Adaptive potential of genomic structural variation in human and mammalian evolution

Abstract

Because phenotypic innovations must be genetically heritable for biological evolution to proceed, it is natural to consider new mutation events as well as standing genetic variation as sources for their birth. Previous research has identified a number of single-nucleotide polymorphisms that underlie a subset of adaptive traits in organisms. However, another well-known class of variation, genomic structural variation, could have even greater potential to produce adaptive phenotypes, due to the variety of possible types of alterations (deletions, insertions, duplications, among others) at different genomic positions and with variable lengths. It is from these dramatic genomic alterations, and selection on their phenotypic consequences, that adaptations leading to biological diversification could be derived. In this review, using studies in humans and other mammals, we highlight examples of how phenotypic variation from structural variants might become adaptive in populations and potentially enable biological diversification. Phenotypic change arising from structural variants will be described according to their immediate effect on organismal metabolic processes, immunological response and physical features. Study of population dynamics of segregating structural variation can therefore provide a window into understanding current and historical biological diversification.

Introduction

Single-nucleotide polymorphisms (SNPs) introduce only single DNA base pair (bp) variation, yet they have been shown to directly alter a number of non-disease phenotypes in mammals, including hair thickness [1], muscle mass [2–4] and locomotion [5], among others. Attempts to determine how SNPs affect complex trait phenotypes is now standard practice, especially in the field of human genetics [6]. Genomic structural variants (SVs) comprise another class of phenotype-shaping genetic variation that has emerged in the past decade, due in large part to the development of high-resolution technologies, such as array comparative genomic hybridization, and advanced computational tools to analyze next-generation sequence data. SVs (used broadly in this review to indicate any structural change to the length or organization of a chromosome of ≥1 bp) encompass a variety of chromosomal alterations such as deletions, insertions, duplications, inversions, translocations and transposable elements. A special subset of SVs, copy number variants (CNVs), refers to loci that differ in integer allelic copy number between individuals and was among the first classes of SVs observed to be polymorphic at many loci in human genomes [7, 8]. Unlike SNPs, which only involve substitution of one nucleotide-pair for another, SVs involve a local sequence gain or loss or reorganization. SNP analysis itself may be considered a saturated mutagenesis experiment in humans, whereby across the whole human population statistically every genomic position that does not affect viability has likely been mutagenized. With structural variation, however, this is almost certainly not the case, due to the enormous breadth of possibility for individual events, including of different SV types, genomic positions and lengths. Thus, SVs have the potential to markedly affect organismal phenotype owing to the range of possible consequences that include gene dosage variation, open reading frame alterations, transcription factor binding site modifications and gain or loss of functional genomic elements.

General surveys of SVs, particularly CNVs, have been conducted in humans [9–12] and other mammalian species [13–27]. In humans, SVs have been shown to affect more base pairs in an average individual than SNPs [28], and CNVs in particular contribute a sizeable amount (nearly 18%) of the variation in lymphoblastoid cell gene expression [29]. Additional studies in humans [30] and mice [31, 32] have confirmed this expression effect in various tissues and cell types, with one study in mice suggesting that CNVs can modify not only gene expression levels but also affect the timing of expression [33]. Multi-allelic CNVs (loci that have more than two segregating structural alleles) have been shown in humans to account for nearly 90% of the variation in gene dosage between individuals, though they represent only a small proportion of all CNVs [34]. Interestingly, copy number of certain genes can rise to >50 in some human populations [35], suggesting positive selective pressure on paralogs acting cumulatively or neofunctionalization of the duplicates, or at least reduced purifying selection. Iskow et al. [36] provide a table of many CNV genes that show apparent evidence of positive selection in humans, yet the functional output of variable copy number is often unclear. Studies in humans have also examined the genomic extent of variable pseudogenes (copies of genes, usually truncated into exon-only sequences relative to their ‘parent’ sequences) [37], and have found a subset that encode chimeric transcripts that might have the potential for cellular function [38]. Studies in humans have also examined the relationship between CNV genes (compared with all genes) and microRNAs; one study found an enrichment for both the number of microRNAs that target CNV genes as well as the number of microRNA binding sites within CNV genes [39]. Another study found that genes targeted by copy-number variable microRNAs tended to have larger expression variability and a greater likelihood of differential expression between tissues and developmental stages [40].

Although general survey studies of SVs are highly informative, finding differences within species at thousands of loci, these studies do not pinpoint specific SVs as leading to particular phenotypes. The vast majority of genomic studies involving particular SVs have focused on disease-associated phenotypes, especially in humans (reviewed by Weischenfeldt et al. [41]). Yet, given the large number of SV events observed in humans that do not appear to cause disease [11, 42], nonpathogenic SVs either alone or aggregated together may have functional impact on phenotype, as changing genomic arrangements may have wide-ranging regulatory or protein consequences. This may be true even if the alteration is small compared with overall genome size, but whether and how this occurs is poorly understood. However, the focus of this review is the growing body of studies that have analyzed how nonpathogenic genomic diversity, associated with specific SVs of large functional effect, shapes organismal phenotype. These studies give insight into the genomic processes that may be shaping species’ interactions with their own members or with their environment. If a particular genomic alteration can produce a phenotypic innovation that facilitates a fitness advantage to the individuals possessing the variation, this enables positive selection on the variant and may set the stage for adaptation in the population. Through selection in a lineage, these alterations could ultimately lead to fixation of the trait in the species and thus alter its evolutionary trajectory. Since this type of genetic variation has also existed throughout biological history, by studying it, we might gain insight into its contribution to phenotypic divergence in past evolution events (see Figure 1 for a model). A greater understanding of the role of SVs in adaptive diversification might also help us to identify potential present-day diversification events. Analysis of segregating (polymorphic) SVs can, therefore, provide insight into the biological diversification potential of a population or a species.

Figure 1.

Model of biological diversification mediated through adaptive structural variants. AMY1 represents the human salivary amylase gene on chromosome 1 that has been shown to undergo adaptive duplication [43]. This schematic model represents DNA sequence that can undergo structural sequence mutation to alter phenotype that produces lineage divergence of a population through positive selective pressure. (A colour version of this figure is available online at: http://bfg.oxfordjournals.org)

In this review, we will demonstrate that naturally occurring genetic variation in humans and other mammals, arising particularly from SVs, is a source of phenotypic potential that could directly enhance organismal fitness in any of three classes of phenotypes: (i) metabolic processes, (ii) immunological response, and (iii) physical features. This categorization provides a powerful framework for how to conceive of evolutionary adaptation, and can likely be applied to other groups of organisms besides mammals. Although connections and interrelationships exist between these three classes of phenotypic adaptation, their partition provides a helpful way to conceive of the processes of evolutionary adaptation.

Metabolic processes

Adaptations at the molecular level that affect fundamental internal physiological processes in cells and tissues, provide possibly the most direct examples of ongoing evolution in a population of organisms. Although many of the early survey studies of structural variation gave clues as to the general metabolic processes affected by SVs, a direct link between an SV and altered internal physiology has been difficult to establish. Studies looking at how overall patterns of SVs, particularly CNVs, affect metabolic variation have identified changes to extracellular processes and signal transduction, such as the NF-κB and MAPK signaling pathways [44, 45]. A literature review by He et al. [46] examined the associations between CNV regions, particularly CYP2D6, that coincide with pharmacologic targets and how those regions can modulate drug response/toxicity. Other studies have shown an enrichment of SVs in olfactory genes in humans [47–49], and a study in cattle identified a deletion polymorphism of olfactory receptor genes including BTA5 [50]. Taken together, these findings suggest CNVs can broadly affect many types of internal metabolite responses that may affect organismal phenotype, leading to overall fitness differences.

Several examples of SVs affecting specific metabolic processes with the potential to positively alter fitness consequences have emerged (Table 1). One classic example is the correlation between copy number of the salivary amylase gene, AMY1 and starch digestion. In the landmark study, our lab was able to positively correlate copy number of AMY1 in humans to salivary amylase protein level [43], and found that individuals from regions with high-starch diets had, on average, more copies of AMY1 than those from populations with low-starch diets. Based on these findings, we have hypothesized that the observed directional selection of AMY1 copy number confirms an adaptive benefit: the more AMY1 copies—and higher AMY1 enzyme levels—a person has, the more rapidly one can break down starch after a meal and thus more efficiently use the sugar source available from a high-starch diet. Because of agricultural production, which tends to yield high-starch food, possessing a greater ability to biochemically break down starch would enable an individual to better use the energy-richness of sugar that farming provides. This would explain why AMY1 was shown to be abundant in various farming populations. More recently, a study demonstrated variable copy number of pancreatic amylase in dogs, from the gene AMY2B [51]. The authors hypothesized that the increase in amylase copy number was owing to metabolic selection occurring in dogs during the domestication process, adapting away from the carnivorous diets of wolves to the high-starch diets given to them by humans.

Table 1.

Metabolic process phenotypes observed from structural variants in mammals

Phenotype	Organism	SV type, extent	Genetic element	Study
Increase in salivary amylase production	Human	Duplication, full	AMY1	Perry (2007) [43]
Increase in pancreatic amylase production	Dog	Duplication, full	AMY2B	Axelsson (2013) [51]
Bone mineral density increases	Human	Deletion, full	APC	Chew (2012) [52]
Lack of urinary testosterone	Human	Deletion, full	UGT2B17	Jakobsson (2006) [53]
Increase in testosterone and estradiol in serum	Human	Deletion, full	UGT2B17	Yang (2008) [54]
Increase in production traits	Cow	Duplication, full	APOL3 and FABP2	Bickhart (2012) [55]
Increase in total merit	Cow	Duplication, full	BTA18	Seroussi (2010) [56]
Higher monounsaturated fatty acid proportion	Cow	Deletion, partial	SREBP-1	Hoashi (2007) [57]; Cecchinato (2012) [58]
In females, more offspring and higher recombination rate	Human	Inversion, partial	900kb located in 17q21.31 locus	Stefansson (2005) [59]
Increase in bull fertility	Cow	Duplication, full	TSPY	Hamilton (2012) [60]
Increase in breeding values of sires	Cow	Duplication, full	KIAA1683	Glick (2011) [61]

A few other examples in humans of SVs operating to modulate metabolic activities have also emerged. One study was able to calculate that Caucasian individuals with a single copy loss of the tumor suppressor APC locus had a net increase in bone mineral density of the forearm (∼8%), spine (13%) and hip (13%) [52]. In a study of men from Sweden and Korea, urinary testosterone concentrations were shown to correlate with UGT2B17 allele count. UGT2B17 codes for an enzyme involved in the metabolism of steroids (among other substrates), and men who had a homozygous deletion of the gene had no or negligible amounts of urinary testosterone compared with men of other genotypes [53]. In another study, homozygous deletion of UGT2B17 was significantly associated with higher serum concentrations of both testosterone and estradiol in Chinese men [54].

SVs have also been shown to modulate metabolic activities in domesticated cattle. From ‘artificial’ selection on beef breeds, genes involved in lipid transport and metabolism, including APOL3 and FABP2, are highly duplicated, suggesting a link between CNVs and production traits [55]. In Holstein cattle in particular, close associations have been reported between copy number within the BTA18 telomeric region and total merit, as well as fat and protein production levels [56]. In Japanese Black cattle, a deletion in an intron of the transcription factor gene SREBP-1 was associated with beef having a higher (1.3%) monounsaturated fatty acid level and a lower (−1.6C) melting point of intramuscular fat [57]. A follow-up study confirmed higher fat content in cattle carrying the non-deletion (longer) allele [58].

An evolutionarily interesting phenotype found to be affected by structural variation is fertility. A study in humans identified a 900 kb inversion on chromosome 17 that exists in two distinct structural segments termed H1 and H2 (now more recently shown to exist in multiple haplotypes [62]), with different population frequencies [59]. In what turned out to be a provocative interpretation, the researchers claimed that the inverted H2 segment is undergoing positive selection in an Icelandic population, owing to their data indicating that carrier females tended to have more children, on average, than noncarriers, and have an overall higher recombination rate. More offspring and the ability to generate more genetic diversity of haplotypes through recombination ‘shuffling’ could be regarded as having adaptive potential owing to likely fitness increases. These inversion polymorphisms, along with other examples, are discussed in more detail in this same issue, page XXX [63]. Increased fertility resulting from structural variation has also been tentatively shown in domesticated cattle, with a positive correlation between copy number of TSPY (testes-specific protein Y-encoded) and bull fertility [60], as well as the number of copies of a specific variant (V1) of the uncharacterized gene KIAA1683 associating with breeding values of sires [61].

Immunological response

The ability to resist predation from harmful microorganisms or having a reduced susceptibility to a particular disease presents a clear fitness advantage. A classic example of immunological adaptation in humans, in this case due to balancing selection, is seen in people who lack a full complement of the α-globin genes, in areas of the world where malaria has high prevalence. Individuals typically have four copies in a diploid genome (αα/αα). For most individuals, deletion of two copies, either in the form (–/αα) or (-α/-α), causes the blood disorder thalassemia and results in a form of anemia, while deletion of three copies results in very severe anemia (zero copies is embryonically lethal). However, the normal complement of four copies increases susceptibility to severe malaria infection; therefore, when living in endemic malaria environments, having two copies of the allele is immunologically advantageous, as these individuals can often avoid the worst malarial and anemic outcomes [64, 65].

Another popular example of purported immunological adaptation in humans is that derived from the copy number of the CCL3L1 gene. The product of this gene is a ligand for the C-C chemokine receptor-5 (CCR5), an important cell-surface protein that acts as a co-receptor for HIV entry into cells. In a controversial paper, Gonzalez et al. (2005) made the claim that humans with a copy number (which varies from 0 to at least 14) lower than the population average had greater susceptibility to HIV-1/AIDS infection [66]. A later study claimed that a higher copy number of CCL3L in rhesus macaques accounted for an appreciable amount (18%) of the variance of time of progression to simian-AIDS, with lower copy number associated with faster disease progression [67]. These studies led some to conclude that higher CCL3L1 copy number may confer some protection against HIV-1 susceptibility, with a meta-analysis of available data supporting this conclusion [68]. However, conflicting findings from various studies make it difficult to determine the overall benefits versus risks of copy number variation for this gene. Some research groups had difficulty replicating the associations between low copy number and increased HIV-1 susceptibility, including a study in a Zimbabwean population [69] and a study in a North American population [70], among others. One study found some evidence that the presence of low CCL3L1 copy number served as slight protection against anemia [71]. Another study found that for people already at a reduced risk for a rare inflammatory disease, Kawasaki disease (KD), due to a deletion in CCR5, individuals who also carried a high copy number of CCL3L1 had even more significantly reduced risk of developing KD [72]. Difficulty interpreting together specific findings such as these should serve as reminders for the research community of the potential danger of broadening particular biological findings from a certain cohort or population to other groups, given the complexities of epistatic interactions resulting from various haplotypes and genetic backgrounds. Usher and McCarroll in this same issue [73] review some of these difficulties.

Copy number of beta-defensin genes, the peptides of which are highly antimicrobial agents, has been implicated in variable immune response in humans. A cluster of at least seven of these genes is known to have diploid copy numbers ranging between two to seven. One study examined meiotic transmission of this region and identified the rate of copy number change in the germ line to be about 0.7% per gamete, making this cluster one of the highest-frequency recurrent CNVs known to date [74]. A study involving 67 populations revealed an unusually high frequency of DEFB103-expressing copies in East Asia, which corresponds to the geographical location of both historical and modern influenza epidemics, which the authors suggest may be the result of selection for resistance to influenza [75]. One older study showed a positive correlation of DEFB4 genomic copy number in humans and the corresponding levels of its mRNA transcript, implicating copy number of this gene as having important immune consequences [76]. This correlation was shown in a cervical cancer study demonstrating that women with five or more copies of DEFB4 had a lower likelihood of developing cervical cancer compared with those with four or less copies [77]. However, as noted above, it is important to be cautious when drawing broad conclusions from highly specific disease studies, especially when involving immune system modulation, as immune-related SVs that appear protective in some contexts might be detrimental in others. A classic example illustrating this is the homozygous 32 bp internal deletion within CCR5 that has been found to be protective against HIV-1 infection [78], yet also increases the risk of symptomatic West Nile virus infection [79]. In the case of DEFB4, a study in Chinese individuals showed that increased copy number led to greater susceptibility of the autoimmune diseases systemic lupus erythematosus (SLE) and ANCA-associated small vasculitis [80]. In a different study, individuals possessing five or more copies of the complement component C4 and three or more copies of C4A, compared with having lower copy numbers, had some protection against developing SLE [81]. Being aware of these nuances when synthesizing multiple disease studies from participants with different genetic backgrounds and then attempting to associate SVs with disease susceptibility is an important challenge for researchers in which to gain mastery.

Several other findings in humans suggest that immunological diversity in general may still be underappreciated. One study in an Irish cohort was able to demonstrate that individuals that carried more FLG intragenic tandem repeats have lower risk of having the inflammatory skin disease eczema [82]. In relation to the human immunoglobulin heavy-chain locus, specifically that of the variable chain, a study using a hydatidiform mole BAC clone resource was able to add over 1 Mb of additional sequence for alternative assembly to the reference genome in the region 14q32.33, including identification of eight CNV-containing haplotypes from a panel of nine genomes [83]. One study surveying the TRIM5 locus, a region encoding a retroviral restriction factor, found this region to be duplicated in one Han Chinese woman (out of 72 individuals) with 12 copies of TRIM genes beyond the study’s other individuals [84]. Additionally, deletion at G176 of TRIM5α was found to be associated with reduced susceptibility to HIV-1 infection, with the truncated protein having enhanced antiviral activity [85]. These studies indicate that the human immune response may be even more diverse than previously known, and suggest that further study be done across populations of different ancestry.

Several other studies have examined how SVs interface with immune system function in nonhuman mammals. A study involving rhesus macaques found that higher copy number of KIR2DL4, an immunoglobulin-like receptor gene expressed in natural killer (NK) cells, resulted in increased interferon-gamma production from NK cells and better retainment of CD4⁺ T cells during SIV infection, which may slow disease progression [86]. A study of Angus cattle identified nearly 300 CNVs that associated with increased resistance to gastrointestinal nematodes; by network analysis, many of these variants fell into regions annotated as immune-related, including the WC1 gene, the product of which is expressed on certain T cells in cattle [87]. Another study in cattle showed that two genes related to pathogen and parasite resistance, CATHL4 and ULBP17, respectively, were highly duplicated in one indicine cow compared with taurine cows [55].

It is apparent from these findings (see Table 2 for a summary) that immune system modulation via structural variation might confer fitness advantages. If selective pressures are strong enough, they could become adaptive or even fixed in a population, thereby permanently altering how its members interact with the environment. For example, if individuals are capable of fending off (or are adapted to) previous external immunological threats, they may be able to explore ecological niches previously unavailable, such as new food sources (plants or animals) that have natural microorganisms associated with them that previously prevented their consumption. Also, mutualistic microbial interactions could eventually result if animals were not susceptible to the potentially negative effects of the microbes’ colonization in the gut or other bodily location.

Table 2.

Immunological response phenotypes observed from structural variants in mammals

Phenotype	Organism	SV type, extent	Genetic element	Study
Reduced susceptibility to malaria infection	Human	Deletion, full	α-globin genes	Flint (1986) [64]; May (2007) [65]
Reduced susceptibility to HIV-1/AIDS	Human	Duplication, full	CCL3L1	Gonzalez(2005) [66]; Liu (2010) [68]
Slower time progression to simian-AIDS	Rhesus macaque	Duplication, full	CCL3L1	Degenhardt (2009) [67]
Slight protection against anemia	Human	Deletion, full	CCL3L1	Carpenter (2012) [71]
Reduced risk for Kawasaki disease	Human	Deletion, partial ; Duplication, full	CCR5 CCL3L1	Burns (2005) [72]
Reduced likelihood of developing cervical cancer	Human	Duplication, full	DEFB4	Abe (2013) [77]
Protective against HIV-1 infection	Human	Deletion, partial	CCR5	Liu et al. (1996) [78]
Reduced risk for developing systemic lupus erythematosus	Human	Duplication, full	C4 and C4A together	Yang (2007) [81]
Reduced risk of having eczema	Human	Duplication, full	FLG intragenic repeats	Brown (2012) [82]
Reduced susceptibility to HIV-1 infection	Human	Deletion, partial	TRIM	Nakajima (2009) [85]
Increased interferon-gamma production; better retainment of CD4+ T cells	Rhesus macaque	Duplication, full	KIR2DL4	Hellmann (2013) [86]
Increased resistance to gastrointestinal nematodes	Cow	Duplication, full	WC1 and ∼300 other genes	Hou (2012) [87]
Increased resistance to pathogens and parasites	Cow	Duplication, full	CATHL4 and ULBP17	Bickhart (2012) [55]

Physical features

Unlike the many examples in humans involving metabolic or immunological response alterations resulting from SVs, very few examples [88, 89] of nonpathogenic physical feature differences caused by SVs have been documented. This could be due to the bias in human studies toward disease phenotypes or to a broader issue of the definition of what is pathological in an organismal context, especially in humans where disease research is more common. Nonetheless, examples in other mammals of seemingly nonpathogenic physical feature alterations are abundant, suggesting that future work should be done investigating human variation in physical features arising from SVs. Coat color variation is a prime example of physical feature variation in nonhuman mammals. SVs involving several genes have been repeatedly linked to different coat coloration outcomes (Table 3). One common example involves the Agouti signaling protein (ASIP) gene whose product serves to regulate mammalian pigmentation by acting as an inverse agonist at melanocortin receptors. Melanocortin-1 receptor (MC1R) is a G protein-coupled receptor whose activation promotes the production of the dark pigment eumelanin; when antagonized by ASIP, cells instead produce the light pigment phaeomelanin [90]. Many studies have identified SVs of ASIP and MC1R that led to multiple pigmentation outcomes. For these instances, the SV was located in a genomic position sufficient to alter either gene regulation or affect the gene product, a change that led directly to coat coloration differences. While many of these examples are due to artificial selection from human breeders, some instances could be suggestive of potential fitness advantages gained by the affected individuals in natural habitats given the right circumstances. For example, rabbits with a deletion in the reading frame of MC1R have altered coat coloration to either black or red, depending on which deletion they carry [91]. Similar darkening of coat color has been observed in both deer mice [92] and gray squirrels [93] and resulted from an SV in ASIP and MC1R, respectively. These color changes could theoretically provide enhanced concealment from predators in terrestrial habitats, especially at night. As well, darkening of coat color may provide other advantages, such as social dominance or mating behavior preferences, as has been seen in a sheep population [125].

Table 3.

Physical feature phenotypes observed from structural variants in mammals

Phenotypes	Organism	SV type, extent	Genetic element	Study
Olive color skin	Human	Duplication, partial	Promoter region of MATP	Graf (2007) [88]
Short stature	Human	Deletion, partial and full	Multiple loci	Dauber (2011) [89]
Red coat or black coat	Rabbit	Deletion, partial	MC1R	Fontanesi (2006) [91]
Darker coat	Deer mouse	Deletion, partial	ASIP	Kingsley (2009) [92]
Black coat	Gray squirrel	Deletion, partial	MC1R	McRobie (2009) [93]
White coat	Sheep	Duplication, full	ASIP and ITCH promoter	Norris (2008) [94]
Darker coat	Sheep	Deletion, partial	ASIP	Gratten (2010) [95]
Grey coat	Sheep	Duplication, full	ASIP	Fontanesi (2011) [96]
White/tan coat	Goat	Duplication, full	ASIP	Fontanesi (2009) [97]
Black coat	Alpaca	Deletion, partial	ASIP	Feeley (2011) [98]
Black coat	Horse	Deletion, partial	ASIP	Rieder (2001) [99]
Saddle tan coat	Dog	SINE insertion, partial	ASIP	Dreger (2011) [100]
Black coat	Rabbit	Insertion, partial	ASIP	Fontanesi (2010) [101]
Red coat	Pig	Insertion, partial	MC1R	Kijas (2001) [102]
Yellow coat	Raccoon dog	Deletion, partial	5′ UTR of MC1R	Han (2012) [103]
Brindle coat	Rabbit	Deletion, partial	MC1R	Fontanesi (2010) [104]
White coat	Horse	Deletion, partial	KIT	Holl (2010) [105]
White coat	Horse	Deletion, partial	KIT	Haase (2011) [106]
Tobiano coat	Horse	Inversion, partial	100 kb downstream of KIT	Brooks (2007) [107]
White coat	Pig	Duplication, partial and full	KIT	Moller (1996) [108]
Grey-roan coat	Pig	Deletion, partial	KIT	Fontanesi (2010) [109]
White coat or white-spotted	Pig	Duplication, full	KIT	Rubin (2012) [110]
Color sidedness	Cow	Translocation, partial	KIT locus on chr 6 with locus on chr 29	Durkin (2012) [111]
Red coat	Cow	Deletion, partial	MSHR	Joerg (1996) [112]
Dilute-colored coat	Cow	Deletion, partial	PMEL	Schmutz (2013) [113]
Splashed white coat	Horse	Insertion, partial	promoter of MITF	Hauswirth (2012) [114]
Black coat	Dog	Deletion, partial	CBD103	Candille (2007) [115]
Brown coat	Pig	Deletion, partial	TYRP1	Ren (2011) [116]
Furnishings trait	Dog	Insertion, partial	3′ UTR of RSPO2	Cadieu (2009) [117]
Rex hair	Rabbit	Deletion, partial	LIPH	Diribarne (2011) [118]
More muscular	Cow	Deletion, partial	MSTN	Grobet (1997) [119]
More muscular and faster	Dog	Deletion, partial	MSTN	Mosher (2007) [120]
More muscular	Horse	SINE insertion, partial	promoter region of MSTN	Petersen (2013) [121]
Greater body mass at birth	Cow	Insertion, partial	3′ UTR of Nfix	Zhou (2012) [122]
Increased body proportions	Pig	Duplication, full	seven genes individually	Chen (2012) [123]
Greater body mass	Rabbit	Deletion, partial	IGF2	Fontanesi (2012) [124]

Physical feature variation resulting from structural variation can also change hair texture in mammals. In a large study of various dog breeds, the presence of an insertion within the 3′ untranslated region (UTR) of RSPO2, which codes for the signaling protein R-spondin-2, was associated with the ‘furnishings’ trait, a distinct facial hair pattern marked by a moustache and eyebrows [117]. In rabbits, deletion of a single nucleotide in an exon of the lipase member H (LIPH) gene is associated with the ‘rex’ hair phenotype (short, soft down hair) [118]. In a follow-up study, it was found that this single-nucleotide deletion, compared with wild-type, reduced by 3-fold the abundance of the LIPH mRNA transcript at both the fetal and adult stages, and reduced the lipase activity of the protein [126]. It is unclear whether this variation has immediate adaptive potential, yet in the wild, if hair-length shortening allows rabbits to sense the external environment differently (temperature, moisture content, etc.), it could alter how they interact with the environment, such as affecting foraging times. Note that this hypothetical SV-induced behavior change could eventually alter whole ecosystem balance if there was any time-dependence to the system—illustrating that given the right selective circumstances, SVs might play an even wider role in the adaptation process beyond that in a single, directly affected species.

Changes in an organism’s body morphology arising from SVs are also seen in several mammalian species (Table 3). A classic example of this is the so-called ‘double-muscled’ phenotype. In domesticated cattle, notable increases in muscle mass have been shown to result from a short deletion in the coding sequence of MSTN, the gene encoding the secreted muscle inhibitory factor myostatin [119]. This phenotype was also observed in dogs carrying a homozygous 2 bp deletion in an exon of MSTN, which leads to a premature stop codon. Even dogs carrying only one copy of this deletion were found, on average, to be statistically more muscular than wild-type dogs and faster in competitive racing events [120]. Similarly, muscle fiber typing studies in Quarter horses showed that insertion of a 5′ SINE in the promoter region of MSTN is associated with altered muscle fiber proportions that favor an optimal alternative gait; this change presumably accounts for the observed strong signal of selection at that locus in the horse lineage [121]. Other examples of genomic alterations leading to body size variation have been found in mammals. In cattle, a short insertion in the Nfix gene, which encodes site-specific binding proteins, is thought to be associated with greater body mass at birth [122]. In pigs, seven CNVs were found as candidate genes for a large number of body morphology phenotypes in categories such as measurement proportions, fat and muscle content and body weight, among others [123]. Similar to pigs, rabbits can also be evaluated for weight, which may be relevant in commercial settings. One study showed an association between a small deletion in the insulin-like growth factor-2 (IGF2) gene and larger weight [124]. Collectively, these body morphology studies stand out as examples of particular fitness enhancements resulting from SVs, owing to the variations directionally increasing animals’ physical presence.

Concluding remarks

In this review, we used examples of naturally segregating genetic variation in humans and other mammals, specifically genomic structural variation, as a way to gain insight into potential routes of biological diversification (see Figure 1 for model). If SVs occur in genomic regions (either individually or in combination) that facilitate phenotypic variability, under certain circumstances this could provide a fitness advantage for the organism and its progeny—ultimately giving these variants the opportunity to become adaptive or even fixed in the population under positive selective pressure. Combined with reproductive barriers either in allopatry or sympatry, these SVs could drive biological diversification of a population or species. By classifying potentially adaptive SVs according to their effects on three categories of phenotypes (metabolic processes, immunological response, and physical features), a conceptual framework can be built to better understand how SVs are operating on a population level and shaping evolutionary trajectories.

We have provided multiple examples (though not exhaustive) of nonpathogenic SV-produced phenotypic change that may have adaptive potential either from known or hypothesized selective pressures. In regards to metabolic processes (Table 1), we have seen how selection for increased enzyme activity (in this case from increased copy number of the human salivary amylase gene, AMY1) allows farming populations better energy utilization during digestion of high-starch foods [43]. Also in humans, we have seen that fertility can be potentially increased from an inversion polymorphism [59]. Immunological response alterations provide more evidence of the potential adaptive value of SVs (Table 2). Whether it be from the classic case in humans of balancing selection among the number of the α-globin genes [64, 65], reduced susceptibility to diseases such as eczema [82] or HIV-1 [85] or pathogen and parasite resistance in cattle [55], immune phenotypes from SV alterations can provide observable selective advantage. And while many examples of physical feature alterations caused by SVs (Table 3) come from artificial selection by human breeders to produce more desirable breeds, these examples of coat color diversity, hair texture variation, and body morphology increases, provide a window into the potentially adaptive value of SVs in the wild given the right selective pressures. These extreme examples from artificial selection (e.g., double-muscling in cattle can also result in obstructed labor/dystocia) are still useful in demonstrating the potential of strong or long-term selection acting on a beneficial trait, and can therefore be seen as evidence of the potential of SVs to become adaptive in the wild given the right selective pressures. Similar traits derived through natural selection could help animals more efficiently avoid predators or obtain greater numbers of prey, or otherwise better compete with members of the same population for food and other resources.

If nonpathogenic SVs alter the direction of how a species interacts with its environment, especially if they enable access to previously inaccessible ecological opportunity (the availability of an incompletely filled biological niche [127]), then a special type of evolution known as adaptive radiation may occur. Adaptive radiation can be defined as an ancestral species giving rise to multiple descendant species each adapted to distinct parts of the environment [128], an evolutionary concept that may be among the clearest demonstrations of Darwin’s ‘principle of divergence’ [129]. An often-recognized precondition for adaptive radiation is that of ecological opportunity. The ability of a species to exploit this opportunity is sometimes mediated through the development of a key innovation: an adjustment in a species’ morphological and/or physiological organization [130]. These genetically heritable novel features could enable a species to access the environment in a previously inaccessible way, thereby laying the groundwork for expansion into new a niche and the possibility of an adaptive radiation. Structural variation could theoretically provide a faster route to this process than SNPs due to SVs having potentially larger phenotypic effect-sizes because of affecting chromosomal structure, sometimes of hundreds or thousands of base-pairs in a single event. Therefore, analysis of structural variation in conjunction with adaptive radiation research could yield fruitful insight into evolutionary processes.

We focused the review on differences between individuals of the same species, not fixed genomic differences between species [131], as potential interspecies differences in genomic architecture make it challenging to compare even similar SVs across species. Studies attempting to explain biological adaptation have traditionally focused on fixed differences between species, yet this can result in parsimony difficulties that arise because of historical events that are inherently not directly testable. However, with the availability of high-throughput technologies that can rapidly assess SVs within multiple individuals of the same species, evolutionary questions can now be addressed by analyzing evolution in real-time using genomic alterations still undergoing flux in a species. In this way, analysis of segregating structural variation can provide much insight into the potential for, and constraint of, evolutionary processes, as well as provide a more probabilistic ascertainment of the steps involved in mammalian evolution.

By narrowing the review to non-disease phenotypes (but aware that secondary deleterious consequences might exist that are not readily identifiable), we highlight the potentially adaptive nature of certain SVs. We attempt to provide examples of only genomic structural variation affecting phenotype, though tightly linked SNPs co-segregating on the same haplotype along with the SV may also be responsible in some cases. These data can provide researchers in the field of mammalian evolution, or biological diversification more generally, insight into the nonpathogenic genomic plasticity range of species. Therefore, our summary also serves as a counterpoint to the many published reports on the effects of deleterious SVs that undergo strong purifying selection. This review may be of interest to those involved in agricultural production and maintenance, as awareness of biological enhancements/differences between livestock may alter breeding programs and result in different facility and land-use strategies. Finally, the examples of natural genetic variation given in this review may be potentially valuable to researchers in the emerging field of genomic engineering, as these examples provide proof-of-principle for viable genomic alterations tolerable in organisms, in some cases in regions not well studied experimentally.

Key points.

• Structural variation can have wide-ranging nonpathogenic phenotypic consequences in humans and other mammals.

• Under the right circumstances, certain structural variants could be under positive selective pressure and therefore adaptive in populations.

• Studying currently segregating structural variation allows insight into past and present biological evolution.

Funding

This material is based on work supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE1144152 (DWR).