic inheritance and its possible implications in terms of evolution have been discussed by Sarkies (2020), and it is, however, important to stress that either inherited or not, the epimutation must first appear due to the genotype, the environment or randomness. In the following section, we will describe the mechanistic aspects of DNA methylation, with the intention of highlighting how these epigenetic modifications might occur due to chance.
Emergence of stochastic variation of DNA methylation
Methylation fidelity during cell replication
The first possible mechanism sustaining stochastic variation of DNA methylation refers to the need of the cell to faithfully replicate its epigenetic information for the functioning of its daughter cells, as it similarly applies to the genetic information. Recent findings, however, emphasize the imperfect fidelity of methylation replication. Single-cell bisulfite sequencing has identified substantial unexplained methylome variation (on average 3.3%) among genetically identical mouse liver cells (Gravina et al. 2015; Gravina et al. 2016). This result is concordant with results by Landan et al. (2012): in an in vitro microevolutionary experiment, they let immortalized fibroblasts divide for 300 generations and found considerable differences in methylomes at the end of the experiment, that they attributed to stochastic effects.
Where does this stochasticity come from? To answer this question, it is interesting to consider the mechanisms that regulate and maintain DNA methylation, especially during cell replication and meiosis. In eukaryotes, DNA methylation is faithfully reproduced through the cooperation of enzymes of the DNA methyltransferase (DNMT) family (de Mendoza et al. 2020). DNMT1 is responsible for replicating DNA methylation patterns during the S phase of the cell cycle, using the copied strand methylation patterns as a template for new methylation on the neosynthesized strand. The question subsequently raised is: how does DNMT1 differentiate between unmethylated and hemimethylated DNA? Various factors have been identified so far, mostly in mice and humans. Key to accurate replication of methylation is the autoinhibition of DNMT1. The enzyme contains a domain that binds to unmethylated DNA, inducing a structural change in the active methyltransferase site (Fig. 2). This autoinhibition explains why DNMT1 does not typically contribute to de novo methylation during replication. Recognition of hemimethylated sites, on the other hand, relies on the presence of external factors, such as ubiquitin-like, containing PHD ring finger domains 1 (UHRF1). This protein increases methylation fidelity in two ways: (1) by recruiting DNMT1 to hemimethylated DNA sequences and (2) by cancelling the autoinhibition of the enzyme. This is done by sterically occupying the target DNA sequence that would otherwise cause the inhibitory structural change in the methyltransferase (Bostick et al. 2007; Avvakumov et al. 2008; Fang et al. 2016; Li et al. 2018 ). Hence, it is the combination of both autoinhibition of DNMT1 in the presence of unmethylated DNA and activation through UHRF1 when a site is methylated that explains the preference of this methyltransferase to hemimethylated DNA (Song et al. 2012). Both mechanisms are therefore important considerations when quantifying methylation fidelity.
Mechanism of DNA methylation replication. (a) Unmethylated site: the autoinhibitory loop of DNMT1 interferes with the enzyme’s own methyltransferase active site, restricting methylation; this interaction is dependent on RFTS (Replication Focus Target Sequence) binding to unmethylated DNA. (b) Hemimethylated site: UHRF1 binds to the methyl group on the parent strand; UHRF1 possesses a domain which binds to the RFTS and subsequently displaces the inhibitory loop from that catalytic domain of DNMT1, allowing methylation to occur. High methylation fidelity is the result of both autoinhibition and displacement of inhibition by UHRF1, these mechanisms may therefore be potential sources of stochastic error in reproducing methylation patterns.
DNMT1 is justifiably called the maintenance DNA methyltransferase (Law and Jacobsen 2010). It is suggested that these interactions, that is recruitment by UHRF1 and preference for hemimethylated DNA, greatly increase methylation fidelity, but rates of ∼95% accuracy are nonetheless estimated for DNMT1 fidelity (Riggs and Xiong 2004; Goyal et al. 2006). Regardless of the strong affinity for hemimethylated DNA, random errors still arise, especially since replication is required over the entire genome.
When investigating seemingly stochastic differences in methylation patterns between cells, it is interesting to not only look at the mechanisms that regulate methylation, but also demethylation. Here, a second class of enzymes is involved: Ten–eleven translocation (TET) enzymes recognize and hydroxylate methyl groups previously added by methyltransferases, from 5mC to 5-hydroxymethylcytosine (5hmC) (Williams et al. 2011). This 5-end cytosine modification impedes recognition and activity by DNMT1 by intruding the catalytic domain of the enzyme. In ensuing replications, these oxidized hemimethylations are not recognized as such by DNMT1, resulting in the passive loss of these marks through replication. They can also be recognized, after which base-pair excision and replacement with unmethylated cytosine takes place (Ji et al. 2014; Weber et al. 2016; reviewed in Wu and Zhang 2017).
De novo methylation during gametogenesis and development
Methylation patterns are replicated over the course of mitosis, but these marks are placed at very specific times during development, namely gametogenesis and early embryogenesis. The patterns are set by other members of the methyltransferase family, namely DNMT3A and DNMT3B, also known as de novo methyltransferases. A first window of susceptibility to stochastic differential methylation could, therefore, lie in the imprinting of primordial germ cells during gametogenesis, silencing certain genes between maternal and paternal genomes. Well-known examples of imprinted genes include Igf2 (DeChiara et al. 1991) and Igf2r (Barlow et al. 1991) (both paternally expressed, respectively, encoding for insulin-like growth factor and IGF2 receptor), as well as the maternally expressed H19 gene, which encodes for a ncRNA (Bartolomei et al. 1991; Peters 2014). To date, it is estimated that around a hundred different genes are imprinted in mammalian genomes (DeVeale et al. 2012; Cassidy and Charalambous 2018). If cells affected by an environmental factor can induce heritable changes in gamete epigenome, as it is the case with certain endocrine disruptors (Santangeli et al. 2016; Brander et al. 2017) such as vinclozolin in rats and mice (Guerrero-Bosagna et al. 2010; Gillette et al. 2018; Skinner et al. 2019; Ben Maamar et al. 2020), then stochastic epimutations may also have the potential to be inherited and cause similar phenotypic changes if they affect promoter regions crucial for development or gamete survival.
Embryogenesis offers another window of susceptibility for stochastic epigenetic alterations. While it is still unclear for nonmammalian vertebrates and for invertebrates (de Mendoza et al. 2020), a second wave of genome-wide demethylation takes place shortly after fertilization in mammals, resetting the epigenetic landscape (leaving intact germline differentially methylated regions set during gametogenesis) and is called the DNA methylation reprogramming. Demethylation of paternal and maternal genomes in the early zygote is achieved through active (likely involving TET enzymes) and/or passive (absence of methylation maintenance through replication) processes (Gu et al. 2011). Once more, errors in demethylation may have important consequences for embryo development and may contribute to phenotypic variation (reviewed in Smallwood and Kelsey 2012). Finally, DNMT3A and DNMT3B methylate the DNA sequence without a pre-existing pattern (as opposed to DNMT1). Although it is not yet precisely known how DNMT3A and DNMT3B selectively methylate genome sequences in male and female germ cells, it has been suggested that target sequences recruit these enzymes and enable new imprinting at specific sites (Wienholz et al. 2010; for a review of DNMTs, see Edwards et al. 2017).
As with mitotic methylation maintenance, many factors have been found to be associated with de novo methylation and maintenance of marks during embryogenesis: ZFP57 (KRAB protein family) for example is expressed in female germline cells and embryos of mammals. This protein is necessary for the methylation of certain imprinted genes such as Snrpn (in oocytes), as well as maintaining imprint methylation during embryogenesis (Li et al. 2008; Shi et al. 2019). TRIM28 is equally essential for maintaining genomic imprinting during embryogenesis in these organisms (Messerschmidt et al. 2012). Both maternal and zygotic TRIM28 are involved in the maintenance of methylation during embryogenesis: the loss of expression of either source of TRIM28 leads to a decrease in methylation fidelity, and the absence of both provokes the loss of imprinting altogether in mutants. These observations highlight the crucial role of this protein and its correct abundance in maintaining epigenetic marks during embryogenesis. In mice, changes in expression of TRIM28 leads to changes in epigenomes during embryogenesis, ultimately leading to variation in adult phenotypes (Whitelaw et al. 2010).
In summary, stochastic variability in DNA methylation patterns can emerge during (1) cell division, (2) gametogenesis, and (3) and embryogenesis. It is evident by now that accurately setting, erasing, and reproducing methylation patterns are complex processes involving a series of interconnected factors. As highlighted recently by Nicholson (2019), the cell is not a perfectly efficient “machine.” On the contrary, it is increasingly evident that stochasticity is ubiquitous in biological processes that are highly dependent on timing and abundances of enzymes, transcription factors, and so forth. That is precisely the point that we have illustrated here: stochasticity is just as likely to play an important role in epigenetic reprogramming as a potential source of phenotypic variation.
Epimutation rate and genetic assimilation
As developed above, several mechanisms could be responsible for the stochastic establishment of epigenetic marks. Nevertheless, the link between these mechanisms and the stochastic between-individuals epigenetic variability is not clear. Moreover, studies that aimed to quantify the rate of such epimutations are rare, and have mainly focused on model organisms, such as Arabidopsis thaliana. Epimutation rates were estimated in this species at 4.46 × 10−4 per generation per haploid genome (Schmitz et al. 2011), which was further confirmed by a second study, providing both forward and backward CG dinucleotide epimutation rates of, respectively, 2.56 × 10−4 and 6.30 × 10−4 (van der Graaf et al. 2015). This finding contrasts with the spontaneous genetic mutation rate of 7 × 10−9 base substitutions per site per generation (Drake et al. 1998). However, quantifying epimutation rates in animals proves to be a great challenge. In birds, methylation profiles were studied between five closely related species of Darwin’s finches (Geospiza fortis, G. fuliginosa, G. scandens, Camarhynchus parvulus, and Platyspiza crassirostris), which provided some estimation of epimutation rate in animals: variable epimutation seems to appear continuously during evolution as there is a correlation between phylogenetic distance and level of differential methylation in distinct finch species. Furthermore, the number of epigenetic changes is greater than mutational change between species, suggesting, as observed in plants, that epimutations occur more frequently than genetic mutations (Skinner et al. 2014). In fish, natural populations of North American stream fishes have been studied for DNA methylation variation (Smith et al. 2016). The authors observed that within a single species, epigenetic diversity is greater than genetic diversity both within and among populations. They showed that populations present greater methylome differentiation than genome and that the former is changing faster than the nucleotide sequence.
Although transgenerational epigenetics is not yet fully understood and the distinction between transgenerational and non-transgenerational epimutations is often difficult (McCarrey 2014), the transmission of heritable phenotypic variation can be advantageous in periodically fluctuating environments. Inheritance is in this case advantageous if the change in environment lasts longer than the generation time, enabling offspring to benefit from the randomly acquired adaptive advantage. The model of Lachmann and Jablonka (1996), furthermore, proposes that optimal epimutation rates are those that match environmental fluctuation rates: a greater increase in epigenetic variability is necessary for a smaller number of generations affected by the change during a cycle.
As epimutations occur at a higher rate than genetic mutations, stochastically occurring epigenetic variability can be a means of producing and maintaining necessary diversity until phenotypes become stably fixed through DNA sequence mutations (i.e., genetic assimilation). Klironomos et al. (2013) suggested that a high rate of epimutations can be beneficial for a population, as it allows for a faster exploration of the fitness landscape and speed of adaptation, compared to solely relying on genetic mutations. This is concordant with the hypothesis that genetic mutations may be followers as well as pilots of evolutionary change (West-Eberhard 2005), especially when considering the possibility that epimutations may cause subsequent mutations by decreasing genome stability (Timp and Feinberg 2013). Stochastic epimutations can be an effective way of increasing fitness: this is particularly relevant in the case of fluctuating environments and the evolution of bet-hedging strategies that possibly use these processes, which is the topic of the following section.
Spontaneous epimutations: A potential mechanism for diversified bet-hedging
Diversified bet-hedging (hereafter bet-hedging) is a “risk-spreading” evolutionary strategy in which a single genotype produces a distribution of phenotypes across offspring, which results in an increased long-term reproductive success (Simons 2011; Honegger and de Bivort 2018). This is mainly observed for organisms living in unpredictably fluctuating environments, in which individuals stochastically express a phenotype of reduced fitness that could be adaptive in the future (Grimbergen et al. 2015). Theoretical studies show that if the variations of the environment are predictable, a sense-and-response system (i.e., phenotypic plasticity) seems to be the best strategy (Botero et al. 2015). For example, if the environmental cues perceived during development are good indicators of what the offspring will face later in life, then phenotypic plasticity may be an effective strategy that is likely to increase fitness. However, the plastic response may be too costly or complex (Levy et al. 2012; Honegger and de Bivort 2018), or may be useless if the environmental cues do not provide reliable information about future adaptive phenotypes (Abley et al. 2016). Instead, diversified bet-hedging enables a species to face highly unpredictable changes, by always presenting different phenotypes in a population (Lenormand et al. 2009; Rajon et al. 2014). By doing so, whatever the conditions are, a portion of the population will be adapted to the current environment (Simons 2011).
Bet-hedging has been found in 16 phyla and most evidence has been so far provided for prokaryotes, chordates, angiosperms, and arthropods (reviewed in Simons 2011). One of the proposed mechanisms underlying this strategy is that the stochastic establishment of epigenetic marks could lead to different phenotypic states across an isogenic line facing the same environment (Levy et al. 2012; Casadesús and Low 2013; Angers et al. 2020). As a potential consequence, genotypes promoting stochasticity could be favored by natural selection. For instance, DNMT1 copies with reduced affinity for hemimethylated sites could be selected for, therefore, increasing the rate of de novo methylation (Castonguay and Angers 2012). However, very little is still known about the nature of the molecular mechanisms underlying the phenotypic variation observed in bet-hedging, particularly in multicellular organisms (Herman et al. 2014; Abley et al. 2016; Honegger and de Bivort 2018). The study of wild populations of Chrosomus eos‐neogaeus by Leung et al. (2016) is the only case of multicellular animals, to our knowledge, in which the methylome has been studied from a bet-hedging perspective. This hybrid fish reproduces clonally via gynogenesis and therefore presents a low level of genetic diversity. Leung’s team highlighted the fact that populations living in unpredictable environments show significantly more stochastic methylation marks than the organisms found in predictable environments, which suggests the selection for systems responsible for the establishment of stochastic epigenetic diversity. Clonal species are excellent models for deciphering the role of stochastic epigenetic variation in natural populations. More broadly, these organisms illustrate that low genetic diversity can be substituted by epigenetic variability, as explained in the next section.
DNA methylation variability as a substitute for low genetic variability
It is well established that poor genetic diversity makes populations vulnerable to environmental fluctuations and leads, when it is the result of inbreeding, to a phenomenon named inbreeding depression (Awad and Roze 2020). Therefore, at least some degree of variability in heritable phenotype is necessary to survive these fluctuations, as genetically poor populations are more susceptible to extinction. However, genetically identical organisms kept in the same environment are not necessarily phenotypically identical. Studies working on highly-inbred, isogenic, monozygotic, or clonal organisms, reared in strictly identical controlled environments illustrate this (Veitia 2005; Peaston and Whitelaw 2006). Initially, evidence was brought by Gärtner (1990) who, working on inbred strains of mice reared in the same environment, reported baseline variability of biological traits, while increasing environmental variability did not increase it. In this situation, stochastic epigenetic variability might be predominant and compensate for low levels of genetic diversity. Three life-history strategies seem to thrive in fluctuating environments despite displaying poor, if not absence of genetic diversity altogether: invasive species, asexually reproducing organisms, and selfing organisms.
Invasive species
First of all, the case of invasive species is relevant when considering the genetic paradox of invasion, in which a population thrives and adapts to a new environment regardless of a significant decrease in genetic diversity associated with passing through a bottleneck (Estoup et al. 2016). Examples of increase in epigenetic diversity among invasive species include the pygmy mussel Xenostrobus secures, the tubeworm Ficopomatus enigmaticus (Ardura et al. 2017), the mussels Mytilus galloprovincialis, and Xenostrobus securis (Ardura et al. 2018), and house sparrows Passer domesticus during its introduction in Kenya (Schrey et al. 2012; Liebl et al. 2013). In the latter study, epigenetic diversity was found to be negatively correlated with genetic diversity, whereas a positive correlation was established between extent of inbreeding and epigenetic diversity. In this context, the invasive marbled crayfish, Procambarus virginalis, can offer further insight (Vogt et al. 2008). This cambarid Crustacean is an all-female apomictic parthenogenetic freshwater species discovered in 1995, and producing genetically identical progeny expressing a variety of phenotypic traits. Despite isogenicity, siblings showed a broad range of social, reproductive, and locomotor behaviors, which cause different individual fitness in the population. Using whole-genome bisulfite sequencing, Gatzmann et al. (2018) revealed that its methylome is highly conserved between the different analyzed tissues and that gene bodies are highly hypomethylated, contrary to what is generally admitted (Zhang et al. 2016). This overall hypomethylation is correlated with a high level of gene expression variability and raises the possibility of an epigenetic mechanism sustaining adaptability of invasive species (Vogt 2018).
Asexually reproducing organisms
Asexually reproducing organisms do not make use of homologous recombination and shuffling associated with meiosis and sexual reproduction to generate genetic diversity. Rather than being evolutionary dead-ends, their success is evident in diverse environments, and both strict and facultative asexuality are widespread across taxa (Vogt 2015). In their review, Verhoeven and Preite (2014) argued that asexual organisms can take advantage of epigenetic variation via two routes: epigenetics-mediated phenotypic plasticity, and heritable variation via stochastic epimutations. An example concerns the asexual snail, Potamopyrgus antipodarum, whose different populations derived from the same clonal lineage, but living in urban or rural lakes. They showed different phenotypic and DNA methylation patterns, supporting the idea that epigenetic variation is associated with phenotypic variation and environmental factors (Thorson et al. 2019).
Epigenetics may here play a role similar to that observed in invasive species, that is, the generation of the necessary phenotypic variability (Goyal et al. 2006; Peaston and Whitelaw 2006). Besides the example of the marble crayfish, parthenogenetic aphids are other invertebrates that produce alternative morphotypes by genetically identical individuals. The genome of pea aphid, Acyrthosiphon pisum, has been sequenced and epigenetic mechanisms have been advanced to explain its phenotypic plasticity (Srinivasan and Brisson 2012). More precisely, it has been proposed that aphid polyphenism may be regulated by DNA methylation of genes involved in the metabolism of the juvenile hormone, which contributes to the regulation of morphotype determination in aphids (Walsh et al. 2010). Daphnia crustaceans are well-established models in ecology and evolutionary biology and are now also used as invertebrate models in epigenetics (Harris et al. 2012). The benefits of studying these species are numerous: they are capable of high levels of phenotypic plasticity in response to environmental changes, their well-known parthenogenetic life cycle reduces the issue of genetic diversity as a confounding factor, and large amounts of data about their ecology and their genomes are available. Recently, Daphnia has been extensively used to study the transgenerational effects of aquatic pollutants or other stressors via epigenetic mechanisms (Jeremias et al. 2018; Ellis et al. 2020). The evolutionary perspectives of the DNA methylation pattern have also been reported for two species and showed the importance of methylation in exons for gene expression regulation (Kvist et al., 2018). Asselman et al. (2015) reported significant genotype, environment, and genotype × environment effects on DNA methylation patterns, pointing toward the importance of epigenetic mechanisms in responding to environmental stressors and to the expression of phenotypes. However, the importance of a stochastic source for epigenetic variance is still to be elucidated.
In vertebrates, a few fish species presenting no genetic diversity have been studied and are promising to determine the role of epigenetic mechanisms in sustaining phenotypic variation in the absence of genetic variation. The all-female Amazon molly, Poecilia formosa, is the result of the hybridization between the sailfin molly Poecilia latipinna and the Atlantic molly Poecilia mexicana. It reproduces gynogenetically and gives birth to isogenic offspring. It shows individuality in its behavior within siblings as reported by Bierbach et al. (2017), suggesting some implication of epigenetic mechanisms. The diploid hybrid C. eos-neogaeus is another good model fish species that reproduces clonaly via gynogenesis. A study of its DNA methylation concluded that epigenetic variation in the absence of genetic variation can be a source of phenotypic variation in natural populations, and that the observed epigenetic variation is the result of both stochastic and environmental epimutations on clones originating from different populations (Massicotte et al. 2011).
The origin of epigenetic variability in a selfing vertebrate species
As developed in the present review, determining the part of stochasticity in the origin of epigenetic variability (along with genetic and environmental components) in animal populations can have important evolutionary implications but is still widely unexplored. The role of spontaneous epimutations in diversifying the phenotype can advantageously be investigated in a fish recently added to the list of model species, the mangrove rivulus Kryptolebias marmoratus. Contrary to Amazon molly or C. eos-neogaeus, it reproduces sexually. However, its reproduction system is unique among the vertebrates as it involves a mixed-mating system, alternating between mating of a male and a hermaphrodite on the one hand, and self-fertilization hermaphrodites on the other hand (Fig. 3) (Tatarenkov et al. 2009; Costa 2013). The exclusive selfing process during several generations gives rise to isogenic strains (Tatarenkov et al. 2012) even if a low level of genetic polymorphism exists (Lins et al. 2018). Importantly, we can find many different isogenic lineages in its natural environment (the red mangroves from Florida to Central America and Brazil) whose individuals are genetically identical within each lineage, but genetically different among the different lineages. This species exhibits a high level of phenotypic variation, both within and among lineages, which makes this fish a useful model to study the role of epigenetics in modeling phenotypes (Fellous et al. 2018). It was suggested that natural variation of selfing rate among populations might be explained by epigenetic regulation of sex ratio in consequence of environmental signals (Ellison et al. 2015). We now know that there is a high epigenetic variation between different isogenic lineages, and methylation differences are greater between genotypes than between environments (Ellison et al. 2015; Berbel-Filho et al. 2019a, 2019b), emphasizing both genetic and environmental origins of epigenetic variation.
The mangrove rivulus, Kryptolebias marmoratus, is a new model fish species which reproduces by androdioecy. In its natural environment, hermaphrodites (foreground) coexist with males (background), while no female has ever been observed. Hermaphrodites commonly self-fertilize and produce offspring that are highly homozygous and isogenic. While the selfing rate is variable, depending on the ratio of males in the population, many isogenic lineages exist in natural populations, each of them being genetically different. This species represents a unique case of a vertebrate capable of producing clonal lineages by sexual reproduction and is a model of choice to investigate the role of epigenetic variation in adaptation and evolution within isogenic lineages and between different genotypes. Photo credit: Frédéric Silvestre.
Moreover, after showing that the DNA methylation reprogramming pattern during early development is longer, and reaches lower levels of global methylation when compared to other species, Fellous et al. (2018) proposed that environmental cues could be integrated into the epigenome during the development, facilitating the expression of new phenotypes in lineages with low genetic diversity. This hypothesis still needs to be tested but it would be in accordance with the general-purpose genotype model, which states that evolutionary success of isogenic lineages is possible in generalist individuals showing a good ability to change their phenotype (Massicotte and Angers 2012). However, even if such environmental epimutations are meaningful, it does not prevent a significant role of stochastic epimutations in the adaptation of this species to its environment. The possibility that this fish species could use a strategy of diversified bet-hedging should not be excluded either, and would emphasize the importance of stochastic epigenetic variability in the adaptation to its environment.
As can be seen, the mangrove rivulus is an attractive model to investigate the origins of epimutations and their respective importance in adaptive evolution. It is worth mentioning that the species is also a great model to explore inbreeding depression in an extreme inbreeding system by comparing the fitness in populations that differ by their rates of selfing. If epigenetic processes have been shown to regulate the inbreeding effect in plants (Vergeer et al. 2012), then the role of epigenetics in inbreeding depression in animal populations has poorly been investigated. For now, there is no evidence that the extreme inbreeding reported in many mangrove rivulus populations is accompanied by inbreeding depression (Kelley et al. 2016; Turko et al. 2018). If this is confirmed, then a potential explanation would be that epigenetic variability would compensate for a low genetic diversity to maintain the fitness of the population, and that stochasticity could be an important source of this variability.
Conclusions
Stochasticity has been found to play a notable role in many biological processes. As a result, further understanding the emergence and subsequent implications of stochasticity in epigenetics has become increasingly relevant. As our understanding of epigenetics gradually grows, it also becomes increasingly evident that the processes involved are complex. Unexplained variation is ubiquitous but still a necessary element to consider. Nevertheless, important limitations exist when investigating stochasticity in this context. For example, it remains difficult to accurately quantify the effect of stochastic epigenetic variability on the phenotype, especially when considering the complexity of development in many organisms. Next, a deeper understanding of epigenetic inheritance across individuals is still necessary, particularly when exploring evolutionary scenarios. Finally, it is important to emphasize the multiplicity of epigenetic mechanisms: while the issue is mainly explored with regards to DNA methylation, histone modification and RNA interference are further mechanisms that must be included in the discussion.
From the symposium “Epigenetic Variation in Endocrine Systems” presented at the annual meeting of the Society for Integrative and Comparative Biology January 3–7, 2020 at Austin, Texas.
Glossary
Bistability – The coexistence of two stable phenotypes within a clonal population due to stochastic fluctuations in the cellular components.
CGIs – A short CpG-rich region of the genome characterized by >0.5 kb of DNA with a G + C content ≥55%, and an observed/expected ratio ≥0.6.
Epigenetics – Mitotically and/or meiotically heritable changes in gene function that is not associated with changes in DNA sequence. It includes three main types of mechanisms: DNA methylation, histone modifications, and ncRNA.
General-purpose genotype model – An evolutionary model suggesting that generalist lineages of asexual organisms could have been selected for their flexible phenotypes.
Genetic assimilation - A process by which a phenotype, initially produced in response to some environmental influence, becomes genetically encoded via selection, and no longer requires the environmental signal for its expression.
Gynogenesis – A form of parthenogenesis in which the sperm serves to trigger the embryogenesis but is absent from the genome of the embryo.
Hemimethylated – The state of a CpG site when only one of the two complementary strands is methylated.
Inbreeding depression – Reduction of fitness in a population resulting from inbreeding.
Isogenicity – All individuals are genetically identical.
Parthenogenesis – Development of an organism from an unfertilized egg cell and consequently, only using maternal genetic information.
Self-fertilization – Mode of reproduction in which fertilization is the result of the fusion of gametes from the same individual.
Stochasticity – Refers to randomly determine processes that cannot be predicted a priori from readily measurable variables.
Transgenerational epigenetic inheritance – Transmission of epigenetic marks from one generation to the next, in the absence of an environmental trigger acting directly on the subsequent generation.
Funding
This study was funded by the Fonds National de la Recherche Scientifique (FRS-FNRS), “Crédit de recherche” number J.0189.20 “Epigenetic diversity origin in rivulus”, and PhD fellowship to V. Chapelle.
References
Author notes
C. Biwer and B. Kawam contributed equally to this work.
