how missing genes interact

3
NEWS AND VIEWS 440 VOLUME 39 | NUMBER 4 | APRIL 2007 | NATURE GENETICS sets a venerable standard for future studies of complex traits. The study illustrates how a population rese- quencing approach of functional elements of a particular gene may accelerate discovery of the genetic basis of complex traits. However, recent studies have highlighted the need to consider synonymous ‘silent’ SNPs as poten- tially functional. For example, a synonymous SNP identified in the multidrug resistance gene, MDR-1, has been shown to affect the timing of cotranslation folding, resulting in an altered protein structure 6 . Similarly, a synonymous SNP found in the gene encod- ing human catechol-O-methyltransferase (COMT) was shown to have a regulatory function, stabilizing the RNA transcript and resulting in a reduced amount of translated protein 7 . Furthermore, variants in promoter regions and the UTR sequences also have the potential to be functional. Analytical methods are needed to predict the functional effects of these non–amino acid–changing variants in order to reap the full benefits of population resequencing of functional elements. Capturing rare variants Using the classical approach of genome-wide scans for linkage, and the more contemporary approach of genome-wide association studies, several notable susceptibility genes for complex traits have recently been reported 2,5,8–15 (Table 1). Common variants in three genes have been associated with the risk for age-related macular degeneration that together explains about half of the estimated heritability of the disease 10,11 . But for a variety of reasons, genome-wide scans using common variants frequently do not iden- tify markers accounting for such a large fraction of the disease. For example, one genome-wide scan identified a common variant (36% fre- quency) in the gene NOS1AP that accounts for 1.5% of the variance in cardiac electrical repo- larization 12 . Another genome-wide scan identi- fied a rare IL23R variant (2% frequency) that confers strong protection against inflammatory bowel disease 8 . In spite of these exciting successes, we are left with the following question: where is the rest of the genetic variance underlying these heritable traits, and would resequencing capture it? It is quite possible that some of this missing varia- tion is accounted for by common variants with very small effect sizes (<1%), some accounted for by rare variants and some accounted for by gene-gene or gene-environment interactions. Current genome-wide scans typically have lim- ited power to capture the association of variants with these characteristics. Although the relative contributions of these various types of vari- ants to complex traits remain to be determined, the work of Cohen and colleagues suggests that rare nonsynonymous variants may be impor- tant and are readily identified by resequencing approaches. Future of resequencing It is noteworthy that targeted resequencing of genes has been successful at finding associa- tions of rare variants with quantitative traits by analyzing the extremes of a population dis- tribution. However, for dichotomous traits, such as disease state, this analysis approach is not possible. Additionally, the approach is currently limited to searching candidate genes; thus, unexpected associations such as those identified by genome-wide scans are not found. In order to take into account the lessons from the ANGPTL4 study and appreciate the role of synonymous variants in gene regulation, it may be time to consider developing meth- odologies for systematic ‘deep’ resequencing and functional analysis of all coding genes in the human genome. This may be considered a formidable step, but just as complex traits have been found to be more complex, resequenc- ing technology is rapidly progressing as well 1 . The current momentum in discovery of genes associated with complex traits has underscored the benefits of more systematic approaches. Ultimately, when genome-wide resequencing is practical and affordable, it will be increas- ingly difficult for the genomic basis of health 16 and disease to be left undetected. COMPETING INTERESTS STATEMENT The authors declare no competing financial interests. 1. Service, R.F. Science 311, 1544–1546 (2006). 2. Cohen, J. et al. Nat. Genet. 37, 161–165 (2005). 3. Cohen, J.C. et al. Proc. Natl. Acad. Sci. USA 103, 1810– 1815 (2006). 4. Cohen, J.C. et al. Science 305, 869–871 (2004). 5. Romeo, S. et al. Nat. Genet. 39, 513–516 (2007). 6. Kimchi-Safarty, C. et al. Science 315, 525–528 (2007). 7. Nackley, A.G. et al. Science 314, 1930–1933 (2006). 8. Duerr, R.H. et al. Science 314, 1461–1464 (2006). 9. Hampe, J. et al. Nat. Genet. 39, 207–211 (2007). 10. Li, M. et al. Nat. Genet. 38, 1049–1054 (2006). 11. Maller, J. et al. Nat. Genet. 38, 1055–1059 (2006). 12. Arking, D.E. et al. Nat. Genet. 38, 644–651 (2006). 13. Grant, S.F. et al. Nat. Genet. 38, 320–323 (2006). 14. Sladek, R. et al. Nature 445, 881–885 (2007). 15. Graham, R.R. et al. Nat. Genet. 38, 550–555 (2006). 16. Nadeau, J.H. & Topol, E.J. Nat. Genet. 38, 1095–1098 (2006). How missing genes interact Clifford Zeyl Epistasis is an interaction among genes that makes the phenotypic effect of an allele dependent on which alleles are present at other loci. Two new genomic studies find abundant epistasis in the yeast genome, with significant implications for the evolution of sex and for the inference of genetic pathways from genomic data. Epistasis is often first encountered in intro- ductory genetics courses as a troubling depar- ture from the familiar 9:3:3:1 mendelian ratio. For example, if two genes encode enzymes that catalyze different steps in the same biochemical pathway, mutations in the upstream gene may block the pathway before the downstream gene can be expressed. This can be used to deduce the sequence of events in such pathways and to reconstruct gene regulatory hierarchies 1 . Epistasis may also be an important factor in such fundamentals as the evolution of sex and speciation 2 . Two new studies tackle epistasis on genomic scales, and both take advantage of a remarkable genetic resource: a large collection of yeast geno- types, each with a single gene deleted. In the first paper, Jasnos and Korona 3 (page 550) use pairwise combinations of deletions to test for a type of epistasis that has been hypothesized to provide an evolutionary advantage for sex and recombination 4 . In the second paper 5 (on page 199 of the February issue), St. Onge et al. demonstrate the use of epistasis to reconstruct the interactions of genes that are involved in a common function. In a simpler world, any mutation could be said to have its own effect on fitness, and Clifford Zeyl is at the Department of Biology, Wake Forest University, Winston-Salem, North Carolina 27109, USA. e-mail: [email protected] © 2007 Nature Publishing Group http://www.nature.com/naturegenetics

Upload: clifford

Post on 21-Jul-2016

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: How missing genes interact

NEWS AND V IEWS

440 VOLUME 39 | NUMBER 4 | APRIL 2007 | NATURE GENETICS

sets a venerable standard for future studies of complex traits.

The study illustrates how a population rese-quencing approach of functional elements of a particular gene may accelerate discovery of the genetic basis of complex traits. However, recent studies have highlighted the need to consider synonymous ‘silent’ SNPs as poten-tially functional. For example, a synonymous SNP identified in the multidrug resistance gene, MDR-1, has been shown to affect the timing of cotranslation folding, resulting in an altered protein structure6. Similarly, a synonymous SNP found in the gene encod-ing human catechol-O-methyltransferase (COMT) was shown to have a regulatory function, stabilizing the RNA transcript and resulting in a reduced amount of translated protein7. Furthermore, variants in promoter regions and the UTR sequences also have the potential to be functional. Analytical methods are needed to predict the functional effects of these non–amino acid–changing variants in order to reap the full benefits of population resequencing of functional elements.

Capturing rare variantsUsing the classical approach of genome-wide scans for linkage, and the more contemporary approach of genome-wide association studies, several notable susceptibility genes for complex traits have recently been reported2,5,8–15 (Table 1). Common variants in three genes have been associated with the risk for age-related macular degeneration that together explains about half of the estimated heritability of the disease10,11. But for a variety of reasons, genome-wide scans

using common variants frequently do not iden-tify markers accounting for such a large fraction of the disease. For example, one genome-wide scan identified a common variant (36% fre-quency) in the gene NOS1AP that accounts for 1.5% of the variance in cardiac electrical repo-larization12. Another genome-wide scan identi-fied a rare IL23R variant (∼2% frequency) that confers strong protection against inflammatory bowel disease8.

In spite of these exciting successes, we are left with the following question: where is the rest of the genetic variance underlying these heritable traits, and would resequencing capture it? It is quite possible that some of this missing varia-tion is accounted for by common variants with very small effect sizes (<1%), some accounted for by rare variants and some accounted for by gene-gene or gene-environment interactions. Current genome-wide scans typically have lim-ited power to capture the association of variants with these characteristics. Although the relative contributions of these various types of vari-ants to complex traits remain to be determined, the work of Cohen and colleagues suggests that rare nonsynonymous variants may be impor-tant and are readily identified by resequencing approaches.

Future of resequencingIt is noteworthy that targeted resequencing of genes has been successful at finding associa-tions of rare variants with quantitative traits by analyzing the extremes of a population dis-tribution. However, for dichotomous traits, such as disease state, this analysis approach is not possible. Additionally, the approach is

currently limited to searching candidate genes; thus, unexpected associations such as those identified by genome-wide scans are not found. In order to take into account the lessons from the ANGPTL4 study and appreciate the role of synonymous variants in gene regulation, it may be time to consider developing meth-odologies for systematic ‘deep’ resequencing and functional analysis of all coding genes in the human genome. This may be considered a formidable step, but just as complex traits have been found to be more complex, resequenc-ing technology is rapidly progressing as well1. The current momentum in discovery of genes associated with complex traits has underscored the benefits of more systematic approaches. Ultimately, when genome-wide resequencing is practical and affordable, it will be increas-ingly difficult for the genomic basis of health16 and disease to be left undetected.

COMPETING INTERESTS STATEMENTThe authors declare no competing financial interests.

1. Service, R.F. Science 311, 1544–1546 (2006).2. Cohen, J. et al. Nat. Genet. 37, 161–165 (2005).3. Cohen, J.C. et al. Proc. Natl. Acad. Sci. USA 103, 1810–

1815 (2006).4. Cohen, J.C. et al. Science 305, 869–871 (2004).5. Romeo, S. et al. Nat. Genet. 39, 513–516 (2007).6. Kimchi-Safarty, C. et al. Science 315, 525–528

(2007).7. Nackley, A.G. et al. Science 314, 1930–1933 (2006).8. Duerr, R.H. et al. Science 314, 1461–1464 (2006).9. Hampe, J. et al. Nat. Genet. 39, 207–211 (2007).10. Li, M. et al. Nat. Genet. 38, 1049–1054 (2006).11. Maller, J. et al. Nat. Genet. 38, 1055–1059 (2006).12. Arking, D.E. et al. Nat. Genet. 38, 644–651 (2006).13. Grant, S.F. et al. Nat. Genet. 38, 320–323 (2006).14. Sladek, R. et al. Nature 445, 881–885 (2007).15. Graham, R.R. et al. Nat. Genet. 38, 550–555 (2006).16. Nadeau, J.H. & Topol, E.J. Nat. Genet. 38, 1095–1098

(2006).

How missing genes interactClifford Zeyl

Epistasis is an interaction among genes that makes the phenotypic effect of an allele dependent on which alleles are present at other loci. Two new genomic studies find abundant epistasis in the yeast genome, with significant implications for the evolution of sex and for the inference of genetic pathways from genomic data.

Epistasis is often first encountered in intro-ductory genetics courses as a troubling depar-ture from the familiar 9:3:3:1 mendelian ratio. For example, if two genes encode enzymes that catalyze different steps in the same biochemical pathway, mutations in the

upstream gene may block the pathway before the downstream gene can be expressed. This can be used to deduce the sequence of events in such pathways and to reconstruct gene regulatory hierarchies1. Epistasis may also be an important factor in such fundamentals as the evolution of sex and speciation2. Two new studies tackle epistasis on genomic scales, and both take advantage of a remarkable genetic resource: a large collection of yeast geno-types, each with a single gene deleted. In the

first paper, Jasnos and Korona3 (page 550) use pairwise combinations of deletions to test for a type of epistasis that has been hypothesized to provide an evolutionary advantage for sex and recombination4. In the second paper5 (on page 199 of the February issue), St. Onge et al. demonstrate the use of epistasis to reconstruct the interactions of genes that are involved in a common function.

In a simpler world, any mutation could be said to have its own effect on fitness, and

Clifford Zeyl is at the Department of Biology, Wake Forest University, Winston-Salem, North Carolina 27109, USA.e-mail: [email protected]

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureg

enet

ics

Page 2: How missing genes interact

NEWS AND V IEWS

NATURE GENETICS | VOLUME 39 | NUMBER 4 | APRIL 2007 441

Figure 1 Two forms of epistasis. (a) Fitness effects of combining two mutations. Three genotypes share a common first mutation. Second mutations may show positive epistasis (green), no epistasis (multiplicative effects, blue) or negative epistasis (purple) with the first. All mutations are assumed to have the same fitness effect when alone. (b) Approximate frequencies of deletion mutation pairs showing negative epistasis (green), no detected epistasis (blue) or positive epistasis (purple) in rich medium (YPD) in ref. 3, and in the same medium without (YPD) and with MMS (YPD + MMS) in ref. 5. All pairings of the three results are significantly different (P < < 0.001).

YPD

(ref. 3)

YPD

(ref. 5)

YPD+ MMS(ref. 5)

Fitn

ess

Number of mutations

0 1 2

Freq

uenc

y

1.0

0.8

0.6

0.4

0.2

0.0

a b

Kim

Cae

sar

the combined effect of two mutations would be predictable from their separate effects. Suppose the fitness costs (s) of two muta-tions are s1 = 0.05 (a 5% fitness reduction) and s2 = 0.1, relative to a wild-type fitness of 1. If one measures fitness as a growth rate, as Jasnos and Korona did, then in the absence of epistasis, these mutations would have additive effects, and the fitness of the double mutant would be 1 – (0.05 + 0.1) = 0.85. For other measures, such as the relative doubling times that St. Onge et al.5 estimated, the mutations would have multiplicative effects if they acted independently, and the fitness of the double mutant would be (1 – 0.05)(1 – 0.1) = 0.855.

Epistasis and sexDepartures from independent effects hold some interesting possibilities, depending on the form they take. Epistasis, represented by ε, is quantified as the difference between the additive or multiplicative prediction and the observed fitness of the double mutant. If the fitness of the double mutant is higher than expected (ε > 0), the epistasis is referred to as positive or alleviating (among other terms). If ε < 0, the epistasis is termed negative or aggra-vating. (Fig. 1a)

The evolutionary success of sex has proven difficult to explain for organisms with dis-tinct male and female roles in reproduction. Females contribute almost all of the resources but only half of their alleles to sexual offspring, so hypothetical females who cloned them-selves would have twice the fitness of sexuals. Among the numerous hypotheses proposed to overcome this twofold disadvantage, the deter-ministic mutation hypothesis4 (DMH) is one of the most prominent. It has the attractions of relying on an inevitable process—harm-ful mutation—and having just two, testable requirements: a harmful mutation rate of >1 per genome per generation, and negative epi-stasis between those mutations. Estimates of mutation rates are scattered about 1 (refs. 6–8), but testing for epistasis has been rare9,10. The two mating types of yeast contribute equally to sexual offspring, so its sexuality is no conun-drum. Rather, Jasnos and Korona used yeast to test for a general pattern that might apply to plants and animals, too. They crossed pairs of single mutants to obtain double mutants and estimated the growth rates of all three along with the wild-type strain. After calculating epistasis, they concluded that overall ε is posi-tive (mean = 0.024)—the opposite direction required to explain sex.

But the average doesn’t tell the whole story. The observed variation in ε among different pairs of mutations is a further obstacle to the DMH11. Approximately 73 of 640 pairings

are positively epistatic, and 19 are negative (Fig. 1b; there is a little statistical uncertainty about the numbers). Interestingly, individual genes do not interact with others in consis-tent ways: there is no correlation between ε for the same gene in different pairings.

The information in epistasisAlthough comprehensive, Jasnos and Korona’s study is somewhat abstract. It gives us the big-gest sample thus far of genome-wide inter-actions, but a more satisfying picture would add the genetic and metabolic mechanisms behind the numbers. St. Onge et al.5 focused on a set of 26 genes that are involved in repair-ing damaged DNA. They measured the fitness of single and double deletion mutants in the presence of the chemical methyl methanesul-fonate (MMS), which causes DNA damage. Working out the genetic pathway was not their primary goal—that was already well understood, although they have probably found a new role for one gene. Rather, they demonstrate a system for screening genomic data to reconstruct such pathways, which until recently required concentrating a great deal of effort on a few genes at a time.

St. Onge et al. started with the established premise that epistatic pairs of genes share some common role. Negative epistasis implies their involvement in separate pathways that feed into the same end result, and positive epistasis suggests roles in the same pathway, as described above. St. Onge et al. observed approximately twice as many significantly epistatic interactions in each direction in the presence of MMS as they did in the same rich medium used by Jasnos and Korona (Fig. 1b), illustrating the functional specificity of these

interactions for the process of DNA repair. They subdivide positively epistatic (allevi-ating) interactions according to the relative fitness of the single and double mutants, and from these subdivisions they make more specific predictions about the genetic basis of those interactions. Perhaps the easiest to visualize are the interactions they labeled “coequal,” in which both single mutants and the double mutant all had indistinguishable fitness. Coequal interactions suggest a very tight functional relationship, such as the for-mation of a single protein complex. If one subunit is inactivated, the complex doesn’t function; if a second subunit is inactivated, it still doesn’t function. The accuracy with which observed epistasis predicted known interactions in the DNA repair pathway sug-gests that this is a promising way to deduce unknown interactions on a genomic scale. In this study, the net direction of epistasis is neg-ative, but again, the observed variation among pairs of deletions (Fig. 1b) argues against the deterministic mutation hypothesis.

The study of epistasis is still in its early stages. In both of the papers reviewed here, haploid mutants were used to dodge the complicating issue of dominance. The interaction between dominance and epistasis is a challenging but biologically important area for future research. Similarly, clean deletions are ideally suited to reconstructing genetic pathways, but they are hardly the random mutations that are of the greatest evolutionary interest. Transposon insertions in Escherichia coli have also shown a mix of positive and negative epi stasis9, which is encouraging for the generality of this result.

Epistasis is a trait of the genome that can itself respond to selection. In an intriguing

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureg

enet

ics

Page 3: How missing genes interact

NEWS AND V IEWS

442 VOLUME 39 | NUMBER 4 | APRIL 2007 | NATURE GENETICS

simulation of gene network evolution, sexual recombination selected for networks in which mutations were negatively epi-static12—the very condition hypothesized to favor sex. Epistatic interactions will have a major role in the rapidly expanding study of gene and protein networks both as a basic property of those networks and as a valu-able tool.

COMPETING INTERESTS STATEMENTThe author declares no competing financial interests.

1. Avery, L.W. & Wasserman, S. Trends Genet. 8, 312–316 (1992).

2. Wolf, J.B., Brodie, E.D. III & Wade, M.J. Epistasis and the Evolutionary Process (Oxford Univ. Press, New York, 2000).

3. Jasnos, L.K. & Korona, R. Nat. Genet. 39, 550–554 (2007).

4. Kondrashov, A.S. Nature 336, 435–440 (1988).5. St. Onge, R.P. et al. Nat. Genet. 39, 199–206 (2007).

6. Lynch, M. et al. Evolution Int. J. Org. Evolution 53, 645–663 (1999).

7. Haag-Liautard, C. et al. Nature 445, 82–85 (2007).8. Keightley, P.D. & Eyre-Walker, A. Science 290, 331–333

(2000).9. Elena, S.F. & Lenski, R.E. Nature 390, 395–398

(1997).10. de Visser, J.A.G.M., Hoekstra, R.F. & van den Ende, H.

Genetics 145, 815–819 (1997).11. Otto, S.P. & Feldman, M.W. Theor. Popul. Biol. 51,

134–147 (1997).12. Azevedo, R.B.R. et al. Nature 440, 87–90 (2006).

The human promoter methylomeDaniel Zilberman

DNA methylation is a heritable epigenetic mark found in a wide range of eukaryotes. By mapping DNA methylation within the majority of human promoters, the authors of a new study uncover intriguing insights into genome evolution, cellular differentiation and potential links to tumorigenesis.

The completion of the human genome has provided a wealth of information about our genetic wiring. However, there is a great deal of information within the chromatin fiber beyond the DNA sequence. Unlike genetic informa-tion, which can be read, but not written, this epigenetic information is actively altered, pro-viding a flexible framework for specifying states of gene activity. DNA methylation, a covalent modification of DNA, is used by organisms ranging from mammals to plants to bacteria to convey epigenetic information1,2. Eukaryotes predominantly methylate cytosines in CpG dinucleotides. Because CpG dinucleotides are symmetric, a DNA methyltransferase can scan newly replicated DNA for hemimethylated CpG sites and fill in the missing methylation. This semiconservative replication, akin to that of the DNA sequence, makes methylation an excel-lent repository of epigenetic information. The first methylome—the profiling of methylation within an entire genome—was recently reported for the plant Arabidopsis thaliana3,4. Now, Weber et al.5 take the closest step yet toward the human methylome by mapping DNA methylation within the majority of human promoters in somatic cells and mature sperm.

A versatile epigenetic mechanismIn eukaryotes, DNA methylation is associated with transcriptional repression, particularly at gene promoters. DNA methylation is impor-tant in a variety of processes, including trans-poson silencing, X chromosome inactivation

and genomic imprinting1. Whether changes in somatic DNA methylation have a role in mam-malian development has been controversial6. Dirk Schübeler’s group has recently developed a method that uses antibodies to 5-methylcyto-sine to enrich methylated DNA7. In the current study, they combined this with high-density DNA microarrays to obtain the methylation profile of human promoters5. To monitor pro-moter activity, the authors used the same arrays

to map sites of RNA polymerase II binding. Their results suggest that differential methyla-tion is not a general mechanism for regulating gene expression, because most inactive pro-moters remained unmethylated. However, the promoters of a small subset of genes were meth-ylated in fibroblasts, but not in sperm, indicating that somatic methylation might contribute to differentiation and development. A particularly intriguing finding was that among differentially

Daniel Zilberman is at the Fred Hutchinson Cancer Research Center, 1100 Fairview Ave. N., Seattle, Washington 98109, USA.e-mail: [email protected]

Figure 1 DNA methylation in development and disease. (a) Methylation of key genes during development ensures a stable differentiated phenotype. (b) A somatic cell undergoes genome-wide hypomethylation, leading to a population of dedifferentiated, precancerous cells. Inappropriate methylation of tumor suppressors in combination with genetic mutations leads to cancer.

Methylation ofgermline promoters

Methylation ofpromoters

Embryonic stem cell Somatic stem cell Differentiated cell

Genomichypomethylation

Promoter hypermethylation and genetic mutations

Differentiated cell Dedifferentiated,precancerous cell

Cancer cell

a

bK

im C

aesa

r

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureg

enet

ics