Extragenomic developmental information: epigenetics and environment
These insights into gene regulation suggest a very different view of the developmental role of an individual's DNA sequence than Watson and Crick could have envisioned. Instead of a coded text consisting of pure bits of information that are faithfully transmitted across generations, the genome now appears as a stunningly dynamic network of biochemical interactions—not a "blueprint" for a determined phenotypic outcome but rather an agile "information management system" (Pagel and Pomiankowski 2008). And yet this system is not bounded by the nucleus but is itself embedded in a larger and even more complex informational context. To the many interacting genomic factors that influence phenotypic expression must be added three further sources of developmental information: epigenetic processes and the organism's cellular and external environments.
Variability and heredity expanded: epigenetic regulation of gene expression
Epigenetic effects at the molecular level can be defined as biochemical mechanisms that shape patterns of gene expression in the absence of any change in nucleotide sequence; these mechanisms act primarily by altering the accessibility of DNA to the transcriptional machinery (C.-T. Wu and Morris 2001; Bateson and Gluckman 2011; see Duncan et al. 2014 for a brief overview).2 Although initially seen as a rare embellishment to Mendelian variation, early whole-genome studies in "model" organisms such as Arabidopsis and in humans revealed molecular epigenetic modifications to be unexpectedly widespread (e.g., Heintzman et al. 2007; X. Zhang et al. 2007). These modifications are now understood to be a fundamentally important source of variation in gene expression. In the past decade, a remarkably exciting, if bewildering, array of epigenetic mechanisms has been uncovered. To date, such mechanisms are known to include DNA methylation and hydroxymethylation (Holliday and Pugh 1975; Cedar and Bergman 2009; P. Jones 2012), histone modifications such as acetylation and phosphorylation (Bannister and Kouzarides 2011), and diverse activities of small RNA molecules (Ha and Kim 2014; K. Morris and Mattick 2014), but the list is rapidly expanding as new phenomena come to light. Although biologists are just beginning to understand these novel regulatory processes, it is clear that the epigenome is an essential source of developmental information within and across generations, in both unicellular and multicellular taxa (Jaenisch and Bird 2003).
Epigenetic mechanisms were initially investigated as regulators of cell differentiation during ontogeny (Reik 2007). In many organisms, methy- lation and histone modifications play an important role in the correct timing of tissue-specific gene expression patterns as ontogeny proceeds (Heard 2013; Boland et al. 2014), and multiple epigenetic effects shape the developing brain in vertebrates (Meaney and Ferguson-Smith 2010). Substantial epigenetic developmental information is evidently transferred to offspring via sperm cells in humans and other mammals, and via pollen in plants (Hammoud et al. 2009; Herrera et al. 2014; Ihara et al. 2014). Epigenetic mechanisms also underlie the process of genomic imprinting, a complex mode
2 Alternatively, the term epigenetics is sometimes used in a broader sense to include the entire environmental and cellular context of development (Waddington 1957; see discussion of alternative definitions in Bird 2007).
of inheritance in which certain genes are expressed in a parent-of-origin-specific manner and which is common in mammals, flowering plants, and insects (MacDonald 2012). The discovery of genomic imprinting demonstrated that gene expression patterns could be inherited based on epigenetic silencing of one set of alleles, depending on the gender of the parent that contributed them, without any change in DNA sequence. In mammals, imprinted genes are often central to fetal development (Gluck- man et al. 2007; Plasschaert and Bartolemei 2014).
In addition to mediating characteristic ontogenetic trajectories, epigenetic changes that affect phenotypes can also be variably induced in individuals, either by specific environmental factors or spontaneously (references in Jaenisch and Bird 2003; Gluckman et al. 2007; Nelson et al. 2012; Duncan et al. 2014; also see Section 1.2.2 on epigenetic mechanisms and environmental response). Although work in natural systems has been limited to date, epigenetic modifications are evidently a source of substantial phenotypic variation in nature that does not arise directly from allelic variants (e.g., Herrera and Bazaga 2010; Lira-Medeiros et al. 2010; Becker et al. 2011; C. Richards, Schrey, et al. 2012). Because epigenetic effects on gene expression can be maintained across mitotic and, in some cases, meiotic divisions, they can be long term, but to varying degrees (Gill et al. 2012; Turck and Coupland 2013; Herrera et al. 2014). These biochemical modifications are well documented to persist through cell lineages within an individual and (when they occur in germ cells or in organisms with no segregation between germ and somatic cells) through several but not all subsequent generations; at some point they are biochemically "reset" (Holliday 1990; Johannes et al. 2009; Burton et al. 2011; Seong et al. 2011). Although such reversals can be experimentally induced by environmental or dietary changes that undo DNA methylation or histone modifications, it is not yet known whether in natural systems such reversals are typically stochastic or are induced by specific conditions, or how long epigenetic modifications persist outside of the laboratory (Klironomos et al. 2013; Duncan et al. 2014; Heard and Martienssen 2014). Even laboratory data on long-term epigenetic persistence are uncertain, since the reported number of generations may simply reflect the duration of an experiment rather than an observed reversal.
Evidence for inherited epigenetic effects on human health has opened a major new avenue for biomedical investigation. Epigenetically mediated effects of poor maternal nutrition on offspring development and physiology may strongly influence the incidence of adult obesity and cardiovascular disease (Gluckman et al. 2009; Jimenez-Chillaron et al. 2012; and references therein); heritable epigenetic variation has been implicated in other disease states, including cancer (Nelson et al. 2012; Varley et al. 2013). In studies with mice, dietary factors such as folate that directly affect the biochemical process of methylation create, in the subsequent generation, paternal effects that are associated with developmental malformations. These effects are caused by diet-induced changes to sperm-cell epi- genomes; such changes can take place either when males are in utero (and the maternal diet is manipulated) or during adult spermatogenesis (Lambrot et al. 2013). Indeed, the prevalence and impact of these nonallelic (and potentially environmentally induced) transgenerational effects on individual health may be a key reason why genome-wide association studies have shown such limited success in identifying DNA sequence variants that in themselves explain the distribution and inheritance of disease phenotypes (Miklos 2005; Manolio et al. 2009; Slatkin 2009; Drong et al. 2012; Mattick 2012; Nelson et al. 2012). More broadly, a great deal remains to be learned about both the induction and resetting of these various effects and their modes of heritable transmission in various systems (Lam- brot et al. 2013; Herman et al. 2014). It has just recently been shown that regulatory RNA molecules released from somatic cells (in this case, human tumor cells grafted into mice) mediate the transfer of information via the bloodstream to germ cells and thence to sperm, thus providing a route for molecular epigenetic information to become heritable in a mammal (Cossetti et al. 2014; see Soubry et al. 2014 for a review of paternal transmission of epigenetic marks).
Although in some cases epigenetic effects on gene expression can be heritable for between a few to several dozen generations or more (Jablonka and Raz 2009; Mattick 2012), their persistence varies because they are also reversible. In other words, they simply "do not adhere to the rules of Men- delian inheritance" (Lemos et al. 2008, 91; Becker et al. 2011; Cossetti et al. 2014). Indeed, epigenetic modifications of phenotypic expression are sometimes considered to be "Lamarckian" because they can be transmitted to subsequent generations after being acquired. Not surprisingly, then, it has taken decades for these molecular effects to be accepted as a part of mainstream genetics. Contemporary awareness of molecular epigenetics has expanded the neo-Darwinian view of DNA sequence as the fundamental mode of inherited developmental information (Jablonka and Lamb 2002; Mattick 2012), placing even the initial phase of gene expression squarely in a dynamic cellular, organismic, and environmental context.
At the mechanistic level, epigenetic modifications shape gene expression by altering protein- gene interactions that determine the accessibility of DNA to the biochemical machinery of gene transcription. DNA molecules exist in the cell nucleus wrapped around core histone proteins in nucle- osomes, the repeating unit of the chromatin complex. Each histone contains many chemical sites available for posttranslational chemical modifications; in the textual language of genetic information, such sites are termed marks. These marks affect the binding activity of regulatory proteins to the DNA and thus specify whether particular regions of DNA are to be activated or silenced in specific cell lineages. Chromatin enriched for acetylated histones is accessible to transcription factors; as a result, genes located in that chromatin can be transcriptionally active, while genes located in "condensed" chromatin (enriched for nonacety- lated histones) are inaccessible to transcription factors and are thus inactive (Jablonka and Lamb 2005; Bannister and Kouzarides 2011). Numerous studies have shown that increased methylation at promoter regions also tends to be associated with gene "silencing" or down-regulation, while methy- lation at intragenic regions can be associated with either repression or activation, depending on the particular sequence in which it occurs (P. Jones 2012; Varley et al. 2013).
As noted above, the effects of chromatin modification mechanisms (i.e., DNA methylation and
histone marks) on gene expression can lead to the expression of disease phenotypes. The potential medical importance of these mechanisms is particularly clear with respect to the malfunction in mitotic regulation that underlies cancer. Many tumor cells show aberrant, cell-heritable epigenetic modifications that silence tumor-suppressing genes (Jablonka and Lamb 2002; West and Johnstone 2014). Since human cancers commonly show widespread changes in DNA methylation patterns in early stages, it is hoped that further methylation studies may lead to new techniques for early cancer detection (Kanai and Hirohashi 2007); drugs that alter histone acetylation are already in use as a treatment for certain cancers, with more under clinical trial (West and Johnstone 2014).
Epigenetic mechanisms may also be a heretofore unrecognized source of selectively important phenotypic variation in natural populations. Indeed, the first naturally occurring morphological mutant to be characterized at the molecular level was determined to be an "epimutation" rather than a DNA sequence change (Cubas et al. 1999). In 1749, the Swedish botanist and pioneering systematist Carl Linnaeus described from his own cottage garden outside Uppsala a radially symmetric mutant of "butter-and-eggs," Linaria vulgaris, a common European herb with characteristic bilaterally symmetric, snapdragon-like flowers (Figure 1.4). This radical change in floral morphology can be traced to the lack of expression of Lcyc, a homolog of the cycloidea gene that regulates dorsiventral asymmetry in the related genus Antirrhinum. The lack of Lcyc expression is due not to a mutation in the gene's DNA sequence but instead to heavy methylation at the Lcyc locus, as this methylation transcriptionally silences the gene. This epimutation is heritable but can occasionally revert to the characteristic flower type after several generations, through methyla- tion changes during development that restore Lcyc gene expression (Cubas et al. 1999). This species of Linaria depends on cross-fertilization by honeybees, so this change in floral form and hence pollinator search image could have substantial fitness consequences—an example of an individually mediated developmental phenomenon of potential evolutionary significance.
Figure 1.4 The common European herb Linaria vulgaris produces characteristic bilaterally symmetric flowers (left panel). A naturally occurring "epimutation" of this species with radially symmetrical flowers (right panel) was first described by Unnaeus in 1749. These dramatically altered plants have the same DNA sequence at the floral symmetry locus Lcycbut a different pattern of chromatin methylation. Image courtesy of John Innes Centre, Norwich, UK.
It has become clear that heredity is mediated at the molecular level not purely by discrete, stably transmitted DNA sequence variants but also by multiple information-altering mechanisms that lend the process an unlooked-for flexibility. Qualitatively new modes of cross-generational gene regulation are continuing to be found, including several that show gene silencing and other epigenetic roles for noncoding RNA (Bernstein and Allis 2005; Mattick and Mehler 2008; Lenhard et al. 2012; Ha and Kim 2014). For example, Arabidopsis mutants that are homozygous for certain point mutations at the hothead locus produce a proportion of progeny with wild-type DNA sequences at this locus: evidently the previous nucleotide sequence is restored by an RNA or DNA "archive" (Lolle et al. 2005). The converse has been shown in mice with respect to Kit, which encodes a key cell surface receptor protein that binds various growth factors and hormones. Mutations at this locus show pleiotropic (and sometimes drastic) effects on both development and gamete production. Studies of a
particular Kit mutation that produces a characteristic white-spotted coat revealed that homozygous wild-type mice can produce offspring with this phenotype, even though they lack the mutant allele. This "paramutation" evidently represents information carried over from a previous heterozygote generation via RNA molecules (present in mouse sperm along with the male's haploid genome) that influence embryonic gene expression (Ras- soulzadegan et al. 2006; Chandler 2007; and references therein). Non-Mendelian inheritance can also be mediated directly by noncoding RNAs, as was recently found in the unicellular ciliate Oxytricha trifallax. In this highly polyploid organism, "RNA- guided recombination" directs the reassembly of coding sequences into stably inherited alternative genome arrangements and regulates chromosome copy number in progeny cells (Nowacki et al. 2008; Yao 2008).
Many genomic sequences that were previously considered "junk" are now known to code for small or "micro" RNAs (and possibly long RNAs as well) that play a regulatory role, for instance by altering enzymatic access to the chromatin by binding to DNA (Koziol and Rinn 2010). These noncoding RNAs can also target and disrupt protein-coding mRNA transcripts to silence gene expression in a tissue- or cell-lineage-specific manner (Ha and Kim 2014; K. Morris and Mattick 2014). Small RNAs found near transcriptional start sites apparently interact with promoters to help regulate transcription, although this process is not yet well understood (Lenhard et al. 2012). Interestingly, noncoding RNAs may carry environmentally induced effects on the phenotype from one generation to the next, including the neurobehavioral effects of social environment. In one recent study, traumatic, unpredictable separation of newborn mice from their mothers altered several aspects of microRNA activity in the pups, including in their hippocampi and other brain structures involved in stress responses. These epigenetic changes were associated with different behavioral responses to aversive conditions such as brightly illuminated maze compartments. When sperm RNA from traumatized males was injected into fertilized wild-type egg cells, these phenotypic effects were reproduced in the F2 generation; this result indicates that RNA can contribute to the transmission of stress-induced traits in mammals (Gapp et al. 2014).
Additional mechanistic complexity arises from the ways epigenetic processes interact with each other to shape gene expression patterns. DNA methylation and histone modifications can jointly regulate transcriptional patterns, and noncoding RNAs can direct DNA methylation to particular sites in the genome to cause highly specific effects on gene activity (Meaney and Ferguson-Smith 2010; Duncan et al. 2014; and references therein). Histone modifications and methylation marks can also alter the three-dimensional conformation of DNA in the nucleus so as to physically influence epistatic interactions among genetic loci (Qu and Fang 2013). These "overlapping and interdependent" epigenetic effects (Bateson and Gluckman 2011, 57) verify that phenotypes emerge from the dynamic interplay of different types of regulatory elements and not simply from the presence or absence of particular DNA sequences. Indeed, the very notion of genes as discrete pieces of developmental information has become open to question (Keller 2000; K. Morris and Mattick 2014).
It is not yet clear how these expanding insights to molecular epigenetics will change our understanding of allelic diversity and its heritability. This understanding must take into account not only the partially heritable effects of interacting epigenetic processes on gene expression but also the reciprocal effects of the genome itself on these epigenetic processes. DNA sequence influences the likelihood and precise location of spontaneous or environmentally induced epigenetic marks: for example, singlenucleotide polymorphisms can alter particular sites so as to block them from being methylated or can change the binding of transcription factors that in turn regulate patterns of methylation (Gutierrez- Arcelus et al. 2013; Teh et al. 2014; and references therein). Such effects of genotype on epigenotype were demonstrated when both types of data were analyzed from a large, ethnically diverse group of newborn babies (Teh et al. 2014). Variation in the infants' genome-wide methylation patterns was strongly associated with in utero factors such as maternal smoking or depression, gestational age, and the baby's birth order, but the precise impact of these environmental factors on methylation varied depending on DNA sequence polymorphisms associated with ethnic group (Teh et al. 2014). The effect of specific epigenetic marks on transcriptional activity can also be sequence dependent (Meaney and Ferguson-Smith 2010). The impact of sequence variation on epigenetic marking and its transmissibil- ity (Herrera et al. 2014) may provide a mechanistic, epigenetic basis for certain aspects of genotype by environment (G x E) interaction (see Section 1.2.2).
As a result of the interplay of genetic and epigenetic variation, the source of heritable variation in individual development is neither genotype alone nor genotype plus independently generated epigenetic modifications but rather a gene regulatory system that arises from the interaction of genotype and epigenotype. As is true for any interactive system, studies of either component alone (i.e., the association of either gene sequence variants or epigenetic variants with phenotypes of interest) will reveal an incomplete picture of the causal dynamics (Meaney and Ferguson-Smith 2010; Teh et al. 2014). Although a great deal remains to be learned about epigenetic mechanisms, it is clear that genetic information is mediated at multiple levels that shape its expression and hence its precise developmental influence. The next step is to situate this intricate gene regulatory system in its environmental context.