TIGS-1210; No. of Pages 9

Review

The origins of mutational robustness Mario A. Fares1,2 1 2

Instituto de Biologı´a Molecular y Celular de Plantas (CSIC-UPV), Valencia, Spain Department of Genetics, Smurfit Institute of Genetics, University of Dublin, Trinity College Dublin, Dublin, Ireland

Biological systems are resistant to genetic changes; a property known as mutational robustness, the origin of which remains an open question. In recent years, researchers have explored emergent properties of biological systems and mechanisms of genetic redundancy to reveal how mutational robustness emerges and persists. Several mechanisms have been proposed to explain the origin of mutational robustness, including molecular chaperones and gene duplication. The latter has received much attention, but its role in robustness remains controversial. Here, I examine recent findings linking genetic redundancy through gene duplication and mutational robustness. Experimental evolution and genome resequencing have made it possible to test the role of gene duplication in tolerating mutations at both the coding and regulatory levels. This evidence as well as previous findings on regulatory reprogramming of duplicates support the role of gene duplication in the origin of robustness. Robustness to mutations and its role in evolution All biological systems are resistant to genetic variation and environmental changes; a property known as robustness [1,2]. That is, despite significant changes in the inputs to a system (e.g., genetic or environmental variation), the outputs (e.g., phenotypes) after the perturbation are equivalent to those before the perturbation of the system. Waddington was amongst the first to realize that developmental programs, such as the development of wings in Drosophila, are generally resilient to minor perturbations caused by environmental changes, including heat stress or osmotic stress, and he called this property canalization [3–5]. Since the work of Waddington, robustness has become synonymous with canalization. Subsequent work has shown that robustness is not restricted to development but is a universal property of many levels of complexity. At the molecular level, systematically mutating bacteriophage T4 lysozyme showed that the enzyme continued to function more than half the time after testing 2015 single amino acid mutations. This robustness likely accounts for the persistence of large protein variability observed in T4 populations [6,7]. In metabolic networks, studies based on flux balance analyses have been performed focusing on gene products from essential metabolic networks of the Corresponding author: Fares, M.A. ([email protected], [email protected]). Keywords: mutational robustness; gene duplication; evolutionary capacitors; experimental evolution; evolvability; regulatory plasticity. 0168-9525/ ß 2015 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tig.2015.04.008

bacterium Escherichia coli, including glycolysis, pentose phosphate, and tricarboxylic acid pathways. Such studies have revealed that the flux could be brought down to levels as low as 15% for enzymes of the pentose phosphate pathway, and 19% in the tricarboxylic cycle acid reactions without compromising optimal cellular growth, suggesting that metabolic networks are largely robust to fluctuations in the reactions substrates or efficiencies [8]. Gene expression is often subject to noise but the level of expression noise depends on the location of the gene in the genome [9,10]. Indeed, the expression of essential genes is robust (i.e., they maintain low noise levels in their expression) in the face of internal cellular factors that often alter gene expression, including the switching of chromatin between open and closed states [9,10]. The robustness of essential genes to factors that introduce expression noise may have a selection basis because essential genes are often clustered and located in genome regions with open chromatin organization [11]. Cells are also robust to single gene deletions [12,13]. Knockout strains of the yeast Saccharomyces cerevisiae exist for 96% of open reading frames, indicating a remarkable tolerance to single gene deletions, although many of these genes are required under certain growth conditions [12].

Glossary Cryptic genetic variation: genetic variants that differ from the most abundant genotypes in the population that are phenotypically silent causing no effects on the fitness of individuals carrying such variation. Evolvability: capacity to generate heritable phenotypic variation that may be adaptive in a particular context [90]. Exaptation: a trait that serves a particular function in the current context but that may serve another function in another context [91]. Genetic interactions: also known as epistasis and refers to the dependent effect of deleting one gene based on the presence of one or more modifier genes. Deletion of two interacting genes from an organism would lead to effects that are significantly more compensatory or aggravating than the multiplicative effects of single gene deletions. Mutational robustness: is the extent to which the phenotype of an organism (i.e., morphology or functional performance) remains constant in spite of mutations to its genotype. Neofunctionalization: another of the fates proposed by the classic view of gene duplication is that, while one copy preserves the ancestral function, the other, devoid of selective pressure, can explore alternative functions and innovate. Paralogs: genes from the same species that share the same ancestry. Partitioning of ancestral functions: when a multifunctional gene (e.g., a gene with a catalytic and regulatory functions) duplicates, both of the copies can mutate asymmetrically in two different functional domains, such that one copy keeps one of the ancestral functions (e.g., regulatory), and the other keeps the complementary ancestral function (e.g., catalytic). Phenotypic plasticity: ability of organisms to change their phenotype, maintaining the same genotype, with changes in the environment [92]. Subfunctionalization: Ohno [46,49] proposed that after gene duplication, asymmetric mutations in the resulting gene copies could lead to a partitioning of ancestral functions.

Trends in Genetics xx (2015) 1–9

1

TIGS-1210; No. of Pages 9

Review At the level of development, there is mounting evidence of robustness to perturbations in the signals that tissues receive extrinsically. For example, during Drosophila oogenesis, follicle development is guided in part by the coordinated growth rate of both the germline and somatic cell lines as well as by extrinsic signals. Despite this tight coordination and communication, a recent study found that mature eggs with the normal size and shape can be produced in the context of aberrant extrinsic signaling, highlighting the overlapping layers of regulation that make process robust [14]. Despite the ubiquity of robustness, the mechanisms that give rise to robustness, specifically mutational robustness (see Glossary), and the question of whether mutational robustness is a selectable, and hence evolvable, trait remain under intense debate [15,16]. Of the various manifestations of robustness, the ability to withstand mutations is of particular interest because it has implications for evolution. Mutational robustness and the origin of innovations are linked through phenotypically neutral variants of a gene in the population. In robust systems, many variants of a gene can be tolerated while maintaining the same phenotype (i.e., these genetic variants are neutral). When these genotypic variants are connected through single mutational steps (i.e., two neighboring genotypes are separated by a single mutation) they

Trends in Genetics xxx xxxx, Vol. xxx, No. x

form a network called genotypic network [2]. In a robust genotypic network, the transition between genotypes is phenotypically silent (i.e., the two connected genotypes encode the same phenotype). However, subsequent mutations in a particular genotypic background may cause new phenotypic manifestations. Within a given population, there will be many genotypic backgrounds, in some of which a new mutation will have a phenotypic effect and in others it will not. The number of different phenotypes emerging from new mutations will increase with the number of neutral genotypes (i.e., possible backgrounds) in the population. Consequently, if there are a large number of neutral genotypes, the population can access more new phenotypes [2,17,18]. Thus, the larger the genotypic network (i.e., the larger the mutational robustness of a population), the higher the potential is to produce novel adaptations [19]. The details however matter: increasing robustness up to intermediate levels increases evolvability, but when a set of possible accessible phenotypes has reached its maximum, adding more genotypes to the genotypic network does not increase evolvability (Box 1). Several studies have shown a relation between the size of the genotypic network and the potential access to different phenotypes. For example, analyses of genotypic networks for protein structures reveal that different regions of the network provide access to different neighboring structures

Box 1. Robustness and evolvability

2

(B)

(C)

(A)

0.9

Evolvability

The link between robustness and evolvability is subject to certain limitations, with intermediate levels of robustness yielding higher evolvability, as a trade-off exists between increasing the diameter of the genotypic network and the overlap between phenotypes accessible by all the genotypes in the network. As this overlap increases, the correlation between robustness and evolvability declines (Figure I) [19]. Take for instance two genotypic networks generated by a number of genotypic backgrounds (Figure IA). Genotypic backgrounds within the same network are connected through single mutations, and mutations in a specific genotypic background that lead to a new genotype from the same genotypic network has no phenotypic effect. Conversely, a mutation that leads to a genotypic background from a different network produces a different phenotype. If we measured evolvability as the ratio between the number of accessible phenotypes (in this case two, Figure IA, green and black) and the number of genotypes in the black network (three genoytpes that lead to black phenotype), this would yield a value of 0.66 (2 phenotypes/3 genotypes = 0.66) in our example. Increasing the robustness of the black network is equivalent to increasing the number of genotypic backgrounds that yield to the same phenotype (Figure IB). New mutations in this wider genotypic network are more likely to provide new phenotypes, each of which is encoded in a different genotype network (Figure IB). For example, in Figure IB, the black phenotype is encoded by two additional genotypes compared to Figure IA, but this increase allows accessing three phenotypes (green, brown, and red) through subsequent mutations. In this case, the evolvability of the black network has increased compared to that in Figure IA (4 phenotypes/5 genotypes = 0.8). Increasing the genotypic network by two additional genotypes (Figure IC) allows certain nodes access to more than one phenotype as a result of exhausting the full phenotypic space, which leads to the overlap between the set of accessible phenotypes for genotypes of the same network. This phenotypic overlap between genotypes means that while the robustness of the system increases its evolvability does not (4 phenotypes/7 genotypes = 0.57). In conclusion, increasing robustness increases evolvability [90–92] as long as the genotypes of the same genotypic network cannot access all possible phenotypes.

0.8 0.7 0.6 0.5 0.4

a

b

c TRENDS in Genetics

Figure I. Intermediate levels of mutational robustness increases evolvability. Neutral genotypic networks are those in which several genotypes (circles) are connected through single mutations and lead to the same phenotype (phenotype is symbolized with the color of the network). A small genotypic network (A) has a low potential to evolve novel phenotypes (e.g., network a presents an evolvability of 0.67). As the network increases in size (B), the number of accessible phenotypes through subsequent single mutations increases disproportionally more (e.g., evolvability of network B has increased 13% by adding two additional genotypes to the network). Large genotypic networks (e.g., high robustness, such as in genotype C), decreases evolvability (network C has decreased its evolvability to 0.50) because genotypes overlap in the space of their accessible phenotypes.

TIGS-1210; No. of Pages 9

Review [18,20]. This phenomenon also seems to act in metabolic networks, where different regions of a genotypic network provide access to different alternative metabolisms [18]. Although the link between the amount of genetic variation in the population and evolvability is clear, it is less clear whether robustness is subject to natural selection. Is robustness a selected trait? One of the most controversial questions surrounding the origin and evolution of robustness is whether robustness is under selection. The evidence for the adaptive evolution of robustness in biological populations remains largely debated [21]. There are a number of possibilities to explain the origin of robustness. Robustness to mutations can be an intrinsic emergent property of a system. Take for instance the case of protein–protein interaction networks, where a majority of nodes have a low number of connections, while a minority of them has a large number of connections. This network geometry provides robustness to the random failure of single nodes, as the probability of failure would be higher for a less-connected node than for a highly connected node since there are more less-connected nodes. Random removal of less-connected nodes would not disrupt the network because it would affect only a low number of links. Robustness to mutations is also thought to be a byproduct of environmental robustness. Such claims stem from the fact that environmental and mutational robustness is correlated in a number of situations. For example, analyses of secondary RNA structures showed high and inherent correlation between environmental and mutational robustness [22]. In a bacterial evolution experiment, the enzymatic activity of tumor endothelial marker 1 (TEM-1) b-lactamase was selected such that it was retained under error-prone transcription (representing the environmental perturbation in this experiment). This gene presented increased mutational robustness in a correlated manner with its environmental robustness [23]. Despite experiments pinpointing the correlation between environmental and mutational robustness, under some circumstances, mutational robustness has an adaptationist origin. For a well-adapted population, further mutations would inevitably lead to deviations from the optimal performance of the population in a selective environment. Robustness to new mutations, whatever the mechanism of robustness was, would be favored by selection. However the evidence of the high correlation of environmental and mutational robustness brings this scenario into question. A number of studies have shown, nevertheless, that under particular conditions, including high mutation rates and large population sizes, robustness may be a selected trait [24,25]. There is also some indirect evidence that mutational robustness is independent of environmental robustness. For example, gene expression polymorphism that provides robustness to genetic (among lines) variation is different from that providing resistance to environmental (within lines) variation [26]. Another study in which E. coli was subjected to both environmental and genetic perturbations showed that mutational and environmental robustness were anticorrelated [27]. Therefore, while there is evidence supporting a neutral origin for robustness, some other evidence pinpoints the role of

Trends in Genetics xxx xxxx, Vol. xxx, No. x

selection in the evolution of robustness. It is plausible, nevertheless, that robustness is a selectable trait under specific conditions. Origins and mechanisms of mutational robustness One area of active debate is whether certain mechanisms or gene products modulate the extent of mutational robustness in a way that is evolutionarily adaptive. A role for such gene products, referred to as evolutionary capacitors, in modulating robustness has been questioned [28]. However, it is clear that increasing the tolerance of a biological system to deleterious mutations increases the genetic reservoir of the population and facilitates the emergence of exaptations. That is, increasing robustness expedites the emergence of innovations by enhancing the ability of the system to generate heritable genetic variation; a property known as evolvability. In this sense, mechanisms that increase robustness have a key role in evolution. Chief among the mechanisms of robustness are those proposed to increase the amount of cryptic genetic variation in the population, including molecular chaperones and gene duplication. The consensus in the field is that when such mechanisms are impaired or disrupted, previously neutral genetic variation is no longer buffered and manifests in the form of novel morphologies or functions, some of which may lead to new adaptations. This relationship between cryptic genetic variation and phenotypic diversification has been revealed in a number of model organisms, including the bacterium E. coli [29–31], the budding yeast S. cerevisiae [32–34], and the nematode worm Caenorhabditis elegans [35,36]. The molecular chaperone heat shock protein (Hsp)90 has been proposed to be a capacitor of evolution by increasing mutational robustness. Hsp90 folds proteins involved in the signal transduction pathway. Disruption or impairment of Hsp90 through heat stress or pharmacological inhibition reveals cryptic genetic variation in a wide range of distant species [37–39]. This phenotypic emerging variation is heritable [37,38] and can lead to novel adaptations in natural populations [40]. Similarly, changes in the light conditions (e.g., light becomes rare) of natural surface populations of the cavefish Astyanax mexicanus can reveal phenotypic variation in the form of eyeless fishes that was silent in the previous luminous conditions allowing the emergence of adaptations to living in cave environments, where eye development is costly and useless [40]. It should be noted, however, that recent work has urged caution when interpreting the phenotypic variation caused by impairing Hsp90, as the relation between impairment of Hsp90 and the phenotypic manifestation of cryptic genetic variation does not indicate a decrease in robustness but may point to simple epistatic interactions between the new genotype (i.e., silencing of Hsp90) and the phenotype regardless of the effect on robustness [28]. By contrast, there is independent evidence supporting the role of Hsp90 as an evolutionary capacitor, as the gene copies of the duplicated kinases that interact with Hsp90 evolve faster than their sister gene copies that do not interact with Hsp90 [41]. GroEL, another molecular chaperone that folds many proteins in bacteria, was shown to increase the tolerance of a system to mutations. Experimental evolution of E. coli 3

TIGS-1210; No. of Pages 9

Review under strong genetic drift effects followed by overexpression of groE, the operon encoding GroEL and its cochaperone GroES, rescued evolved E. coli bacteria with declining fitness [42]. Moreover, the role of GroEL in expediting evolution is also supported by the finding that proteins that require GroEL for folding evolve at a higher rate (i.e., the ratio of non-synonymous-to-synonymous rates was higher) than proteins that can fold independently of GroEL [43,44]. Another study took advantage of the capacity of GroEL to buffer mutations to evolve four enzymes with weak specificities for alternative substrates in the presence of high concentrations of GroEL under mutational drift. Enhancing GroEL concentrations allowed the folding of enzyme variants carrying destabilizing mutations in their core, which increased enzyme specificity for alternative substrates [45]. Another way of originating robustness and increasing genetic variability in the population is through gene redundancy; a phenomenon mostly promoted by gene duplication [46]. Genetics redundancy often loosens selective constraints on redundant genes and allow them to acquire new genotypes. Therefore, genetic redundancy can originate robustness to mutations. Gene duplication has been credited with enormous importance in generating new functions and leading to large biological innovations. Determining the mechanism through which duplicated genes persist in the genomes and generate novel functions has been an enduring goal of evolutionary biology. However, the genetic, and hence functional, redundancy of duplicate genes makes organisms carrying them the target of purifying selection. The paradox of the persistence of duplicates in the genomes has gone unresolved for many decades. The controversial role of genetic redundancy in mutational robustness Genetic redundancy means that inactivation of one gene has little or no effect on the phenotype of the organism due to compensation by one or more other genes. Redundancy has two possible meanings. First, one part of a system can be removed without any phenotypic consequence owing to the compensation of the deletion by other parts with which it has no apparent functional overlap (Box 2). Examples of

Box 2. Distributed robustness In contrast to robustness through genetic redundancy, distributed robustness can exist in the absence of functional or genetic redundancy [93]. This kind of robustness is prominent in metabolic pathways, in particular those in which chemical reactions can allosterically inhibit the enzyme catalyzing the first reaction of the pathway. Furthermore, robustness can be shared or distributed between different pathways in metabolism, such that blockage or inhibition of one pathway can be phenotypically silent if another pathway yields a main metabolite of the silenced one, in spite of the fact that the two pathways may share no functions or enzymes. This type of redundancy is also found in developmental biology. For example, distributed redundancy arises from the crosstalk between the Ras and Notch pathways, which contribute to specifying cell types in development [94]. The activation of one of these two pathways and inhibition of the other is fundamental to determining the cell type. When Ras is activated, it leads to the degradation of Notch, whereas high Notch activity inhibits Ras pathway activity [95–98]. 4

Trends in Genetics xxx xxxx, Vol. xxx, No. x

this type of redundancy are prevalent among metabolic pathways, in which the removal of a reaction is compensated for by redirecting substrate fluxes through other reactions and quickly adapting to the new metabolic context. A case in point is that of the adaptive growth of the bacterium E. coli MG1655 to glycerol as a sole carbon source instead of glucose [47]. Similar results have been observed in studies adapting E. coli growth to multiple other carbon sources [48]. Second, redundancy is also achieved when two parts of a system back each other up by their partial or complete functional overlap. A number of studies support the role of genetic backup in robustness but others do not. To shed some light on the controversy, in this paper, I specifically and strictly refer to the second case of genetic redundancy, which is mainly represented by duplicated genes. In cases where gene duplication generates two identical gene copies, one of the copies may perform the required function while the other copy is likely to accumulate deleterious mutations [46,49]. It is therefore believed that genetic redundancy cannot be evolutionarily stable because the gene copy bearing deleterious mutations is likely to be removed by purifying selection. In support of this idea, following the whole-genome duplication (WGD) roughly 100 MYA in Saccharomyces species, 92% of the duplicates returned to single copy genes owing to the nonfunctionalization and erosion of their sister gene copies, and in many other cases one of the gene copies is silent under specific conditions, such as glucose starvation [50– 52]. This is not always the case, however, for example, some organisms, in particular plants, carry >30% of their genomes in duplicate [53,54]. This high retention may be simply due to the functional divergence between the two gene copies followed by strong and differential selective constraints on each copy [55–58]. Divergence of gene duplicates requires, nevertheless, long evolutionary times and can hardly account for the persistence of duplicated genes in the genome during the first million years following duplication. Moreover, a number of lines of evidence support a significant and preserved redundancy between the copies of a duplicated gene. For example, deletion of singletons in S. cerevisiae has larger effects on fitness than deletion of duplicates, and an estimated 25% of duplicates in S. cerevisiae show evidence of functional compensation between gene copies [59]. This functional compensation has also been demonstrated using genetic interaction studies in S. cerevisiae, in which the deletion of two copies of a duplicated gene has a larger effect than the multiplicative single deletion effects [60]. In humans, duplicates are less likely to harbor known disease mutations than singletons [61]. Finally, duplicates have been shown to be robust to transient knockdowns in C. elegans [62]. Duplicate genes that arise through WGD have been found to persist more often than those generated by small-scale duplications (SSDs), particularly for genes with many interactions. This is because WGDs do not upset the stoichiometric balance of the cell [63–65]. A number of other factors may help in the preservation of duplicated genes, including increased gene dosage [57], partitioning of ancestral functions between the gene copies [66–68], and functional divergence between the duplicate

TIGS-1210; No. of Pages 9

Review copies [69]. Notwithstanding cases in which expression and functional divergence have been detected for gene copies, some duplicated genes show substantial functional redundancy despite their ancient origin. For example, the duplicate gene pair SIR3 and ORC1 originated from the WGD in Saccharomyces yeast species, but SIR3 is involved in gene silencing and ORC1 functions as part of the origin recognition complex. However, their ortholog in Saccharomyces kluyveri, a species branching off prior to the WGD, functionally replaces both gene copies [70]. Contrary to these results, other studies have shown a limited contribution of gene duplicates to mutational robustness in laboratory [71] or natural conditions [72]. These disparate observations highlight the need for tractable systems to address the question of the role of genetic redundancy. Previous models have shown that redundancy can be maintained, or even favored by selection, if the duplicates overlap in a subset of functions and the mutation rate of one gene copy is higher than that of its sister gene copy [24]. The biological system also matters. On the one hand, functional compensation of duplicate (paralogous) genes has been shown to play an important role in the genetic robustness of yeast and nematodes, in which the deletion of one gene copy had little to no effect on the fitness of the cells [59,62,73]. This result is also supported by the strong synergistic deleterious effects of double knockout experiments in yeast of paralogs derived from WGD [74,75]. On the other hand, there is no evidence for functional compensation between duplicated genes in mammals. For example, the deletion of a duplicated gene has on average as large an effect on the fitness of mice as the deletion of a single-copy gene [76,77]. The situation in mammals may be more complex, however, as ancient duplicated genes, developmental duplicated genes, and genes duplicated through WGD show lower functional compensation than other duplicates [78]. Empirical evidence of the role of genetic redundancy in mutational robustness Studies on gene knockouts in mammals and yeast have revealed a complex, and perhaps conflicting, relationship between mutational robustness and gene duplication. One possible way to reconcile these data is to test whether the rates of evolution, calculated as the levels of amino acid divergence between orthologous protein sequences from different species, are higher for duplicates than for singletons. This test is limited, however, by the fact that genome sequences observed today comprise mixed signatures of adaptive, slightly deleterious, and neutral mutations, making it difficult to reach firm conclusions about the tolerance to deleterious mutations in duplicates genes [79]. Instead, the robustness provided by duplicated genes can be tested through forcing the fixation of deleterious mutations in a population and interrogating the genomes for the distribution of such mutations among duplicated and singleton genes. This kind of test can be performed through experimental evolution approaches (Box 3), which provide detailed information on the dynamics of mutations in the genome. A recent study took advantage of the laboratory evolution of independent experimental lines derived from the same ancestral haploid population of

Trends in Genetics xxx xxxx, Vol. xxx, No. x

Box 3. Experimental evolution Evolution occurs over long time scales, mostly beyond our lifespan. Thus, most studies of evolution rely on inferential approaches to test hypotheses. These evolutionary hypotheses are often controversial because evolution cannot be observed while it is happening. One empirical approach, however, is to use model organisms to reproduce evolutionary processes or phenomena in laboratorycontrolled experiments. Microbes, including bacteria and yeast, have been particularly useful for such endeavors owing to their short generation times, high mutation rates, and the possibility of growing enormous population sizes in small laboratory spaces. One typical evolution experiment is that looking for adaptations by subjecting populations growing in minimum media to daily transfers for thousands of generations in constant environments or environments with constant fluctuations (Figure I). For example, in one evolution experiment, the authors found thermoresistant mutations emerging in a population of the bacterium Escherichia coli that evolved over 15–20 days, corresponding to 140 generations of E. coli [99]. An advantage of performing evolution experiments with microbes is that a detailed ‘fossil’ record can be built by freezing microbes at different steps of the experiment that can be thawed again, allowing for assessment of their fitness in comparison with their last common ancestor (Figure I)[100].

Ancestral cell

ti gener.

tn gener.

Head-to-head compeon TRENDS in Genetics

Figure I. Experimental evolution of microbes. Many replicate populations derive from a single microbial cell. The evolution experiment proceeds by daily transferring a single cell to the new environment (flask). In the figure example, five independent lines of evolution started and evolved for many generations (tn gener.). At specific points of the evolution experiment (e.g., ti), Competition essays can be performed by mixing ancestral populations with evolved ones at equal proportions and letting cells compete head-to-head.

S. cerevisiae under strong genetic drift effects, thereby forcing the accumulation of deleterious mutations in the cell [80] (Figure 1). The founder of the original population was a haploid S. cerevisiae cell lacking the repair gene msh2 (Figure 1A). After 2200 generations, most of the mutations accumulated in the evolved S. cerevisiae genomes were deleterious, while there was evidence of compensatory mutations in the genome (Figure 1B). Duplicated genes showed higher fixation rates of mutations in their coding regions than singleton genes (Figure 1C). Remarkably, regulatory regions upstream of duplicated genes were orders of magnitude more enriched for disrupting mutations than the regulatory regions upstream of singleton genes (Figure 1C). 5

TIGS-1210; No. of Pages 9

Review

Trends in Genetics xxx xxxx, Vol. xxx, No. x

S. cerevisiae (haploid, Δmsh2)

(A)

Growth

440 gen.

µa

Time (generaons)

Time

660 gen.

µb

1.1K gen.

µc

1.5K gen.

µd

1.9K gen.

µe

2.2K gen.

µf

(B)

Rate growth (µ)

x x

Stress condions

Fitness decline Compensatory evoluon

TRENDS in Genetics

Generaons (C)

Duplicates 7.15%

Singletons 5.5%

SNPs coding

26.69%

18.78%

SNPs regulatory

TRENDS in Genetics

Figure 1. Duplicate genes give rise to coding and regulatory mutational robustness. A recent study [80] tested the tolerance of the yeast Saccharomyces cerevisiae to mutations in coding and regulatory regions of ancient duplicated and singleton genes. (A) A single ancestral population of a haploid strain of S. cerevisiae lacking the repair gene msh2 was bottlenecked in five independent evolution lines for 2200 generations of the yeast. Growth curves where built at 440, 660, 1100 (1.1K), 1.5K, 1.9K, and 2.2K generations of the yeast for each line. The growth rate at exponential phase (slope of the exponential phase: m) was calculated for each lineage. (B) Growth of the different lineages declined with the number of generations as a result of the fixation of deleterious mutations in bottlenecked populations but this decline plateaued up and acquired a positive slope at 1.5K generations, indicating compensatory evolution owing to antagonistic epistasis. (C) Genomes of the five lines were resequenced at the end of the evolution experiment and the distribution of nonsynonymous nucleotide polymorphisms (i.e., SNPs causing amino acid changes in the encoded protein sequence) and SNPs in regulatory regions was determined. Duplicates showed greater proportion of SNPs than singletons in coding and regulatory regions, with the strongest difference being in regulatory SNPs. Abbreviation: SNP, single nucleotide polymorphism.

The strong robustness to mutations in the regulatory regions of duplicated genes might be explained by the idea that high regulatory robustness could increase the phenotypic plasticity of the population and lead to adaptations to stress conditions. In a population of yeast growing in lowstress conditions (Figure 2), mutations in regulatory regions of one gene copy could lead to novel genotypes. 6

Lenient condions

Figure 2. Gene duplication and regulatory partial divergence provides redundancy and robustness against perturbations. Cells symbolize budding yeast. Arrows are genes, circles refer to transcription factor-binding motifs in gene promoters. Motifs of the same color regulate in the same manner while those with different color have diverged from the ancestral state. Because of the regulatory redundancy generated after gene duplication, mutations in regulatory promoter regions are neutral. Many of the individuals in a population will bear duplications that will diverge leading to regulatory phenotypes that are neutral. Others, however, will bear mutations in regulatory regions that are preadaptive to stress conditions (e.g., environmental challenges), such that in lenient conditions one of the copies is expressed and the other silent (silent copies are indicated with a red cross) whereas stress conditions will lead to the reverse outcome.

This would often lead to gene silencing, which would be tolerated for one gene copy, provided the sister copy compensates for it. A few novel genotypes, however, can generate genetic variants that differ in their expression patterns from the ancestral preduplication gene. Such variation may be preadaptive to a specific stress condition. If these stress conditions alternate with non-stress conditions, then each gene copy can perform optimally in both of the conditions without compromising the performance of the required function. This expression divergence is particularly frequent amongst WGD duplicates in yeast; following WGD, many genes exhibit partitioned expression such that one copy is involved in stress-response pathways while the other is not [57,81]. An interesting example of this partitioning is that of the proteins cell division cycle (CDC)19 and pyruvate kinase (PYK2) encoded by a duplicated gene, both of which catalyze the last reaction of the glycolytic pathway that converts phosphoenol pyruvate to pyruvate. Under normal glucose-rich conditions, the gene CDC19 is induced by the upstream metabolite fructose-16-bisphosphate, while the PYK2 gene is silenced. When the cell is starved of glucose, there is not enough fructose-1-6bisphosphate to activate CDC19, but PYK2 becomes active [82]. The mode of gene duplication (WGD or SSD) has also been proposed to play a key role in the fate of duplicate genes and their role in robustness [83]. In the yeast evolution experiment, researchers observed higher tolerance to mutations in the coding regions of genes arising through

TIGS-1210; No. of Pages 9

Review Box 4. Robustness through transcription control reprogramming The mechanisms through which one copy of a duplicated gene functionally backs up its sister gene copy such that, when one copy is knocked down the other kicks in, has been a long-standing problem in evolutionary biology. Intuitively, one might expect that paralogs with more similar expression profiles should backup one another to a larger extent than those with less similar profiles. Examination of coexpression profiles for duplicate genes in a long list of experiments representing different conditions led, however, to the counterintuitive conclusion that duplicates of S. cerevisiae with fewer coexpression profiles of their gene copies are more dispensable than those with more similar expression profiles [87]. To shed light on this counterintuitive result, the authors analyzed the overlap in transcription factor-binding motifs between gene copies. They found that genes sharing some of their regulatory motifs with their paralogous gene copies allowed the cell to perform optimally in both normal conditions and conditions in which one of the genes is silenced, as the other kicks in. They proposed an elegant mechanistic explanation of this: when a gene copy is silenced, its sister copy undergoes regulatory reprogramming such that it uses the appropriate regulatory motif to rescue the affected cell. This mechanism links mutational robustness to evolvability, because the primary cellular function encoded in the duplicated gene remains unperturbed despite the silencing of one gene copy due to the redundancy in the regulatory motifs. This robustness also allows one gene copy to acquire novel regulatory motifs, diverging in its expression from its sister copy, without undergoing nonfunctionalization because selection to keep the ancestral function favors redundancy, hence robustness.

SSD than through WGD [80]. This higher robustness of S. cerevisiae to mutations in SSDs raises the possibility that SSDs are more likely to lead to novel functions than WGDs. Accordingly, SSDs tend to exhibit a higher number of genetic interactions and interact with genes encoding less-related functions than WGDs and singletons [84]. Interestingly, despite the higher functional divergence between the copies of SSDs as compared to those from WGDs, the former presents larger redundancy than the latter (e.g., SSD gene copies share more interacting functions than WGD gene copies do) [84]. Therefore, SSDs may have more potential to evolve novel functions than WGDs because

Trends in Genetics xxx xxxx, Vol. xxx, No. x

they are less subjected to stoichiometric balance constraints and more robust to mutations [84–86]. The tolerance of yeast to mutations in regulatory regions of duplicate genes seems to be insensitive to the mechanism of duplication. This may be because both the subfunctionalization, a phenomenon more frequent after WGD, and neofunctionalization, frequent after SSD, require robustness in regulatory regions of genes. In agreement with a previous model in which redundancy is an important requisite for innovation through mutational robustness [84], redundancy at the regulatory level can provide an exploratory arena for transcription control reprogramming of duplicates [87–89] (Box 4). Concluding remarks To understand the robustness of a biological system we need to be able to determine what factors gave rise to it. Whether robustness to mutations is adaptive, a byproduct of other intrinsic or extrinsic system properties, or an emergent property of the biological system itself remains a tantalizing mystery. However, genetic redundancy seems to be largely correlated with robustness and may play an important role in its origins. The link between redundancy and the origin of robustness remains nevertheless part of a long list of questions about the mechanisms giving rise to this property (Box 5). Gene duplication is clearly linked to the origin of novel functions through preadaptive mutations that drift to fixation in the populations owing to the existence of backup gene copies, negating the need to invoke selection to explain their origin. Although researchers have classically linked gene duplication to innovation, such a link requires finding the factors that allow the persistence of genes in duplicate for long evolutionary periods. A number of recent studies show, however, that these links can be mechanistically established if the persistence of duplicates is the result of robustness to mutations that emerges as a property of redundant systems (e.g., gene duplication) and that provides opportunity for achieving novel phenotypes.

Box 5. Outstanding questions A number of questions concerning the origin and evolution of robustness remain unresolved, of which I highlight here a small sample. To resolve these questions, novel approaches will be needed including theoretical studies, evolution experiments, and large-scale metabolic approaches.  What factors contribute to robustness in the face of perturbations? One of the main questions is whether robustness is an emergent property of biological systems or if there are particular factors that dramatically increase robustness to perturbations. In particular, novel empirical studies would make it possible quantifying the contribution of genetic redundancy and distributed robustness to mutational robustness.  At what level of robustness is evolvability maximized? The phenotypic plasticity of different biological systems is likely to react differently to increasing levels of robustness. For example, it remains unknown whether the link between robustness and evolvability holds true for all organisms and at all levels or if it is dependent on the system under study.  What is the best measure of robustness? To determine the factors that increase robustness, robustness needs to be quantified in its various forms; be it through genetic redundancy or distributed robustness. Quantifying robustness is a

long-standing challenge often because the behavior of perturbations is uncharacterized. For example, robustness to mutations cannot be measured simply by counting the number of mutations to which a trait remains unperturbed because of the epistatic interactions between these mutations. To what extent do mutations interact epistatically is unknown: many mutations cause no effect in one genomic background but may be lethal in other genomic backgrounds.  How can we quantify evolvability? In addition to measuring robustness, to understand the role of robustness in phenotypic plasticity and evolvability, a reliable measure of evolvability is also required. There are two difficulties associated with quantifying evolvability of a system: (i) we have little knowledge about the parameters defining the phenotypic space of a genotypic network; and (ii) we do not know the shape of the fitness space for all possible reachable phenotypes.  Is robustness an evolvable trait? To answer this question we need to determine whether robustness is an intrinsic property of complex systems, a byproduct of other types of selected traits, or a trait under selection. Answering these questions is still at the core of ongoing discussions. 7

TIGS-1210; No. of Pages 9

Review Robustness provides phenotypic plasticity only if variation can be cryptically generated and maintained in a given population, a phenomenon expedited by gene duplication. Therefore understanding robustness through genetic redundancy will shed considerable light on the role of duplication in innovation and on how robustness fuels the remarkable phenotypic plasticity of biological systems. Acknowledgments I thank Dr. Christina Toft for her critical reading of the manuscript and insightful discussions. I am also grateful to the editor for her guidance during the preparation of this manuscript. This study is supported by a grant from the Spanish Ministerio de Economı´a y Competitividad (BFY2009-12022) and a grant from Science Foundation Ireland (12/IP/ 1673) to MAF.

References 1 Wagner, A. (2005) Circuit topology and the evolution of robustness in two-gene circadian oscillators. Proc. Natl. Acad. Sci. U.S.A. 102, 11775–11780 2 Wagner, A. (2012) The role of robustness in phenotypic adaptation and innovation. Proc. Biol. Sci. 279, 1249–1258 3 Waddington, C.H. (1959) Canalization of development and genetic assimilation of acquired characters. Nature 183, 1654–1655 4 Waddington, C.H. (1953) Genetic assimilation of an acquired character. Evolution 7, 9 5 Waddington, C.H. (1956) Genetic assimilation of the bithorax phenotype. Evolution 10, 13 6 Rennell, D. et al. (1991) Systematic mutation of bacteriophage T4 lysozyme. J. Mol. Biol. 222, 67–88 7 Sinha, N. and Nussinov, R. (2001) Point mutations and sequence variability in proteins: redistributions of preexisting populations. Proc. Natl. Acad. Sci. U.S.A. 98, 3139–3144 8 Edwards, J.S. and Palsson, B.O. (2000) Robustness analysis of the Escherichia coli metabolic network. Biotechnol. Prog. 16, 927–939 9 Raj, A. et al. (2006) Stochastic mRNA synthesis in mammalian cells. PLoS Biol. 4, e309 10 Raser, J.M. and O’Shea, E.K. (2005) Noise in gene expression: origins, consequences, and control. Science 309, 2010–2013 11 Batada, N.N. and Hurst, L.D. (2007) Evolution of chromosome organization driven by selection for reduced gene expression noise. Nat. Genet. 39, 945–949 12 Giaever, G. et al. (2002) Functional profiling of the Saccharomyces cerevisiae genome. Nature 418, 387–391 13 White, J.K. et al. (2013) Genome-wide generation and systematic phenotyping of knockout mice reveals new roles for many genes. Cell 154, 452–464 14 Vachias, C. et al. (2014) Tight coordination of growth and differentiation between germline and soma provides robustness for drosophila egg development. Cell Rep. 9, 531–541 15 Hartman, J.L. and 4th et al. (2001) Principles for the buffering of genetic variation. Science 291, 1001–1004 16 Masel, J. and Siegal, M.L. (2009) Robustness: mechanisms and consequences. Trends Genet. 25, 395–403 17 Wagner, A. and Wright, J. (2007) Alternative routes and mutational robustness in complex regulatory networks. Biosystems 88, 163–172 18 Wagner, A. (2011) Genotype networks shed light on evolutionary constraints. Trends Ecol. Evol. 26, 577–584 19 Draghi, J.A. et al. (2010) Mutational robustness can facilitate adaptation. Nature 463, 353–355 20 Ferrada, E. and Wagner, A. (2010) Evolutionary innovations and the organization of protein functions in genotype space. PLoS ONE 5, e14172 21 Hermisson, J. and Wagner, G.P. (2004) The population genetic theory of hidden variation and genetic robustness. Genetics 168, 2271–2284 22 Ancel, L.W. and Fontana, W. (2000) Plasticity, evolvability, and modularity in RNA. J. Exp. Zool. 288, 242–283 23 Goldsmith, M. and Tawfik, D.S. (2009) Potential role of phenotypic mutations in the evolution of protein expression and stability. Proc. Natl. Acad. Sci. U.S.A. 106, 6197–6202 8

Trends in Genetics xxx xxxx, Vol. xxx, No. x

24 Nowak, M.A. et al. (1997) Evolution of genetic redundancy. Nature 388, 167–171 25 Montville, R. et al. (2005) Evolution of mutational robustness in an RNA virus. PLoS Biol. 3, e381 26 Fraser, H.B. and Schadt, E.E. (2010) The quantitative genetics of phenotypic robustness. PLoS ONE 5, e8635 27 Cooper, T.F. et al. (2006) Effect of random and hub gene disruptions on environmental and mutational robustness in Escherichia coli. BMC Genomics 7, 237 28 Richardson, J.B. et al. (2013) Histone variant HTZ1 shows extensive epistasis with, but does not increase robustness to, new mutations. PLoS Genet. 9, e1003733 29 Freddolino, P.L. et al. (2012) Newly identified genetic variations in common Escherichia coli MG1655 stock cultures. J. Bacteriol. 194, 303–306 30 Freddolino, P.L. et al. (2012) Fitness landscape transformation through a single amino acid change in the rho terminator. PLoS Genet. 8, e1002744 31 Freddolino, P.L. and Tavazoie, S. (2012) Beyond homeostasis: a predictive-dynamic framework for understanding cellular behavior. Annu. Rev. Cell Dev. Biol. 28, 363–384 32 True, H.L. and Lindquist, S.L. (2000) A yeast prion provides a mechanism for genetic variation and phenotypic diversity. Nature 407, 477–483 33 Jarosz, D.F. and Lindquist, S. (2010) Hsp90 and environmental stress transform the adaptive value of natural genetic variation. Science 330, 1820–1824 34 Halfmann, R. et al. (2012) Prions are a common mechanism for phenotypic inheritance in wild yeasts. Nature 482, 363–368 35 Duveau, F. and Felix, M.A. (2012) Role of pleiotropy in the evolution of a cryptic developmental variation in Caenorhabditis elegans. PLoS Biol. 10, e1001230 36 Milloz, J. et al. (2008) Intraspecific evolution of the intercellular signaling network underlying a robust developmental system. Genes Dev. 22, 3064–3075 37 Queitsch, C. et al. (2002) Hsp90 as a capacitor of phenotypic variation. Nature 417, 618–624 38 Rutherford, S.L. and Lindquist, S. (1998) Hsp90 as a capacitor for morphological evolution. Nature 396, 336–342 39 Sangster, T.A. et al. (2004) Under cover: causes, effects and implications of Hsp90-mediated genetic capacitance. Bioessays 26, 348–362 40 Rohner, N. et al. (2013) Cryptic variation in morphological evolution: HSP90 as a capacitor for loss of eyes in cavefish. Science 342, 1372–1375 41 Lachowiec, J. et al. (2015) Hsp90 promotes kinase evolution. Mol. Biol. Evol. 32, 91–99 42 Fares, M.A. et al. (2002) Endosymbiotic bacteria: groEL buffers against deleterious mutations. Nature 417, 398 43 Bogumil, D. and Dagan, T. (2010) Chaperonin-dependent accelerated substitution rates in prokaryotes. Genome Biol. Evol. 2, 602–608 44 Williams, T.A. and Fares, M.A. (2010) The effect of chaperonin buffering on protein evolution. Genome Biol. Evol. 2, 609–619 45 Tokuriki, N. and Tawfik, D.S. (2009) Chaperonin overexpression promotes genetic variation and enzyme evolution. Nature 459, 668–673 46 Ohno, S. (1970) Evolution by Gene Duplication, Springer Verlag 47 Ibarra, R.U. et al. (2002) Escherichia coli K-12 undergoes adaptive evolution to achieve in silico predicted optimal growth. Nature 420, 186–189 48 Barve, A. and Wagner, A. (2013) A latent capacity for evolutionary innovation through exaptation in metabolic systems. Nature 500, 203–206 49 Ohno, S. (1999) Gene duplication and the uniqueness of vertebrate genomes circa 1970-1999. Semin. Cell Dev. Biol. 10, 517–522 50 Wolfe, K.H. and Shields, D.C. (1997) Molecular evidence for an ancient duplication of the entire yeast genome. Nature 387, 708–713 51 Castillo-Davis, C.I. et al. (2002) Selection for short introns in highly expressed genes. Nat. Genet. 31, 415–418 52 Pal, C. et al. (2001) Highly expressed genes in yeast evolve slowly. Genetics 158, 927–931 53 Blanc, G. and Wolfe, K.H. (2004) Widespread paleopolyploidy in model plant species inferred from age distributions of duplicate genes. Plant Cell 16, 1667–1678

TIGS-1210; No. of Pages 9

Review 54 Cui, L. et al. (2006) Widespread genome duplications throughout the history of flowering plants. Genome Res. 16, 738–749 55 Blanc, G. and Wolfe, K.H. (2004) Functional divergence of duplicated genes formed by polyploidy during Arabidopsis evolution. Plant Cell 16, 1679–1691 56 Scannell, D.R. and Wolfe, K.H. (2008) A burst of protein sequence evolution and a prolonged period of asymmetric evolution follow gene duplication in yeast. Genome Res. 18, 137–147 57 Conant, G.C. and Wolfe, K.H. (2008) Turning a hobby into a job: how duplicated genes find new functions. Nat. Rev. Genet. 9, 938–950 58 Fares, M.A. et al. (2006) Rate asymmetry after genome duplication causes substantial long-branch attraction artifacts in the phylogeny of Saccharomyces species. Mol. Biol. Evol. 23, 245–253 59 Gu, Z. et al. (2003) Role of duplicate genes in genetic robustness against null mutations. Nature 421, 63–66 60 VanderSluis, B. et al. (2010) Genetic interactions reveal the evolutionary trajectories of duplicate genes. Mol. Syst. Biol. 6, 429 61 Hsiao, T.L. and Vitkup, D. (2008) Role of duplicate genes in robustness against deleterious human mutations. PLoS Genet. 4, e1000014 62 Conant, G.C. and Wagner, A. (2004) Duplicate genes and robustness to transient gene knock-downs in Caenorhabditis elegans. Proc. Biol. Sci. 271, 89–96 63 Freeling, M. and Thomas, B.C. (2006) Gene-balanced duplications, like tetraploidy, provide predictable drive to increase morphological complexity. Genome Res. 16, 805–814 64 Makino, T. and McLysaght, A. (2010) Ohnologs in the human genome are dosage balanced and frequently associated with disease. Proc. Natl. Acad. Sci. U.S.A. 107, 9270–9274 65 Lynch, M. and Conery, J.S. (2000) The evolutionary fate and consequences of duplicate genes. Science 290, 1151–1155 66 Force, A. et al. (1999) Preservation of duplicate genes by complementary, degenerative mutations. Genetics 151, 1531–1545 67 Barkman, T. and Zhang, J. (2009) Evidence for escape from adaptive conflict? Nature 462, E1 discussion E2–3 68 Des Marais, D.L. and Rausher, M.D. (2008) Escape from adaptive conflict after duplication in an anthocyanin pathway gene. Nature 454, 762–765 69 He, X. and Zhang, J. (2005) Rapid subfunctionalization accompanied by prolonged and substantial neofunctionalization in duplicate gene evolution. Genetics 169, 1157–1164 70 van Hoof, A. (2005) Conserved functions of yeast genes support the duplication, degeneration and complementation model for gene duplication. Genetics 171, 1455–1461 71 Ihmels, J. et al. (2007) Backup without redundancy: genetic interactions reveal the cost of duplicate gene loss. Mol. Syst. Biol. 3, 86 72 Plata, G. and Vitkup, D. (2014) Genetic robustness and functional evolution of gene duplicates. Nucleic Acids Res. 42, 2405–2414 73 Guan, Y. et al. (2007) Functional analysis of gene duplications in Saccharomyces cerevisiae. Genetics 175, 933–943 74 DeLuna, A. et al. (2008) Exposing the fitness contribution of duplicated genes. Nat. Genet. 40, 676–681 75 Musso, G. et al. (2008) The extensive and condition-dependent nature of epistasis among whole-genome duplicates in yeast. Genome Res. 18, 1092–1099 76 Liang, H. and Li, W.H. (2007) Gene essentiality, gene duplicability and protein connectivity in human and mouse. Trends Genet. 23, 375–378 77 Liao, B.Y. and Zhang, J. (2007) Mouse duplicate genes are as essential as singletons. Trends Genet. 23, 378–381

Trends in Genetics xxx xxxx, Vol. xxx, No. x

78 Makino, T. et al. (2009) The complex relationship of gene duplication and essentiality. Trends Genet. 25, 152–155 79 Lynch, M. and Katju, V. (2004) The altered evolutionary trajectories of gene duplicates. Trends Genet. 20, 544–549 80 Keane, O.M. et al. (2014) Preservation of genetic and regulatory robustness in ancient gene duplicates of Saccharomyces cerevisiae. Genome Res. 24, 1830–1841 81 Conant, G.C. and Wolfe, K.H. (2006) Functional partitioning of yeast co-expression networks after genome duplication. PLoS Biol. 4, e109 82 Boles, E. et al. (1997) Characterization of a glucose-repressed pyruvate kinase (Pyk2p) in Saccharomyces cerevisiae that is catalytically insensitive to fructose-1,6-bisphosphate. J. Bacteriol. 179, 2987–2993 83 Lynch, M. et al. (2001) The probability of preservation of a newly arisen gene duplicate. Genetics 159, 1789–1804 84 Fares, M.A. et al. (2013) The roles of whole-genome and small-scale duplications in the functional specialization of Saccharomyces cerevisiae genes. PLoS Genet. 9, e1003176 85 Hakes, L. et al. (2007) Specificity in protein interactions and its relationship with sequence diversity and coevolution. Proc. Natl. Acad. Sci. U.S.A. 104, 7999–8004 86 Carretero-Paulet, L. and Fares, M.A. (2012) Evolutionary dynamics and functional specialization of plant paralogs formed by whole and small-scale genome duplications. Mol. Biol. Evol. 29, 3541–3551 87 Kafri, R. et al. (2005) Transcription control reprogramming in genetic backup circuits. Nat. Genet. 37, 295–299 88 Koonin, E.V. (2005) Paralogs and mutational robustness linked through transcriptional reprogramming. Bioessays 27, 865–868 89 Hurst, L.D. and Pal, C. (2005) Dissecting dispensability. Nat. Genet. 37, 214–215 90 Masel, J. and Bergman, A. (2003) The evolution of the evolvability properties of the yeast prion [PSI+]. Evolution 57, 1498–1512 91 Gould, S.J. and Vrba, E.S. (1982) Exaptation – a missing term in the science of form. Paleobiology 8, 12 92 Price, T.D. et al. (2003) The role of phenotypic plasticity in driving genetic evolution. Proc. Biol. Sci. 270, 1433–1440 93 Felix, M.A. and Wagner, A. (2008) Robustness and evolution: concepts, insights and challenges from a developmental model system. Heredity 100, 132–140 94 Giurumescu, C.A. et al. (2006) Intercellular coupling amplifies fate segregation during Caenorhabditis elegans vulval development. Proc. Natl. Acad. Sci. U.S.A. 103, 1331–1336 95 Shaye, D.D. and Greenwald, I. (2002) Endocytosis-mediated downregulation of LIN-12/Notch upon Ras activation in Caenorhabditis elegans. Nature 420, 686–690 96 Chen, N. and Greenwald, I. (2004) The lateral signal for LIN-12/Notch in C. elegans vulval development comprises redundant secreted and transmembrane DSL proteins. Dev. Cell 6, 183–192 97 Berset, T.A. et al. (2005) The C. elegans homolog of the mammalian tumor suppressor Dep-1/Scc1 inhibits EGFR signaling to regulate binary cell fate decisions. Genes Dev. 19, 1328–1340 98 Yoo, A.S. et al. (2004) Crosstalk between the EGFR and LIN-12/Notch pathways in C. elegans vulval development. Science 303, 663–666 99 Bennett, A.F. and Lenski, R.E. (2007) An experimental test of evolutionary trade-offs during temperature adaptation. Proc. Natl. Acad. Sci. U.S.A. 104 (Suppl 1), 8649–8654 100 Ostrowski, E.A. et al. (2008) The genetic basis of parallel and divergent phenotypic responses in evolving populations of Escherichia coli. Proc. Biol. Sci. 275, 277–284

9