Is the Term ‘Functional Redundancy’ Functionally Redundant?

Abstract

When no phenotypic effects can be detected following knock-out experiments, genes are typically described as being ‘functionally redundant’. Examples of this so-called functional redundancy are ubiquitous in the scientific literature, across all kingdoms of life. However, functional redundancy remains a challenge to explain as redundant genes should not be maintainable by natural selection, and thus their persistence in the genome represents an apparent Darwinian paradox. Numerous models and hypotheses have been proposed to explain the persistence of genetic redundancy, but many appear to have significant shortcomings. This review seeks to evaluate these models and hypotheses, and to examine whether the widespread use of the term ‘functional redundancy’ is truly justified.

Introduction

Genetic redundancy refers to the phenomenon of functionally redundant genes within an organism’s genome (Pickett & Meeks-Wagner, 1995; Pérez-Pérez et al., 2009), whereby two or more genes overlap in function (Cooke et al., 1997; Vavouri et al., 2008). Therefore, an absence of one of these genes produces no measurable phenotypic effect, as its function is compensated for by the other gene(s). Many so-called redundant genes have been discovered through knock-out and knock-down experiments, where a target gene’s function is disrupted. This is achieved either through removal of the target gene from the genome (e.g., via homologous recombination (Rong & Golic, 2001)), or through disrupting its expression using RNA interference (RNAi). Researchers then measure for any phenotypic effect that this might have on the organism and its fitness (Skarnes et al., 2011). However, this apparent functional redundancy is believed to constitute a problem in evolution, as it appears to undermine the power of natural selection due to the invariable persistence of redundant DNA within the genome. The purpose of this review is to: explore examples of functional redundancy discussed in the scientific literature; to examine the models explaining its persistence within the genome; and to critically analyse the evidence presented for and against its occurrence. From this, we may understand the reality and extent to which genetic redundancy truly occurs, and address whether the available evidence justifies the extensive use of the term ‘functional redundancy’ by scientists today.

There are numerous instances in the literature of genes that overlap in function, and knock-out/down experiments on these genes tend to produce no observable phenotypic effects. In fact, experimental evidence would suggest that genetic redundancy is typical in most biological systems studied thus far, present across all kingdoms of life (Kafri et al., 2009). For example, in the gram-negative bacterium Rhodospirillum rubrum, the genes encoding for Polyhydroxyalkanoate (pha) polymerases (such as phaC1, phaC2, and phaC3) have arisen through gene duplication. Yet each gene in this small network shares an overlap in function with one another (Jin & Nikolau, 2012). Similarly, genetic redundancy is also thought to be abundant in the plant kingdom, and is believed to have played an important role in the evolution of certain plant species. E.g., in Glycine max (soybean), the phytochrome A gene (phyA) has undergone multiple duplications (Liu et al., 2008) that have led to a functional overlap in its daughter genes. When one of these genes is knocked-out, its function is preserved through gene compensation by corresponding paralogues. The researchers posit that the presence of genetic redundancy in this species, whilst protecting against any potential deleterious mutations, has also played a role in the evolution of photoperiod insensitivity (the physiological reaction of plants to periods of light and dark), which is important in determining seedling emergence in soybean.

Examples of genetic redundancy in animals are also believed to be prevalent. For example, through RNAi experiments in the nematode worm Caenorhabditis elegans, researchers have identified large portions of the genome that show little to no detectable phenotypic effect when suppressed (Conant & Wagner, 2004). This effect is smallest/non-existent in the most closely related genes (i.e., paralogues that have undergone the least divergence). Examples of genetic redundancy/functional overlap appear to be ubiquitous in animals in all contexts examined thus far, from signalling and metabolic pathways to developmental mechanisms. Chemosensory genes in Drosophila have been shown to arise through multiple gene duplication events (McBride et al., 2007), and this vast network of genes is believed to possess redundant genes due to functional overlaps and sequence similarities between many components. Alternatively, the abl gene and its corresponding homologues (Dsrc29A, Dsrc64B and fps) in Drosophila exhibit functional redundancy in signalling transduction pathways and play a broad but integral role in Drosophila development (Hoffmann, 1991).

There are also many examples of genetic redundancy in the budding yeast Saccharomyces cerevisiae (Vavouri et al., 2008). For instance, two regulatory genes of Ty-transcription in S.cerevisiae, SPT23 and MGA2, show considerable homology. Single mutations of these genes have modest effects at most, but an absence of both causes severe deleterious effects on cell growth (Zhang et al., 1997). These two genes are therefore believed to possess shared function, and in the absence of one the other appears to compensate. Furthermore, through analysis of annotated open reading frames (ORFs), functional profiling in S.cerevisiae has revealed that mutations in ~83% of their entire genome produces no obvious phenotype, so long as the yeast is grown in a rich medium (Giaever et al., 2002). Other examples in fungi demonstrate how functional redundancy is favoured by stabilising selection, as it generates ecological plasticity in mechanisms for nitrate transport in the paralogues nrtA and nrtB in Aspergillus nidulans, thereby facilitating greater fitness in variable environments with different soil types (Unkles et al., 2001).

Yet, we must remember that complex phenotypes are mostly polygenic (Yang et al., 2013), and entire networks of genes are responsible for various traits. E.g., multiple networks are believed to be involved in the development of teeth in vertebrates (Fraser et al., 2010), with each network comprising of many genes. A useful analogy to functional redundancy is the Internet, as both the genome and the Internet are examples of error-tolerating scale-free networks (Pastor-Satorras & Vespignani, 2001; Barabási & Oltvai, 2004; Albert, 2005). The Internet is comprised of millions of networked computers, and is able to withstand knocking out/down of several of its components without impairing its function. As such, although it may cause a measurable deleterious effect, knocking out huge swathes of computers is unlikely to cause the collapse of the entire system (Pastor-Satorras & Vespignani, 2001). However, the analogy is not perfect as the Internet possesses no ‘master switch’ (i.e., it has no centralised components), whereas transcription factors and the Hox family in animals can be thought of as ‘master switches’ in the genome, as their absence would cause the entire biological system to collapse. Yet it still nevertheless serves to illustrate the point that there are many interdependent components all acting in concordance to produce a durable phenotype.

Examples of these gene networks are abundant in the literature, as they possess promising explanatory power for the complexity observed in many traits (Brazhnik et al., 2002). Many networks are thought to have arisen through gene duplications, such as the chemosensory network in Drosophila discussed above, as well as the all-important Hox genes in animals (McGinnis & Krumlauf, 1992). Other examples can often be traced back to whole-genome duplication events. Polyploidy is common in plants (Pérez-Pérez et al., 2009; De Smet & Van de Peer, 2012), and is thought to enable the ‘rewiring’ of new genes to take on a novel function, often augmenting greater complexity. The vertebrate genome has also undergone two whole-genome duplications (Donoghue & Purnell, 2005). This is thought to have been pivotal in the evolution of vital gene networks such as the globin superfamily of genes in vertebrates (Hoffmann et al., 2012), giving rise to integral metabolic products such as haemoglobin, myoglobin, and cytoglobin.

Genetic redundancy therefore represents an apparent Darwinian paradox, as it would appear to defy natural selection. There are numerous examples of daughter genes assuming new roles following gene (or whole-genome) duplications. Why, therefore, are there apparently abundant examples of genetic redundancy between paralogues in the genomes of a plethora of species? One should expect redundant DNA to do either one of two things: either it should mutate and assume a new function rendering it advantageous or essential; or, since it is redundant, it should eventually mutate out of existence as there are no obvious selection pressures acting to maintain it. There may even be selection pressures against redundant DNA, as it is costly and wasteful to transcribe. Natural selection optimises genome sizes to be efficient and small, as DNA is costly to replicate (Denver et al., 2004). Does this suggest that the ‘redundant’ genes in question might in fact actually be functional? One might envisage that genes may share functional overlap, but to suggest that they are completely redundant and serve no purpose within the genome may be somewhat premature. It is therefore important to discuss whether these genes truly do have a selective advantage or function (and if so, how they are being maintained by natural selection), and to ponder whether the term ‘functional redundancy’ is sufficiently accurate to warrant its widespread use in the scientific literature.

Does Genetic Redundancy Truly Exist and Can it be Maintained?

There is a wealth of research that attempts to elucidate the origin and purpose of genetic redundancy. One process by which genetic redundancy is thought to arise is following gene duplication (Zhang, 2003). When a gene duplicates, two identical daughter genes are generated. Therefore, knocking down/out one of these copies will have no observable effect on the organism being studied, as the other gene will continue that function. Alternatively, convergent evolution of two unrelated genes may also lead to an overlap in function (Conant & Wagner, 2004). But maintaining this functional redundancy is where the apparent Darwinian conundrum lies. The evidence for genetic redundancy is commonly divided between the fields of population genetics and molecular genetics, and they will be critically analysed as such here.

Population Genetics. There have been several models and hypotheses put forward in order to explain the apparently ubiquitous nature of functional redundancy. One hypothesis that has been proposed is that gene duplication can improve genetic robustness (Mendonça et al., 2011). If a specific gene fails, then other genes (either paralogues or genes that have evolved through convergent evolution such as homeoplasies) will compensate for the loss of function. Highly variable environments require a robust genetic architecture that is able to withstand environmental conditions that exist in a permanent state of flux. Consequently, one may predict that if variable environments require highly robust genomes with high degrees of genetic redundancy, organisms in predictable environments will show reduced genomes with fewer redundant genes. And this is exactly what research has shown: obligate parasites and other symbionts living in predictable host-dependent environments possess significantly fewer redundant genes (and therefore lower genetic robustness) than their free-living counterparts. For example, Mendonça et al. (2011) compared free-living parasites such as strains of Escherichia coli and Mycobacterium tuberculosis with obligate endosymbiotic parasites such as strains of Buchnera aphidicola and Mycobacterium leprae. The researchers found that the obligate endosymbionts exhibited reduced genomes with lower genetic robustness and functional redundancy, with fewer paralogues due to selective gene loss. Mendonça et al. conclude that as a consequence of a more constant environment, there is weaker selection for genetic robustness. Functional redundancy is wasteful and more costly to maintain, and is therefore less common. However, it has been argued that the hypothesis for improved genetic robustness arising from gene duplication is inadequate to fully explain genetic redundancy, as the robustness observed in some species (e.g., yeast) is partially caused by single-copy genes, as opposed to solely paralogous duplicates (Conant & Wagner, 2004). This implies convergent evolution between unrelated genes, meaning that there must be active selection pressures in some species selecting in favour of functional overlap.

Nowak et al. (1997) have used mathematical models to describe different scenarios under which two redundant/overlapping loci (A and B) can stably coexist in a population. The first model proposes equal mutation rates and effectiveness of both loci A and B, but this is unrealistic and is not discussed in great depth by the researchers. In the second model, locus A works at an efficacy of 1, but locus B works at a slightly lower efficacy. However, locus A is also subject to a higher mutation rate than B. Thus, under this model locus B acts as a ‘genetic buffer’ (Kafri et al., 2008), to act in the case of failure of locus A (Fig.1A). The two loci are able to coexist in a population under this scenario, as locus B is maintained as a failsafe. The third model proposes pleiotropy of genetically redundant genes, e.g., apparent functional redundancy of one gene may only be true for a single function, but the gene may also be responsible for other traits elsewhere (Fig.1B). If the gene is expressed in other networks, then it will be maintained by natural selection. So functional overlap may exist between the two genes in question, but one or both may have additional roles, and thus a deleterious effect will arise if they are disturbed. This scenario is also likely to arise if two unrelated genes convergently evolve the same function in the genome, perhaps explaining the presence of functionally redundant single-copy genes. The fourth scenario suggested by Nowak et al. is that if locus A irregularly fails to function in development, this is easily overcome through compensation by locus B (again, see Fig.1A). Locus B therefore acts as a failsafe much like the scenario proposed in model 2, but this irregularity in function is not heritable. Instead, it is a facet of developmental errors that may be caused by a feature of its structure, such as its position on the chromosome.

However, there are many criticisms that one might envisage in response to Nowak’s models. For example, Nowak et al. fail to discuss how two genes with different efficiencies and mutation rates adopt the same or a similar function initially in model 2, bypassing the issue of how this scenario might arise to begin with. Their models rely on pre-existing genetic redundancy, and only deal with its maintenance. The researchers do however recognise caveats in their first model regarding the unlikelihood of equal mutation rates and efficiency, yet their second model may also be too simplistic. Genes tend to have variable optimums at different stages of an organism’s life cycle, so a pair of functionally redundant genes may actually be differentially regulated in a spatiotemporal manner, with one being used for the desired metabolic pathway in one area of the body, and the other gene in another area of the body. Alternatively, they both may operate in the same location, but perhaps at different stages of development, thereby complementing each other. Indeed, Kafri et al. (2009) note how two paralogues may actually develop new roles in an expression circuit, regulating the expression levels of one another. One might also criticise the untestable nature of some of Nowak’s models. In Nowak’s first model discussing the persistence of a newly duplicated paralogue, the researchers assert that the average lifetime of redundancy is predicted to be around 107 generations. Testing 10 million generations is far from plausible, even in bacterial colonies with the fastest generation times.

Yet, many of these redundant genes have indeed been shown to be ancient, with the vast majority analysed in some studies discovered to be 100-150 million years old (with some even being over one billion years old, arising before the divergence of eukaryotes such as yeast, plants, and animals) (DeLuna et al., 2008; Vavouri et al., 2008), so they cannot be a mere transient facet of gene duplication. They must somehow be maintained in the genome for them to persist for millions of generations, especially considering that some ancient paralogues still remarkably show high sequence similarity (Nishida & Gotoh, 1993). Some models consider the effects that both mutation rates and population sizes might have on this  maintenance of genetic redundancy (Conant & Wagner, 2004; Vavouri et al., 2008). It has been proposed that functional redundancy can only persist in very large populations with low genetic drift (thereby preventing allele fixation), or if mutation (and thus failure) rates of redundant genes are high. High mutation rates may prolong functional redundancy and delay divergence/gene loss in smaller populations, but this is not evolutionarily stable as deleterious mutations will eventually accumulate and rise to fixation in smaller populations (Conant & Wagner, 2004). Very recently, Saito et al. (2013) have proposed an evolutionary model that encompasses genotype-phenotype mapping (GPM) to explain the presence of functionally redundant genes in populations. They discuss how genetic redundancy may indeed be disadvantageous to a population, as it reduces fitness sensitivity to mutations, thereby allowing deleterious mutations to accumulate. Yet they recognise that at the individual level, individuals with higher genetic robustness are better equipped to face environmental fluctuations (e.g. Mendonça et al., 2011). Their model seeks to address this imbalance between the population and individual by incorporating the complex nature of the relationship between the genotype and phenotype. Biological systems are error-tolerant scale-free networks (Wolf et al., 2002; Barabási & Oltvai, 2004; Albert, 2005), and under Saito et al.’s model, redundancy may only evolve under the proviso that the GPM is complex. According to their models, genetic redundancy in simpler GPMs tend to break down, perhaps explaining why simpler organisms (such as parasites) typically exhibit lower redundancy.

As briefly discussed, a popular theory surrounding the maintenance of genetic redundancy is that functionally redundant genes act as a ‘buffer zone’ to cope with internal molecular noise and stochastic mutation (Fig.1A) (Nowak et al., 1997; Wagner, 2000; Kafri et al., 2008; Mendonça et al., 2011). This buffer zone is hypothesised to operate as a failsafe against any potential deleterious mutations, and from an engineer’s perspective is entirely logical. However, although this hypothesis is attractive, a genetic buffer zone for the purposes of protection against mutation simply cannot be maintained by natural selection. Even though a buffer zone might be advantageous in the long term, natural selection has no foresight and a buffer zone fails to provide any immediate benefits. At best, the selection pressures would be too weak to maintain identical genes in the genome that would otherwise be free to mutate over time. Models that incorporate a genetic buffer zone often stipulate a high mutation rate in one of the functionally redundant genes (see Nowak et al. 1997), yet this means such a gene would be lost through an accumulation of heritable deleterious mutations. Indeed, functionally redundant non-coding DNA (such as some short tandem repeats) tend to have high mutation rates due to low selection pressures (Schlötterer et al., 1998), which is exactly what should be expected of ‘junk’ portions of the genome. This issue may be overcome, however, when one considers the persistence of redundant genes across neighbouring populations. Stabilising selection may maintain two redundant genes at high frequencies if each gene possesses a higher selective advantage than the other in different environments (as discussed in Unkles et al., 2001). Heterozygosity in a single organism may therefore be achieved by coincidence of the two functionally overlapping genes being brought together by drift. Nonetheless, this is quite different to selection maintaining a genetic buffer zone as a precautionary failsafe against mutation.

Molecular Genetics. As well as failing to provide any immediate advantage, redundant DNA may actually be disadvantageous and selected against, as it is costly to replicate and transcribe DNA, especially if that DNA is wasteful and bears no purpose. As previously discussed, selection favours smaller genome sizes for optimal efficiency, as replicating DNA expends energy (Denver et al., 2004). This is evidenced by the fact that obligate endosymbionts tend to have smaller genome sizes, as natural selection has opted for efficiency with a reduced phenotype (Andersson & Kurland 1998). Genome streamlining has also been shown to occur in species other than endosymbionts, such as the free-living oceanic bacterium Pelagibacter ubique (Giovannoni et al., 2005). Hence, selection is expected to remove wasteful portions of the DNA, acting to attain a more concentrated genome. It may be argued in response, however, that parasitic elements of DNA (such as endogenous retroviruses (ERVs)) are able to persist within the genome, so perhaps natural selection is too weak to operate at this level. Yet viral DNA has evolved over time in an evolutionary arms race to overcome natural cell defences in order to replicate itself (Stoye, 2012), and some viral DNA may actually be functional, such as the ERVs discovered to play an integral role in placental formation in humans (Blond et al., 2000; Muir et al., 2004). Furthermore, examples of intragenomic conflict demonstrate the power of natural selection in its capabilities to counteract ultra-selfish elements (Burt & Koufopanou 2004). Natural selection is also powerful enough to relocate loci in the genome and cluster co-functional genes together, thereby giving rise to neighbourhood continuity (Boutanaev et al., 2002; Pál & Hurst, 2003; Meadows et al., 2010). This concentration/clustering of non-homologous functionally-related genes has been discovered in all species studied thus far (Lee & Sonnhammer, 2003), indicating the power of natural selection at the molecular level. Therefore, selection is more than capable of relocating costly and wasteful portions of the genome to non-essential areas more prone to mutation, or perhaps even of removing them altogether.

Moreover, Qian et al. (2010) have suggested that dosage compensation may exist, but that this is not a case of true redundancy. Instead, they propose that functional overlap is a distinctive example of subfunctionalisation, but without the functional divergence. For example, the researchers demonstrated that following a specific gene duplication in mammals (i.e., mice and humans), each daughter copy underwent reduced expression in order to readjust expression levels to normal. This is thought to arise to avoid saturation of gene products, as over-expression of genes can often have negative ramifications on fitness (Papp et al., 2003). Consequently, expression of both genes is reduced, resulting in expression levels that are comparable to those prior to the duplication event (Fig.2). Carrying both these genes will now be advantageous, as knocking out one will result in sub-optimal gene expression. However, as there will still likely be some degree of expression from a single copy (albeit reduced expression), its effect on the phenotype may be negligible and difficult to detect. Yet it still constitutes a less favourable level of gene expression that would normally be required for optimum fitness, and it would be disadvantageous for an organism to be hemizygous. This suggests a speculative explanation for some of the ‘functionally redundant’ genes reported in the literature (Qian et al., 2010), yet a number of assumptions must be met in order for this model to work. For example, assumptions include an availability of promoters to transcribe both new daughter genes, which may not always be the case. It also assumes saturation of gene products to be deleterious, when in some circumstances it may be beneficial (e.g., in the synthesis of particular metabolites).

Evidence for the functional redundancy of certain genes may also be brought into question. Whilst there are myriad examples of apparent redundancy when knock-out experiments are conducted, one may highlight potential problems in the methodologies of these experiments. Many experiments are conducted in a laboratory environment, with nutrient-rich mediums or highly favourable conditions/environments for the organisms being studied. Moreover, most experiments are conducted on a mere snapshot of an organism’s life cycle in more complex multicellular species. As such, reporting the removal of a particular gene as having no measurable phenotypic effect may be rather premature, as (A) these genes may actually serve a functional role in an environment other than the laboratory; or (B) the gene of interest may not prove essential at that particular stage the researchers are studying, but may be necessary earlier on in development or at a later stage such as during reproduction. Many studies may have fallen into the trap of concluding functional redundancy in experiments that use optimal laboratory conditions. The example discussed earlier by Giaever et al. (2002) does just this in discussing how mutations in approximately 83% of the yeast genome caused no obvious phenotype when grown in a rich medium. However, Hillenmeyer et al. (2008) reappraised this in an elegant experiment involving large-scale knock-outs of the genes in S.cerevisiae and subjecting the colonies to over a thousand chemical assays. This procedure was designed to emulate the potential environmental variations that yeast may experience. Overall, the researchers found that 97% of the gene deletions they generated produced a measurable effect on culture growth. The researchers concluded that almost all genes in yeast are necessary for optimum growth across many different environments, and one may speculate over the remaining 3%.

Consequently, this reasoning may be extended to many of the ‘redundant’ genes described in the literature. Such genes may not actually be redundant after all, but only non-essential in the limited scope of the experiments they were studied in. This is evident through the pleiotropic nature of many genes, the polygenic nature of many complex phenotypic traits, and the regulatory factors controlling genes that are dependent upon a host of spatiotemporal variables. A pleiotropic gene may not exhibit function in the trait that the researchers are measuring, but may have a subtle effect in another genetic pathway regulated at specific intervals in an organism’s life cycle. Alternatively, genes that cooperate in polygenic traits may be interdependent, but the effect of a single gene deletion may only be measurable if that trait is under stress in some way. This has even been hinted in the literature by some of models discussed above, e.g. it has been suggested that variable environments select in favour of genetic robustness, and may provide the selection pressures necessary to maintain functional overlap between genes (Mendonça et al., 2011). Variable environments tend to subject an organism to different stresses, and thus genes that may appear redundant in the laboratory could be otherwise essential under stressful conditions.

Future Directions

As discussed, there are many issues with the methods presently used in functional genomics, which may have led to erroneous identification of genetically ‘redundant’ genes. New technology and techniques may be pivotal in elucidating the relationship between the genotype and phenotype by facilitating more powerful and accurate ways of determining a gene’s function. Next generation sequencing (NGS) is perhaps the most relevant example through its capabilities to profile entire genomes in a matter of days (Mardis, 2008; Schuster, 2008). Modern sequencing methods, such as Illumina and SOLiD, can achieve high-throughput sequencing (Morozova & Marra, 2008) to accurately map multiple genomes from a species, allowing cross-comparisons within a population. This so-called ‘power sequencing’ (Graveley, 2008) therefore has the power to further address some of the issues we face surrounding genetic redundancy, with a profound potential to enhance our understanding of the relationship between genotype and phenotype (Mardis, 2008). For example, this potential may be realised through experiments utilising chromatin immunoprecipitation (ChIP) analysis, as sequencing methods are capitalised upon to identify target sequences of DNA fragments pulled through from protein interactions (Schmid & Bucher, 2007; Shendure & Ji, 2008). This will facilitate a greater understanding of transcriptional regulation in identifying the function of specific protein-coding genes. Understanding the regulation of gene expression may well be the key to understanding gene function, and advances in NGS can only aid in this field.

Focus on the transcriptome may also assist further research into functional genomics, and can be achieved by utilising NGS of cDNA as opposed to genomic DNA (Morozova & Marra, 2008). Microarrays are commonly utilised to accomplish this by binding cDNA to short oligonucleotide sequences allowing efficient and visual representation of gene expression from target cells/areas (Schena et al., 1995). The use of microarrays, whilst falling short of yielding direct information on function, can and has nonetheless been used in a functional context to identify target genes for a specific phenotype. For example, sequencing cDNA from different rose cultivars and generating ESTs specific to each ecotype can be used in tandem with microarrays to accurately identify genes that strongly correlate with scent production (Guterman et al., 2002). Common techniques employed in reverse genetics may then be used to mutate these candidate genes to determine whether they truly do have an effect on scent production, thereby providing empirical data for function. Correlating genotypes with phenotypes in this context has the potential to link genes whose function is otherwise unknown with potential phenotypes that it may influence. Future studies on genetic redundancy might therefore consider measuring for potential correlations between ‘functionally redundant’ sequences and particular phenotypes.

ORF finders have also proven to be an exceptional utility in the arsenal of geneticists, who may scan entire genome sequences to identify potential genes of interest (Rombel et al., 2002). Researchers may annotate these ORFs with information regarding the potential protein it may encode, or where the introns and exons may lie. Homology can easily be drawn between species, and many studies on the function of particular genes utilise ORF finders (e.g. see: Unkles et al., 2001; Giaever et al., 2002; Muir et al., 2004; Fievet et al., 2012). However, they do have significant limitations that should not be overlooked. ORF finders merely provide a starting point for research on a particular sequence, and should not be relied upon too heavily. For example, they can routinely generate false positives (Rombel et al., 2002; Larsen & Krogh, 2003), and as such should be used with caution in functional genomics. Thus, knocking-out ORFs identified by an ORF finder may in fact prove ineffectual due to erroneous identification, perhaps leading to invalid conclusions concerning function and redundancy.

Research from slightly further afield may also help to shed light on the nature of genetic redundancy. The ENCODE project has been working over the last decade to investigate the function of non-coding DNA, and 30 publications last September in Science, Nature, and elsewhere wholly challenged the notion of ‘junk DNA’ (Dunham et al., 2012). Research has shown that sequences previously dismissed simply as being ‘junk’ actually bear an integral role in transcriptional regulation (Muro et al., 2011; Wang et al., 2011) and that structural organization of the genome plays a fundamental role in determining gene expression (and hence the overall phenotype of an organism (Conaway, 2012)). Thus, non-coding regions previously dismissed as being functionally redundant have been demonstrated to be crucial in the evolution of certain species. It is plausible that the same is true for so-called redundant genes, which may have been prematurely dismissed as being functionally redundant.

Functional redundancy has to be considered in both the contexts of population and molecular genetics, and here the evidence for each has been critically analysed. Numerous models have been proposed to explain the persistence of apparently functionally redundant genes, however, many tend to overlook the short-sighted nature of natural selection. For instance, many discuss the idea of a genetic ‘buffer zone’ acting as a long-term failsafe against any potential deleterious mutations, yet such a buffer zone provides no short-term advantages. Yet, similar hypotheses regarding genetic robustness are perhaps more promising, with genetic redundancy being favoured in variable and unpredictable environments.

Moreover, few models actually deal with how genetic redundancy might arise to begin with, with most tending to focus on duplication events. This is one area that future studies should seek to address, perhaps through studying how functional overlap has convergently evolved between unrelated genes. Additionally, the designs of some experimental methods used to examine gene function have significant limitations that may have led to erroneous conclusions regarding the role of certain genes. Laboratory experiments are far from perfect, and future research should consider studying organisms in natura, perhaps to F1 and F2 generations, to strengthen experimental validity. As a consequence, genes that appear to be functionally redundant may eventually be shown to have a purpose after all, and in light of appreciating their potential importance, we may begin to understand deeper aspects of the genome that go otherwise unexplored.

Conclusions

Is the term ‘functional redundancy’ functionally redundant? In the light of recent advances in genetics and the arguments put forward here, we should reconsider our classification of certain genes as ‘functionally redundant’ as conclusions regarding gene function (or a lack thereof) may be somewhere premature. Examples of gene compensation and functional overlap are certainly ubiquitous in nature, and there are numerous models explaining their persistence in the genome. However, discussions on this topic often appear to conflate gene compensation and functional overlap with the term functional redundancy. Yet, I believe that in these circumstances the term functional redundancy is somewhat of a misnomer as it implies that a gene has no purpose in the DNA. This may lead to confusion for the layman, and has even be hijacked by creationists who use it as ‘evidence’ that natural selection cannot occur, as apparently functionless portions of the DNA are able to persist in a largely invariable state within the genome (e.g., see http://creation.com/genetic-redundancy). Therefore, I propose that as scientists we should instead aspire to use the existing and more accurate terminology (such as functional overlap and gene compensation) when describing this phenomenon to ensure greater clarity. I suspect that as our appreciation of the nature of the genome proceeds to grow, and we begin to recognize the vast complexity and scale of the networks present, we will continue to find more and more examples of functional overlap. Understanding these intricate networks may hold the key to a greater comprehension of the interactions between genotype and phenotype, and continuing to use terms such as functional redundancy may only hinder our progress.

Literature Cited

Albert, R. (2005) Scale-free networks in cell biology. Journal of cell science, 118(21), pp.4947–57.

Andersson, S. & Kurland, C. (1998) Reductive evolution of resident genomes. Trends in microbiology, 6(7), pp.263–8.

Barabási, A.L. & Oltvai, Z.N. (2004) Network biology: understanding the cell’s functional organization. Nature Reviews Genetics, 5(2), pp.101–13.

Blond, J.L. et al. (2000) An envelope glycoprotein of the human endogenous retrovirus HERV-W is expressed in the human placenta and fuses cells expressing the type D mammalian retrovirus receptor. Journal of Virology, 74(7), pp.3321–9.

Borger, P. (2008) Evidence for the Design of Life: Part 1 – Genetic Redundancy [online]. Creation, Creation Ministries International. Available from: http://creation.com/genetic-redundancy [Accessed 29th April 2013]

Boutanaev, A. et al.  (2002) Large clusters of co-expressed genes in the Drosophila genome. Nature, 420(6816), pp.666–9.

Brazhnik, P., De la Fuente, A. & Mendes, P. (2002) Gene networks: how to put the function in genomics. Trends in biotechnology, 20(11), pp.467–72.

Burt, A. & Koufopanou, V. (2004) Homing endonuclease genes: the rise and fall and rise again of a selfish element. Current opinion in genetics & development, 14(6), pp.609–15.

Conant, G.C. & Wagner, A. (2004) Duplicate genes and robustness to transient gene knock-downs in Caenorhabditis elegans. Proceedings of the Royal Society B: Biological Sciences, 271(1534), pp.89–96.

Conaway, J.W. (2012) Introduction to theme “Chromatin, epigenetics, and transcription”. Annual review of biochemistry, 81, pp.61–4.

Cooke, J. et al. (1997) Evolutionary origins and maintenance of redundant gene expression during metazoan development. Trends in Genetics, 13(9), pp.360–4.

DeLuna, A. et al. (2008) Exposing the fitness contribution of duplicated genes. Nature Genetics, 40(5), pp.676–81.

Denver, D.R. et al. (2004) High mutation rate and predominance of insertions in the Caenorhabditis elegans nuclear genome. Letters to Nature, 430(7000), pp.679–82.

Donoghue, P.C.J. & Purnell, M. (2005) Genome duplication, extinction and vertebrate evolution. Trends in ecology & evolution, 20(6), pp.312–9.

Dunham, I. et al. (2012) An integrated encyclopedia of DNA elements in the human genome. Nature, 489(7414), pp.57–74.

Fievet, B.T. et al. (2012) Systematic genetic interaction screens uncover cell polarity regulators and functional redundancy. Nature cell biology, 15(1), pp.103–12.

Fraser, G.J. et al. (2010) The odontode explosion: the origin of tooth-like structures in vertebrates. BioEssays : news and reviews in molecular, cellular and developmental biology, 32(9), pp.808–17.

Giaever, G. et al. (2002) Functional profiling of the Saccharomyces cerevisiae genome. Nature, 418(6896), pp.387–91.

Giovannoni, S.J. et al. (2005) Genome streamlining in a cosmopolitan oceanic bacterium. Science, 309(5738), pp.1242–5.

Graveley, B.R. (2008) Power Sequencing. Nature methods, 453(7199), pp.1197–1198.

Guterman, I. et al. (2002) Rose Scent : Genomics Approach to Discovering Novel Floral Fragrance – Related Genes. The Plant Cell, 14(10), pp.2325–38.

Hillenmeyer, M.E. et al. (2008) The chemical genomic portrait of yeast: uncovering a phenotype for all genes. Science, 320(5874), pp.362–5.

Hoffmann, F.G., Opazo, J.C. & Storz, J.F. (2012) Whole-genome duplications spurred the functional diversification of the globin gene superfamily in vertebrates. Molecular biology and evolution, 29(1), pp.303–12.

Hoffmann, F.M. (1991) Drosophila abl and genetic redundancy in signal transduction. Trends in Genetics, 7(11), pp.351–5.

Jin, H. & Nikolau, B.J. (2012) Role of genetic redundancy in polyhydroxyalkanoate (PHA) polymerases in PHA biosynthesis in Rhodospirillum rubrum. Journal of bacteriology, 194(20), pp.5522–9.

Kafri, R. et al. (2008) Preferential protection of protein interaction network hubs in yeast: evolved functionality of genetic redundancy. Proceedings of the National Academy of Sciences of the United States of America, 105(4), pp.1243–8.

Kafri, R., Springer, M. & Pilpel, Y. (2009) Genetic redundancy: new tricks for old genes. Cell, 136(3), pp.389–92.

Larsen, T.S. & Krogh, A. (2003) Easy-Gene – a prokaryotic gene finder that ranks ORFs by statistical significance. BMC Bioinformatics, 4(21).

Lee, J.M. & Sonnhammer, E.L.L. (2003) Genomic gene clustering analysis of pathways in eukaryotes. Genome research, 13(5), pp.875–82.

Liu, B. et al. (2008) Genetic redundancy in soybean photoresponses associated with duplication of the phytochrome A gene. Genetics, 180(2), pp.995–1007.

Mardis, E.R. (2008) The impact of next-generation sequencing technology on genetics. Trends in genetics, 24(3), pp.133–41.

McBride, C.S., Arguello, J.R. & O’Meara, B.C. (2007) Five Drosophila genomes reveal nonneutral evolution and the signature of host specialization in the chemoreceptor superfamily. Genetics, 177(3), pp.1395–416.

McGinnis, W. & Krumlauf, R. (1992) Homeobox genes and axial patterning. Cell, 68(2), pp.283–302.

Meadows, L. a et al. (2010) Neighbourhood continuity is not required for correct testis gene expression in Drosophila. PLoS biology, 8(11), p.e1000552.

Mendonça, A.G., Alves, R.J. & Pereira-Leal, J.B. (2011) Loss of genetic redundancy in reductive genome evolution. PLoS computational biology, 7(2), p.e1001082.

Morozova, O. & Marra, M. (2008) Applications of next-generation sequencing technologies in functional genomics. Genomics, 92(5), pp.255–64.

Muir, A., Lever, A. & Moffett, A. (2004) Expression and functions of human endogenous retroviruses in the placenta: an update. Placenta, 25(Suppl A), pp.S16–25.

Muro, E.M., Mah, N. & Andrade-Navarro, M. (2011) Functional evidence of post-transcriptional regulation by pseudogenes. Biochimie, 93(11), pp.1916–21.

Nishida, E. & Gotoh, Y. (1993) The MAP kinase cascade is essential for diverse signal transduction pathways. Trends in biochemical sciences, 18(4), pp.128–31.

Nowak, M. et al. (1997) Evolution of genetic redundancy. Nature, 388(6638), pp.167–71.

Pál, C. & Hurst, L.D. (2003) Evidence for co-evolution of gene order and recombination rate. Nature genetics, 33(3), pp.392–5.

Pastor-Satorras, R. & Vespignani, A. (2001) Epidemic Spreading in Scale-Free Networks. Physical Review Letters, 86(14), pp.3200–03.

Pérez-Pérez, J.M., Candela, H. & Micol, J.L. (2009) Understanding synergy in genetic interactions. Trends in genetics, 25(8), pp.368–76.

Pickett, F.B. & Meeks-Wagner, D.R. (1995) Seeing double: appreciating genetic redundancy. The Plant cell, 7(9), pp.1347–56.

Qian, W. et al. (2010) Maintenance of duplicate genes and their functional redundancy by reduced expression. Trends in genetics, 26(10), pp.425–30.

Rombel, I.T. et al. (2002) ORF-FINDER: a vector for high-throughput gene identification. Gene, 282(12), pp.33–41.

Rong, Y.S. & Golic, K.G. (2001) A targeted gene knockout in Drosophila. Genetics, 157(3), pp.1307–12.

Saito, N., Ishihara, S. & Kaneko, K. (2013) Evolution of Genetic Redundancy : The Relevance of Complexity in Genotype-Phenotype Mapping. arXiv preprint, 1, pp.1–5.

Schena, M. et al. (1995) Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science, 270(5235), pp.467–70.

Schlötterer, C. et al. (1998) High mutation rate of a long microsatellite allele in Drosophila melanogaster provides evidence for allele-specific mutation rates. Molecular biology and evolution, 15(10), pp.1269–74.

Schmid, C.D. & Bucher, P. (2007) ChIP-Seq data reveal nucleosome architecture of human promoters. Cell, 131(5), pp.831–2; author reply 832–3.

Schuster, S.C. (2008) Next-generation sequencing transforms today’s biology. Nature Methods, 5(1), pp.16–18.

Shendure, J. & Ji, H. (2008) Next-generation DNA sequencing. Nature biotechnology, 26(10), pp.1135–45.

Skarnes, W.C. et al. (2011) A conditional knockout resource for the genome-wide study of mouse gene function. Nature, 474(7351), pp.337–42.

De Smet, R. & Van de Peer, Y. (2012) Redundancy and rewiring of genetic networks following genome-wide duplication events. Current opinion in plant biology, 15(2), pp.168–76.

Stoye, J.P. (2012) Studies of endogenous retroviruses reveal a continuing evolutionary saga. Nature reviews. Microbiology, 10(6), pp.395–406.

Unkles, S.E. et al. (2001) Apparent genetic redundancy facilitates ecological plasticity for nitrate transport. The EMBO journal, 20(22), pp.6246–55.

Vavouri, T., Semple, J.I. & Lehner, B. (2008) Widespread conservation of genetic redundancy during a billion years of eukaryotic evolution. Trends in genetics, 24(10), pp.485–8.

Wagner, A. (2000) Robustness against mutations in genetic networks of yeast. Nature genetics, 24(4), pp.355–61.

Wang, X. et al. (2011) The long arm of long noncoding RNAs: roles as sensors regulating gene transcriptional programs. Cold Spring Harbor perspectives in biology, 3(1), p.a003756.

Wolf, Y.I., Karev, G. & Koonin, E. V (2002) Scale-free networks in biology: new insights into the fundamentals of evolution? BioEssays : news and reviews in molecular, cellular and developmental biology, 24(2), pp.105–9.

Yang, J. et al. (2013) Ubiquitous Polygenicity of Human Complex Traits: Genome-Wide Analysis of 49 Traits in Koreans S. M. Williams, ed. PLoS Genetics, 9(3), p.e1003355.

Zhang, J. (2003) Evolution by gene duplication: an update. Trends in Ecology & Evolution, 18(6), pp.292–8.

Zhang, S. et al. (1997) Genetic redundancy between SPT23 and MGA2: regulators of Ty-induced mutations and Ty1 transcription in Saccharomyces cerevisiae. Molecular and Cellular Biology, 17(8), pp.4718–29.