Using asexual vertebrates to study genome evolution and animal physiology: Banded (Fundulus diaphanus) x Common Killifish (F. heteroclitus) hybrid lineages as a model system

Abstract Wild, asexual, vertebrate hybrids have many characteristics that make them good model systems for studying how genomes evolve and epigenetic modifications influence animal physiology. In particular, the formation of asexual hybrid lineages is a form of reproductive incompatibility, but we know little about the genetic and genomic mechanisms by which this mode of reproductive isolation proceeds in animals. Asexual lineages also provide researchers with the ability to produce genetically identical individuals, enabling the study of autonomous epigenetic modifications without the confounds of genetic variation. Here, we briefly review the cellular and molecular mechanisms leading to asexual reproduction in vertebrates and the known genetic and epigenetic consequences of the loss of sex. We then specifically discuss what is known about asexual lineages of Fundulus diaphanus x F. heteroclitus to highlight gaps in our knowledge of the biology of these clones. Our preliminary studies of F. diaphanus and F. heteroclitus karyotypes from Porter's Lake (Nova Scotia, Canada) agree with data from other populations, suggesting a conserved interspecific chromosomal arrangement. In addition, genetic analyses suggest that: (a) the same major clonal lineage (Clone A) of F. diaphanus x F. heteroclitus has remained dominant over the past decade, (b) some minor clones have also persisted, (c) new clones may have recently formed, and iv) wild clones still mainly descend from F. diaphanus ♀ x F. heteroclitus ♂ crosses (96% in 2017–2018). These data suggest that clone formation may be a relatively rare, but continuous process, and there are persistent environmental or genetic factors causing a bias in cross direction. We end by describing our current research on the genomic causes and consequences of a transition to asexuality and the potential physiological consequences of epigenetic variation.


| G ENER AL INTRODUC TION
Sexual reproduction, defined as "the union of two gametes and genomes" (Lehtonen, Jennions, & Kokko, 2012), is the predominant form of reproduction in eukaryotes (reviewed by Maynard-Smith, 1978;Otto, 2009). This prevalence is surprising, because sexual reproduction has immediate fitness costs that can be avoided with asexual reproduction (reviewed by Lehtonen et al., 2012). Therefore, a major goal of evolutionary biology is to better understand why sex originally evolved and is maintained; this is often done by quantifying and comparing the costs and benefits of sexual versus asexual reproduction across different evolutionary time-scales and environmental parameters using theoretical models (reviewed by Otto, 2009), laboratory-based studies (reviewed by Sharp & Otto, 2016), and field studies (reviewed by Neiman, Meirmans, Schwander, & Meirmans, 2018).
As environment-specific factors are predicted to be important in the maintenance of sex, further comparisons of unisexual clonal lineages to their sexual congeners in the wild will be critical for evaluating hypotheses for why sex prevails in nature (Neiman, Lively, & Meirmans, 2017;Neiman et al., 2018). Beyond helping unravel the mystery of why so many eukaryotes reproduce sexually, natural clonal lineages have also been used to determine the genetic mechanisms by which sexual reproduction breaks down and asexual fertile eggs are produced (reviewed by Avise, 2012;Neaves & Baumann, 2011;Neiman, Sharbel, & Schwander, 2014). Furthermore, asexual clones can be used to address other biological questions that require genetically identical replicate individuals. In such studies, natural clonal vertebrates may be preferred over inbred/isogenic genetically identical, laboratory-bred lines because they have evolved in the wild, so represent real evolutionary outcomes, unlike laboratory strains that might have low fitness in natural environments (reviewed by Laskowski, Doran, Bierbach, Krause, & Wolf, 2019).
The goal of this paper is not to discuss the evolutionary paradox of sex, but to highlight how studying naturally occurring, asexual vertebrate lineages can provide insight into the mechanisms by which genomes evolve and epigenetic modifications influence vertebrate physiology. In the first section of this paper, we review general trends in asexually reproducing vertebrates, including the cellular and molecular mechanisms leading to asexual reproduction, and the known genetic and epigenetic consequences of the loss of sex. We then review the benefits of using asexual fishes as model organisms and discuss what is known about Fundulus diaphanus x F. heteroclitus clonal lineages, including new data from our laboratories. We end by outlining our current research using these asexual fish to study the genomic causes and consequences of a transition to asexuality and the effects of epigenetic variation on phenotype. For readers interested in the broader topic of asexuality in multicellular organisms, including the causes and consequences of this mode of reproduction in invertebrates and plants, we suggest reviews by Neiman et al. (2014), Neiman et al. (2017), Schwander, Marais, and Roze (2014), and Schmidt, Schmid, and Grossniklaus (2015).

| INTRODUC TI ON TO A S E XUAL REPRODUC TI ON IN VERTEB R ATE S
The majority of vertebrates reproduce sexually, but there are about 100 known asexual lineages of fish, reptiles, and amphibians (Avise, 2008(Avise, , 2015Neaves & Baumann, 2011;Vrijenhoek, Dawley, Cole, & Bogart, 1989). The lack of avian or mammalian unisexual clonal lineages is hypothesized to be the result of developmental and genetic constraints in these endothermic lineages, such as the developmental lethality associated with errors in mammalian genomic imprinting (Kono et al., 2004) and the bird sex determination system (female heterogamety may prevent successful asexual reproduction; see Engelstädter, 2008).
Asexual lineages of vertebrates are, to the best of our knowledge, almost exclusively the result of interspecific hybridization between parental species that tend not to be sister-taxa (reviewed by Avise, 2015;Laskowski et al., 2019;Neaves & Baumann, 2011).
This observation, in combination with data showing that the proportion of unreduced gametes increases and hybrid fecundity decreases with parental species divergence, led Moritz et al. (1989) to suggest the "balance hypothesis." This hypothesis predicts that asexual clone formation is a form of reproductive isolation that occurs when reproductive incompatibilities have accumulated to the extent where normal gametogenesis is disrupted, but hybrids are still viable and asexually fertile (Vrijenhoek, 1989). In particular, divergence between parental species may interfere with homologous chromosome alignment, homology search and subsequent crossovers during meiosis in F1 hybrids (Dion-Côté & Barbash, 2017), which may increase the probability of a transition to asexuality by destabilizing meiosis (Janko et al., 2018).
The ability to form these asexual lineages is also hypothesized to be phylogenetically constrained, such that only certain taxa have the genetic and developmental machinery required to form fertile clones (reviewed by Engelstädter, 2008). The specific genetic and cellular mechanisms leading to the formation of most unisexual vertebrates are still under study (see Section 2.1 "Cellular and molecular mechanisms of asexual reproduction in vertebrates"), but evidence that hybridization is critical for the initiation of clonality in vertebrates continues to mount (Janko et al., 2018;Neiman et al., 2014).

| Cellular and molecular mechanisms of asexual reproduction in vertebrates
Asexuality has evolved independently many times in vertebrates (Avise, 2015). Consequently, the mechanisms underlying the modes of egg production and activation of embryonic development vary across-and sometimes even within-lineages. Here, we discuss the most common forms of egg production ( Figure 1) and development initiation ( Figure 2) in asexual vertebrates, while considering ploidy maintenance and the genetic impacts of different modes of asexuality.
F I G U R E 1 Modes of asexual egg production. This figure is inspired and adapted from Neaves & Baumann, (2011) and Neiman et al., (2014). Different shades indicate homologous chromosomes. F0: parental generation, F1: progeny , R: DNA replication, M: mitosis (upper panel) and meiosis (lower panel). PB: polar body. Upper panel: (a) Under mitotic egg production, the cell divides according to regular mitosis, and both ploidy and heterozygosity are maintained. Progeny are perfect clones of their mother. Lower panel: Asexual egg production by a modified version of meiosis may yield to variable results. (b) Under endoduplication, the germ cell undergoes an additional round of DNA replication prior to entering meiosis (R 1 , R 2 ). Duplicate chromosomes recombine together and therefore heterozygosity is maintained and the progeny are perfect clones. (c) Under oogonial fusion (which has not been observed in vertebrates in nature), premeiotic oogonia fuse, and sister chromosomes recombine as in endoduplication, also resulting in perfect clonality and the maintainenance of heterozygosity. (d) Under fusion, the early meiotic program is maintained (including meiotic recombination between homologous chromosomes) and the maturing egg fuses with either of the first polar bodies (central fusion) or the secondary polar body (terminal fusion). Progeny heterozygosity and genetic makeup will vary depending on the exact product that fuses with the oocyte. (e) Under gametic duplication, haploid gametes are produced following meiosis, and their genome is subsequently duplicated. Genetic makeup may vary (depending on whether a recombinant sister chromatid ends up into the egg) but heterozygosity is completely lost. (f) Finally, hybridogenesis is a particular case in which the paternal genome is eliminated prior entry into meiosis. The resulting progeny are thus perfect half clones (no recombination can occur) but differ in their paternal complement

| Modes of asexual egg production
If eggs are produced mitotically, ploidy and genetic makeup are readily maintained. However, mitotic egg production is not known to occur in asexual vertebrates (Neiman et al., 2014). Thus, asexual vertebrate eggs likely all undergo some modified form of meiosis.
Meiosis is a reductional cell division, so a frequent prerequisite to asexual reproduction in vertebrates is the production of unreduced eggs (i.e., same ploidy as the mother-either diploid or polyploid).
These eggs may, or may not, be genetically identical to their mother, depending on whether recombination has occurred between sister (see next paragraph) or homologous chromosomes, and the specific mechanism used to maintain ploidy (see Engelstädter, 2008;Neaves & Baumann, 2011;Neiman et al., 2014 for more detailed explanations and diagrams).
Under endoduplication (also termed endoreplication), which appears to be the most frequent mechanism used by asexually reproducing vertebrates to maintain ploidy, germ cells undergo an additional round of DNA replication prior to entering meiosis (Neaves & Baumann, 2011). This means that the cell entering meiosis has twice the amount of DNA there would be in a sexually reproducing organisms ( Figure 1); this process has been reported in a number of vertebrates, including fishes in the family Cobitidae (Arai & Fujimoto, 2013;Dedukh et al., 2020;Itono et al., 2006;Juchno, Arai, Boron, & Kujawa, 2017) and the genus Poeciliopsis (Cimino, 1972b).
After endoduplication, meiotic recombination occurs between duplicate chromosomes ("sister chromosomes"), instead of homologous chromosomes ( Figure 1b). Sister chromosomes are predicted to align on the metaphase plate in meiosis I (Kuroda, Fujimoto, Murakami, Yamaha, & Arai, 2018), resulting in an egg that is an exact genetic and chromosomal copy of its mother. Other processes besides endoduplication may also yield unreduced eggs, but these have been poorly characterized in vertebrates at present. For example, endoduplication could be substituted by oogonial fusion (oogonia are premeiotic germ cells), a process that would be genetically equivalent, but has not been reported in vertebrates (Figure 1c; Neaves & Baumann, 2011). As well, Newton et al. (2016) found that cells entering meiosis in parthenogenetic whiptail lizards were tetraploid, similarly to a closely related sexual species, meaning that the additional round of DNA replication may be a stochastic event that occurs during, and not before, meiosis. Note that under hybridogenesis, a form of "semi-asexual reproduction" (see Section 2.1.2 "Development initiation in asexual vertebrates"), reduced eggs are produced by premeiotic exclusion, a mechanism by which the paternal chromosome complement is excluded, yielding reduced, nonrecombinant eggs ( Figure 1f). These eggs are subsequently fertilized, and the paternal genome incorporated, to generate a diploid embryo, such as in some Poeciliopsis and Squalius lineages (Alves, Coelho, & Collares-Pereira, 2001;Cimino, 1972a;Schultz, 1977;reviewed in Lamatsch & Stöck, 2009).
There are other potential mechanisms that may produce eggs with a ploidy similar to the mother's somatic cells, which have only been reported in invertebrates to date (Figure 1d,e; central fusion, terminal fusion and gamete duplication; reviewed by Neaves & Baumann, 2011;Neiman et al., 2014;Suomalainen, Saura, & Lokki, 1987). To understand other types of ploidy restoration and their genetic consequences, it is important to remember that (a) female meiosis is characterized by the production of a single egg, and degeneration of the other three products (i.e., polar bodies), and (b) meiotic recombination normally occurs between homologous chromosomes (not sister chromosomes), such that recombinant chromosomes bear new allele combinations. Under central fusion, the egg fuses with a polar body bearing non-sister chromatids, so the resulting embryo may, or may not, be genetically identical to their mother depending on whether recombinant chromatids end up in the egg. Under terminal fusion, meiotic products bearing sister chromatids postrecombination fuse, yielding an embryo that is genetically different from their mother (resulting in a substantial decrease of heterozygosity). For both terminal and central fusion, fusion between two polar bodies to give rise to the embryo is also a possibility but has not been reported in vertebrates (Stenberg & Saura, 2009). Finally, under gamete duplication (Figure 1e), there is no fusion of meiotic products, and ploidy is instead restored by duplication of the egg's genetic material, giving rise to an entirely homozygous embryo.

| Development initiation in asexual vertebrates
Among vertebrates, the mode of development initiation during asexual reproduction comes in two major flavors ( Figure 2): (a) parthenogenesis (spontaneous development initiation), and (b) gynogenesis (requires sperm activation of development). There are also two systems that integrate elements of both asexual and sexual reproduction: hybridogenesis and kleptogenesis (Avise, 2008).
Hybridogenesis requires sperm activation and incorporates the paternal genetic complement, without recombination with the maternal genome, for just one generation (see Figure 2). For this reason, hybridogenesis is often referred to as hemiclonal reproduction (Avise, 2008). Kleptogenesis, a particular reproductive mode where unisexual females appear to sporadically "steal" male genomic DNA from congener species, is found in some salamanders in the genus Ambystoma (Bogart, Bi, Fu, Noble, & Niedzwiecki, 2007). Since kleptogenesis appears to be restricted to these salamanders, we will not discuss it further in this review focused upon fishes (see Avise, 2008 for further information).
Parthenogenesis is an umbrella term describing the development of an individual from an unfertilized egg (Maynard-Smith, 1978). As far as we know, parthenogenesis in vertebrates is restricted to the production of females without the contribution of a male (i.e., thelytoky; Avise, 2015). All known parthenogenetic vertebrates are squamates, mainly lizards (Avise, 2008). If eggs were to be produced mitotically (i.e., apomixis), there would be no ploidy reduction and the resulting progeny would be genetically identical to their mother.
As eggs from asexual vertebrates are produced meiotically (i.e., automixis), the genetic makeup of the progeny may, or may not, be identical to their mother, depending on the exact cellular and molecular mechanisms involved in egg production (see previous section).
For example, in Aspidoscelis whiptail lizards, ploidy and heterozygosity are respectively maintained by first undergoing two premeiotic replication cycles (i.e., premeiotic endoduplication) followed by the strict pairing of sister chromosomes, and recombination among sister chromosomes (Lutes, Neaves, Baumann, Wiegraebe, & Baumann, 2010). This sister chromosome pairing maintains elevated heterozygosity and linkage disequilibrium, as opposed to canonical meiosis where homologous chromosomes pair and recombine (Lutes et al., 2010).
Gynogenesis is similar to parthenogenesis with the key difference being that females produce unreduced eggs that still require sperm, which leads to the activation of Ca 2+ signaling pathways initiating embryo development (Stricker, 1999). Sperm may also contribute a centrosome, an organelle essential for chromosome segregation (Engelstädter, 2008;Lamatsch & Stöck, 2009). Sperm normally comes from a closely related sexually reproducing species, but the paternal complement is not incorporated in the developing embryo; ploidy is therefore maintained (Avise, 2008). This mode of unisexual reproduction has been reported in more than half of asexually reproducing vertebrates, including several fish and amphibian lineages (Avise, 2015). For example, some female Chrosomus eos-neogaeus hybrids (formerly known as Phoxinus) produce unreduced eggs that must be activated by C. eos sperm (Dawley, Jack Schultz, & Goddard, 1987). However, gynogenesis is not always maintained in Chrosomus hybrids, as in some cases the paternal genetic complement is incorporated, yielding triploid individuals (reviewed by Lamatsch & Stöck, 2009). In other cases, haploid eggs are produced by triploid female Chrosomus eos-neogaeus hybrids and fertilized by C. eos sperm, yielding diploid fish with fully eos nuclear genomes and neogaeus mitochondrial genomes (i.e., cybrids, Goddard, Megwinoff, Wessner, & Giaimo, 1998). Engaging in the risky aspects of sex, without any of the proposed benefits, makes the persistence of gynogenesis especially interesting from an ecological and evolutionary perspective (Schlupp, 2005). There is also evidence that at least one vertebrate, the cyprinid Squalius alburnoides, may reproduce by androgenesis sometimes mitochondrial chromosomes, so is the "flip-side of gynogenesis" (reviewed by Lehtonen, Schmidt, Heubel, & Kokko, 2013).
Females that reproduce by hybridogenesis produce reduced eggs carrying the maternal chromosome complement that has not undergone recombination; the paternal complement is eliminated from the egg before meiosis is initiated (Lafond, Hénault, Leung, & Angers, 2019;Lavanchy & Schwander, 2019). Eggs are fertilized by sperm, and the new paternal complement is subsequently incorporated in the diploid organism but ejected from the germ line. Thus, hybridogens are effectively hemi-clones, as only the maternal genome is transmitted to the progeny, but not entirely asexual organisms, strictly speaking, as adults have genomes from two parents.
As with gynogens, sperm is needed to release arrest of meiosis II and activate embryonic development, and might also contribute a centrosome (Lamatsch & Stöck, 2009). Hybridogenesis is well documented in fish of the genera Poeciliopsis, in which the formation of a unipolar spindle precluding meiosis leads to the transmission of strictly the maternal complement to the oocyte, a mechanism referred to as premeiotic exclusion (Cimino, 1972a). Therefore, meiosis initiates with a single chromosome complement, and cell division in meiosis I is suppressed (Cimino, 1972a). The oocyte is haploid, carrying the maternal complement, and diploidy is restored upon fertilization by incorporating the new father's genetic material (Cimino, 1972a).

| Ploidy
Whether polyploidy is a cause or consequence of asexual reproduction has been widely debated (reviewed by Avise, 2009).
However, multiple lines of evidence, including cytonuclear data and the presence of asexual taxa with the same ploidy level as sexual parents, suggest that polyploidy is normally a consequence of hybrid-induced asexuality, especially in vertebrates (reviewed by Avise, 2008;Neiman et al., 2014). Despite this evidence, it remains extremely difficult to disentangle the cause(s) from the consequence(s) of asexuality, hybridization and polyploidization (e.g., Lundmark & Saura, 2006). If polyploidy is a consequence of asexual reproduction, the prevalence of polyploids among asexual organisms may suggest polyploidy is advantageous (e.g., by increasing allelic diversity at immune genes, e.g., King, Seppälä, & Neiman, 2012) or simply better tolerated in asexual organisms (e.g., because of reduced constraints with respect to sex determination; Stenberg & Saura, 2009).
Asexual vertebrate polyploids are generally thought to have evolved via a "primary-hybrid route", in which diploid, hybrid mothers first produce unreduced eggs, and then additional events led to polyploid lineages (Schultz, 1969;reviewed in Avise, 2008reviewed in Avise, , 2015Lamatsch & Stöck, 2009). Polyploids (typically triploids) often occur in sperm-dependent asexual lineages (but exceptions exist, e.g., in lizards of the genus Aspidoscelis), where eggs are frequently exposed to sperm. There are two general mechanisms by which "primary hybrid" polyploid asexual lineages may form: genome addition and genome duplication (reviewed in Avise, 2008).
Under the "genome-addition hypothesis," unreduced eggs (AB) are produced and paternal genome is subsequently added (A' or B'), yielding a triploid embryo (ABA' or ABB'). Under the "genome-duplication hypothesis," an equational meiotic division (i.e., meiosis II, so sister chromatids do not segregate) is suppressed, yielding unreduced, highly homozygous eggs (AA or BB) which are subsequently fertilized (by A' or B') to give rise to a triploid embryo (AAB' or A'BB; note AAA' and BBB' could theoretically be formed but are not found in nature; Avise, 2008). A major difference between genome addition and genome duplication is that the heterozygosity of the homospecific chromosome set will be similar to the parental species in the former and close to null in the latter (except for de-novo mutations).

| Elevated heterozygosity
As almost all known vertebrate asexual lineages are hybrids, they have elevated heterozygosity compared to parental species (note that this is assuming orthologous sequences are present in both parental species; if not, individuals would have elevated hemizygosity). What is perhaps more surprising is that this heterozygosity is generally retained (Warren et al., 2018), suggesting that, in asexual vertebrates, meiosis is usually modified in ways that maintain heterozygosity. While mitotic parthenogenesis, endoduplication and oogonial fusion maintain heterozygosity, central and terminal fusion would lead to a progressive loss of heterozygosity, and gamete duplication would result in an immediate loss of heterozygosity. As mitotic parthenogenesis has never been reported in vertebrates, egg production in asexually reproducing vertebrates is thought to proceed mostly by either premeiotic endoduplication or oogonial fusion (Neaves & Baumann, 2011). Altogether, these observations suggest that elevated heterozygosity may be advantageous in asexual vertebrates, perhaps by buffering deleterious recessive mutations.

| Deleterious mutation accumulation
Under most types of asexual reproduction, meiotic recombination among homologous chromosomes is disrupted, so linkage disequilibrium is complete. Thus, new allele combinations are not tested and deleterious alleles can no longer be purged from a lineage (Felsenstein, 1974;Hill & Robertson, 1966). As a result, deleterious alleles tend to accumulate, a phenomenon termed Muller's ratchet (Keightley & Otto, 2006;Muller, 1964). Mutations are also predicted to accumulate due to the reduced effective population size of asexual lineages (Balloux, Lehmann, & Meeus, 2003;Orive, 1993).
The adaptive potential of asexual lineages is thus thought to be limited, leading to the classic prediction that asexual lineages should be short-lived, evolutionary dead ends (Maynard-Smith, 1978). Genetic, and more recently, genomic studies, have revealed that asexual lineages do tend to accumulate more deleterious mutations (Bast et al., 2018;Hartfield, 2016). Alternatively, allelic gene conversion (i.e., mitotic recombination, the process by which a sequence replaces the homologous sequence) can accelerate the spread of beneficial mutations in asexual lineages (Mandegar & Otto, 2007) and counteract the accumulation of deleterious mutations, a phenomenon that has been observed in ancient bdelloid rotifers (Flot et al., 2013).
Similarly, a recent genomic study on the gynogenetic Amazon molly (Poecilia formosa) revealed a global lack of degeneration in this rather ancient asexual lineage (~500,000 generations, Warren et al., 2018).
One explanation could be that paternal introgression ("stealing" of DNA fragments from the sexually reproducing male fertilizer, similarly to kleptogenetic salamanders) is more frequent than expected, thus mitigating deleterious mutation accumulation (Warren et al., 2018). Overall, while we do have solid theoretical foundations predicting deleterious mutation accumulation in asexual lineages, empirical data remains equivocal. More studies are clearly needed to understand the persistence of asexual lineages despite their predicted genomic decay.

| Transposable element load
Transposable elements (TEs) are selfish genetic elements that can multiply within a genome and colonize new genomes, thus increasing in frequency in a population, without providing any benefit to their host (Doolittle & Sapienza, 1980;Orgel & Crick, 1980). These properties lead to the hypothesis that asexual lineages should have a reduced TE load (Hickey, 1982), as, in the absence of sex, TEs can no longer colonize new genomes, and their evolutionary fate is then tightly linked to the survival of their host lineage. This prediction has received support in yeast, where Bast, Jaron, Schuseil, Roze, and Schwander (2019) found that asexual reproduction appears to select for an increased excision rate of TEs, leading to a reduced TE load over time. In asexually reproducing vertebrates, two observations complicate this prediction: (a) TEs tend to accumulate in nonrecombining regions of the genome (Bachtrog, 2003; Charlesworth, Jarne, (b) TE de-repression (i.e., reactivation or mobilization) has repeatedly been observed in interspecific crosses (Arkhipova &

| Genome architecture
If meiosis is lost, then global genome organization (e.g., the karyotype and gene synteny) and other structures important for meiosis (e.g., meiotic proteins) may degenerate, as a result of reduced constraints and relaxed selection, respectively (Judson & Normark, 1996). In particular, ectopic recombination should be less deleterious in asexual lineages in which constraints on homologous chromosome alignment during meiosis are gone (due to sister chromosome alignments instead of homologous chromosome alignment during meiosis 1). One might then predict asexual lineages to have highly rearranged genomes. Indeed, this has been observed in asexual nematodes closely related to Caenorhabditis elegans that all have a single chromosomes (compared to 5-7 chromosomes in related sexually reproducing species; Fradin et al., 2017). The genome of one of these species has been sequenced, revealing a highly rearranged single chromosome with high level of heterozygosity (note that the two alleles are now on the same chromosome) that arose by the fusion of the six ancestral chromosomes (Fradin et al., 2017). A similarly highly rearranged genome, incompatible with conventional meiosis, has also been documented in ancient bdelloid rotifers (Flot et al., 2013), and high rates of genome rearrangement have also been reported in parthenogenetic aphids (Blackman, Spence, & Normark, 2000). Yet, Majtánová et al. (2016) reported no increased rate of chromosomal change in asexual Cobitis fish. The question of whether asexual reproduction yields to an increased rate of genome rearrangements in vertebrates therefore remains open. sequence and may influence phenotype by suppressing or facilitating the expression of associated genes (Feil & Fraga, 2012). Epigenetic variation spans a spectrum from being fully genetically controlled to being fully autonomous from DNA sequence variation, and in this review we focus only on fully DNA-dependent and fully autonomous epigenetic mechanisms for simplicity (see Richards, 2006 for a further discussion). From an evolutionary standpoint, DNA-dependent epigenetic variation acts as all other genetically based changes (Richards, 2006), and may result from the mechanisms discussed in the previous section (see Section 2.2. "The genomic consequences of asexuality"). However, autonomous epigenetic modifications may further explain phenotypic variability, and potentially heritability, beyond what can be understood by studying genetics alone, so better characterizing the causes and consequences of this type of epigenetic change has become a current focus in evolutionary biology (Verhoeven et al., 2016).

| The effect of asexual reproduction on vertebrate epigenetic variation
Prior to discussing the potential effects of asexuality on autonomous epigenetic variation, it is important to describe the types of non-DNA-dependent epigenetic modifications that may occur.
Autonomous epigenetic modifications can be classified as randomly animals have evolved to sense and respond appropriately to environmental variability (Feil & Fraga, 2012). However, targeted epimutations can also have a negative effect on fitness, as evidenced by the environment-induced epigenetic changes leading to human diseases (Cavalli & Heard, 2019). On the other hand, higher rates of untargeted or stochastic epimutations are predicted to be similar to random mutations and be associated with bet-hedging; stochastic epigenetic modifications may be induced by stressful environments or unpredictable environments (Vogt, 2017).
Epigenetic changes can also be classified based upon their persistence. Some epigenetic modifications may persist through mitosis (intragenerational inheritance), through meiosis to the subsequent generation (intergenerational inheritance or meiotic epigenetic inheritance), and even through meiosis to future generations that have never experienced the stimulus leading to the initial epigenetic change (transgenerational epigenetic inheritance; Best et al., 2018;Skvortsova et al., 2018). However, the likelihood of transmission decreases from intra-to inter-to transgenerational inheritance, as substantial epigenetic resetting occurs during gamete formation and development in most eukaryotes (reviewed by Burggren, 2015;Feng, Jacobsen, & Reik, 2010). In this section, we separately discuss the immediate (intragenerational effects) and longer term (inter-and transgenerational effects) of asexuality on epigenetic modifications and clarify when we discuss DNA-dependent or autonomous epimutations.

| Immediate effects of asexual reproduction (within a generation)
Many of the immediate epigenetic effects of a transition to asexual reproduction in vertebrates likely stem from the lineage-forming interspecific hybridization event. As noted in the preceding section on "the genomic consequences of asexuality," novel parental genome interactions in hybrids may lead to DNA-dependent changes in epigenetic marks, including a decrease in DNA methylation that can initiate transposable element derepression (e.g., O'Neill et al., 1998). Studies in plants suggest that other epigenetic markers may also vary after hybridization-induced asexuality, such as histone acetylation and methylation (e.g., Jiao et al., 2018;Madlung & Wendel, 2013;Shi, Zhang, Ko, & Chen, 2015 Thus, a wider range of epigenetic marks such as histone modifications should be examined in vertebrates of similar ploidy to clarify the potential effects of hybridization, and hybridization-induced asexuality on epigenetic variation. A transition to asexuality may also increase the relative importance of phenotypic plasticity resulting from epigenetic variation as a mechanism leading to population persistence in a variable environment (Angers, Castonguay, & Massicotte, 2010;Castonguay & Angers, 2012;Verhoeven & Preite, 2014). Phenotypic plasticity, which can be defined as the ability of "individual genotypes to produce different phenotypes when exposed to different environmental conditions" (Pigliucci, Murren, & Schlichting, 2006) (Massicotte & Angers, 2012;Massicotte et al., 2011). In particular, environment-directed epimutations lead to most variation among predictable environments, while randomly occurring epimutations predominated in more variable environments . These data match theoretical predictions that environment-directed epimutations contribute more to phenotypic plasticity and randomly occurring epimutations contribute more to diversifying bet-hedging . To determine if epigenetic modifications make a larger contribution to beneficial plasticity in asexual versus sexual fish, we require further information about epimutation variation in clones compared to sexual parents and must test the effects of candidate epimutations on phenotypes and fitness.

| Longer term effects (over generations)
The genomes of most asexual clones do not undergo recombination among homologous chromosomes, so cannot purge deleterious mutations or combine new potentially beneficial genetic combinations (as reviewed in Section 2.2.3 "Deleterious mutation accumulation").
Thus, clonal lineages with higher beneficial phenotypic plasticity (general purpose genotype; Baker, 1965), may outcompete other lineages with lower beneficial plasticity in variable environments (Lynch, 1984). Epigenetic differences among clonal lineages are one potential mechanism generating this variation in plasticity. Thus, in variable environments, clonal selection may result in the dominance of lineages with higher epigenetic variation, which may be either DNA-based or autonomous (Verhoeven & Preite, 2014). However, it is not yet clear if this clonal selection results in a higher level of epigenetic variation in successful asexual lineages compared to sexual congeners.
Changes in meiosis during gamete production in clones could also influence the stability of epigenetic markers if clones modify the normal resetting of epigenetic marks that occurs during this process (Verhoeven & Preite, 2014

| Asexual fish lineages as model organisms
Naturally This is because experiments testing the factors leading to variation of autonomous epigenetic modifications must control for background genetic variation, which can be done by using clone mates (reviewed by Best et al., 2018;Bossdorf, Richards, & Pigliucci, 2008;Hu & Barrett, 2017;Laskowski et al., 2019;Verhoeven & Preite, 2014;Vogt, 2017). For these studies, parthenogenic or gynogenetic animals are required, because sibling hybridogens will genetically differ due to the incorporation of meiotically produced sperm from sexual fathers (reviewed by Avise, 2008;Laskowski et al., 2019).
Alternatively, hybridogens might be quite useful for studies testing for allele-specific effects on DNA-dependent epigenetic regulation in a common maternal genetic background (Laskowski et al., 2019).
While all unisexual vertebrate lineages can be used to study the genomic causes and consequences of asexuality and the factors influencing epigenetic variation, ray-finned fishes may be especially tractable experimental animals (Figure 3; reviewed by Best et al., 2018;Franěk et al., 2019;Laskowski et al., 2019;Vrijenhoek, 1994). This is because most asexual fishes are oviparious and have external fertilization, with the exception of asexual poecillids (Figure 3; Avise, 2008;Lamatsch & Stöck, 2009). So, when testing the factors leading to autonomous epigenetic change, it will be possible to more tightly control other, potentially confounding, early-life environmental factors than in viviparous species with parental care (reviewed by Best et al., 2018;Laskowski et al., 2019;Verhoeven & Preite, 2014;Vogt, 2017). As well, the use of oviparous animals allows detection of transgenerational epigenetic effects a generation earlier than in viviparous animals. This is because pregnant, viviparous females (F0 generation), contain the F1 generation, which also contains the germ cells that contribute to the F2 generation, so these will be directly exposed to any epigenetic inducer. Therefore, true transgenerational effects cannot be detected until the F3 generation (reviewed by Best et al., 2018). However, in oviparous animals, transgenerational effects can be detected in F2's, facilitating studies of epigenetic stability (reviewed by Best et al., 2018).
All well-characterized clonal fish are hybridogens or gynogens that require sperm from a parental species or another congener to activate embryonic development ( Harrington, 1961), and congener Kryptolebias ocellatus (Tatarenkov, Lima, Taylor, & Avise, 2009), are highly inbred, almost entirely homozygous, and thus essentially clonal. Therefore, the mangrove rivulus shares many of the benefits of asexual clonal lineages as models in which to examine the factors influencing autonomous epigenetic variation (Berbel-Filho et al., 2019;Fellous et al., 2018). Mangrove rivulus also have some additional perks, because crosses can be made to conduct quantitative genetic studies and facilitate the mapping of genotype to phenotype (Kanamori et al., 2016;Kelley et al., 2016). It is quite possible there are still undiscovered asexual fish lineages, which may be identified by features such as a highly female skewed sex ratios or a high number of F1 hybrids without evidence of backcrossing (Avise, 2008;Beukeboom & Vrijenhoek, 1998). Furthermore, for experimental questions where natural clonal lineages are not required or tractable, there are many artificial fish asexual and isogenic lines currently available, and the potential for more to be created (e.g., Franěk et al., 2019;Shao et al., 2018;Spivakov et al., 2014).

| Asexual, clonal lineages of banded (F. diaphanus) and Common Killifish (F. heteroclitus)
The and relatively low-cost and easy care in the laboratory (reviewed by Burnett et al., 2007;Dawley, 1992  this data to representative clones from 2004-2007 (see Data S1 for further information). We found that the same major clonal lineage has remained dominant over the past decade (Clone A; Figure 5). We also found evidence of what may be subsequent mutations in this lineage (Data S1; Clones A.2 to A.5) and that some of the less common lineages have also persisted over the past decade in Porter's Lake ( Figure 5, Clones B-K). Finally, we found some potential new clonal lineages ( Figure 5, Clones L-P) not detected in -2007(Hernández Chávez & Turgeon, 2007Mérette et al., 2009 (Dawley et al., 1999;Fritz & Garside, 1974a), suggesting that prezygotic factors, such as mate choice, and not intrinsic genetic incompatibilities, are likely contributing to this bias in cross direction in the wild.
The majority of hybrids found in Porter's Lake and the Saint Mary's River sites also appear to be F1 clones, but there is evidence that a small proportion of sexual hybrids might exist (Dawley et al., 1999;Hernández Chávez & Turgeon, 2007;Mérette et al., 2009).
In vitro crosses of Porter's Lake fish resulted in offspring of both sexes, and males survived to adulthood, but it is not yet known if these offspring are capable of sexual reproduction (Fritz & Garside, 1974a A common prediction for clonal lineages is that the reduction in population size and lack of recombination will lead to mutational meltdown, making them "evolutionary dead ends" (Lynch, Bürger, Butcher, & Gabriel, 1993). However, this process may take 10 4 -10 5 generations to manifest (Lynch & Gabriel, 1990), and the high heterozygosity found in most hybrid clonal vertebrate genomes may buffer against the lack of new genetic diversity from meiotic recombination (Warren et al., 2018). The exact time of origin of the F. diaphanus x F. heteroclitus clonal lineages is not known, but they might have formed as recently as 70 years ago in Porter's Lake when this lake was first connected to the ocean, allowing the more freshwater tolerant F. diaphanus and brackish/salt water-inhabiting F. heteroclitus to come into contact (Fritz & Garside, 1974b). Population genetic analysis of the mitochondrial genome also suggests that asexual hybrid origin is recent, as clonal haplotypes are at the edges of the mitochondrial network (Hernández Chávez & Turgeon, 2007). If these hybrids formed in their current location, they must be younger than ~12,000 years, as these regions of Nova Scotia were covered by the Wisconsinian glaciers until this point (Shaw et al., 2006). Further population genetic analysis is needed to definitively age clones, as has been done for other species [e.g., Poeciliopsis, (Quattro, Avise, & Vrijenhoek, 1992;Stöck et al., 2010)], but these populations are predicted to be too young to be experiencing mutational meltdown (Lynch & Gabriel, 1990). While measures of fitness have not been compared between parental species and the F1 clones, the F1 hybrids seem to have parasite loads (King, 2009) and salinity tolerances (Jonah, 2019) intermediate to that of their parent species, suggesting there is no major hybrid dysfunction in adult clonal females, beyond their asexuality. Nor does there appear to be any major heterosis or increased plasticity compared to parental species (Jonah, 2019), as would be predicted if the dominant clone was a true general-purpose genotype (Baker, 1965).

| Current research-What are the genomic causes and consequences of a transition to asexuality?
According to the "balance hypothesis," asexual reproduction can evolve when species have accumulated enough divergence to disrupt meiosis, yet not enough to compromise viability of their hybrids (Janko et al., 2018;Moritz et al., 1989). Therefore, we are working to identify the proximate mechanisms leading to asexual lineage formation in Fundulus hybrids. Specifically, we are interested in understanding how sequence divergence and global genome architecture may interact to destabilize meiosis in Fundulus hybrids. Disentangling the consequences of asexual reproduction in vertebrate hybrids is a tricky endeavor. The Fundulus system, with the likely presence of both sexually and asexually reproducing hybrids, allows us to use a comparative framework; this may be an extremely powerful way to distinguish between the consequences of hybridization alone and the combination of hybridization and asexual reproduction.

| How does genome architecture differ between F. heteroclitus and F. diaphanus?
Genome architecture can be defined as "the totality of non-random arrangements of functional elements in the genome" (Koonin, 2009), in which we include the karyotype (chromosome number, arrangement and centromere position), the amount and distribution of repetitive elements (e.g., transposable elements and satellite repeats), as well as gene synteny. Genome architecture directly influences many processes of interest to evolutionary biologists, such as recombination rate patterns, adaptive introgression and speciation (Lynch, 2007). As genome architecture is directly linked to chromosome alignment during meiosis, it is critical that we gain a better sense of genome organization in Fundulus to understand hybridization outcomes and the transition to asexuality.
Fundulidae have rather conserved karyotypes (typically 2N = 46-48, NF = 48-52;Arcement & Rachlin, 1976;Chen, 1970, Chen, 1971), yet F. diaphanus has a genome size that is ~10% larger than F. heteroclitus (Dawley, 1992). This suggests a higher repetitive element load in F. diaphanus, which could result from reduced selection efficiency against selfish genetic elements (assuming F. heteroclitus has a larger effective population size, being a more widely distributed marine species, Reid et al., 2017). We will perform a comparative study of the genomes of F. heteroclitus and F. diaphanus, with a particular emphasis on repetitive elements, which are known to correlate well with genome size in eukaryotes (Dufresne & Jeffery, 2011;Lynch & Conery, 2003).  Symonová et al., 2013). It will also be interesting to contrast these results with that seen in other species of Fundulidae, such as F. grandis, which also hybridize with F. heteroclitus, but produce fertile, sexually reproducing F1 offspring (Barbas & Gilg, 2018).

| Do sexually reproducing hybrids exist and do specific genotype combinations result in asexual reproduction?
Long thought to be a maladaptive process, hybridization is now recognized as a frequent source of beneficial genetic novelty in animals (Abbott et al., 2013;Harrison & Larson, 2014;Runemark, Vallejo-Marin, & Meier, 2019). Thus, sexually reproducing Fundulus hybrids may provide a conduit for new adaptive alleles to enter the gene pools of the parental species (e.g., Oziolor et al., 2019). For example, higher freshwater tolerance in Northern versus Southern populations of F. heteroclitus (Scott et al., 2004), is hypothesized to be the result of adaptive introgression of "freshwater tolerance alleles" from the freshwater-preferring F. diaphanus (Hernández Chávez & Turgeon, 2007). A potentially confounding factor for the classification of hybrids from Porter's Lake might be the action of mutational processes and gene conversion at microsatellite loci, which would either increase or decrease heterozygosity, respectively. Genomic data should help to exclude such technical issues, clarify the presence of sexually reproducing hybrids in Porter's Lake, and determine if sexual hybrids act as an introgression vehicle for the transfer of novel alleles into either of the parental species (e.g., Edelman et al., 2019). If the presence of sexually reproducing hybrids is confirmed, the Fundulus spp. system could also be an excellent comparative system to study the genetic factors that result in the production of sexual versus asexual hybrid offspring.
As noted in Section 3.2 "Asexual, clonal lineages of Banded (F. mitochondrial haplotype, it seems likely that this asymmetry is a consequence of some ecological factors (e.g., female mate choice) and not due to major cytonuclear incompatibilities. We are currently carrying out controlled laboratory crosses to investigate female mate choice, male-male competition, and salinity tolerance in conspecific and reciprocal hetero-specific crosses to determine the reasons for this bias in hybridization direction.

| How is meiosis modified in Fundulus asexual hybrids?
Asexually reproducing vertebrates produce their eggs through some e.g., Neaves and Bauman, 2011;Newton et al., 2016). We aim to use a combination of cytogenetic approaches to investigate the mode of egg production in asexual hybrids compared to parental species.
Specifically, we will use histological sections of ovaries combined with semi-quantitative DAPI staining to determine the DNA content of germ cell throughout egg formation (Newton et al., 2016). This should allow us to determine at which stage the genome is duplicated, therefore maintaining ploidy in these clonal lineages. In addition, we will develop immunofluorescence techniques that, in combination with cytogenetic approaches, should allow us to determine whether sister chromosomes pair and recombine (Kuroda et al., 2018;Pereira et al., 2013), thus maintaining heterozygosity in asexual Fundulus lineages (Hernández Chávez & Turgeon, 2007).

| Do asexual lineages have higher rates of genome rearrangements and repetitive DNA accumulation?
As discussed in Section 2.2 "The genomic consequences of asexuality," meiosis modification releases constraints on global genome  (Chen, 1971). a, acrocentric; sm, sub-metacentric; st, sub-telocentric architecture as homologous chromosomes are no longer required to align during meiosis I ( Flot et al., 2013;Fradin et al., 2017). In addition, transposable elements, especially if mobilized as a result of hybridization, may serve as substrate for recombination between nonhomologous loci, thus resulting in chromosome rearrangements (Deininger, Moran, Batzer, & Kazazian, 2003). Asexual lineages may also lose male-specific genes or genes involved in meiosis that are no longer needed (Yin et al., 2018, but see Warren et al., 2018, or display high rates of transposable element excision (Bast et al., 2019) that would result in genome size reduction over time (this is especially true in fishes that contain high loads of "cut-and-paste" DNA transposons; Symonová & Suh, 2019  see Section 2.2.4 "Transposable element load"). We might expect hybrids to have a larger genome, but, in nematodes, reproductive mode shifts have been associated with genome shrinkage, probably due to the loss of male-specific genes (Yin et al., 2018). Combined with cytometry, sequencing data will allow us to test these predictions in asexually reproducing Fundulus. The comparison between asexually and sexually reproducing hybrids will be of the upmost importance to disentangle the consequences of hybridization and asexual reproduction.

| Do asexual lineages have higher rates of deleterious mutation accumulation?
Asexual organisms are predicted to accumulate deleterious mutations due to the lack of recombination (Hill & Robertson, 1966), reduced effective population size (Balloux et al., 2003;Orive, 1993) and Muller's ratchet (Felsenstein, 1974;Muller, 1964), a phenomena that has been confirmed in asexually reproducing Daphnia, freshwater snails and stick insects (Bast et al., 2018;Neiman, Hehman, Miller, Logsdon, & Taylor, 2010;Paland & Lynch, 2006). A particularity of the clonal Fundulus system is that these are likely evolutionary "young" clonal lineages (potentially <70 years old, see Section 3.2 "Asexual, clonal lineages of Banded (F. diaphanus) and Common Killifish (F. heteroclitus)"; Fritz & Garside, 1974a). While this means Fundulus spp. clones in Porter's Lake might not have had enough time to accumulate deleterious mutations, it potentially allows us to witness the very first deleterious mutations that accumulate, even before fixation. We will test for nonsynonymous, frameshift and nonsense mutation accumulation in RNA-sequencing and wholegenome sequencing datasets generated to answer other questions.
Preferential expression of one parental allele over the other may also represent a mechanism to mitigate deleterious mutation expression (see following section).

| How is the transcriptome remodeled in asexually reproducing hybrids?
Gene expression differences may lead to significant phenotype changes even between closely related species (Pavey, Collin, Nosil, & Rogers, 2010). In hybrids, the interaction of two divergent genomes can lead to novel patterns of gene expression that can generate novel phenotypes (e.g., "hybrid vigor"; Landry, Hartl, & Ranz, 2007) and even underlie reproductive isolation (Mack & Nachman, 2017). Recent work investigating gene expression changes across five asexual stick insects species has revealed repeated convergent changes of genes involved in meiosis in the reproductive tract (Parker et al., 2019a). More surprisingly, repeated masculinization (increased expression of male-biased genes) was also observed in these stick insect species (Parker et al., 2019b). We will ask whether and how transcriptional networks are remodeled in asexual Fundulus hybrids. We are particularly interested in testing for parental allele bias among expressed genes as a mechanism to mitigate deleterious allele accumulation, which can be tested by comparing wild asexual individuals to controlled crosses. In addition, it will be exciting to look at gene expression profiles in ovaries as well as in tissues that may regulate important responses to environmental stressors, such as the gills, which must respond to changes in salinity as a primary site of osmoregulation. Finally, these transcriptomic datasets will allow us to test directly for transposable element derepression in these asexual hybrids.

| Current research-What epigenetic mechanisms lead to variation in animal performance?
A major aim of comparative physiology is to determine the mechanisms that contribute to differences in animal performance within and among individuals, populations and species in the wild (Mykles, Ghalambor, Stillman, & Tomanek, 2010). One mechanism that we still have much to learn about is the potential contributions of autonomous epigenetic variation to differences in animal function and fitness (reviewed by Burggren, 2015;Burggren & Crews, 2014;Hu & Barrett, 2017;Verhoeven et al., 2016). In particular, we need to learn more about: (a) how specific environmental factors influence the epigenome, (b) how the resulting epigenetic variation influences phenotype, performance and fitness and, (c) the stability of epigenetic inheritance (Hu & Barrett, 2017;Skvortsova et al., 2018;Verhoeven et al., 2016). In this section, we discuss how studies in clonal killifish hybrids might contribute to these research questions.

| How does the environment influence the epigenome?
To learn more about how the environment may influence autonomous epimutations, it is critical to control for DNA-dependent epigenetic effects (Bossdorf et al., 2008;Laskowski et al., 2019). Indeed, inbred lines of vertebrates have been the model of choice for medical physiologists testing the links between autonomous epigenetic modifications and disease (Cavalli & Heard, 2019). However, ecological physiologists interested in animal performance and fitness in the wild may prefer to study animals that are "the products of natural selection" (Laskowski et al., 2019;Vogt, 2017

| How does epigenetic variation influence phenotype and performance?
Over the past decades, studies in genetic model organisms have shown that epigenetic modifications can lead to changes in gene expression that may translate to differences in organismal phenotypes (Feil & Fraga, 2012). However, there is still much to learn about the linkages between specific epigenetic marks and phenotype, even in genetic model organisms (Cavalli & Heard, 2019;Duncan et al., 2014). Furthermore, studies in natural populations have mainly focused upon genome wide changes in DNA methylation (Hu & Barrett, 2017;Verhoeven et al., 2016), and have rarely studied epigenetic variation known to be autonomous (but see Berbel-Filho et al., 2019;Fellous et al., 2018;Thorson et al., 2017).
Thus, we cannot yet accurately predict the effects of specific autonomous epimutations on phenotype in natural populations, and know much less about the phenotypic effects of other types of epimutation beyond DNA methylation in wild animal populations (Hu & Barrett, 2017).
Linking specific epimutations to phenotype will be difficult, as many epigenetic changes will occur in concert, so establishing causation will require more than correlational studies. Even in clonal lineages which have no differences in DNA-dependent epigenetic modifications, the plethora of autonomous epigenetic modification will be hard to tease apart. One option will be to conduct epigenetic editing of candidate sites and then test for effects on gene regulation and candidate phenotypes (Josipović et al., 2019;Liu et al., 2016). Furthermore, as we learn more about how specific epigenetic marks might influence gene expression and phenotypes in natural vertebrate populations, a later goal should be to study combined profiles of different types of epigenetic marks (e.g., DNA methylation, specific forms of histone modifications, and small RNAs) that may interact to influence phenotype (Burggren, 2015). In terms of the Fundulus asexual lineages, a particularly interesting question is if epigenetic factors (DNA-dependent or autonomous) contribute to the differences in the production of sexual versus asexual hybrid offspring in different populations (Hernández Chávez & Turgeon, 2007).

| How stable are epigenetic marks?
To determine the evolutionary role of autonomous (non-DNA-dependent) epigenetic modifications in vertebrates, we must first gain a better understanding of the persistence of epigenetic marks (Lind & Spagopoulou, 2018;Skvortsova et al., 2018;Verhoeven et al., 2016). In the past decade, it has become clear that certain epigenetic modifications produced as a result of environmental factors in plants and some invertebrates can persist, and continue to modify gene expression and physiology long after the initial stressor has ended (reviewed by Hollick, 2017;Quadrana & Colot, 2016;Serobyan & Sommer, 2017;Skvortsova et al., 2018). Furthermore, these modifications can contribute to local adaptation (Schmid et al., 2018).
However, the extent to which epigenetic modifications are transmitted across generations in vertebrates, which undergo more substantial germline epigenetic reprogramming, is less clear (Heard & Martienssen, 2014;Hu & Barrett, 2017;Lim & Brunet, 2013;Perez & Lehner, 2019;Radford, 2018;Skvortsova et al., 2018). Recent studies suggest that the answer to this question may be taxa specific: Anne-I was a postdoctoral fellow in the Bernatchez laboratory from 2013-2016; this was a very exciting time, as there were at least eight postdoctoral fellows, fifteen graduate students and three technicians working with Louis! This group included diverse array of biologists conducting research in population genetics, evolutionary biology and molecular ecology (Louis' core research themes), along with some physiology, molecular biology, genomics, and conservation biology. Furthermore, these scientists were interested in questions in both pure and applied research and came from a range of nationalities and first languages. While I had not left Canada (my home country), my time in the Bernatchez Lab and Quebec City felt like a truly international experience. I was able to experience the bumps of learning a new language and culture in an intellectually fertile, supportive and fun working environment. The ability to form teams of interdisciplinary researchers and provide them with the tools and freedom to develop new ways of thinking is one of Louis' many skills. I will forever be amazed at how he was able to track progress on the dozens of projects in his lab, keep up with editorial responsibilities, and organize parties-he is a force to be reckoned with! The working relationships and friendships developed in asexuals compared to sexuals will allow evolutionary biologists to test the hypothesis that epigenetic modifications play an important role in facilitating asexual hybrid lineage persistence (Castonguay & Angers, 2012;Leung et al., 2016;Massicotte et al., 2011;Verhoeven and Preite, 2014;Vogt, 2017). Comparing the extent of autonomous and DNA-dependent transgenerational epigenetic inheritance among more (e.g., Clone A) and less dominant killifish clonal lineages in Porter's Lake (e.g., Clones B-L) may also clarify the contributions of each type of epigenetic variation in clonal lineage success.

| SUMMARY
The field of comparative physiology has been strongly influenced by Krogh's Principle, that "for a …number of problems there will be some animal of choice, or a few such animals, on which it can be most conveniently studied" (Krogh, 1929). As comparative physiologist work to understand the importance and mechanisms of epigenetic regulation on phenotypic variability (reviewed by Burggren & Crews, 2014), the ability to use genetically identical clones may be an ideal "Kroghian model" to control for the potentially interacting effects of genetic variation while maintaining ecological relevance (Laskowski et al., 2019). We hope that studying clonal Fundulus spp.
that derive from Common Killifish, a well-established model species in comparative animal physiology (reviewed by Burnett et al., 2007;Crawford et al., 2020), will provide us with an opportunity to merge a deep knowledge of physiological mechanisms with potential underlying epigenetic variation.
Furthermore, fish of the family Fundulidae have become a key model system in which to study the genetic basis for adaptation (e.g., Brennan et al., 2018;Crawford et al., 2020;Oziolor et al., 2019) and speciation (e.g., Barbas & Gilg, 2018;Fuller, McGhee, Schrader, 2007) in natural populations. By combining genomic and cytogenetic data from F. diaphanus x F. heteroclitus clones with currently available genomic resources for the Fundulidae clade (e.g., Johnson et al., 2020;Reid et al., 2017), we hope to better understand the mechanisms by which genomes are regulated and evolve in response to hybridization and asexual reproduction.

ACK N OWLED G EM ENTS
We would like to thank Dr. Louis Bernatchez for welcoming us both into his laboratory and encouraging our personal research interests, which led to this collaboration (see Box 1

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The microsatellite genotypes from this study are provided as Appendix S1.