This page has been archived and is no longer updated
Exploring genetic interactions and networks with yeast
Author: Charles Boone
Keywords
Keywords for this Article
Add keywords to your Content
Save
|
Cancel
Share
|
Cancel
Revoke
|
Cancel
Rate & Certify
Rate Me...
Rate Me
!
Comment
Save
|
Cancel
Flag Inappropriate
The Content is
Objectionable
Explicit
Offensive
Inaccurate
Comment
Flag Content
|
Cancel
Delete Content
Reason
Delete
|
Cancel
Close
Full Screen
"Genome sequencing and large-scale genetic analyses have unmasked the enormous scale of genetic interac- tions in biological systems 1,2 . A key challenge now is to understand how genes function as networks to carry out and regulate cellular processes. Many recent insights into genetic interactions and networks have emerged from studies using the yeast Saccharomyces cerevisiae, in which powerful functional genomic tools allow systematic analyses 3,4 , revealing both novel interacting components and key properties of the genetic networks in which they participate. A general understanding of the topology of genetic-interaction networks, which is rapidly being gained for yeast, has a wider importance, because similar networks are expected to underlie the relationship between genotype and phenotype in outbred populations in which combinations of specific alleles determine the fitness of individuals. In terms of human disease, numerous modifiers and enhancers contribute to complex genetic disorders, but the topol- ogy of the underlying networks is largely unknown. Thus, mapping genetic networks in model organ- isms such as yeast provides an important framework for studying genetic interactions in more complex systems. Here we provide a detailed discussion of the tools that have allowed genetic interactions to be so exten- sively mapped in S. cerevisiae and the insights that they provide into the structure and function of genetic networks in this organism. We then examine how this knowledge can be applied more widely to gain an under- standing of gene networks in complex traits, including human disease. Enhancement genetics: synthetic lethality Large-scale genetic analyses reveal that mutations in most eukaryotic genes have little discernable effect. For example, systematic gene deletion in S. cerevisiae, discussed in detail below, produced a remarkable result: only ~20% of yeast genes are essential for viability when deleted individually in haploids growing in standard laboratory conditions 5,6 . Recent systematic analyses revealed a measurable growth phenotype under at least one condition for virtually every yeast gene deletion 7,8 . Nonetheless, the ability of most deletion mutants to grow under optimal conditions reflects the robustness of biological circuits and cellular buffering against genetic variation, underscoring a key property of biological net- works: their resilience to attack at a single node 9,10 . Synthetic enhancement genetics can be used to exam- ine how mutations in two genes interact to modulate a phenotype. Essentially, synthetic enhancement screens represent an application of Fisher?s definition of epistasis (BOX 1) ? in this case, a double mutant shows an unex- pected, non-multiplicative phenotype, the most dramatic being inviability. Early genetic investigations using the fruitfly found that some pairwise combinations of mutant alleles were inviable, whereas singly, the same alleles were viable 11,12 , a phenomenon termed synthetic lethality (BOX 2). Yeast geneticists embraced the tools of synthetic enhancement to assist in functional analyses 13 (reviewed in REF. 14). However, synthetic enhancement combina- tions are infrequent in the large combinatorial sea of possible pairs of genes, and finding interacting partners for a given gene has required the development of sensitive and selective screening methods 15,16 . *Banting & Best Department of Medical Research and Terrence Donnelly Centre for Cellular & Biomolecular Research, University of Toronto, 160 College Street, Toronto M5S 3E1, Canada. ? Department of Biology, McGill University,1205 Docteur Penfield, Montreal H3A 1B1, Quebec, Canada. Correspondence to C.B. or B.J.A. e-mails: charlie.boone@utoronto.ca; brenda.andrews@utoronto.ca doi:10.1038/nrg2085 Synthetic enhancement The situation in which a mutation in one gene exacerbates the phenotypic severity of a mutation in a second gene. Synthetic lethality The situation in which two genes that are non-essential when individually mutated cause lethality when they are combined as a double mutant. Exploring genetic interactions and networks with yeast Charles Boone*, Howard Bussey ? and Brenda J. Andrews* Abstract | The development and application of genetic tools and resources has enabled a partial genetic-interaction network for the yeast Saccharomyces cerevisiae to be compiled. Analysis of the network, which is ongoing, has already provided a clear picture of the nature and scale of the genetic interactions that robustly sustain biological systems, and how cellular buffering is achieved at the molecular level. Recent studies in yeast have begun to define general principles of genetic networks, and also pave the way for similar studies in metazoan model systems. A comparative understanding of genetic- interaction networks promises insights into some long-standing genetic problems, such as the nature of quantitative traits and the basis of complex inherited disease. REVIEWS NATURE REVIEWS | GENETICS VOLUME 8 | JUNE 2007 | 437 Haploinsufficiency The situation in diploid cells in which heterozygous mutants that produce a reduced amount of functional gene product can be less robust than the wild type to perturbations that affect essential functions. Tetrad analysis The four haploid cells that are produced by an individual meiosis in budding yeast are referred to as a tetrad. The tetrad is enclosed in a sac called an ascus. Tetrad analysis involves the isolation and analysis of the haploid meiotic spores of individual asci for the segregation of genetic markers. Functional genomic tools for systematic genetics Compiling genetic interactions case by case as a by-product of directed biological studies is highly informative. However, genomics allows genetic net- works to be built systematically. Only in this way can a complete genetic network be mapped (a goal that is still far from being achieved for any organism) and its full explanatory potential realized. Many technological platforms and tools have been created for large-scale functional analysis in S. cerevisiae 17 . Deletion-mutant collection. By 2001, a deletion allele was available for each yeast gene 5,6 . In these deletion strains, the entire target gene is replaced with a kanamy- cin-resistance marker 18 plus two unique 20-bp flanking barcodes (FIG. 1a). In this way, the abundance of each mutant can be quantified from a mixed population using a barcode microarray (FIG. 1b). The yeast gene-deletion set is a key resource for large- scale and systematic genetics. The collection includes ~6,000 heterozygous diploid strains, each of which is deleted for a single copy of a specific gene in the S288c genetic background. Deletion alleles for all S. cerevisiae genes are represented and, apart from a few hundred haploinsufficient genes (~3%), the heterozygous mutants grow normally on a rich medium 8 . Tetrad analysis of the heterozygous strains identified ~1,000 deletion mutants that failed to grow as haploid meiotic progeny, thereby defining the S. cerevisiae essential gene set and creating a set of ~5,000 viable haploid deletion-mutant strains 5 . Mating of these mutants generated a set of ~5,000 homozygous diploid mutants, which carry a deletion of both alleles of each gene. As the roster of ORFs has been revised, largely through sequencing of evolutionarily related yeast species 19,20 , the deletion-mutant set has been correspondingly updated 21 . Essential gene mutant collections. Conditional alleles of the ~1,000 essential S. cerevisiae genes are required to enable systematic genetic analysis. There is value in gen- erating a variety of collections of essential gene alleles, as they are likely to provide complementary information in systematic function and genetic interactions of this important gene set. An extensive set of promoter-shutoff strains, in which an essential gene is placed under the control of a tetracycline (tet)-repressible promoter, has been constructed 22 . In these strains, the endogenous promoter of an essential gene is replaced with one that binds a tet-repressible transcriptional activator, which is expressed constitutively. Both the engineered essen- tial gene and the tet-responsive activator are linked to selectable markers, for ease of use in genetic analysis. Temperature-sensitive (ts) conditional alleles of essential genes have been used traditionally for study- ing essential processes such as cell-cycle control and secretion. The recently introduced ?heat-inducible degron system? provides a simple way to systematically generate ts alleles of essential genes 23 . An Arg-Dhfr(ts) protein, a ts variant of dihydrofolate reductase, carrying an amino (N)-terminal arginine (Arg) residue (a desta- bilizing residue according to the N-end rule), functions as a heat-activated degron, resulting in destruction of the tagged protein at 37�C. Large collections of degron alleles of essential genes have been made and subjected to phenotypic analysis 24 . In addition, ts alleles for ~50% of essential genes have been collected, and these are being integrated into the same strain background as the deletion collection (C.B. and B.J.A., unpublished observations). Hypomorphic allele collections can also be constructed systematically. For example, replacing the 3? UTR of an essential gene with a selectable marker often leads to lower transcript levels and a resultant phenotype 25 ? a method known as DAmP (decreased abundance by mRNA perturbation). Comprehensive gene-overexpression libraries. The complete set of yeast genes has been cloned into several yeast vectors that allow expression under the control of the strong galactose-inducible GAL1 promoter, Box 1 | Epistasis A review by Philips from 1998 describes the early literature on epistasis, much of which refers to the fruitfly, Drosophila melanogaster 77 . The language of genetic interactions has been profoundly influenced by these early studies and has led to two related but distinct meanings of the term epistasis, both of which derive from the quantitative analysis of double-mutant phenotypes and are relevant to large- scale mapping and interpretation of genetic networks. One view of epistasis derives from the world of statistical genetics. Fisher referred to deviations from the expected quantitative combination of independently functioning genes as ?epistacy? 78 , a concept that has been adopted by quantitative geneticists to describe a range of genetic interactions. The Fisher definition is quite general and inclusive, and encompasses any phenotype of a given double mutant that cannot be anticipated by simply combining its component single-locus effects multiplicatively. In other words, in the absence of a genetic interaction, the fitness of a double-mutant is expected to be the product of the individual fitness of the corresponding single mutants. For example, consider a yeast strain that carries a mutation in gene A, conferring a defective response, and consequent increased sensitivity, to the DNA-damaging agent methyl methanesulphonate (MMS), with a 20% growth-rate reduction compared with a wild-type strain at the same dose of MMS. Likewise, mutant B shows an MMS sensitivity with a 10% growth-rate reduction. The double mutant, however, grows 90% slower than the wild type in the presence of MMS, such that the genetic combination causes a much more severe phenotype than expected for the combination of the mutant B allele within the mutant A genetic background (0.8 � 0.9 = 0.72, or a 28%-reduced growth rate). One interpretation of this type of genetic interaction is that both genes might be involved in DNA repair but occur in separate pathways, such that the cell can tolerate loss-of-function mutations in either pathway but not both. The second definition of epistasis derives from the pioneering work of Bateson, who coined the term to explain genetic interactions that alter single Mendelian gene effects. The Bateson definition is familiar to classical and molecular geneticists, who typically use epistasis to describe situations in which the activity of one gene masks effects at another locus, allowing inferences about the order of gene action. As a simple example, consider the yeast transcriptional activator SWI5 ? mutation of SWI5 results in a failure to express the HO endonuclease gene that is required for mating-type switching. The swi5 mutant phenotype is suppressed by loss-of-function mutations in the SIN3 gene, which encodes a transcriptional repressor 79 . According to the Bateson definition, SIN3 is epistatic to SWI5, because its mutation masks defects in SWI5. This observation allows an inference to be made about a pathway (in this case, that the SIN3 product lies downstream of the SWI5 product in a common pathway). Classical examples of Bateson-type epistasis analysis include studies of signalling pathways that control the yeast cell cycle 80 and pheromone responses (reviewed in REF. 81), development in the nematode worm Caenorhabditis elegans 82 and sex determination in D. melanogaster 83 . REVIEWS 438 | JUNE 2007 | VOLUME 8 www.nature.com/reviews/genetics C2 C1 C2 c1 C1 c2 c1 c2 Essential function Essential function Essential function Essential function Pathway A A1 A2 A3 Pathway B B1 B2 B3 Essential biological function a Between-pathway genetic interactions b Within-pathway genetic interactions Wild type Viable Viable LethalCell proliferation N-end rule Relates the in vivo half-life of a protein to the identity of its N-terminal residue. In eukaryotes, the N-end rule pathway is part of the ubiquitin system. Hypomorphic Describes an allele that carries a mutation that causes a partial loss of gene function. Synthetic genetic array analysis A robotic procedure that is used to create, select and systematically examine the growth phenotypes of yeast double-mutant haploid strains. Pinning The use of hand-held or robotic tools, which are composed of small floating pinheads, to replicate yeast colonies to different media for genetic tests (typical formats include 96, 384, 768 and 1,536 pinheads per replica tool). typically resulting in protein overproduction. Partial but significant collections of genes have been con- structed encoding GAL1-regulated proteins, which are either untagged or carry a carboxy (C)-terminal Flag epitope 26,27 . Complete collections of genes have also been generated that encode proteins tagged with N-terminal glutathione S-transferase (GST)?histidine 6 (His6) or C-terminal His6?HA?ZZ 28,29 . Although over- expression and tagging of proteins is valuable, there can be limitations with such collections owing to dosage and functional issues. To obviate such limitations, ordered libraries of full-length genes under the control of their native promoters are under construction by our group and others. Methods for systematic genetic analysis in yeast Synthetic genetic array (SGA) analysis. In its simplest form, synthetic genetic array analysis 30 involves a series of replica-pinning procedures, in which mating and meiotic recombination are used to convert an input array of single mutants into an output array of double mutants (FIG. 2). SGA has been used extensively for syn- thetic-lethal screening of non-essential genes involved in many cellular functions 2 . The final transfer step (FIG. 2e?f) results in an ordered array of double-mutant haploid strains, the growth rates of which can be quan- titatively assessed 25 . Essential-gene mutant collections can also be used both as queries and as input arrays in an SGA screen Box 2 | Mechanisms of synthetic-lethal interactions What is the mechanistic basis for synthetic-lethal interactions? Because our knowledge of cellular functions is incomplete, we often do not understand why particular double mutants show a synthetic-lethal phenotype. However, possible mechanisms depend on the characteristics of the interacting alleles. For example, if both mutations occur in non-essential genes and are null alleles, the common interpretation is that the genes function in parallel pathways that impinge on a shared essential function (part a; thin lines indicate potential genetic interactions). This is often referred to as the ?between-pathway? model and typically reflects bidirectional genetic redundancy, in that each pathway compensates for defects in the other 14,32,55 . More elaborate mechanisms can be understood from a more detailed knowledge of gene function and pathway circuitry, such as a synthetic-lethal interaction that reflects ?unidirectional compensation?, whereby one pathway normally prevents a potentially harmful cellular event that can be corrected by another pathway 32 . A pertinent example involves the oxidative-stress response system, which precludes the accumulation of reactive oxygen species and protects the cell from DNA damage. By this mechanism, functional DNA repair pathways can compensate for defects in the oxidative-stress response system, but not the reverse. The specific case of synthetic genetic interactions involving duplicated genes or paralogues is also of interest. Here recent systematic studies revealed that patterns of genetic interactions are divergent between duplicates, suggesting that paralogous genes maintain functional specificity 84 . Conversely, distant paralogues encoding metabolic genes can show synthetic interactions, indicating that the product of the evolved copy of the duplicated gene might retain sufficient activity to mask the loss of the conserved copy 85 . For essential genes, in which single null mutations are lethal, conditional or hypomorphic alleles can be used to evaluate synthetic phenotypes. In these cases, interpretation is more complex, because interactions can occur ?within pathways? as well as between pathways. In the within-pathway model (part b; conditional mutations are indicated by an altered protein shape and a lower-case ?c?), synthetic lethality indicates that both genes function in the same essential pathway, the function of which is diminished by each mutation 14,86 . In this context, synthetic lethality can result from mutations in genes that affect the same stage of the pathway; for instance, when mutations weaken interactions between subunits of a protein complex so that two mutations disrupt complex formation altogether, or render its function below the viability threshold. Biologically compelling examples of this type of interaction are seen in the yeast secretion system; most so-called SEC genes are essential, but synthetic-lethal interactions between sec mutants are highly specific for genes that are involved in the same stage of the system 87 , and also occur among protein-complex subunits (for example, the exocyst complex 88 ). REVIEWS NATURE REVIEWS | GENETICS VOLUME 8 | JUNE 2007 | 439 Hybridization of labelled barcodes to a DNA microarray a b Competitive growth of a deletion-mutant pool in the presence of a growth inhibitory drug Barcode amplification and labelling UP DNkanR ORF UP DNkanR kanR kanR kanR kanR kanR kanR kanR Suppression The situation in which a mutation in one gene counteracts the effects of a mutation in another, so that the phenotype of the double mutant is more like that of the wild type. to generate networks that focus on essential genes. A proof-of-principle study generated a network of 567 interactions, 386 of which occur between 286 essential genes 1 . The use of expanded collections should soon incorporate all essential genes in the global genetic- interaction map (see below). Diploid-based synthetic lethality analysis with microarrays (dSLAM). As an alternative to visualiz- ing colonies in an array format, the barcodes that are associated with each deletion mutant enable quantifi- cation of each double mutant in a mixed population. The dSLAM method takes advantage of this barcode approach 31 (FIG. 3). Analysis of the barcode representa- tion in each population, by hybridization to a barcode microarray, provides a measure of the relative fitness of the double mutants and identifies potential syn- thetic interactions. dSLAM has been used to define a network of genes involved in maintaining genome integrity 32 . Synthetic dosage-suppression and lethality. Other types of synthetic genetic interaction are powerful for navi- gating genetic pathways, and have recently been incor- porated into systematic platforms. Dosage-suppression analysis, in which mutants are screened for phenotypic suppression using a library of overexpressed genes, has augmented pathway analysis in yeast. In a typical dosage-suppression screen, a mutant that carries a ts allele of an essential gene is transformed with a genomic library, which is carried on a multicopy plasmid, at a growth-permissive temperature. The transformants are then screened for dosage suppressors at a restrictive temperature. For example, using a conditional allele of the cell-division cycle gene CDC28 that is defective only at the G2?M transition of the cell cycle, a screen for dosage suppressors identified a set of G2-specific B-type cyclins 33 . Hundreds of such dosage suppressors are known and have broadly contributed to our under- standing of functional pathways 34 . In a conceptually reciprocal approach, dosage- lethality screens exploit features of both dosage-suppres- sion and synthetic-lethal screens to identify interacting proteins. Synthetic dosage lethality (SDL) derives from the idea that increasing levels of a protein might have no effect on the growth of an otherwise wild-type strain, but might cause a phenotype ? such as lethality ? in a mutant strain in which the activity of an interacting protein is reduced 35,36 . For example, SDL defined a broad range of interacting mutations involving components of the yeast kinetochore and the origin recognition complex (ORC) 35,37 . Current overexpression libraries have recently been arrayed so that SGA-based manipulations allow the introduction of any specific query mutation into a collection of ~6,000 yeast strains, each of which carries a unique gene-overexpression plasmid. This method allows rapid assessment of gene-overexpression pheno- types in any mutant background of interest. In addition to examining loss-of-function phenotypes associated with deletion-mutant alleles, overexpression alleles enable the exploration of gain-of-function phenotypes to augment gene-function analysis 38 . Conversely, SDL can be assessed by scoring for an enhanced-fitness defect that is due to gene overexpression in any mutant background. As proof-of-principle, a deletion allele of PHO85, which encodes a cyclin-dependent kinase, was crossed to a gene-overexpression array, reveal- ing 65 SDL interactions 38 , several of which involve in vivo substrates for the kinase (see below for more discussion). Figure 1 | The yeast deletion collection and parallel analysis. a | Construction strategy for the yeast deletion-mutant collection. Each yeast ORF is replaced with a ?deletion cassette? that consists of an antibiotic-resistance marker, kanR (which confers resistance to kanamycin), and two unique 20-nt molecular barcodes (?uptag? (UP) and ?downtag? (DN)). Each barcode is flanked by common primer sites (indicated by coloured half-arrows). Incorporation of the cassette into the yeast genome is accomplished through homologous recombination of 45-bp regions of homology upstream and downstream of the yeast ORF. b | Parallel analysis of large pools of deletion mutants. Populations of pooled mutant cells, each marked with unique molecular barcodes, are grown in the presence or absence of a growth-inhibitory drug. Genomic DNA is extracted from the pool of mutants, and barcodes that represent each strain are amplified by PCR using common primers that are labelled with fluorescent markers Cy3 or Cy5. Drug-sensitive mutants are identified by competitive hybridization of the barcode PCR products to a microarray that contains oligonucleotides corresponding to each barcode, giving a quantitative read-out of the representation of each mutant in a mixed population. REVIEWS 440 | JUNE 2007 | VOLUME 8 www.nature.com/reviews/genetics a/? Mating Sporulation MATa haploid selection (can1?::MFA1pr?HIS3) Double-mutant selection kanR selection a b c d e f Deletion mutationsWild-type allelesCAN1 can1?::MFA1pr?HIS3 MAT? query MATa xxx? Haploinsufficiency. Other genetic interactions that reflect gene-dosage effects can be crucial for cellular and devel- opmental homeostasis. In diploids, haploinsufficiency can arise when a mutation in one copy of an allelic pair reduces the amount of functional gene product to a point at which a phenotype is produced. Classically, a heterozygote is viewed as the wild type (that is, the mutant phenotype is recessive), and this is the case for most enzyme-coding genes 39 . However, for human transcription factors, over 65% of disease-causing mutations are dominant, and might reflect a haploinsufficient phenotype 39 . Haploinsufficiency can be particularly significant in the context of environ- mental or chemical interactions and has been exploited extensively to link inhibitory bioactive molecules to their targets, as heterozygote target-gene deletion mutants are often hypersensitive when compared with wild-type cells owing to their reduced target-gene dosage 40,41 . The combination of two heterozygous mutations might lead to a genetic interaction in which the diploid hemizygote double mutant shows an extreme synergistic phenotype, such as synthetic lethality. This combinato- rial double-mutant effect has been referred to as complex haploinsufficiency 42 . A screen of 4,800 complex hemizy- gote yeast strains, in which an actin-null allele was com- bined with the non-essential gene-deletion collection, identified 208 genes showing deleterious complex hap- loinsufficient (CHI) interactions and many of the double mutants showed actin-based morphology defects. Thus, CHI genetic-interaction screens can provide extensive functional information if carried out on a global scale. Quantitative mapping of epistatic relationships Synthetic methodologies allow a quantitative assessment of the relative fitness of double-mutant meiotic progeny. This means that, in addition to Fisher?s general idea of epistasis, other more specific ones, including Bateson?s classical definition in which one allele masks the effects at another locus (BOX 1), can be examined globally. In the Fisher model, the double-mutant growth rate should devi- ate from the expected multiplicative value that is associ- ated with the combined single-mutant phenotypes, and this can potentially be examined in detail. In particular, so-called aggravating interactions, in which the double- mutant fitness is lower than expected, might reflect sepa- rate but compensatory pathways. Synthetic-lethal double mutants obviously deviate from the multiplicative; how- ever, synthetic slow-growing double mutants with fitness rates that are less than either single mutant but equal to the expected multiplicative double-mutant fitness would not be scored as showing a genetic interaction. Using Fisher?s quantitative definition of epistasis may be important for identifying true interactions and thereby revise genetic networks that have not applied this model 43 . In contrast to aggravating interactions, so-called alleviating interactions occur when the double-mutant fitness is greater than expected, such as cases in which the fitness defect of a double mutant is no greater than for either of the single mutants. This often occurs when genes function in the same non-essential pathway or complex. Indeed, a quantitative analysis of an SGA interaction map 44 that focused on genes involved in endoplasmic reticulum (ER) to Golgi transport seems to support this idea, because genes in the same path- way deviated from the expected multiplicative dou- ble-mutant phenotype and displayed a level of fitness resembling the single-mutant phenotypes 25 . Thus, genes Figure 2 | The synthetic genetic array (SGA) methodology. a | A MAT? strain carries a query mutation linked to a dominant selectable marker (represented as a filled black circle), such as the nourseothricin-resistance marker natMX, and the SGA reporter can1?1::MFA1pr?HIS3 (in which MFA1pr?HIS3 is integrated into the genome such that it deletes the ORF of the CAN1 gene, which normally confers sensitivity to canavinine). This query strain is crossed to an ordered array of MATa deletion mutants (xxx?). In each of these deletion strains, a single gene is disrupted by the insertion of a dominant selectable marker, such as the kanamycin-resistance (kanR) module (the disrupted gene is represented as a filled blue circle). b | The resultant heterozygous diploids are transferred to a medium with reduced carbon and nitrogen to induce sporulation and the formation of haploid meiotic spore progeny. c | Spores are transferred to a synthetic medium that lacks histidine, which allows for selective germination of MATa meiotic progeny because these cells express the SGA reporter can1?1::MFA1pr?HIS3. To improve this selection, canavinine, which selects for can1?1 and kills CAN1 cells, is included in the selection medium. d | The MATa meiotic progeny are transferred to a medium that contains kanamycin, which selects for single mutants, equivalent to the original array mutants and double mutants. e,f | An array of double mutants is selected on a medium that contains both nourseothricin and kanamycin. REVIEWS NATURE REVIEWS | GENETICS VOLUME 8 | JUNE 2007 | 441 a Construction of the haploid-convertible pool b dSLAM Convertible a/? pool a/? pool Convertible a/? pool Amplification of barcodes Hybridization to DNA microarray Deletion mutationsWild-type allelesCAN1 can1?::MFA1pr?HIS3 MATa haploid selection (MFA1pr?HIS3) Sporulation Control pool (single mutants) Experimental pool (double mutants) Transform CAN1L::MFA1pr?HIS3::CAN1R Transform query mutation in the same pathway share alleviating interactions with each other. Exploring systems and pathways with quantitative and qualitative interaction maps. Several recent studies have widened the types of genetic interaction that can be identified in yeast. Drees 45 and colleagues defined a range of interactions by enumerating all possible ?greater than?, ?less than? and ?equal to? relationships among single- and double-mutant invasive growth phenotypes. They also scored for nine general types of epistatic interaction, including aggravating and alleviating types, but also for less familiar ones that were not previously considered. For example, in an ?asynthetic? interaction, a double mutant and its corresponding single mutants all have the same deviant phenotype, which is a specific subset of alleviating interactions. This broad analysis revealed that genetic interactions can occur frequently and allowed construction of elaborate interaction networks. In a theoretical analysis, Segre and colleagues exam- ined the predicted fitness of a double mutant under a multiplicative model and showed that, in addition to aggravating (antagonistic) synthetic effects, alleviating (buffering) interactions that ameliorate the effects of a mutation in double-mutant combinations are common among the genes involved in intermediary metabolism 46 . As with synthetic interaction studies, they found that alleviating interactions tend to be the same for related groups of genes, revealing functional or modular cluster- ing. Comparisons between aggravating and alleviating effects revealed that, for most functional groups, interac- tions were either largely aggravating or largely alleviat- ing, but not mixed, an asymmetrical feature that they termed ?monochromatic?. In another study, 650 double-deletion strains were made, corresponding to all possible pairings of 26 dele- tions that confer sensitivity to the DNA-damaging agent methyl methanesulphonate (MMS) 43 . The fitness of each strain was measured and examined with respect to the multiplicative neutral model. In the presence of MMS, approximately one-third of the unique double mutants that were tested were found to deviate from the multi- plicative model, corresponding to both aggravating and alleviating combinations. Distinct forms of alleviating interactions were identified, and those that were asym- metrical were used to infer pathway order corresponding to the classical Bateson definition of epistasis. Properties of genetic networks Genetic networks are complex but functionally coherent. Analysis of the large but still incomplete yeast genetic network offers a glimpse at its size and structure. From a set of SGA screens, a network of ~1,000 genes and ~4,000 interactions was generated 2 . The number of genetic interactions averaged 34 in each screen for non- essential genes 2 , with screens that were focused on essen- tial genes exhibiting fivefold more interactions 1 . From these studies, we estimate that a global network will contain ~200,000 synthetic-lethal interactions. To put this number in context, there are ~1,000 essential genes in yeast, for which a single mutation leads to a lethal Figure 3 | Diploid-based synthetic lethality analysis with microarrays (dSLAM). a | The first step in this method is the construction of a haploid-convertible heterozygous diploid pool. A haploid selection synthetic genetic array (SGA) reporter, which includes sequences that flank the endogenous CAN1 locus (CAN1L?LEU2? MFA1pr?HIS3?CAN1R), is transformed into a pool of heterozygous diploid deletion mutants to replace one copy of CAN1 in each mutant. In each of these deletion strains, a single gene is disrupted by the insertion of a kanamycin-resistance (kanR) module (the disrupted gene is represented as a filled blue circle), which is tagged with unique barcodes, and a wild-type copy of the same gene. Transformants are selected on plates and then pooled for genetic-interaction screens. b | For dSLAM, a query mutation that is linked to the URA3 selectable marker (represented as a filled black circle) is introduced into the pool of haploid-convertible heterozygous diploid strains by high-efficiency integrative transformation. Haploid single-mutant (control) or double-mutant (experimental) pools are selected after sporulation, through germination of spores on a medium that lacks histidine and selection for the relevant alleles. Genomic DNA samples are isolated from both pools and used as templates for PCR amplification of the tags, during which they are labelled with fluorescent dyes (Cy5 for the single-mutant pool and Cy3 for the double-mutant pool). Microarray analysis of these dye-labelled tags reveals the synthetic interaction between each of the corresponding deletion alleles with the query mutation. REVIEWS 442 | JUNE 2007 | VOLUME 8 www.nature.com/reviews/genetics phenotype, but there are 200-fold more ways to generate a similar phenotype through a digenic synthetic-lethal interaction. This finding indicates that digenic interac- tions might underlie many inherited phenotypes, and begins to explain why the analytical power of single-gene effects on many phenotypes has been so limited. For both non-essential 2,32 and essential genes 1 , genetic interactions tend to occur among functionally related genes (FIG. 4), although interactions of essential genes cor- respond to a broader functional range. So, the set of inter- actions that are observed for a particular query gene can be suggestive of its function, with the position of a gene in a genetic-interaction network being highly predictive of its molecular role. For example, when a deletion allele of BNI1, which functions in actin-based polarized secre- tion and spindle orientation 47 , was screened against all viable gene-deletion mutants, most of the interacting genes had roles in cell polarity and spindle orientation (annotated as ?mitosis? in FIG. 4). By contrast, the genetic interactions for SGS1, which encodes a DNA helicase, were largely associated with roles in DNA synthesis and repair (FIG. 4). The small world of genetic interactions. The current syn- thetic genetic network for yeast has two properties that are shared by networks as diverse as the World Wide Web and protein?protein interaction maps 48 . First, the connec- tivity distribution broadly follows a power-law distribu- tion, containing many genes with few interactions and a few genes with many interactions 2 . Highly connected ?hub genes? are likely to be more important for fitness than less connected genes, because random mutations in organisms that lack these genes are more likely to be asso- ciated with a fitness defect. Indeed, yeast hub genes that are conserved in humans could be potential targets for anti-cancer drugs, because cancer cells often carry a large mutation load making them more susceptible to chemical perturbation, and therefore may be killed preferentially when network hubs are attacked 49 . Second, the genetic network seems to be an example of a small-world network in which the length of the shortest path between a pair of vertices or nodes tends to be small (that is, the network has a short characteristic path length) and local neighbourhoods tend to be densely connected. The genetic network that was mapped by Tong et al. 2 has a short characteristic path length of 3.3, which is consistent with a small-world network 48 . The topology of the genetic network also exhibits dense local neighbourhoods, as the immediate neighbours of a gene, its genetic-interaction partners, also tend to interact with one another 2 . The dense neighbourhood characteristic of small-world net- works is of particular interest because it can be exploited to predict interactions, as previously shown for protein? protein interactions 50 . Thus, if all the yeast genes are placed on a relatively sparse genetic network ? that is, a network that contains most or all the genes with a small subset of their interactions ? most interactions should be efficiently identified by testing for interactions among genes that share interaction partners (in the same neigh- bourhood). Indeed, when the immediate neighbours of three query genes, SGS1, RAD27 or BIM1 were tested for interactions with one another, ~20% of the tested potential interactions were confirmed 2 , and were highly enriched compared with the 1% observed for the average query gene against all SGA-tested gene pairs. Genetic networks reveal gene functions Relationship between the physical-interaction and the genetic-interaction maps. Large-scale analysis of genetic networks has revealed a relationship between the physi- cal-interaction and the genetic-interaction networks. The physical-interaction map, generated by large-scale two-hybrid 51,52 or affinity purification followed by mass spectrometry identification 26,43,53,54 , provides a view of the gene products that assemble into soluble protein com- plexes and function together as biochemical machines. Rather than physical information, the genetic-interaction map provides functional information, largely identify- ing gene products that operate in functionally related pathways. Although genetic interactions overlap with protein?protein interactions more often than expected by chance, such overlap is relatively rare, occurring at a frequency of less than 1% (REF. 2). Neither the genetic- nor the physical-interaction map has been deeply sampled so far, and the overlap between the maps might increase. Nonetheless, a large overlap between the two is not expected as far as genes that encode components of non-essential pathways are concerned, because physical interactions should occur among the pathway components but synthetic-lethal interactions would be precluded by definition (BOX 2; FIG. 5a). However, synthetic-lethal interactions are expected among the components of essential pathways and, in this case, physical and genetic interactions might overlap (BOX 2; FIG. 5b) ? these are so-called within- pathway interactions 55 . Regardless, essential genes often buffer numerous different pathways 1 , and therefore most interactions for these genes occur between pathways and show no overlap with physical interactions (FIG. 5b). Because most genetic interactions do not overlap with physical interactions, the two types of interaction are said to be largely orthogonal 55?57 . Nevertheless, the genetic-interaction map is rich in physical-interaction information. For example, the set of interacting genes that is associated with a particular query is often enriched for all of the genes encoding the components of a func- tionally related pathway or complex. This makes sense, because if the activity of a particular pathway or complex is required in the absence of function of the query gene, then genes encoding all of the important components of that pathway or complex should be identified in the synthetic-lethal screen. Because a given query gene often shows in the order of ~30 different interactions, and most pathways contain only a handful of genes, each gene seems to buffer numerous other pathways. Precise biochemical functions can be deciphered from genetic-interaction maps because genes with prod- ucts that function in the same pathway or complex often show a similar pattern of genetic interactions 2 . Indeed, clustering algorithms or other measures of shared genetic- interaction patterns, such as the congruency score 57 , can be used to identify genes encoding components Nodes In typical network diagrams, genes or proteins are represented as nodes, whereas the connections between the nodes are termed edges. Clustering algorithms Algorithms that group together objects that are ?similar?; objects belonging to other clusters are ?dissimilar?. Clustering algorithms have been used extensively to view large collections of biological data, such as microarray expression profiles and genetic-interaction data. Congruency score A numerical ranking of the degree of partner sharing in a network. REVIEWS NATURE REVIEWS | GENETICS VOLUME 8 | JUNE 2007 | 443 Cell polarity Cell-wall maintenance Cell structure Mitosis Chromosome structure DNA synthesis and repair Unknown Others VPS29VPS29 RPS23ARPS23A RPS18BRPS18B VAM7VA REM50REM50 RPL16ARPL16A FPR1FPR1 CLB4CLB4 NBP2NBP2 CIN2CIN2 YGL211wGL211w YGL217cGL217c KIP3KIP3 CSM3CSM3 DDC1DDC1 XRS2XRS2 RAD57RAD5 RAD17RAD17 RAD24RAD24 RAD51RAD5 RAD55RAD55 EX01EX CAC2CA MRE11MRE11 CTF4TF4 HST3HST3 HST1HST1 RPL27ARPL27A DOC1DOC1 RPS30BRPS30B YNL171cYNL171c ESC2ESC2 RPL24ARPL24A YBR094wYBR094w YNL218wYNL218w RRM3RRM3 RNR1RNR1WSS1WSS1 SLX1SLX1 SLX4SLX4 ASF1ASF1 SWE1SWE1 YGL250wGL250w YDR018cYDR018c DEP1DEP1 SR09SR09 ARP2ARP2 YER083cYER083c CHS6CHS6 HOC1HOC1 SPF1SPF1 KRE1KRE1 VRP1VRP1 PEA2PEA2 ARC40ARC40 CCT3CC TFP3TFP3 CIK1CIK1 GLO3GLO3 SEC22SEC22 ARC18ARC18 RAS2RAS2 SDS3SDS3 YLR235cYLR235c BBC1BBC1 BEM4BEM4 BEM1BEM1 AST1AST1 SMI1SMI1 PAC1PA YDR149cYDR149c ASE1ASE1 ARP1ARP1 JNM1JNM1 DYN1DY DYN2DY SNC2SNC2 YKR047wYKR04 YLR190wYLR190w YNL119wYNL119w YBL051cYBL051c YHR111wYHR111w YPT6YPT6 GIN4GIN4 ELM1ELM1 CLA4CLA4 YBL062wYBL062w CHS7CHS7 SKT5SKT5 CHS3CHS3 BNI4BNI4 BCK1BCK1 SLT2SLT2 DRS2DRS2 PCL1PCL1 NAP1NAP1 VPS28VPS28 BNR1BNR1 SHS1SHS1 YMR299cYMR299c NIP100NIP100 TUS1TUS1 CYK3CYK3 BUD6BUD6 NUM1NUM1 PAC11C11 FAB1FA BNI1BNI1 SLA1SLA1 SAC6SAC6 ELP3ELP3 PAC10C10 CAP2CA CAP1CA GIM5GIM5 GIM3GIM3 GIM4GIM4 ELP2ELP2 YMLO95c-AYMLO95c- CHS5CHS5 YKE2YKE2 BEM2BEM2 MYO5MYO5 RIM101RIM101 SUM1SUM1 SAP155SAP155 RUD3RUD3 MNN11MNN11 SEC66SEC66 CPR7CPR7 SHE4SHE4 ILM1ILM1 STE24TE24 YLR111wYLR111w RVS167RV RVS161RV 61 PRK1PRK1 UTH1UTH1 SAC7SAC7 BTS1BTS RGD1RG POL32POL3 SGS1SGS1 PUB1PUB1 ESC4ESC4 TOP1OP1 SAE2SAE2 HPR5HPR5 MMS4MMS4 MUS81MUS81 RAD50RAD50 SIS2SIS2 SOD1SOD1YDJ1YDJ1 LYS7LY YPR116wYPR116w YLR352wYLR352w HPC2HPC2 FYV11FYV11 RAD52RAD5 RAD9RAD9 RAD27RAD27 BUB1BUB1 BUB2BUB2 BUB3BUB3 MAD2MAD2 MAD3MAD3 BFA1BFA1 BIK1BIK1 CHL4CHL4 MCK1MCK1 SLK19SLK19 BIM1BIM1 MCM22MCM22 MCM21MCM21 ARP6ARP6 IML3IML3 CTF8TF8 CTF19TF19 DCC1DCC1 PHO23PHO23 SAP30SAP30 PPZ1PPZ1 INP52INP5 YTA7YTA7 KEM1KEM1 IES2IES2 VID22VID22 AOR1OR1 MRC1MRC1 YBR095cYBR095c YLR386wYLR386w YNL170wYNL170w YPL017cYPL017c RTT103T10 YLR381wYLR381w RAD54RAD54 MAD1MAD1 KAR9KAR9 MON1MON1 YDL063cYDL063c RTG2RT RTG3RT CPR6CPR6 Figure 4 | A yeast genetic-interaction network, as determined by synthetic genetic array (SGA) analysis. A genetic-interaction network was obtained by identifying synthetic-lethal or synthetic-sick interactions using SGA analysis. Genes are represented as nodes (shown as circles), and interactions are represented as edges (shown as lines) that connect the nodes: 291 interactions and 204 genes from eight different SGA screens are shown. Deletion-mutant alleles of BNI1, RAD27, SGS1, BBC1, NBP2, BIM1 and temperature-sensitive conditional alleles of ARP2 and ARP40 were crossed to the set of ~5,000 viable yeast deletion mutants and scored for synthetic-lethal or synthetic-sick double-mutant interactions. All interactions were confirmed by tetrad analysis, with 8?14 tetrads examined in each case. The genes are coloured according to their cellular roles as annotated by the Yeast Proteome Database (YPD) (see the BIOBASE web site). Modified with permission from REF. 30 ? (2001) American Association for the Advancement of Science. of that pathway or complex (FIG. 6a). For example, on the basis of genetic-interaction patterns, CSM3 was linked to the S-phase replication checkpoint pathway and DYN3 (also known as YMR299c) was linked to the dynein?dynactin pathway 2 . From an extensive analysis of the DNA-integrity network in yeast, 16 functional modules or mini-pathways were identified on the basis of global patterns of genetic interactions 32 . Ultimately, the combination of the global genetic-interaction map and the physical-interaction map can be simplified by representation as a higher-order network in which the nodes represent complexes and pathways rather than individual genes, and the edges represent a collection of numerous synthetic genetic interactions that are associated with the individual genes of the pathway or complex 55,57,58 . Deciphering enzyme target relationships from genetic networks. Because synthetic-lethal interactions often identify pathways that buffer one another, genetic-interaction maps are useful for predicting enzyme?substrate relationships. For example, if a gene encoding a kinase is identified in a synthetic-lethal screen, then genes encoding upstream activators and downstream targets of the kinase might also be found in the genetic-interaction profile from the same query. Indeed, a synthetic-lethal screen with a CLA4 query mutation identified both the gene encoding a p21- activated kinase, STE20, and the formin gene BNI1, the product of which is postulated to be activated by the Ste20 kinase 59 . By contrast, SDL can be particularly useful for iden- tifying proteins that are negatively regulated by specific REVIEWS 444 | JUNE 2007 | VOLUME 8 www.nature.com/reviews/genetics A2 A1 A3 C1 C2 C3 Essential complex Non-essential complex Non-essential complex Non-essential complex A2 A1 A3 B2 B3 B1 a b Isogenic Strains or organisms that share identical genotypes. Gene association studies Studies that assess whether genotype frequencies are different between two groups that differ in phenotype. enzymes. For example, if a kinase normally negatively regulates a particular substrate, then overproduction of that substrate in the relevant kinase mutant back- ground might overwhelm the ability of the cell to cope with inappropriate regulation of a significant biological pathway. Indeed, of the 65 synthetic dosage interactions that were observed for the kinase gene PHO85 (REF. 38), four substrates of Pho85 (Pho4, Gsy1, Gsy2 and Gcn4) were identified, each of which is negatively regulated by Pho85 phosphorylation. Challenges for the future Synthetic lethality, population genetics and complex inherited human disease. Yeast genetic-interaction studies involve an inbred isogenic strain under a single set of growth conditions. However, in human popula- tions the issues of an outbred population with high levels of genetic polymorphism and variable envi- ronmental conditions add considerable complexity. The Kruglyak group used yeast to address the issue of polymorphism in genetic interactions 60,61 . Using varia- tions in transcript expression levels between two yeast strains as ?endophenotypes? for QTL analysis, they examined the polymorphic alleles that were involved in the variation. Having identified a primary locus that functioned as a modulator of a given transcript or set of transcripts, they carried out a second search to identify any interacting secondary loci. Such locus pairs were estimated to be responsible for the variation that is seen among some 57% of transcripts. Importantly, 67% of the secondary loci that they identified had effects that were undetectable when assessed singly, the detec- tion of which required the two-step search 60,61 . Because this strategy requires the identification of a primary locus on the basis of its individual effect on transcript level, it cannot be used to examine the frequency of pairs of polymorphic alleles that are singly undetect- able but interact to affect transcript levels. Identifying such interacting loci remains a huge problem in all systems, including humans. To identify candidate interacting alleles in complex disease, it is useful to have a detailed understanding of the genetic polymorphisms in a population so that they can be assessed as contributing allelic components in gene association studies. For humans, this idea has led to the generation of the human HapMap (see the International HapMap Project web site), a database that includes most of the common polymorphisms that are present in the human population 62 . An extension of this idea would be to sequence and compare the genomes of affected and unaffected relatives for a given disease. Although this goal remains unattainable, recent work in yeast approaches it: hybridization of DNA from yeast strains to highly overlapping whole-genome DNA microarrays now allows the global detection of polymorphisms to a single nucleotide resolution 63 . The application of such technology in deciphering the genomic basis of complex phenotypes has been dem- onstrated 64 and, although challenging, the extension of such an approach to more complex systems, including humans, can be contemplated. Extrapolating from yeast: network conservation and prediction. Is the yeast genetic network likely to be a good comparative model for such networks in metazo- ans? The creation of RNAi libraries to target all predicted genes in metazoan models and human genomes offers the potential for genome-wide analysis in complex sys- tems. RNAi screens have been used to systematically identify the genes involved in many biological processes in Caenorhabditis elegans, and in fly and mammalian cell lines 65,66 , and screens to examine double-mutant interactions in metazoan systems are now underway. Focused analyses of interactions between genes involved in DNA repair and posterior patterning in the C. elegans embryo have already uncovered novel genes and genetic Figure 5 | Relationships between genetic and protein interactions for complexes. Interactions are shown for complexes, but the same principles apply to pathways. a | Genetic interactions between two non-essential complexes. Two complexes (A and B) comprising proteins that are encoded by non-essential genes are shown. Protein?protein interactions are indicated by contact between proteins (represented as coloured circles), whereas genetic interactions are indicated by black lines. Genetic interactions occur among the mutant alleles of the genes, but for representational purposes are shown here in the context of the proteins within the complexes. In this model, the two non-essential complexes impinge on the same essential pathway and buffer one another (as shown in FIG. 1a); therefore, genetic interactions occur between the two complexes, but do not occur for components within a particular complex. The genetic- interaction pattern that is associated with each component of the complex is identical; that is, the genes that encode B1, B2 and B3 each show genetic interactions with the genes encoding A1, A2 and A3, and the reverse is also true. b | Genetic interactions that occur within an essential complex, and between an essential and a non-essential complex. The proteins in complex C are each encoded by essential genes. In this model, complex C is buffered by the activity of complex A and thus genetic interactions occur between each component of the two complexes as well as between the genes that encode complex C components (within- pathway interactions). REVIEWS NATURE REVIEWS | GENETICS VOLUME 8 | JUNE 2007 | 445 Deletion array 1? 2? 3? 4? 5? 6? 7? 8? 9? 6? 7? 8? 9?2? 3? 4? 5? Gene A Gene B Gene C Gene D Chemical?genetic interaction Synthetic-lethal interaction 1? Query gene Deletion array Gene X Drug Drug Viable Phenotype Viable Lethal Viable Viable Lethal Synthetic genetic interaction Dynein?dynactin pathway Membrane trafficking DNA-replication checkpoint Sister-chromatid cohesion DNA-damage checkpoint Microtubule dynamics DNA-replication checkpoint and sister-chromatid cohesion Membrane traffic Deletion array Compound X c a Drug targetGene X Phenotypeb Drug target BUB3 ASE1 KA R9 CLB4 KI P3 Y GL217C MON1 BNI1 BIM1 KA R3 CIN1 CIN8 CSM 3 TO F 1 MR C 1 CT F 1 8 DC C1 CT F 8 CT F 4 KEM1 CO G 5 CO G 6 CO G 7 GO S1 TL G 2 SET2 DYN1 PAC11 DYN3 DYN2 JNM1 PAC1 NIP100 NUM1 BFA1 ARL1 ARL3 GYP1 CSM3 MRC1 TOF1 ELG1 CTF18 DCC1 CTF8 CTF4 DDC1 RAD9 RAD24 RAD52 Query genes ARP1 Figure 6 | Hierarchical clustering of genetic and chemical-genetic interactions. a | Two-dimensional hierarchical clustering of synthetic genetic interactions, as determined by synthetic genetic array (SGA) analysis. A small subset of the genetic interactions mapped by Tong et al. 2 is shown. The hierarchical clustering algorithm organizes the query and array genes into sets that show similar patterns of genetic interactions (shown as red squares), thereby grouping together components of specific functional pathways and complexes. Large-scale mapping of genetic interactions provides a genetic-interaction phenotype for each gene, and clustering analysis orders genes into pathways and complexes. b | The left panel shows a chemical?genetic interaction, in which a deletion mutant, which lacks the product of the deleted gene (represented by a black X), is hypersensitive to a normally sublethal concentration of a growth-inhibitory compound. The right panel shows a synthetic-lethal genetic interaction in which two single deletions individually lead to viable mutants but are not viable in a double-mutant combination. Gene-deletion alleles that show chemical? genetic interactions with a particular compound should also be synthetically lethal or sick when combined with a mutation in the compound target gene. c | Comparison of a chemical?genetic profile to a compendium of genetic interaction (synthetic lethal) profiles should identify the pathways and targets that are inhibited by drug treatment. A hypothetical example is shown. Deletion mutants 3, 5, and 6 are hypersensitive to compound X, and a mutation in query gene A leads to a fitness defect when combined with deletion alleles 1, 2, 3 and 4. Here the chemical?genetic profile of compound X resembles the genetic profile of gene B, identifying the product of gene B as a putative target of compound X. Part a modified with permission from REF. 2 ? (2004) American Association for the Advancement of Science. interactions for both processes 67,68 . More recently, large- scale RNAi mapping of genetic interactions for signalling and transcriptional-regulatory pathways in C. elegans uncovered ~350 genetic interactions. Again, both known and novel signalling components were identified. Despite its currently modest size, the C. elegans genetic network recapitulates the topology of genetic networks in yeast 69 , suggesting that a general network structure is conserved in eukaryotes. Indeed, some general principles of genetic networks for model organisms have already been shown to extend to human genetics, with individual anecdotal examples of complex inherited human diseases seem- ing to act through dense local neighbourhoods of interactions that resemble the yeast network topology. For example, the Bardet?Beidl syndrome is caused by interactive defects in genes involved in the assembly and function of the centrosome 70 . However, even in experimentally tractable organisms, the generation of genetic-interaction data is labour- intensive, and comprehensive interaction maps are some way off for most biological systems. It is therefore important to continue efforts to understand the topology of networks in simpler systems so that predictions about more complex organisms can be attempted. Prediction of genetic interactions in yeast following mapping of the local topology around gene pairs has provided candidate interacting pairs that are enriched for synthetic interac- tions 71 . Predictions that come from comparative genom- ics are also useful; for example, knowledge of interactions in both yeast and Drosophila melanogaster has assisted in identifying candidate interactions among orthologous genes in C. elegans 72 . Similar approaches should also pro- vide useful predictions for candidate interactions in less accessible systems. One implication of the high degree of interconnected- ness of the yeast network is that, assuming that human gene networks show the same properties, uncovering the genetic basis of disease susceptibility in humans will be a huge challenge. Many mutant alleles that have no discern- able individual effect could contribute to combinatorial synthetic effects that cause disease. Chemical genomics and genetic-interaction networks. To help to gain a global understanding of complex biological processes, small molecules can serve as a powerful coun- terpart to gene mutations as rapid and reversible modu- lators of gene activity. The use of such chemical probes on a genome-wide scale is called ?chemical genomics? and is well suited for use in yeast, in which simple assays for cell fitness are available. In principle, deletion of a gene that encodes the target of an inhibitory compound should cause cellular effects that are similar to inhibition of the target by drug treatment. If so, crossing a target deletion mutation into the set of ~5,000 viable yeast dele- tion mutants by SGA, and scoring the resultant double mutants for reduced fitness, should generate a set of syn- thetic-lethal interactions for the gene target that resembles the chemical?genetic interaction profile of its inhibitory compound (FIG. 6b,c). In a proof-of-principle study 73 , the chemical?genetic profiles of five different compounds were found to be highly similar to the genetic-interaction REVIEWS 446 | JUNE 2007 | VOLUME 8 www.nature.com/reviews/genetics profiles of the target gene or genes in the target pathway. In general, a comprehensive compendium of global genetic-interaction profiles should allow the targets of growth-inhibitory compounds to be identified. Given that most genes are non-essential, and that proper cell function reflects an interconnected robust- ness, both gene?drug and drug?drug combinations that inhibit such cellular systems should be investigated for therapeutic intervention. For example, an understanding of synthetic-lethal genetic interactions might enable the identification of compounds that target specific pathways and selectively kill cells with defined mutant genotypes in cancer pathways 74 . In fact, it is well established that combinations of molecules can provide highly effective drug regimens 75,76 . A systematic way to identify drugs that could have synergistic effects is by selecting pairs of drugs affecting targets that are themselves syntheti- cally lethal. For example, consider two drugs that target different essential gene products that show a synthetic genetic interaction. For essential genes, a synthetic-lethal interaction is often detected by assaying for the lethality of a double mutant that carries conditional alleles of the essential genes at what is normally a permissive tempera- ture for each of the single mutants. In this scenario, syn- thetic drug combinations that target each of the essential genes should act synergistically by working together at lower minimum inhibitory concentrations (MICs) than if used singly. Most importantly, this type of combina- tion therapy is not limited to essential gene products but, rather, encompasses the entire synthetic-lethal genetic- interaction network. Although there are only 1,000 essential target genes in S. cerevisiae, we estimate that there are ~200,000 synthetic digenic combinations. Thus, by using combinations of drugs that cut at the Achilles? heel of cell function, we can find a 200-fold-wider repertoire of drugs that work in a way that exploits a fundamental weakness of cellular networks. Conclusion Grasping an understanding of genetics through pheno- type can be a slippery task, as phenotypes seldom reflect the function of just one gene. In modern genetics, an iso- genic background allows a focus on the phenotype and function of individual genes, and this has been a useful initial strategy. However, lurking just below the surface is a complexity that we must face. For example, every intensively studied organism shows strain-background differences that everyone notices and almost everyone ignores. Indeed, the genetics we are taught is clearly a simplified and limited view of the nature of human vari- ation. A hard reality, then, is that most phenotypes are not caused by alleles of a single gene. Even most pheno- types that are thought to have their basis in a single gene have, on further study, been found to vary under the influence of many modifying genes. Quantitative traits and most inherited human diseases fall into this abyss of complexity, and progress in our understanding of them has been difficult and slow. Here we have examined how geneticists have begun to grapple with such genetic complexity. The advent of genomics and global gene catalogues, coupled with a growing understanding of the properties of biological networks, has facilitated the study of genetic interactions through double-mutant combinations in systems like yeast, for which there is the technology to manipulate and analyse large numbers of crosses. The resulting genetic networks strongly reflect function, with genetic interactions clustering as functional modules in dense local neighbourhoods. Furthermore, these networks emphasize the deep intrinsic buffering of cellular function through redundant or overlapping pathways. However, a minority of genes are essential, and these define hubs of activity that can in some cases extend beyond a given functional module to influence and even coordinate multiple cellular processes. It is no wonder, given this interactional complexity, that single genes rarely specify a phenotype in its entirety. The outlines of a yeast genetic network are now apparent, but a compel- ling case can be made for a deeper and more complete exploration of this model system as an exemplar for more complex eukaryotes. As useful as it is in defining general genetic principles, yeast only pioneers the way by validating the usefulness of such genetic analysis. We anticipate a growing flood of genetic-interaction networks from model organisms, including chordates such as zebrafish and mice. Such studies should progressively sharpen the outlines of our own, human genetic-interaction network space, and move beyond a comprehension of single-gene effects to a deeper understanding of our inheritance, including our susceptibility to environmental insult, and the basis of collective inherited disorders. 1. Davierwala, A. P. et al. The synthetic genetic interaction spectrum of essential genes. Nature Genet. 37, 1147?1152 (2005). This paper describes the first major application of SGA analysis to mapping of genetic-interaction networks among essential genes, and reveals that they seem to act as highly connected hubs on the network. 2. Tong, A. H. Y. et al. Global mapping of the yeast genetic interaction network. Science 303, 808?813 (2004). This study describes large-scale mapping of synthetic-lethal genetic interactions in yeast by SGA analysis. The results highlight the utility of genetic networks for discovering gene function and define the topology and general properties of genetic networks. 3. Hughes, T. R., Robinson, M. D., Mitsakakis, N. & Johnston, M. The promise of functional genomics: completing the encyclopedia of a cell. Curr. Opin. Microbiol. 7, 546?554 (2004). 4. Dolinski, K. & Botstein, D. Changing perspective in yeast research nearly a decade after the genome sequence. Genome Res. 15, 1611?1619 (2006). 5. Giaever, G. et al. Functional profiling of the Saccharomyces cerevisiae genome. Nature 418, 387?391 (2002). A landmark paper that describes the construction and use of the yeast deletion-mutant collection. 6. Winzeler, E. A. et al. Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 285, 901?906 (1999). 7. Hillenmeyer, M. E. et al. The chemical genomic portrait of the cell reveals a phenotype for all genes. (Submitted). 8. Deutschbauer, A. et al. Mechanisms of haploinsufficiency revealed by genome-wide profiling in yeast. Genetics 169, 1915?1925 (2005). 9. Hartman, J. L., Garvik, B. & Hartwell, L. Principles for the buffering of genetic variation. Science 291, 1001?1004 (2001). An excellent opinion piece that explores how eukaryotic genomes are buffered against genetic and environmental insults and outlines how synthetic-lethal interaction maps can be used to understand the relationship between genotype and phenotype. 10. Hartwell, L. H. Yeast and cancer. Biosci. Rep. 22, 373?394 (2002). 11. Dobzhansky, T. Genetics of natural populations, XIII: recombination and variability in populations of Drosophila pseudoobscura. Genetics 31, 269?290 (1946). REVIEWS NATURE REVIEWS | GENETICS VOLUME 8 | JUNE 2007 | 447 12. Sturtevant, A. H. A highly specific complementary lethal system in Drosophila melanogaster. Genetics 41, 118?123 (1956). 13. Novick, P., Osmond, B. C. & Botstein, D. Suppressors of yeast actin mutants. Genetics 121, 659?674 (1989). One of the first yeast papers to describe synthetic- lethal genetic interactions, with useful references to the early D. melanogaster literature. 14. Guarente, L. Synthetic enhancement in gene interaction: a genetic tool come of age. Trends Genet. 9, 362?366 (1993). 15. Bender, A. & Pringle, J. R. Use of a screen for synthetic lethal and multicopy suppressor mutants to identify two new genes involved in morphogenesis in Saccharomyces cerevisiae. Mol. Cell. Biol. 11, 1295?1305 (1991). This manuscript describes the first use of a yeast colony sectoring assay as screen for synthetic lethal genetic interactions. 16. Basson, M. E., Moore, R. L., O?Rear, J. & Rine, J. Identifying mutations in duplicated functions in Saccharomyces cerevisiae: recessive mutations in HMG-CoA reductase genes. Genetics 117, 645?655 (1987). 17. Suter, B., Auerbach, D. & Stagljar, I. Yeast-based functional genomics and proteomics technologies: the first 15 years and beyond. Biotechniques 40, 625?644 (2006). 18. Wach, A., Brachat, A., Pohlmann, R. & Philippsen, P. New heterologous modules for classical or PCR-based gene disruptions in Saccharomyces cerevisiae. Yeast 10, 1793?1808 (1994). 19. Kellis, M., Patterson, N., Endrizzi, M., Birren, B. & Lander, E. S. Sequencing and comparison of yeast species to identify genes and regulatory elements. Nature 423, 241?254 (2003). 20. Cliften, P. et al. Finding functional features in Saccharomyces genomes by phylogenetic footprinting. Science 301, 71?76 (2003). 21. Kastenmayer, J. P. et al. Functional genomics of genes with small open reading frames (sORFs) in S. cerevisiae. Genome Res. 16, 365?373 (2006). 22. Mnainmeh, S. et al. Exploration of essential gene functions via titrable promoter alleles. Cell 118, 31?44 (2004). 23. Dohmen, R. J. & Varshavsky, A. Heat-inducible degron and the making of conditional mutants. Methods Enzymol. 399, 799?822 (2005). 24. Kanemaki, M., Sanchez-Diaz, A., Gambus, A. & Labib, K. Functional proteomic identification of DNA replication proteins by induced proteolysis in vivo. Nature 423, 720?724 (2003). 25. Schuldiner, M. et al. Exploration of the function and organization of the yeast early secretory pathway through an epistatic miniarray profile. Cell 123, 507?519 (2005). 26. Ho, Y. et al. Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature 415, 180?183 (2002). 27. Butcher, R. A. et al. Microarray-based method for monitoring yeast overexpression strains reveals small-molecular targets in the TOR pathway. Nature Chem. Biol. 2, 103?109 (2006). 28. Zhu, H. et al. Global analysis of protein activities using proteome chips. Science 293, 2101?2105 (2001). 29. Gelperin, D. M. et al. Biochemical and genetic analysis of the yeast proteome with a movable ORF collection. Genes Dev. 19, 2816?2826 (2005). 30. Tong, A. H. Y. et al. Systematic genetic analysis with ordered arrays of yeast deletion mutants. Science 294, 2364?2368 (2001). This paper describes development of the SGA method and its application to synthetic-lethal genetic-interaction mapping. The study also provides the first glimpse of a genetic-interaction network. 31. Pan, X. et al. A robust toolkit for functional profiling of the yeast genome. Mol. Cell 16, 487?496 (2004). A study that describes the development of dSLAM, a transformation-based method of creating double mutants that provides a barcode microarray read- out for synthetic-lethal genetic interactions. 32. Pan, X. et al. A DNA integrity network in the yeast Saccharomyces cerevisiae. Cell 124, 1069?1081 (2006). This work describes the application of dSLAM analysis to the study of genes involved in DNA synthesis and repair, and genome integrity. 33. Surana, U. et al. The role of CDC28 and cyclins during mitosis in the budding yeast S. cerevisiae. Cell 65, 145?161 (1991). 34. Reguly, T. et al. Comprehensive curation and analysis of global interaction networks in Saccharomyces cerevisiae. J. Biol. 5, 11 (2006). 35. Kroll, E. S., Hyland, K. M., Hieter, P. & Li, J. J. Establishing genetic interactions by a synthetic dosage lethality phenotype. Genetics 143, 95?102 (1996). 36. Measday, V. & Hieter, P. Synthetic dosage lethality. Methods Enzymol. 350, 316?326 (2002). 37. Measday, V. et al. Systematic yeast synthetic lethal and synthetic dosage lethal screens identify genes required for chromosome segregation. Proc. Natl Acad. Sci. USA 102, 13956?13961 (2005). 38. Sopko, R. et al. Mapping pathways and phenotypes by systematic gene overexpression. Mol. Cell 21, 319?330 (2006). 39. Veitia, R. A. Exploring the etiology of haploinsufficiency. Bioessays 24, 175?184 (2002). 40. Lum, P. Y. et al. Discovering novel modes of action for therapeutic compounds unsing a genome-wide screen of yeast heterozygotes. Cell 116, 121?137 (2004). 41. Giaever, G. et al. Chemogenomic profiling: identifying the functional interactions of small molecules in yeast. Proc. Natl Acad. Sci. USA 101, 793?798 (2005). 42. Haarer, B., Viggiano, S., Hibbs, M. A., Troyanskya, O. G. & Amberg, D. C. Modeling complex genetic interactions in a simple eukaryotic genome: actin displays a rich spectrum of complex haploinsufficiencies. Genes Dev. 21, 148?159 (2007). 43. Krogan, N. et al. Global landscape of protein complexes in the yeast Saccharomyces cerevisiae. Nature 440, 637?643 (2006). 44. Collins, S. R., Schuldiner, M., Krogan, N. & Weissman, J. S. A strategy for extracting and analyzing large-scale quantitative epistatic interaction data. Genome Biol. 7, R63 (2006). This manuscript describes a method for generating quantitative genetic-interaction data sets using SGA analysis. 45. Drees, B. L. et al. Derivation of genetic interaction networks from quanitative phenotype data. Genome Biol. 6, R38 (2005). 46. Segre, D., Deluna, A., Church, G. M. & Kishony, R. Modular epistasis in yeast metabolism. Nature Genet. 37, 77?83 (2005). This study outlines the general concept that the expected phenotype of a double mutant is a multiplicative combination of two single mutants, and that scoring of deviations from this expected value generates genetic networks to describe functional relationships among metabolic pathways. 47. Evangelista, M. et al. Bni1p, a yeast formin linking Cdc42p and the actin cytoskeleton during polarized morphogenesis. Science 276, 118?122 (1997). 48. Barabasi, A. L. & Bonabeau, E. Scale-free networks. Sci. Am. 288, 60?69 (2003). 49. Kamb, A. Mutation load, functional overlap, and synthetic lethality in the evolution and treatment of cancer. J. Theor. Biol. 223, 205?213 (2003). 50. Goldberg, D. S. & Roth, F. P. Assessing experimentally derived interactions in a small word. Proc. Natl Acad. Sci. USA 100, 4372?4376 (2003). 51. Ito, T. et al. A comprehensive two-hybrid analysis to explore the yeast protein interactome. Proc. Natl Acad. Sci. USA 98, 4569?4574 (2001). 52. Uetz, P. et al. A comprehensive analysis of protein?protein interactions in Saccharomyces cerevisiae. Nature 403, 623?627 (2000). 53. Gavin, A. C. et al. Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature 415, 141?147 (2002). 54. Gavin, A. C. et al. Proteome survey reveals modularity of the yeast cell machinery. Nature 440, 631?636 (2006). 55. Kelley, R. & Ideker, T. Systematic interpretation of genetic interactions using protein networks. Nature Biotech. 23, 561?566 (2005). 56. Bader, G. D. et al. Functional genomics and proteomics: charting a multidimensional map of the yeast cell. Trends Cell Biol. 13, 344?356 (2003). 57. Ye, P. et al. Gene function prediction from congruent synthetic lethal interactions in yeast. Mol. Syst. Biol. 1, 2005.0026 (2005). 58. Zhang, L. V. et al. Motifs, themes and thematic maps of an integrated Saccharomyces cerevisiae interaction network. J. Biol. 4, 6 (2005). 59. Goehring, A. S. et al. Synthetic lethal analysis implicates Ste20p, a p21-activated protein kinase, in polarisome activation. Mol. Biol. Cell 14, 1501?1516 (2003). 60. Brem, R. B. & Kruglyak, L. The landscape of genetic complexity across 5,700 gene expression traits in yeast. Proc. Natl Acad. Sci. USA 102, 1572?1577 (2005). 61. Brem, R. B., Storey, J. D., Whittle, J. & Kruglyak, L. Genetic interactions between polymorphisms that affect gene expression in yeast. Nature 436, 701?703 (2005). 62. Consortium, I. H. A haplotype map of the human genome. Nature 437, 1299?1320 (2005). 63. Gresham, D. et al. Genome-wide detection of polymorphisms at nucleotide resolution with a single DNA microarray. Science 311, 1932?1936 (2006). 64. Perstein, E. O., Ruderfer, D. M., Roberts, D. C., Schreiber, S. L. & Kruglyak, L. Genetic basis of individual differences in response to small- molecule drugs in yeast. Nature Genet. 39, 496?502 (2007). 65. Moffat, J. & Sabatini, D. M. Building mammalian signalling pathways with RNAi screens. Nature Rev. Mol. Cell Biol. 7, 177?187 (2006). 66. Echeverri, C. J. & Perrimon, N. High-throughput RNAi screening in cultured cells: a user?s guide. Nature Rev. Genet. 7, 373?384 (2006). 67. Baugh, L. R. et al. Synthetic lethal analysis of Caenorhabditis elegans posterior embryonic patterning genes identified conserved genetic interactions. Genome Biol. 6, R45 (2005). 68. van Haaften, G., Vastenhouw, N. L., Nollen, E. A., Plasterk, R. H. & Tijsterman, M. Gene interactions in the DNA damage-response pathway identified by genome-wide RNA-interference analysis of synthetic lethality. Proc. Natl Acad. Sci. USA 101, 12992?12996 (2004). 69. Lehner, B., Crombie, C., Tischler, J., Fortunato, A. & Fraser, A. G. Systematic mapping of genetic interactions in C. elegans. Nature Genet. 38, 896?903 (2006). This study describes the first large-scale mapping of synthetic genetic networks in a metazoan, generated by feeding hypomorphic C. elegans mutants arrays of bacteria that expressed dsRNAi molecules targeting specific signalling pathways. 70. Badano, J. L., Teslovich, T. M. & Katsanis, N. The centrosome in human disease. Nature Rev. Genet. 6, 194?205 (2005). 71. Wong, S. L. et al. Combining biological networks to predict genetic interactions. Proc. Natl Acad. Sci. USA 101, 15682?25687 (2004). The first paper to show that functional genomics data sets can be used to predict genetic interactions. 72. Zhong, W. & Sternberg, P. Genome-wide prediction of C. elegans genetic interactions. Science 311, 1481?1484 (2006). 73. Parsons, A. B. et al. Integration of chemical-genetic and genetic interaction data links bioactive compounds to cellular targets and pathways. Nature Biotech. 22, 62?69 (2004). This work describes how synthetic-lethal genetic-interaction maps function as a key for deciphering chemical-genetic maps, providing a means of linking compounds to their target pathways. 74. Sharom, J. R., Bellows, D. S. & Tyers, M. From large networks to small molecules. Curr. Opin. Chem. Biol. 8, 81?90 (2004). 75. Keith, C. T., Borisy, A. A. & Stockwell, B. R. The identification of combinations of molecules can result in highly effective drug regimens. Nature Rev. Drug Discov. 4, 71?78 (2003). 76. Borisy, A. A. et al. Systematic discovery of multicomponent therapeutics. Proc. Natl Acad. Sci. USA 100, 7977?7982 (2003). 77. Phillips, P. C. The language of gene interaction. Genetics 149, 1167?1171 (1998). A wonderful review of the language that is used to describe genetic interactions, including terms like epistasis, with historical context. 78. Fisher, R. A. The correlations between relatives on the supposition of Mendelian inheritance. Trans. R. Soc. Edinb. 52, 399?433 (1918). 79. Sternberg, P., Stern, M. J., Clark, I. & Herskowitz, I. Activation of the yeast HO gene by release from muliple negative controls. Cell 48, 567?577 (1987). 80. Hartwell, L. H., Culotti, J., Pringle, J. R. & Reid, B. J. Genetic control of cell division cycle in yeast. Science 183, 46?51 (1974). REVIEWS 448 | JUNE 2007 | VOLUME 8 www.nature.com/reviews/genetics 81. Sprague, G. F. Jr & Thorner, J. W. in The Molecular and Cellular Biology of the Yeast Saccharomyces: Gene Expression Vol. 2 (eds Jones, E. W., Pringle, J. R. & Broach, J. R.) 657?744 (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, 1992). 82. Avery, L. & Wasserman, S. Ordering gene function: the interpretation of epistasis in regulatory hierarchies. Trends Genet. 8, 312?316 (1992). 83. Baker, B. S. & Ridge, K. A. Sex and the single cell. I. On the action of major loci affecting sex determination in Drosophila melanogaster. Genetics 94, 383?423 (1980). 84. Ihmels, J., Collins, S. R., Schuldiner, M., Krogan, N. & Weissman, J. S. Backup without redundancy: genetic interactions reveal the cost of duplicate gene loss. Mol. Syst. Biol. 3, 86 (2007). 85. Harrison, R., Papp, B., Pal, C., Oliver, S. G. & Delneri, D. Plasticity of genetic interactions in metabolic networks of yeast. Proc. Natl Acad. Sci. USA 104, 2307?2312 (2007). 86. Forsburg, S. L. The art and design of genetic screens: yeast. Nature Rev. Genet. 2, 659?668 (2001). 87. Kaiser, C. A. & Schekman, R. Distinct sets of SEC genes govern transport vesicle formation and fusion early in the secretory pathway. Cell 61, 723?733 (1990). 88. Finger, F. & Novick, P. Synthetic interactions of the post-golgi sec mutations of Saccharomyces cerevisiae. Genetics 156, 943?951 (2000). Acknowledgements Genetic interaction network projects in the Andrews and Boone laboratories are supported by grants from the Canadian Institutes of Health Research (CIHR) and Genome Canada through the Ontario Genomics Institute. C.B. is an International Scholar of the Howard Hughes Medical Institute and holds a Canada Research Chair in Functional Genomics. We would like to thank Amy Hin Yan Tong and Michael Costanzo for help with the figures and comments on the manuscript. Competing interests statement The authors declare no competing financial interests. DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query. fcgi?db=gene BIM1 | BNI1 | CAN1 | CDC28 | CLA4 | CSM3 | DYN3 | GAL1 | HO | PHO85 | RAD27 | SGS1 | SIN3 | SWI5 OMIM: http://www.ncbi.nlm.nih.gov/entrez/query. fcgi?db=OMIM Bardet?Beidl syndrome FURTHER INFORMATION Andrews lab: http://www.utoronto.ca/andrewslab Boone lab: http://www.utoronto.ca/boonelab BIOBASE: http://www.biobase-international.com International HapMap Project: http://www.hapmap.org Saccharomyces genome database: http://www.yeastgenome.org The general repository for interaction datasets: http://www.thebiogrid.org Access to this links box is available online. REVIEWS NATURE REVIEWS | GENETICS VOLUME 8 | JUNE 2007 | 449 "
Add Content to Group
|
Bookmark
|
Keywords
|
Flag Inappropriate
share
Close
Digg
Facebook
MySpace
Google+
Comments
Close
Please Post Your Comment
*
The Comment you have entered exceeds the maximum length.
Submit
|
Cancel
*
Required
Comments
Please Post Your Comment
No comments yet.
Save Note
Note
View
Public
Private
Friends & Groups
Friends
Groups
Save
|
Cancel
|
Delete
Please provide your notes.
Next
|
Prev
|
Close
|
Edit
|
Delete
Genetics
Gene Inheritance and Transmission
Gene Expression and Regulation
Nucleic Acid Structure and Function
Chromosomes and Cytogenetics
Evolutionary Genetics
Population and Quantitative Genetics
Genomics
Genes and Disease
Genetics and Society
Cell Biology
Cell Origins and Metabolism
Proteins and Gene Expression
Subcellular Compartments
Cell Communication
Cell Cycle and Cell Division
Scientific Communication
Career Planning
Loading ...
Scitable Chat
Register
|
Sign In
Visual Browse
Close
Comments
CloseComments
Please Post Your Comment