2.1. Genes and alleles

Genes are segments of DNA coded for the production of RNA molecules and proteins. Transcription is the process in which the DNA molecule is used (by a cell) as a template to produce messenger RNA (mRNA), which in turn serves as a template for protein synthesis (translation). Persons possess two copies of each gene in the cells of their bodies (except for sperm and egg cells, red blood cells, and genes located on the X and Y chromosomes in males), one copy of which comes from the mother and one from the father. Each of the copies of a gene is called an allele of the gene. Alleles of a gene from the two parents may be identical, or they may differ slightly, sometimes differing by only a single nucleotide substitution (nucleotides are the basic subunits that make up the DNA molecule). These minor structural differences between alleles can be found in many different configurations or versions. When two or more versions of an allele for a single gene occur in greater than 1% of a given population, it is referred to as a polymorphism. A version of an allele that occurs in less than 1% of the population is called a mutation. When both of the alleles for a given gene are identical, then one is homozygous for that gene; when the alleles differ, one is heterozygous.

Individual alleles are inherited in a Mendelian manner, which can be best illustrated by considering the inheritance of single-gene or monogenic disorders (e.g., cystic fibrosis, sickle cell anemia), a discrete class of disorders caused by mutations on one or both alleles of the same gene. If the disorder is dominant, then only one mutated allele is necessary to present the disease phenotype; if the disorder is recessive, then both alleles must have mutations. When a monogenic disorder is completely penetrant (i.e., those who present the mutated allele[s] always have the disease phenotype) and the gene mutation known, then based upon the genotype of the parents and whether or not the disorder is dominant or recessive, the odds that the child will inherit the disorder (the risk factor) can be calculated precisely according to the simple rules of Mendelian inheritance. Because the vast majority of human traits are not caused by variations on one or two alleles, they are not inherited in a Mendelian manner.

2.2. Heritability studies

Heritability is defined as the proportion of phenotypic variance attributable to genotypic variance in a given population at a given time. According to the prevailing methodological approach, the heritability of a given trait is ascertained by comparing phenotypic concordances and discordances between subjects relative to their presumed degree of genetic similarity. For example, monozygotic (MZ) twins, inasmuch as they are derived from the division of a single fertilized egg cell, are assumed to share 100% of their segregating genes (genes shared by descent) – that is, they are genetically identical. Dizygotic (DZ) twins, on the other hand, are derived from two separate fertilized egg cells and share on average 50% of their segregating genes, the same percentage as non-twin siblings. Assuming that the environments of MZ and DZ twins do not differ in any systematic ways that would affect the trait under consideration – the so-called trait-specific equal environment assumption (Guo Reference Guo2001) – greater phenotypic concordance between MZ twins as opposed to DZ twins is taken as evidence of a genetic component. The same principles underlie family and adoption studies: Siblings (or singletons) share on average 50% of their segregating genes; half siblings share 25%; and adopted siblings do not share any segregating genes with their (non-adopted) siblings.

One of the simplest and most widely used modeling techniques for estimating heritability partitions total phenotypic variation for a trait (V _P) into a contribution attributable to genotypic variation (V _G) and a contribution attributable to environmental variation (V _E): V _P = V _G+V _E. This model assumes additive genetic variance (A) – that is, independent genetic effects are fully transmitted from parent to offspring, and each allele adds a fixed amount to the genetic component of a given trait. As many researchers in behavior genetics acknowledge, the genetic contributors to variation in behavioral phenotypes likely do not act in this simple manner (Plomin Reference Plomin2004), and additional terms can be added for non-additive genetic effects including interactions among alleles at a single locus (dominance-recessiveness), interactions among alleles at a different locus (epistasis), and interactions between alleles and the environment (gene by environment [G × E] interaction).

2.3. Gene association studies

Gene association studies typically involve the search for statistically significant differences in the prevalence of a polymorphism between individuals who possess a trait of interest and those without the trait. In a candidate-gene association study, an association is proposed between a particular allelic variant and a given phenotype. Suppose that gene A comes in two common forms (polymorphisms), allele A₁ and allele A₂. A claim for an association in a candidate-gene association study generally takes the following form: Those with allele A₁ are more likely to display behavior X than those with allele A₂ (or vice versa); or, those homozygous for gene A (alleles A₁A₁ or A₂A₂) are more likely to display behavior X than those who are heterozygous (alleles A₁A₂),Footnote ¹ or vice versa. Candidate-gene association studies are hypothesis driven, the hypothesis depending upon an assumption concerning the manner in which the protein for which the candidate gene is encoded affects behavior. The hypothesis is commonly tested in behavior genetics by the use of large data sets that contain both a certain amount of genetic data as well as behavioral data (usually in the form of subject responses to a questionnaire). Genome-wide association studies are non-hypothesis driven and involve sequencing all or most of the entire genome, typically of hundreds or thousands of individuals who possess a given phenotype. The researcher looks for any single nucleotide polymorphisms that those who exhibit the trait of interest possess in statistically significant numbers relative to those without the trait. Recently, the use of genome-wide studies has expanded to include a variety of other common (and uncommon) forms of DNA variation, including many of those we consider below.

4.1. Transposable elements

One striking finding to emerge from the completion of the human genome sequencing project is that an astonishing ~48% of the genome is composed of transposable elements, repetitive mobile DNA sequences – “jumping genes” – dispersed throughout the genome (Gibbs Reference Gibbs2003). Although there are many different kinds of transposable elements, they are divided into two classes according to their movement strategies. Class I elements are made up of DNA transposons, composing 3% of the human genome, which move as segments of DNA by a cut-and-paste mechanism, “cutting” themselves out of their current genomic location and “pasting” themselves into a new location (Craig et al. Reference Craig, Craigie, Gellert and Lambowitz2002). Class II elements are retrotransposons, which compose ~45% of the genome and move by a replicative copy-and-paste mechanism: They copy themselves to RNA, and the original DNA copy is maintained at the same location. The RNA copy is then reverse-transcribed into DNA – that is, RNA→DNA (the standard direction of transcription in the synthesis of proteins being DNA→RNA→proteinFootnote ² ), and the DNA is inserted into the genome at a new location. Hence, these elements expand in number as they retrotranspose, leading to an increase in genomic DNA content and a change in DNA sequence and structure at the region of insertion (Goodier & Kazazian Reference Goodier and Kazazian2008).

Conventional wisdom holds that transposable elements are “selfish” DNA parasites living within the genomes of organisms in a manner analogous to viruses (Dawkins Reference Dawkins2006; Orgel & Crick Reference Orgel and Crick1980). And indeed, the activity of transposable elements can disrupt gene transcription and cause a variety of pathogenic conditions (more than 70 known diseases involve heritable and de novo retrotransposition events (Cordaux & Batzer Reference Cordaux and Batzer2009). Over time, mutations and truncations of transposable elements have rendered many of them inactive. Class I elements – transposons – are, for the most part, inert fossil remains of ancient elements in human DNA (Lander Reference Lander2001). And many Class II elements – retrotransposons – have been rendered incapable of retrotransposition; that is, they are transpositionally incompetent, which means they can no longer move about the genome, copying and inserting themselves (Gilbert et al. Reference Gilbert, Lutz, Morrish and Moran2005), while those that remain transpositionally competent are often associated with pathogenesis and normally immobilized by epigenetic gene-silencing mechanisms (Callinan & Batzer Reference Callinan, Batzer and Volff2006). The one-sided view of retrotransposons as wholly destructive DNA parasites or as “junk DNA,” however, is giving way to a much more complex picture.

Recent evidence indicates that three types of retrotransposons retain the capacity to retrotranspose (Ostertag & Kazazian Reference Ostertag and Kazazian2001; Ostertag et al. Reference Ostertag, Goodier, Zhang and Kazazian2003; Rowold & Herrera Reference Rowold and Herrera2000). A base pair is a pair of conjoined DNA bases (the steps of the DNA spiral staircase, i.e., double-helix) and is used as a unit of DNA measurement. A long interspersed nucleotide element (LINE-1 or L1), is a transposition-competent retrotransposon greater than or equal to 7,000 base pairs in length (= 7 kilobases). L1 is the most widespread class of transposable elements in mammals, constituting ~20% of mammalian genomic DNA. While most L1 sequences are transpositionally incompetent, ~100–150 retrotransposons in the human genome retain the capacity to retrotranspose and of these, 10% are classified as highly active or hot (Brouha Reference Brouha2003). Alu elements are members of the class of short (100–300 base pairs) interspersed nucleotide elements (SINEs). Alu elements are a class of primate-specific Footnote ³ retrotransposons constituting the most prevalent repetitive element in the human genome (more than one million in number) and accounting for ~11% of genomic DNA. The human genome contains on average 2,000–3,000 transposition competent Alu elements (unlike L1s, which can autonomously retrotranspose, Alu elements require certain functional products of L1s to retrotranspose). SVA elements are hybrid transposable elements, ranging from 700 to 4,000 base pairs and combining features of both short and long interspersed nucleotide elements (though like Alu elements, they are non-autonomous). Approximately 3,000 active SVA elements have been identified in the human genome, and they are believed to be the youngest primate specific family of retrotransposons (Hancks & Kazazian Reference Hancks and Kazazian2010; Wang et al. Reference Wang, Xing, Grover, Hedges, Han, Walker and Batzer2005).

Consider first transpositionally incompetent L-1s, Alu elements, and SVAs. That a retrotransposon is transpositionally incompetent does not mean that is has no effect upon gene transcription. Diverse DNA regions are involved in the process of gene transcription, including promoter and regulatory regions, exons and introns, stop and start sites, as well as histone proteins, and transpositionally incompetent transposable elements can affect each of these:

1. 1. A promoter is a region of DNA where specific proteins – transcription factors – bind to initiate the process of gene transcription. Transpositionally incompetent L1s that act as alternate promoters have been identified thus far for over 40 human protein-coding genes (Matlik et al. Reference Matlik, Redik and Speek2006). The activity of an alternate promoter can impact gene transcription in a number of ways, including overexpression: higher rates of gene transcription than would occur with only a single promoter, resulting in higher levels of protein synthesis.

2. 2. A regulatory region is a segment of DNA where transcription factors and other regulatory proteins preferentially bind, and genome-wide scans have identified some 23,000 candidate regulatory regions derived from transpositionally incompetent L1s (Faulkner et al. Reference Faulkner, Kimura, Daub, Wani, Plessy, Irvine and Carninci2009).

3. 3. Human genes typically contain several DNA sequences that code for amino acids (the building blocks of proteins) known as exons, interspersed with several introns, non-coding regions, and transpositionally incompetent L1s and Alu elements can function as both exons and introns (Muotri et al. Reference Muotri, Marchetto, Coufal and Gage2007; Zhang & Chasin Reference Zhang and Chasin2006).

4. 4. Transpositionally incompetent L1s and Alu elements can function as stop sites, which are DNA regions that indicate the termination point for transcription of a particular DNA sequence (Muotri et al. Reference Muotri, Marchetto, Coufal and Gage2007; Zhang & Chasin Reference Zhang and Chasin2006). About 250,000 retrotransposon-derived transcription start sites, which are DNA sequences where gene transcription commences, have been identified (Faulkner et al. Reference Faulkner, Kimura, Daub, Wani, Plessy, Irvine and Carninci2009).

5. 5. DNA is organized into units known as nucleosomes, each of which consists of a segment of DNA of ~145–150 base pairs wound around a histone protein core (Richmond & Davey Reference Richmond and Davey2003). Recent research indicates that Alu elements play a critical role in the positioning of nucleosomes (Tanaka et al. Reference Tanaka, Yamashita, Suzuki and Nakai2010), and the configuration of histones plays a critical role in the extent to which any given gene can be transcribed.

Hence, retrotransposons probably have a key influence upon the transcriptional output (the transcriptome) of the mammalian genome (Faulkner Reference Faulkner2011).

In contrast to transpositionally incompetent retrotransposons, transpositionally competent L1s, Alu elements, and SVAs are continually expanding in number in the human genome through ongoing germline retrotransposition (Batzer & Deininger Reference Batzer and Deininger2002; Ostertag & Kazazian Reference Ostertag and Kazazian2001). Some new (i.e., de novo) germline (hence heritable) L1 and Alu retrotransposition insertions have been generated so recently that they are found in only a single individual and are designated private de novo insertions (Conrad et al. Reference Conrad, Keebler, Depristo, Lindsay, Zhang, Casals and Awadalla2011; Mills et al. Reference Mills, Andrew Bennett, Iskow and Devine2007). Humans harbor a large genetic load of recent transposon insertions along with several million fixed (transpositionally incompetent) insertions: Estimates suggest that up to 600 million of these private germline insertions have been generated in human genomes throughout the world (Cordaux & Batzer Reference Cordaux and Batzer2009; Mills et al. Reference Mills, Andrew Bennett, Iskow and Devine2007; Xing et al. Reference Xing, Zhang, Han, Salem, Sen, Huff and Jorde2009). This represents a significant mutagenesis of the genome, collectively equivalent to one insertion for every five base pairs of chromosomal DNA, and these mutations are expected to influence a wide range of human phenotypes (Iskow et al. Reference Iskow, McCabe, Mills, Torene, Pittard, Neuwald and Devine2010). The ability of transposable elements to move within the genome gives them an intrinsic propensity to affect genome evolution through the creation of new DNA sequences and structures and ultimately, to affect the evolution of species (Cordaux & Batzer Reference Cordaux and Batzer2009; Rebollo et al. Reference Rebollo, Horard, Hubert and Vieira2010).

Transpositionally incompetent retrotransposons, the positions of which are fixed within the genome, can be inherited, like other alleles, according to the principles of Mendelian inheritance. Such does not appear to be the case regarding active retrotransposons. One unusual mode of active retrotransposon inheritance involves the inheritance not of DNA, but of RNA. It will be recalled that L1s retrotranspose by copying themselves to an RNA molecule that is then reverse transcribed (RNA→DNA) to form a copy of the original L1 retrotransposon. Kano et al. (Reference Kano, Godoy, Courtney, Vetter, Gerton and Ostertag2009) reported that the RNA of L1 elements is abundant in both germ cells and embryos and that L1 RNA transcribed in male or female germ cells can be carried over through fertilization and integrate in daughter cells during embryogenesis. This RNA is then reverse transcribed creating a de novo L1 retrotransposition insertion in the genome of offspring. In other words, active L1s exhibit a non-Mendelian RNA-based inheritance.

De novo somatic retrotransposition also occurs postconception in the somatic cells of the zygote. The activity of L1s capable of retrotransposition is regulated by the epigenome (Ostertag & Kazazian Reference Ostertag and Kazazian2001) which regulates the expressional activity of genes by a number of processes, one of which is DNA methylation (see sect. 5). During gamete formation and in early embryogenesis, short waves of demethylation in a region of the L1 that serves as a promoter of expression allows L1s to escape epigenetic silencing (Beraldi et al. Reference Beraldi, Pittoggi, Sciamanna, Mattei and Spadafora2006), and research suggests that activated retrotransposons play a critical role in the development of the human brain.

Neural precursor cells have the capacity to develop into any kind of neuron (out of an estimated 10,000 different kinds) as well as other nervous system cells. Coufal et al. (Reference Coufal, Garcia-Perez, Peng, Yeo, Mu, Lovci and Gage2009) demonstrated that human neural precursor cells, whether derived from fetal brains or from cultured embryonic stem cells, support the retrotransposition of an introduced human L1, and cells into which L1s were introduced generated distinct types of functional neurons. Using advanced DNA analysis on biopsied human heart, skin, liver, and brain cells, Coufal et al. detected more L1s in the genome of adult human brain cells than in the genomes of heart or liver cells from the same individual – as many as 100 extra copies per neuron tested relative to other organs, and they estimated 80–800 new insertions per cell in some brain regions. This is consistent with both somatic retrotransposition and with neural precursor cells having undergone more retrotransposition events than other tissues.

Applying high-throughput sequencing techniques to study somatic retrotransposition in human brain tissue, Baillie et al. (Reference Baillie, Barnett, Upton, Gerhardt, Richmond, De Sapio, Brennan, Rizzu, Smith, Fell, Talbot, Gustincich, Freeman, Mattick, Hume, Heutink, Carninci, Jeddeloh and Faulkner2011) identified 7,743 L1 insertions, 13,692 Alu insertions, and 1,350 SVA insertions in the hippocampus and caudate nucleus of three healthy individuals. A number of key loci were found to contain somatic L1 insertions, including dopamine receptors and serotonin neurotransmitter transporters. Baillie et al. also identified a disproportionate number of intronic L1 insertions, which is noteworthy inasmuch as introns are the protein-coding loci of DNA, and determined that genes containing intronic insertions were twice as likely to be differentially overexpressed (i.e., overtranscribed) in the brain. As they note of their findings: “The hippocampus seems to be predisposed to somatic L1 retrotransposition, which is intriguing given that [it] is a main source of adult neurogenesis” (Baillie et al. Reference Baillie, Barnett, Upton, Gerhardt, Richmond, De Sapio, Brennan, Rizzu, Smith, Fell, Talbot, Gustincich, Freeman, Mattick, Hume, Heutink, Carninci, Jeddeloh and Faulkner2011, pp. 3–4). For more on the hippocampus and neurogenesis, see section 4.1.1.

Given that each retrotransposition event that occurs in a cell results both in an increase in nuclear DNA (nDNA) and a change in DNA sequence, the discovery of 80 new retrotransposon insertions in some neurons, 800 in others, and more in the brain than in other tissues, entails somatic mosaicism, the existence in one and the same individual of two or more distinct genomes. It has been a long-standing dogma that humans have identical copies of DNA – identical genomes – in the nucleus (i.e., nDNA) of all the cells of their bodies (with the usual exceptions):

As the author of this quote notes, they are not. In fact, as we shall see, all of the genetic phenomena to be considered in this article contribute, in one way or another, to widespread somatic mosaicism.

4.2. Copy number variations

The single nucleotide polymorphism (SNP) is the most familiar form of DNA variation. Approximately 10 million SNPs have been identified in the human population, and many more are being discovered with the determination of each personal genome sequence (Wheeler et al. Reference Wheeler, Srinivasan, Egholm, Shen, Chen, McGuire and Rothberg2008). Until recently, SNPs were thought to be the predominant form of genomic variation and to account for most normal, as well as abnormal, phenotypic variation in human phenotypes with a genetic component. A recent and important development in human molecular genetics has been the discovery of the ubiquitousness of a variety of structural variations (SVs) in DNA – deletions, insertions, duplications, and inversions – as well as more complex multisite variants of varying sizes that change the chromosomal architecture (Redon et al. Reference Redon, Ishikawa, Fitch, Feuk, Perry, Daniel Andrews and Hurles2006). It is now apparent that human genomes differ more as a consequence of structural variation than from differences in single nucleotides (i.e., SNPs) (Alkan et al. Reference Alkan, Coe and Eichler2011; Conrad et al. Reference Conrad, Pinto, Redon, Feuk, Gokcumen, Zhang and Hurles2010; Li et al. Reference Li, Zheng, Luo, Wu, Zhu, Li and Wang2011).

The most extensively studied of these SVs are submicroscopic copy number variations (CNVs), defined as stretches of DNA at least 1,000 base pairs (1 kilobase) long and extending up to several million base pairs, that are either deleted or present in multiple copies relative to a model normal genome (Feuk et al. Reference Feuk, Carson and Scherer2006; Iafrate Reference Iafrate2004; Redon et al. Reference Redon, Ishikawa, Fitch, Feuk, Perry, Daniel Andrews and Hurles2006). Many CNVs discovered thus far have a population frequency of greater than 1%, in which case they are referred to as copy number polymorphisms. In 2006, Redon et al. (Reference Redon, Ishikawa, Fitch, Feuk, Perry, Daniel Andrews and Hurles2006) constructed a first-generation CNV map of the human genome, showing that 2,900 genes, or 10% of the total number of genes in the human genome, are encompassed by CNVs. The average size of the CNVs was 250,000 base pairs. Since the average nuclear gene is 27,000–60,000 base pairs long (Alberts et al. Reference Alberts, Johnson, Lewis, Raff, Roberts and Walter2002), many of the CNVs were composed of multiple copies (or deletions) of entire genes, in some cases exceeding 12 copies of a single gene, challenging the dogma that individuals possess two copies of each gene (i.e., two alleles) in their nDNA.

The functional impact of CNVs on gene transcription has been demonstrated across the full range of biology (Conrad et al. Reference Conrad, Pinto, Redon, Feuk, Gokcumen, Zhang and Hurles2010; Hurles et al. Reference Hurles, Dermitzakis and Tyler-Smith2008). They can alter the expression of multiple genes located in the region of variable copy number simultaneously – approximately half of CNVs identified thus far overlap protein-coding genes (Sebat et al. Reference Sebat, Lakshmi, Troge, Alexander and Young2004) – allowing expression levels to be higher or lower than could be achieved by a single copy gene (Hastings et al. Reference Hastings, Lupski, Rosenberg and Ira2009; Stranger Reference Stranger2007). And while CNVs can affect the activity of neighboring genes, they can also have an equivalent effect at a distance: There is now substantial evidence that regulatory elements of genes can reside up to a million base pairs or more away (Nobrega et al. Reference Nobrega, Ovcharenko, Afzal and Rubin2003).

CNVs can be inherited via the germline in the manner of SNPs, while exhibiting mutation rates from 100 to 10,000 times greater across the human genome (Zhang et al. Reference Zhang, Gu, Hurles and Lupski2009). This accelerated rate of mutation may help explain several unique characteristics of CNVs. First, CNVs comprise a large percentage of the genetic differences between humans and apes, and appear to be a major driving force in evolution, especially in the rapid evolution that has occurred, and continues to occur, within the human and great ape lineages (Bailey et al. Reference Bailey, Kidd and Eichler2008). Second, CNVs display extremely high interindividual variation: Studies indicate that the total amount of DNA sequence variation involving CNVs between two normal subjects is actually higher than for SNPs (Korbel Reference Korbel2007). Third, and even more surprising, this variation appears to be intraindividual as well as interindividual. In other words, CNVs are another source of somatic mosaicism.

For example, in an analysis of 34 tissue samples from healthy subjects, Piotrowski et al. (Reference Piotrowski, Bruder, Andersson, Diaz de Stahl, Menzel, Sandgren, Poplawski, von Tell, Crasto, Bogdan, Bartoszewski, Bebok, Krzyzanowski, Jankowski, Partridge, Komorowski and Dumanski2008) observed at least six CNVs affecting a single organ of one or more tissues of the same subject. These findings were recently confirmed by Rodríguez-Santiago et al. (Reference Rodríguez-Santiago, Malats, Rothman, Armengol, Garcia-Closas, Kogevinas and Perez-Jurado2010), who, as part of a genome-wide survey of mosaic genomic variation, analyzed blood or buccal DNA samples of 1,991 adult individuals. They found copy number mosaicism in 1.7% of tissue samples from the same individual, and these variants were present in anywhere from 17% to 98% of all cells from the same tissue. Liang et al. (Reference Liang, Conte, Skarnes and Bradley2008) demonstrated that CNVs involving gains or losses of millions of base pairs occur during mitotic divisions of mouse embryonic stem cells during routine culture involving relatively few cellular divisions. The frequency and extent of these genomic changes in embryonic stem cells suggest that most somatic tissues will be mosaics composed of variants of the zygotic genome (Dear Reference Dear2009).

If copy number somatic mosaicism occurs intraindividually in healthy individuals during embryogenesis, then since MZ twins begin as a single individual (a single zygote), somatic mosaicism should occur between healthy MZ twins as well, resulting in inter-MZ twin genetic discordance. Bruder et al. (Reference Bruder, Piotrowski, Gijsbers, Andersson, Erickson, Diaz de Stahl, Menzel, Sandgren, von Tell, Poplawski, Crowley, Crasto, Partridge, Tiwari, Allison, Komorowski, van Ommen, Boomsma, Pedersen, den Dunnen, Wirdefeldt and Dumanski2008) confirmed this in a study of 29 pairs of MZ twins using peripheral blood-derived DNA samples. The study included 10 phenotypically unselected normal MZ twins and 19 twin pairs both concordant and discordant for Parkinson's disease, parkinsonianism, and Lewy body dementia. They discovered large-scale CNVs among all the twins, healthy and diseased, with an estimated frequency of up to 10% variation per twin pair. Footnote ⁵

It should be emphasized that these variations were found in a single tissue (i.e., blood). Given somatic mosaicism, the actual total percentage is likely much higher, and it is for this reason that Bruder et al. (Reference Piotrowski, Bruder, Andersson, Diaz de Stahl, Menzel, Sandgren, Poplawski, von Tell, Crasto, Bogdan, Bartoszewski, Bebok, Krzyzanowski, Jankowski, Partridge, Komorowski and Dumanski2008, p. 766) note: “It is likely that the confirmed CNVs shown here represent the ‘tip of the iceberg’ of all CNVs that are actually present in the studied twins.”

Finally, it is important to keep in mind that CNVs are not the only kind of structural variation common in the human genome. For example, there are more than one million microsatellites, variable numbers of 1–6 base pair repeats that account for 3% of the DNA sequence (Weber et al. Reference Weber, David, Heil, Fan, Zhao and Marth2002); about 150,000 minisatellites and variable number tandem repeats, polymorphic sequences containing 20–50 copies of 6–100 base pair repeats (Naslund et al. Reference Naslund, Saetre, von Salome, Bergstrom, Jareborg and Jazin2005); and about one million deletions and insertions of DNA base pairs (Weber et al. Reference Weber, David, Heil, Fan, Zhao and Marth2002). Therefore, CNV mosaicism itself likely represents the tip of the mosaic iceberg.

4.3. Aneuploidy

Humans possess 23 pairs of chromosomes, with each parent providing one chromosome in each pair, a condition referred to as diploidy (excepting male sex chromosomes: males contain a single X and a single Y chromosome). Any variation in the number of chromosomes, either more or less, is typically associated with pathogenesis. For example, trisomies are characterized by an extra chromosome (e.g., trisomy 21, or Down syndrome, and trisomy 18, or Edward syndrome); monosomies are characterized by the complete or partial absence of a chromosome (e.g., Turner syndrome, characterized by a single X chromosome instead of an XX or XY). Despite this association with pathogenesis, there is a surprisingly high degree of germline and somatic chromosomal mosaicism in healthy individuals.

It is estimated that 2–5% of healthy male sperm exhibit one or another form of aneuploidy, with significant intra- and interindividual variations (Tempest et al. Reference Tempest, Ko, Rademaker, Chan, Robaire and Martin2009; Templado et al. Reference Templado, Vidal and Estop2011). Approximately 2.26% of healthy sperm exhibits disomy (sperm should exhibit monosomy), while the estimated rate of 2× disomy is 4.5%. Longitudinal analysis in which sperm samples were obtained from healthy males, ages 18–32, every 6 months for 18 months, revealed significant time dependent intraindividual fluctuations in both the total amount of aneuploidy and in the chromosomes affected (Tempest et al. Reference Tempest, Ko, Rademaker, Chan, Robaire and Martin2009). Studies have associated increases in sperm aneuploidy frequency with a variety of environmental factors, including increases in consumption of caffeine, alcohol, drugs, and smoking; increased exposure to pesticides and endocrine compounds; and infection or illness (Härkönen Reference Härkönen2005; Robbins et al. Reference Robbins, Elashoff, Xun, Jia, Li, Wu and Wei2005; Shi et al. Reference Shi, Ko, Barclay, Hoang, Rademaker and Martin2001). In oocytes of healthy females, the rate of aneuploidy is much higher: Cytogenetic analysis places the aneuploidy rate at anywhere from 20% to 25% (Templado et al. Reference Templado, Vidal and Estop2011).

Distinguishing between germline and somatic chromosomal mosaicism is complicated because some chromosomal abnormalities, although the result of an abnormality in the parental gamete, are not considered inherited in the usual sense of that term. For example, a common form of trisomy 21 is the result of an error during meiosis – that is, during the formation of the maternal ovum. Because this error affects only one or several ova, the child who develops trisomy 21 has not inherited the mother's genotype (which is normal) but rather has developed from a defective ovum (Antonarakis et al. Reference Antonarakis, Lyle, Dermitzakis, Reymond and Deutsch2004; Hassold & Hunt Reference Hassold and Hunt2001). Hence, this form of aneuploidy is not considered germline aneuploidy. By contrast, germline aneuploidy has been thought to be a rare cause of aneuploidy in the human population, but recent evidence from cytological and population studies suggests otherwise. Molecular and cytogenic analyses have revealed that approximately 5% of young couples with a Down syndrome child show evidence of trisomies on chromosomes 13 and 19 (Kovaleva & Tahmasebi-Hesari Reference Kovaleva and Tahmasebi-Hesari2007). Most recently, direct analysis of fetal ovarian premeiotic, meiotic, and stromal cells (i.e., connective tissue) in normal offspring showed low-level trisomy 21 aneuploidy in every sample tested (Delhanty Reference Delhanty2011; Hulten et al. Reference Hulten, Patel, Westgren, Papadogiannakis, Jonsson, Jonasson and Iwarsson2010; Kovaleva Reference Kovaleva2010). This has led to the suggestion that the apparently universal trisomy 21 germline mosaicism in females may account for the relatively high occurrence of Down syndrome in the human population. As Delhanty (Reference Delhanty2011, p. 136) notes, “Based upon this evidence, germinal or gonadal mosaicism is likely to make a significant contribution to aneuploidy in the human population.”

Recent conservative estimates place the overall percentage of aneuploid neural cells in the normal adult brain at an astonishing 10%, involving monosomy, trisomy, tetrasomy, polyploidy (greater than four chromosomes), and uniparental disomy, that is, two copies of a chromosome from one parent (Iourov et al. Reference Iourov, Vorsanova and Yurov2006; Reference Iourov, Vorsanova, Liehr and Yurov2009; Rehen Reference Rehen, Yung, McCreight, Kaushal, Yang, Almeida, Kingsbury Cabral, McConnell, Anliker, Fontanoz and Chun2005). This is an indication of widespread chromosomal somatic mosaicism in the human brain (see Table 1). Given an estimated 100 billion neurons in the adult brain, this yields a rough (conservative) estimate of 10 billion neurons and 100–500 billion glial cells (neural cells that do not transmit electrical impulses but play an essential role in neuronal structure and function) with one or another form of chromosomal aneuploidy. It is estimated that roughly 28% of embryonic neural precursor cells exhibit chromosomal aneuploidy in one form or another (Iourov et al. Reference Iourov, Vorsanova, Liehr and Yurov2009). This chromosomal diversity appears to result from a high frequency of stochastic postzygotic chromosome mutations in somatic cells, whereby neuroprogenitor cells in numerous regions of the embryonic brain display cell division defects in normal cells that result in aneuploid adult cell progeny.

Table 1.

Percentage of whole chromosome 21 gain and loss in the normal brain varies within and among individuals

Note: Lymphocytes and Down syndrome cells were used to validate the counting criteria. Adapted from Rehen et. al. (Reference Rehen, Yung, McCreight, Kaushal, Yang, Almeida, Kingsbury Cabral, McConnell, Anliker, Fontanoz and Chun2005, p. 2178). Reprinted with permission.

Various lines of evidence indicate that brain tissues may be more prone to aneuploidy than other tissues (Rehen Reference Rehen, Yung, McCreight, Kaushal, Yang, Almeida, Kingsbury Cabral, McConnell, Anliker, Fontanoz and Chun2005). Mature aneuploid neurons are functionally active and integrated into brain circuitry, showing distant axonal connections (Kingsbury Reference Kingsbury2005). One likely result of this is neuronal signaling differences caused by altered gene expression, as documented in mammalian neural cells (Kaushal et al. Reference Kaushal, Contos, Treuner, Yang, Kingsbury, Rehen and Chun2003). Thus, a network composed of intermixed diploid and aneuploid neurons might produce unique signaling properties distinct from a network composed purely of diploid cells (Pavelka et al. Reference Pavelka, Rancati, Zhu, Bradford, Saraf, Florens and Li2010; Westra et al. Reference Westra, Rivera, Bushman, Yung, Peterson, Barral and Chun2010). Once again, the highly stochastic nature of chromosomal aneuploidy and resulting chromosomal mosaicism ensures that MZ twins will have differences in their neuronal chromosomes and that siblings will depart from a presumed 50% possession of the parental nuclear genome. MZ twins discordant for aneuploidy have been recognized for several decades: Discordant MZ twins have been reported for monosomy X, trisomy 1, trisomy 13, trisomy 21, and uniparental disomy (Gilgenkrantz & Janot Reference Gilgenkrantz and Janot1983; Nieuwint et al. Reference Nieuwint, Van Zalen-Sprock, Hummel, Pals, Van Vugt, Van Der Harten and Madan1999; Rogers et al. Reference Rogers, Voullaire and Gold1982; Uchida et al. Reference Uchida, deSa and Whelan1983; Weiss et al. Reference Weiss, Loevy, Saunders, Pruzansky, Rosenthal and Opitz1982).

4.4. Mitochondrial DNA

We are accustomed to discussing the human genome – the familiar double-stranded helix that exists in identical form in all the cells of our bodies. Humans, however, do not have a genome, but at least two entirely distinct genomes, differing not only in structure, but also in the manner in which they are inherited.

Mitochondria are intracellular organelles, small membrane-enclosed structures within the cell, in which the end product of the breakdown of glucose in cells is processed to form the primary source of cellular energy, adenosine triphosphate (ATP). Hence, mitochondria are commonly characterized as the powerhouses of the cell (McBride et al. Reference McBride, Neuspiel and Wasiak2006). However, mitochondria also play a central role in a number of critical cellular and metabolic processes, including cellular proliferation; apoptosis or programmed cell death (cellular suicide), a process aimed at destroying a physiologically unwanted cell (Desagher & Martinou Reference Desagher and Martinou2000); the regulation and homeostasis of intracellular calcium, which acts as an intracellular signal involved in numerous cellular processes including cellular expression and metabolism; fertilization and embryonic development (Cao & Chen Reference Cao and Chen2009); DNA repair (DNA is constantly exposed to endogenous and exogenous agents that generate DNA lesions and induce DNA instability; see Gredilla et al. Reference Gredilla, Bohr and Stevnsner2010); aging (Salvioli et al. Reference Salvioli, Bonafe, Capri, Monti and Franceschi2001); the regulation of steroid hormone synthesis in the adrenal cortex, including estrogen, testosterone, and cortisol (Almahbobi et al. Reference Almahbobi, Williams and Hall1992); synthesis of heme, one of the components of hemoglobin (Atamna Reference Atamna2004); and detoxification of ammonia from the liver (Jackson et al. Reference Jackson, Beaudet and O'Brien1986). Mitochondria also play a critically important role in the brain and are involved in the regulation of brain function, including synaptic plasticity and brain size (Chada & Hollenbeck Reference Chada and Hollenbeck2004).

Mitochondria possess their own genome composed of mitochondrial DNA (mtDNA), a circular structure of double-stranded DNA located within the mitochondrion itself and separate from the more familiar genome located in the cell's nucleus, that is, nuclear DNA (nDNA). Mitochondrial DNA exhibits a number of properties that distinguish it from nDNA. First, mtDNA is not inherited in Mendelian fashion, but rather, it is inherited from the mother; that is, it is exclusively transmitted by the oocytes. Second, mtDNA exhibits polyploidy, numerous multiple copies of the complete mitochondrial genome in a single cell that occur in variable numbers both within individual mitochondria, which possess 2–10 copies of mtDNA each, and in individual cells, which vary significantly in the amount of mitochondria, and hence mtDNA, that they contain, on the basis of cell-type (Clay Montier et al. Reference Clay Montier, Deng and Bai2009). For example, there are 1,075–2,794 copies of mtDNA per cell in muscle cells, 1,200–10,800 in neurons, and up to 25,000 in liver cells. Studies have shown differences in the mtDNA content of different oocytes from the same female ranging from 11,000–903,000 mtDNA molecules per oocyte (MayPanloup et al. Reference MayPanloup, Chretien, Malthiery, Reynier and Justin2007). Third, mtDNA exhibits heteroplasmy, the occurrence of allelic differences between the multiple copies of mtDNA in the same individual that often segregate differentially in different tissues (Pfeiffer et al. Reference Pfeiffer, Lutz-Bonengel, Pollak, Fimmers, Baur and Brinkmann2004). Fourth, the mtDNA mutation rate is anywhere from ~9 to 25 times greater than that for nDNA, and mutations are commonly inherited both somatically and via the germline (Lynch et al. Reference Lynch, Koskella and Schaack2006).

The association between human mtDNA mutations and phenotype is significantly complicated by the dual features of polyploidy and heteroplasmy. Due to heteroplasmy, not all copies of mutated mtDNA will exist in all of the copies of the mtDNA genome; due to polyploidy, the number of copies of the mtDNA genome possessing the mutation will vary in different cells and tissues of the body and may be preferentially segregated in certain tissue types. In the case of diseases known to be associated with a mtDNA mutation, whether or not the phenotype is expressed depends upon whether the mtDNA is localized in certain cells and tissues of the body, and if so, which cells and tissues, and whether the amount of mutated mtDNA surpasses a certain threshold level necessary for the appearance of the phenotype (Taylor & Turnbull Reference Taylor and Turnbull2005).

Because oocytes from the same mother can differ dramatically in the amount of mtDNA they contain and likewise in the amount of mutated mtDNA they contain, the amount of mtDNA any given sibling possesses is a result of whatever maternal oocyte he or she happened to develop from, as well as a host of other stochastic processes that occur during embryogenesis. Hence, siblings will not possess 50% of their mother's mtDNA, and although MZ twins develop from a single fertilized oocyte, mtDNA is stochastically partitioned with the first mitotic cell division. Suppose we are concerned with the heritability of a trait associated with a certain mtDNA mutation. It is possible for one MZ twin to receive more than half mutant mtDNA molecules while the other twin receives only a tiny fraction, depending on how the twins divide from each other and how much mutant mtDNA happens to be on each side of the division. Even with an even division, the mutant mtDNA in one twin may end up in cells that eventually die during normal development, while the mutant mtDNA in the other twin may end up in cells that differentiate into brain tissue (Clay Montier et al. Reference Clay Montier, Deng and Bai2009).

5.1. Epigenetics and monozygotic twins

Monozygotic twins exhibit significant intertwin epigenetic differences, and these differences may play an important role in intertwin phenotypic differences (Fraga et al. Reference Fraga, Ballesta, Paz, Ropero and Setien2005; Kaminsky et al. Reference Kaminsky, Tang, Wang, Ptak, Oh, Wong, Feldcamp, Virtanen, Halfvarson, Tysk, McRae, Visscher, Montgomery, Gottesman, Martin and Petronis2009; Poulsen et al. Reference Poulsen, Esteller, Vaag and Frage2007). MZ intertwin differences in DNA methylation and histone modification profiles have been reported in whole-genome-wide scans, select tissues samples, and specific genomic regions (Boks et al. Reference Boks, Derks, Weisenberger, Strengman, Janson, Sommer and Ophoff2009; Fraga et al. Reference Fraga, Ballesta, Paz, Ropero and Setien2005; Kaminsky et al. Reference Kaminsky, Tang, Wang, Ptak, Oh, Wong, Feldcamp, Virtanen, Halfvarson, Tysk, McRae, Visscher, Montgomery, Gottesman, Martin and Petronis2009; Wong et al. Reference Wong, Caspi, Williams, Craig, Houts, Ambler, Moffit and Mill2010). In a comparison of gene expression of 3-year-old and 50-year-old MZ twins, Fraga et al. (Reference Fraga, Ballesta, Paz, Ropero and Setien2005) identified a fourfold increase in epigenetic discordance in older versus younger twins. And in a recent study, Ollikainen et al. (Reference Ollikainen, Smith, Joo, Kiat Ng, Andronikos, Novakovic and Craig2010) analyzed DNA methylation levels in five different tissues from 56 MZ and 35 DZ newborns. In examining individual CpG dinucleotides, they found up to 82% discordance in DZ twin pairs and up to 54% in MZ twin pairs. They note that their findings “demonstrate that the intrauterine period is a sensitive time for the establishment of epigenetic variability in humans, with implications for the effects of maternal environment in addition to genetics on the development of the newborn epigenome and potentially for programming of later disease risk” (Ollikainen et al. Reference Ollikainen, Smith, Joo, Kiat Ng, Andronikos, Novakovic and Craig2010, p. 4176). (For more on epigenetics and the intrauterine environment, see sect. 7.)

5.2. Behavioral epigenetics

There is growing evidence for the role of epigenetic modifications in all aspects of normal neural development and functioning including perception, memory, cognition and learning, emotion, and neural and behavioral plasticity (Allen 2008; Champagne Reference Champagne2010b; Crews Reference Crews2008; Day & Sweatt Reference Day and Sweatt2011; Gao Reference Gao2008; Keverne & Curley Reference Keverne and Curley2008; Kuss & Chen Reference Kuss and Chen2008; Mehler Reference Mehler2008; Meza-Sosa et al. Reference Meza-Sosa, Valle-García, Pedraza-Alva and Pérez-Martínez2012; Molfese Reference Molfese2011; Nelson et al. Reference Nelson, Kiriakidou, Mourelatos, Tan, Jennings, Xie and Wang2010; Roth et al. Reference Roth, Roth and Sweatt2010). Epigenetic disregulation may contribute to abnormal gene expression in a range of neuropsychiatric disorders and neurodegenerative diseases, including autism, schizophrenia, depression, and Alzheimer's disease (Abdolmaleky et al. Reference Abdolmaleky, Yaqubi, Papageorgis, Lambert, Ozturk, Sivaraman and Thiagalingam2011; Akbarian Reference Akbarian2010; Coppieters & Dragunow Reference Coppieters and Dragunow2011; Gruber Reference Gruber2011; Nelson et al. Reference Nelson, Wang and Rajeev2008; Paslakis et al. Reference Paslakis, Bleich, Frieling and Deuschle2011; Grafodatskaya et al. Reference Grafodatskaya, Chung, Szatmari and Weksberg2010).

6.1. Three objections

7.1. Neogenetics and prenatal stress

There is growing evidence that epigenetic programming by the prenatal maternal environment is an important mechanism in such phenotypic changes, involving processes similar to those implicated in postnatal epigenetic reprogramming by the maternal environment (Lillycrop & Burdge Reference Lillycrop and Burdge2011; Lillycrop et al. Reference Lillycrop, Phillips, Torrens, Hanson, Jackson and Burdge2008; Mesquita et al. Reference Mesquita, Wegerich, Patchev, Oliveira, Leão, Sousa and Almeida2009; Oberlander et al. Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008; Weaver et al. Reference Weaver, D'Alessio, Brown, Hellstrom, Dymov, Sharma and Meaney2007). In animal models, exposure to chronic variable stress during the first trimester is associated with hypermethylation of the GR promoter in the hypothalamus, reduced methylation of the CRF promoter in the amygdala, increased HPA reactivity as measured by stress-induced corticosterone levels, and increased stress associated behaviors and anhedonia (Mueller & Bale Reference Mueller and Bale2008). Prenatal stress has also been associated with significant changes in specific miRNAs in the brains of male offspring and reduced anogenital distance and adult testes weight, which is consistent with decreased testosterone exposure in utero (Morgan & Bale Reference Morgan and Bale2011). In a study of 82 pregnant women, Oberlander et al. (Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008) found increased levels of GR neonatal promoter methylation in infants born to depressed mothers and increased HPA stress reactivity at age 3 months in response to visual novel stimuli.

Offspring of female rats placed on a protein deficient diet throughout gestation exhibit elevated hepatic GR gene transcription associated with decreased DNA methylation, and decreased methylation and increased transcription of the peroxisomal proliferator-activated receptor (PPAR) gene, which codes for proteins that act as transcription factors (Mueller & Bale Reference Mueller and Bale2008; Oberlander et al. Reference Oberlander, Weinberg, Papsdorf, Grunau, Misri and Devlin2008), as well as epigenetic modifications to the renin-angiotensin system, a hormonal system that regulates blood pressure (Bogdarina et al. Reference Bogdarina, Welham, King, Burns and Clark2007; Goyal et al. Reference Goyal, Goyal, Leitzke, Gheorghe and Longo2010). These epigenetic changes have been associated with, among other phenotypes, offspring hypertension, insulin resistance, and obesity (Lillycrop & Burdge Reference Lillycrop and Burdge2011).

A well-known example of epigenetic/phenotypic reprogramming via maternal prenatal diet concerns Agouti mice. Mice carrying the dominant Agouti allele A^vy develop a complex set of traits collectively referred to as the yellow obese or yellow mouse syndrome, characterized by yellow fur, obesity, insulin resistance, and hyperglycemia (Miltenberger et al. Reference Miltenberger, Mynatt, Wilkinson and Woychik1997). Inbred A^vy rodents raised in identical environments nonetheless exhibit significant variation in the severity of their agouti symptoms due to epigenetic differences. Mice with high levels of CpG methylation of the A^vy allele exhibit fewer and milder symptoms (a phenotype referred to as pseudoagouti): The higher methylation levels reduce the levels of transcription of the abnormal A^vy allele and consequently, the severity of the symptoms (Jirtle & Skinner Reference Jirtle and Skinner2007). When pregnant agouti mice are fed a diet supplemented with methyl donors (folate, choline, and vitamin B12), the phenotype of the offspring is shifted away from yellow to pseudoagouti (Dolinoy Reference Dolinoy2008; Waterland & Jirtle Reference Waterland and Jirtle2003; Wolff et al. Reference Wolff, Kodell, Moore and Cooney1998). The methyl supplements that the pup receives from the mother in the womb reprogram the pup's epigenome, effectively silencing the abnormal Agouti gene. Agouti offspring whose mothers receive no methyl supplementation exhibit no corresponding phenotypic shift.

Other neogenetic phenomena are likely involved in maternal programming in stressful prenatal environments. For example, in rats, prenatal stress has been shown to be associated with an increase in oxidative damage to mtDNA in fetal neural stem cells, leading to changes in their differentiation into distinct classes of neurons in the hippocampus (Cai et al. Reference Cai, Zhu, Li, Fan, Jia, Bai and Liu2007; Song et al. Reference Song, Zheng, Li, Jia, Suo, Cai and Zhu2009). Physiological cell stressors and viral infection can demethylate Alu elements, enabling them to retrotranspose (Li et al. Reference Li, Spearow, Rubin and Schmid1999). Transposed elements can confer stress-inducibility to genes in their proximity (Naito et al. Reference Naito, Cho, Yang, Campbell, Yano, Okumoto and Wessler2006) or protect those genes against stress via a small-RNA-mediated mechanism (Hilbricht et al. Reference Hilbricht, Varotto, Sgaramella, Bartels, Salamini and Furini2008). L1 mRNA and proteins can accumulate in stress granules, providing a cellular reservoir that could be quickly activated in response to environmental stimuli (Krichevsky & Kosik Reference Krichevsky and Kosik2001). Moreover, androgenic steroids and steroid-like compounds involved in embryonic brain development, stress responses, and behavior can induce L1 activity (Zuloaga et al. Reference Zuloaga, Puts, Jordan and Breedlove2008). A burst of retrotransposon activity may aid in rapidly generating genetic diversity in changing environments and in mammalian brain development, and maternal stress could impact L1 mobility in newly born neurons during embryonic development (Faulkner Reference Faulkner2011; Marchetto et al. Reference Marchetto, Gage, Muotri and Ulrich2010; Singer et al. Reference Singer, McConnell, Marchetto, Coufal and Gage2010). Hence, maternal diet and stress could affect the activation of transposable elements via reprogramming of the embryonic epigenome.

7.2. Twins and the prenatal environment

8.1. Multifactorialism

Most human traits with a genetic component are multifactorial (or complex), that is, they are polygenic, involving the biochemical products of hundreds and even thousands of genes (see below) interacting with each other and the environment in complex ways. Widespread epistasis, pleiotropy, the epigenome, retrotransposons, CNVs, aneuploidy, mtDNA, and alternative splicing render such traits even more complex. Consider the following:

1. 1. Height is a highly heritable trait: 80% of the variation in height in a given population is attributable to genetic factors. A new, exceptionally large study involving full genome scans (genome-wide association studies, or GWAS) of more than 180,000 individuals identified 180 genomic regions that influence adult height (Lango et al. Reference Lango, Hana Estrada, Lettre, Berndt, Weedon, Rivadeneira and Hirschhorn2010). The variants on these 180 genes, however, explain only 10% of the heritable phenotypic variation in height in a given population. Current estimates are that anywhere from 1,485 to more than 7,244 polymorphisms are necessary to explain 45% of the variance of height.Footnote ¹⁵

2. 2. The largest genetic effects that have been identified for common psychological disorders via GWAS account, all together, for less than 1% of genetic variance (Franke et al. Reference Franke, Neale and Faraone2009; Plomin & Davis Reference Plomin and Davis2009). As Plomin and Davis (Reference Plomin and Davis2009, p. 63) note:

   GWA studies suggest that for most complex traits and common disorders genetic effects are much smaller than previously considered.… This finding [of small genetic effects] implies that hundreds of genes are responsible for the heritability of behavioural problems in childhood, and that it will be difficult to identify these genes of small effect.

3. 3. Researchers performed complete genome-wide gene expression and GWA scans of 40 lines of inbred Drosophila melanogaster (fruit flies), and their analysis implicated at least 266 unique candidate genes associated with natural variation in aggressive behavior (Edwards et al. Reference Edwards, Rollmann, Morgan and Mackay2006; Reference Edwards, Ayroles, Stone, Carbone, Lyman and Mackay2009a; Reference Edwards, Zwarts, Yamamoto, Callaerts and Mackay2009b). The candidate genes were involved in a broad spectrum of biological processes, including vision, olfaction, learning and memory, and the development and function of the nervous system, as well as basic cellular processes including transcription, protein modification, and mitosis, indicating that the single alleles involved in aggression have pleiotropic effects on multiple traits. Furthermore, the findings are consistent with extensive epistasis. At the same time, the heritability of aggressive behavior in Drosophila is relatively low (~0.1). Expressing the genetic and environmental variances of aggressive behavior as genetic and environmental coefficients of variation (CV _G and CV _E , respectively), they found that CV _G = 23.2 and CV _E = 71.9. Thus, the low heritability is not due to a lack of segregating genetic variation, which is abundant, but to a high level of environmental variance.

Given this, we would not expect to find among common gene variants single gene effects that confer a sufficiently high odds ratio to be predictive of a phenotype: If a common polymorphism conferred a high odds ratio, it would have to be the principal contributor to the overall prevalence of the phenotype (Galvan et al. Reference Galvan, Ioannidis and Dragani2010; Hirschhorn et al. Reference Hirschhorn, Lohmueller, Byrne and Hirschhorn2002; Ioannidis et al. Reference Ioannidis, Trikalinos and Khoury2006; Zondervan & Cardon Reference Zondervan and Cardon2004). The discovery that many genes of small effect size contribute to traits known to be highly heritable (such as height) helps toward explaining the so-called missing heritability problem (Manolio et al. Reference Manolio, Collins, Cox, Goldstein, Hindorff, Hunter, McCarthy, Ramos, Cardon, Chakravarti, Cho, Guttmacher, Kong, Kruglyak, Mardis, Rotimi, Slatkin, Valle, Whittemore, Boehnke, Clark, Eichler, Gibson, Haines, Mackay, McCarroll and Visscher2009). This refers to the fact that despite thousands of genome-wide and candidate-gene association studies, very few genes have been identified that are reliably (i.e., consistently) associated with complex phenotypes (Need & Goldstein Reference Need and Goldstein2010). In the words of Conrad et al. (Reference Conrad, Pinto, Redon, Feuk, Gokcumen, Zhang and Hurles2010), to date, GWAS have left a “heritability void.” Retrotransposition, CNVs, aneuploidy, mtDNA, and the epigenome, doubtless all contribute to missing heritability (Slatkin Reference Slatkin2009). It is unlikely however, that methylation on a single gene, or a single CNV, or a single retrotransposon insertion will be the principal genetic contributor to a complex trait any more than an SNP.

These phenomena are also likely part of the answer to another absence of correlation, namely, that between human phenotypic complexity and genome size (e.g., humans possess fewer genes than corn). This lack of correlation is referred to as the C-paradox (the C-value refers to the amount of DNA contained in a gamete, or one-half the amount of DNA contained in the nucleus of a somatic cell) (Biemont & Vieira Reference Biemont and Vieira2006). Most of the differences in genome size between species reside in the non-coding parts (such as the 45% of the human genome that is composed of transposable elements). For example, the human genome is composed of 98% non-coding regions of DNA (International Human Genome Sequencing Consortium 2004), whereas the fruit fly, Drosophila melanogaster, has a very compact genome with fewer sequences of this kind (Clark et al. Reference Clark, Eisen, Smith, Bergman, Oliver, Markow and MacCallum2007; Dowsett & Young Reference Dowsett and Young1982; Hoskins et al. Reference Hoskins, Smith, Carlson, Bernardo Carvalho, Halpern, Kaminker and Karpen2002; Rebollo et al. Reference Rebollo, Horard, Hubert and Vieira2010). This is highly suggestive, especially if we consider that two major classes of retrotransposons are primate specific (see sect. 4.1).

9.1. Maternal effects II

Adaptive phenotypic plasticity is apparent in plants, invertebrates, amphibians, reptiles, fishes, birds, and mammals. What unites almost all species is the centrality of maternal effects – or maternal programming – as a mechanism of phenotypic plasticity (Bernardo Reference Bernardo1996; Maestripieri & Mateo Reference Maestripieri and Mateo2009; Mousseau & Fox Reference Mousseau and Fox1998; Wade Reference Wade, Mousseau and Fox1998; Zhang et al. Reference Zhang, Bagot, Parent, Nesbitt, Bredy, Caldji and Meaney2006). Maternal effects are the effects of a mother's phenotype upon the phenotype of offspring and are distinguished from direct genetic effects, that is, the phenotypic effect upon offspring of inheriting maternal genes (Mousseau & Fox Reference Mousseau and Fox1998). Offspring behavioral plasticity enables the mother to adjust the phenotype of offspring in response to the environment she inhabits and, in doing so, in effect transmit to them information about the environment they will inhabit. If the mother's adjustments to the environment are adaptive, and if the environment is stable across generations, that is, if the cues from the mother's environment are a good predictor of the environment in which offspring will find themselves, then the offspring's phenotypic adjustments are adaptive (Badyaev & Oh Reference Badyaev and Oh2008; Qvarnstrom & Price Reference Qvarnstrom and Price2001). A high degree of phenotypic plasticity may also imply that sometimes maternal effects can be maladaptive. For example, if the environment changes in relevant respects, maternal modifications of offspring phenotype may be maladaptive. Therefore, maternal effects may also provide a mechanism by which maladaptive phenotypic traits are transmitted across generations (Gluckman et al. Reference Gluckman, Hanson, Bateson, Beedle, Law, Bhutta and West-Eberhard2009; Mastripieri & Mateo Reference Mastripieri, Mateo, Mastripieri and Mateo2009).

Consider a few examples across species:

1. 1. Storm and Lima (Reference Storm and Lima2010) studied differences in the behavior of offspring of gravid crickets from two different environments. In one environment, the gravid mothers were regularly exposed to a predator, the wolf spider Hogna helluo, while the other environment was predator free. The offspring of Hogna-exposed gravid mothers displayed increased antipredator immobility in response to Hogna chemical cues and significantly higher survival rates in the presence of Hogna relative to the offspring of non-exposed mothers. According to Storm and Lima (Reference Storm and Lima2010, p. 383), this was an example of “a transgenerational maternal effect in antipredator behavior that takes the form of a ‘warning’ about predators that female fall field crickets Gryllus pennsylvanicus transmit to their offspring.”

2. 2. Female birds are able to alter many aspects of egg composition, including nutrients, hormones, antioxidants, immunoglobulins, and even embryo sex, in response to food availability, levels of sibling competition, and the quality of their mates (Adkinsregan et al. Reference Adkinsregan, Ottinger and Park1995; Groothuis & Schwabl Reference Groothuis and Schwabl2008; Horváthová et al. Reference Horváthová, Nakagawa and Uller2012; von Engelhardt et al. Reference von Engelhardt, Carere, Dijkstra and Groothuis2006). Such maternal effects can result in the influence of a particular environmental factor on phenotypic development persisting across a number of generations, even if the factor itself has altered.

3. 3. In mammals, including humans, any imbalance of nutrient intake, such that one or more nutrients is limiting, will produce adaptive responses in the developing fetus (Bellinger et al. Reference Bellinger, Lilley and Langley-Evans2004; Reference Bellinger, Sculley and Langley-Evans2006; Godfrey et al. Reference Godfrey, Sheppard, Gluckman, Lillycrop, Burdge, McLean and Hanson2011; Wiedmeier et al. Reference Wiedmeier, Joss-Moore, Lane and Neu2011). These metabolic adaptations ensure the immediate survival of the fetus in a less than optimal fetal environment, while long-term modifications to organ structure, hormone responsiveness, or gene expression may predispose to metabolic disorders in later life (Godfrey et al. Reference Godfrey, Sheppard, Gluckman, Lillycrop, Burdge, McLean and Hanson2011; Lillycrop & Burdge Reference Lillycrop and Burdge2011; Tamashiro & Moran Reference Tamashiro and Moran2010). This is precisely what we saw in examining the prenatal environment of twins (sect. 7.2) in which an adaptive response – growth restriction – ensures the survival of twins in a suboptimal prenatal environment.

4. 4. Numerous studies have shown that offspring olfactory cues and food preferences in both humans and other mammals can be shaped by the maternal prenatal diet in utero and the postnatal environment during lactation (Becques et al. Reference Becques, Larose, Gouat and Serra2010; Bilkó et al. Reference Bilkó, Altbäcker and Hudson1994; Cooke & Fildes Reference Cooke and Fildes2011; Hausner et al. Reference Hausner, Nicklaus, Issanchou, Molgaard and Moller2010; Hepper Reference Hepper1995; Reference Hepper1988; March et al. Reference March, Abate, Spear and Molina2009; Pedersen & Blass Reference Pedersen and Blass1982; Schaal et al. Reference Schaal, Orgeur and Arnould1995; Reference Schaal, Marlier and Soussignan2000; Semke et al. Reference Semke, Distel and Hudson1995; Simitzis et al. Reference Simitzis, Deligeorgis, Bizelis and Fegeros2008; Smotherman Reference Smotherman1982). For example, a recent study with mice examined whether maternal exposure to the artificial sweetener acesulfame-K (AK) during pregnancy or lactation affected the sweet preference of adult offspring (Zhang et al. Reference Zhang, Chen, Liu, Zhan, Quan, Qin and Deng2011). At 8 weeks after birth, preference scores for AK and sucrose were significantly elevated among exposed – both in utero and via nursing – as opposed to non-exposed offspring. The researchers were also able to correlate maternal ingestion of AK with AK in both amniotic fluid and breast milk. Human infants whose mothers consumed, for example, carrots, anise, garlic, and fruits during breastfeeding (Mennella & Beauchamp Reference Mennella and Beauchamp1999; Mennella et al. Reference Mennella, Jagnow and Beauchamp2001; Schaal et al. Reference Schaal, Marlier and Soussignan2000) (Cooke & Fildes Reference Cooke and Fildes2011), or who were fed vanilla-flavored (Gerrish & Mennella Reference Gerrish and Mennella2001) or vegetable-flavored (Mennella et al. Reference Mennella, Kennedy and Beauchamp2006) formula showed marked preferences for these foods and flavors. Furthermore, studies have shown that breast-fed, as opposed to formula-fed, infants are more receptive to diverse food flavors (Forestell & Mennella Reference Forestell and Mennella2007; Maier et al. Reference Maier, Chabanet, Schaal, Leathwood and Issanchou2007; Reference Maier, Chabanet, Schaal, Leathwood and Issanchou2008). This may be due to the greater variety of flavors to which an infant is exposed via mother's milk (which can transmit a variety of flavors from the mother's diet).

Adversity during perinatal development can forecast an increased level of demand in the environment the offspring will occupy. Under such conditions, the animal's best interest is to enhance its behavioral (e.g., vigilance, fearfulness) and endocrine (HPA and metabolic) responsiveness to stress (Champagne et al. Reference Champagne, Francis, Mar and Meaney2003a). These responses promote detection of potential threat, avoidance learning, and metabolic/cardiovascular responses that are essential under the increased demands of the stressor. In regard to maternal rearing behavior, under high-risk conditions, when the probability of extended periods of growth and survival are low, the optimal strategy is to maximize the number of offspring through accelerated mating, increasing the chances that at least some offspring will survive to reproductive maturity (Cameron et al. Reference Cameron, Shahrokh, Del Corpo, Dhir, Szyf, Champagne and Meaney2008). Moreover, since adverse environments are characterized by high, unavoidable risks and thus increased mortality, parental investment in offspring quality may be futile (Gangestad & Simpson Reference Gangestad and Simpson2000). Such conditions favor a shift in reproductive investment toward quantity. In contrast, more favorable environmental conditions favor greater investment in individual offspring at the cost of mating, since offspring quality predicts successful competition for available resources and reproductive fitness.

We saw (sect. 5.2.1) how maternal rearing behavior can shape the rearing behavior of offspring through epigenetic reprogramming. Further studies with rodents have shown that the female offspring of low-LG mothers exhibit, in addition to changes in levels of estrogen receptor α (ERα) expression, increased sexual receptivity; increased levels of plasma luteinizing hormone, which regulates a number of aspects of the female menstrual cycle; increased levels of fetal testosterone at embryonic day 20; and accelerated puberty compared to offspring of high-LG mothers (Cameron et al. Reference Cameron, Shahrokh, Del Corpo, Dhir, Szyf, Champagne and Meaney2008; Reference Cameron, Soehngen and Meaney2011; Jia et al. Reference Jia, Tai, An and Zhang2011; Sakhai et al. Reference Sakhai, Kriegsfeld and Francis2011). These changes were associated with earlier puberty and earlier and increased sexual activity in the female offspring of low-LG versus high-LG mothers. There is evidence that human reproductive strategies in adverse environments exhibit a similar dynamic. Studies have shown that stressful early life conditions, including family relations, are associated with earlier age at menarche (Chisholm et al. Reference Chisholm, Quinlivan, Petersen and Coall2005; Ellis Reference Ellis2004; Ellis & Garber Reference Ellis and Garber2000; Graber et al. Reference Graber, Brooks-Gunn and Warren1995; Moffitt et al. Reference Moffitt, Caspi, Belsky and Silva1992; Saxbe & Repetti Reference Saxbe and Repetti2009), and earlier age at menarche is associated with earlier onset of sexual behavior and reproduction (Andersson-Ellström et al. Reference Andersson-Ellström, Forssman and Milsom1996; De Genna et al. Reference De Genna, Larkby and Cornelius2011). In other words, being raised in a stressful environment is associated with accelerated mating.

9.2. Human exceptionalism?

On the basis of several highly influential studies of twins raised apart conducted in the 1980s, researchers concluded that when it comes to behavioral concordances, the shared rearing environment has an effect statistically indistinguishable from 0 (Plomin & Daniels Reference Plomin and Daniels1987). Genetic factors were almost wholly responsible for behavioral concordances. Rearing environmental influences were instead largely a cause of discordance (i.e., were non-shared):

This absence of shared environmental influence led some behavioral geneticists to conclude that to the extent that parents impact child outcomes, they do so at a child-specific, rather than a family-wide, level:

Plomin's point, however, is not that all environmental influences are non-shared, but that “most environmental influence for most traits is non-shared” (Plomin Reference Plomin2011, p. 583). Although, as behavior geneticists have themselves noted, such a finding challenges the validity of a vast amount of psychological research (McGue & Bouchard Reference McGue and Bouchard1998; Bouchard Reference Bouchard2004), nonetheless, it has been widely accepted (Burt Reference Burt2009). Maternal effects, prenatal and postnatal, are phenotypic concordance-producing effects; or in the language of behavior genetics, they are shared environmental effects. As Mateo (Reference Mateo, Maestripieri and Mateo2009, p. 134) notes: “Parental effects lead to parent-offspring resemblance, which can be adaptive if offspring encounter similar social and environmental features as adults.” For example, all of the studies cited earlier concerning postnatal epigenetic reprogramming of offspring via maternal rearing behavior demonstrated the transmission of maternal behavior on both the molecular and phenotypic level; that is, offspring came to resemble their (rearing) mothers. In doing so, they also came to resemble each other. Of course, offspring may differ in the degree or manner in which specific behavioral phenotypes are shaped by the perinatal environment due to any number of genetic, epigenetic, and micro-environmental differences (such as fetal position), or on the basis of sex. These same differences, and countless others, entail that the maternal environment can simultaneously be a cause of phenotypic discordance. Nonetheless, particularly in animal studies where the typical litter may contain up to 10 pups, the concordance-producing effects of the maternal environment on offspring behavior are obvious (Champagne Reference Champagne2010a; Champagne et al. Reference Champagne, Francis, Mar and Meaney2003a).

Were it true that there were no (or minimal) maternal effects upon the behavior of human offspring, then humans would indeed be exceptional among all animal taxa. What this would entail is that humans possessed less behavioral phenotypic plasticity, were less capable of being behaviorally shaped by maternal effects, than rodents. Any assumption of the absence of maternal effects, both prenatally and postnatally, seems doubly counterintuitive given that maternal effects appear to play a larger role in the evolutionary dynamics and adaptation of mammals than in any other animal taxa (Fusco & Minelli Reference Fusco and Minelli2010; Mastripieri & Mateo Reference Mastripieri, Mateo, Mastripieri and Mateo2009; Uller Reference Uller2008). As Chevrud and Wolf (Reference Chevrud, Wolf, Maestripieri and Mateo2009, pp. 13–14) note:

As mammals, humans have extended and intimate maternal associations, both pre- and postnatal, lasting many years. Yet, in the absence of maternal effects, what is this commitment for, involving as it does an enormous expenditure of maternal time and energy? Just as pressing is the question as to what possible adaptive advantage could accrue to the human species by lacking the plasticity to be behaviorally shaped by maternal effects to better meet the demands of its environment, when such a lack is clearly maladaptive. To see just how maladaptive, consider the following three examples:

1. 1. A gravid mother rabbit consumes throughout the perinatal period a berry that is nutritious and grows in abundance in an environment that has an otherwise limited food supply (Bilkó et al. Reference Bilkó, Altbäcker and Hudson1994; Semke et al. Reference Semke, Distel and Hudson1995). Given minimal maternal effects, her offspring have no predilection for the berries. Postweaning, they die of starvation.

2. 2. A dam inhabits a very dangerous environment beset by numerous predators. She has high levels of circulating corticosterone and behaviorally displays heightened reactions to stress. She devotes less time to maternal care and more time to being vigilant against predators. Given minimal maternal effects, her offspring, postweaning, exhibit modest stress responses and moderate levels of fearfulness. They are eaten in short order.

3. 3. The major histocompatibility complex (MHC) is a set of molecules on the surface of cells that play a central role in the immune system and self- or non-self-recognition. The genes coded for MHC molecules are among the most polymorphic coding loci known among vertebrates (Klein Reference Klein1986). Differences in the MHC are associated with different odors, and studies in both animals and humans have indicated a mating preference for individuals whose odor is associated with dissimilar histocompatibility genes (Havlicek & Roberts Reference Havlicek and Craig Roberts2009; Wedekind & Furi Reference Wedekind and Furi1997). It is hypothesized that MHC dissociative mating preferences function to produce disease-resistant, MHC-heterozygous offspring, or to reduce inbreeding, or both (Apanius et al. Reference Apanius, Penn, Slev, Ruff and Potts1997; Brown & Eklund Reference Brown and Eklund1994; Potts & Wakeland Reference Potts and Wakeland1993). Cross-fostering studies with animals have shown that histocompatibility scent preferences are determined by the scent of the rearing, not the biological mother (Ihara & Feldman Reference Ihara and Feldman2003; Penn & Potts Reference Penn and Potts1998). Given minimal maternal effects, offspring are not affected by their mother's scent one way or another. They inbreed and produce a line of disease-ridden offspring that will die out in several generations.