Systematic Genetic Analysis with Ordered Arrays of Yeast Deletion Mutants

See allHide authors and affiliations

Science  14 Dec 2001:
Vol. 294, Issue 5550, pp. 2364-2368
DOI: 10.1126/science.1065810


In Saccharomyces cerevisiae, more than 80% of the ∼6200 predicted genes are nonessential, implying that the genome is buffered from the phenotypic consequences of genetic perturbation. To evaluate function, we developed a method for systematic construction of double mutants, termed synthetic genetic array (SGA) analysis, in which a query mutation is crossed to an array of ∼4700 deletion mutants. Inviable double-mutant meiotic progeny identify functional relationships between genes. SGA analysis of genes with roles in cytoskeletal organization (BNI1,ARP2, ARC40, BIM1), DNA synthesis and repair (SGS1, RAD27), or uncharacterized functions (BBC1, NBP2) generated a network of 291 interactions among 204 genes. Systematic application of this approach should produce a global map of gene function.

For S. cerevisiae deletion mutations have been constructed for all ∼6200 known or suspected genes, identifying ∼1100 essential yeast genes and resulting in 5100 viable haploid gene-deletion mutants, with over 30% of the genes remaining functionally unclassified (1). These findings highlight the capacity of yeast cells to tolerate deletion of a substantial number of individual genes, perhaps reflecting the molecular mechanisms that evolved to buffer the phenotypic consequences of genetic variation (2, 3). Due to the high degree of genetic redundancy in yeast, the functions of thousands of yeast genes remain obscure.

Redundant functions can often be uncovered by synthetic genetic interactions, usually identified when a specific mutant is screened for second-site mutations that either suppress or enhance the original phenotype. In particular, two genes show a “synthetic lethal” interaction if the combination of two mutations, neither by itself lethal, causes cell death (4, 5). Synthetic lethal relationships may occur for genes acting in a single biochemical pathway or for genes within two distinct pathways if one process functionally compensates for or buffers the defects in the other (2). Synthetic lethal screens have been used to identify genes involved in cell polarity, secretion, DNA repair, and numerous other processes (6–9). Despite the utility of this approach, just one or two different interactions are typically identified in a single screen (2). Saturation is rarely achieved presumably because some genes are refractory to forward mutagenesis and because subsequent analysis of synthetic lethal mutations and gene cloning is limited by practical constraints.

To enable high-throughput synthetic lethal analysis, we assembled an ordered array of ∼4700 viable yeast gene-deletion mutants (1) and developed a series of pinning procedures in which mating and meiotic recombination are used to generate haploid double mutants (Fig. 1). A query mutation is first introduced into a haploid starting strain, of mating type MATα, and then crossed to the array of gene-deletion mutants of the opposite mating type, MAT a. Sporulation of resultant diploid cells leads to the formation of double-mutant meiotic progeny. TheMATα starting strain carries a reporter,MFA1pr-HIS3, that is only expressed inMAT a cells and allows for germination ofMAT a meiotic progeny (10), which ensures that carryover of the diploid parental strain and/or conjugation of meiotic progeny does not give rise to false-negative interactions. Both the query mutation and the gene-deletion mutations were linked to dominant selectable markers to allow for selection of double mutants. Final pinning results in an ordered array of double-mutant haploid strains whose growth rate is monitored by visual inspection or image analysis of colony size. We refer to this procedure as synthetic genetic array (SGA) analysis.

Figure 1

Synthetic genetic array methodology (16). (i) A MATα strain carrying a query mutation (bni1Δ) linked to a dominant selectable marker, such as the nourseothricin-resistance marker natMX that confers resistance to the antibiotic nourseothricin, and anMFA1pr-HIS3 reporter is crossed to an ordered array ofMAT a viable yeast deletion mutants, each carrying a gene deletion mutation linked to a kanamycin-resistance marker (kanMX). Growth of resultant heterozygous diploids is selected for on medium containing nourseothricin and kanamycin. (ii) The heterozygous diploids are transferred to medium with reduced levels of carbon and nitrogen to induce sporulation and the formation of haploid meiotic spore progeny. (iii) Spores are transferred to synthetic medium lacking histidine, which allows for selective germination of MAT a meiotic progeny because these cells express the MFA1pr-HIS3 reporter specifically. (iv) The MAT a meiotic progeny are transferred to medium that contains both nourseothricin and kanamycin, which then selects for growth of double-mutant meiotic progeny.

To demonstrate proof of concept for SGA analysis, we tested a deletion of the BNI1 gene (bni1Δ) against a test array of gene-deletion mutants including bnr1Δ, a known synthetic lethal with bni1Δ (11) (Fig. 2A). BNI1 and BNR1both encode members of the highly conserved formin family and control the assembly of actin cables, which guide myosin motors that coordinate polarized cell growth and spindle orientation (12). As anticipated, the double-mutant cells at the bnr1Δ positions failed to grow, forming a residual colony with a reduced size relative to that of the control. Because the resultant double mutants are created by meiotic recombination, gene deletions that are genetically linked to the query mutation form double mutants at a reduced frequency; moreover, when the query mutation is identical to one of the gene deletions within the array, double mutants cannot form. Thus, cells at the bni1Δ position failed to grow under double-mutant selection. Even within the limited subset of genes in this pilot experiment, we detected previously unknown synthetic genetic interactions between bni1Δ and gene deletions ofCLA4 and BUD6 (AIP3) (Fig. 2A). Cla4 is a kinase involved in actin patch assembly and regulation of the cell cycle–dependent transition from apical to isotropic bud growth (13, 14); Bud6 forms a complex with Bni1 and actin to control actin cable assembly and cell polarity (15). Tetrad analysis confirmed that both thebni1Δ bnr1Δ and bni1Δcla4Δ double mutants were inviable and that thebni1Δ bud6Δ double mutant was associated with a slower growth rate or “synthetic sick” phenotype, reflecting reduced fitness of the double mutant relative to the respective single mutants (Fig. 2B).

Figure 2

Final double-mutant array and tetrad analysis for SGA synthetic lethal analysis with bni1Δ and a test array (16). (A)bni1Δ::natR cells were crossed to a test array containing 96 deletion mutants, each arrayed in quadruplicate in a square pattern. bnr1Δ was duplicated within the array. The final array that selects for growth of thebni1Δ double mutants is shown. Synthetic lethal/sick interactions lead to the formation of residual colonies (yellow circles) that were relatively smaller than the equivalent colony on the wild-type control plate. Synthetic lethal/sick interactions were scored with bnr1Δ, cla4Δ, and bud6Δ. When the query mutation was identical to one of the gene deletions within the array, double mutants could not form because haploids carry a single copy of each allele; therefore, bni1Δ appeared synthetic lethal with itself. (B) Tetrad analysis of meiotic progeny derived from diploid cells heterozygous for bni1Δ and either bnr1Δ, cla4Δ, orbud6Δ. Tetratypes (T) contain one double-mutant spore; nonparental ditypes (NPD) contain two double-mutant spores; and parental ditypes (PD) lack double-mutant spores. The spores were micromanipulated onto distinct positions on the surface of agar medium and then allowed to germinate to form a colony.bni1Δ bnr1Δ and bni1Δcla4Δ double mutants are inviable and therefore fail to form a colony, whereas bni1Δ bud6Δ double mutants showed a synthetic slow growth (sick) phenotype (yellow arrows). The genetic make-up of the double mutants was inferred by replica plating the colonies to medium containing nourseothricin, which selects for growth of bni1Δ::natRcells, and kanamycin, which selects for growth of thebnr1Δ, cla4Δ, and bud6Δ gene-deletion mutants.

We next screened a bni1Δ query strain against an array of 4672 different viable deletion strains. For high-throughput automated screening, we designed a robotic system for manipulation of high-density yeast arrays (16). To ensure reproducibility within a screen and to facilitate visual scoring, we arrayed the deletion strains in pairs, at 768 strains per plate. We scored 67 potential synthetic lethal/sick interactions, 51 (76%) of which were confirmed by tetrad analysis. At the current stage of development, SGA analysis yields a substantial number of false-positives; however, these can be reduced by repeated screening or removed through confirmation of the interactions by tetrad analysis, which is straightforward because the tetrads can be obtained directly from the array of sporulated diploids (17). To group the identified genes by function, we assembled a list of their cellular roles as defined by the Yeast Proteome Database (YPD) (18). The BNI1 interactions were highly enriched for genes with roles in cell polarity (20%), cell wall maintenance (18%), and mitosis (16%). Pathways critical for the fitness ofbni1Δ cells were revealed by multiple interactions with subsets of genes involved in bud emergence (BEM1,BEM2, and BEM4), chitin synthase III activity (CHS3, SKT5, CHS5,CHS7, and BNI4), mitogen-activated protein (MAP) kinase pathway signaling (BCK1 andSLT2), the cell cycle–dependent transition from apical to isotropic bud growth (CLA4, ELM1,GIN4, and NAP1), and the dynein/dynactin spindle orientation pathway (DYN1, DYN2, PAC1,PAC11, ARP1, JNM1, NIP100). We identified 8 of the 11 previously known bni1Δ synthetic lethal/sick interactions (BNR1, SLT2,BCK1, DYN1, ARP1, PAC1,NIP100, ASE1) (18). Of the three that were not identified, CDC12 and PKC1 are not contained in our deletion set, whereas cells lacking HOF1grow very slowly and are apparently beyond the sensitivity of the assay (17). In total, we discovered 43 previously unknown synthetic genetic interactions for bni1Δ, including 9 genes of unclassified function.

Because genetic interactions can be represented as binary gene-gene relationships, multiple SGA screens should generate a network of genetic interactions that depicts the functional relationships between genes and pathways. The SGA synthetic lethal data set was first imported into the Biomolecular Interaction Network Database (BIND) (19), then formatted with BIND tools (16) and exported to the Pajek package (20), a program originally designed for the graphical analysis of social interactions. The network shown in Fig. 3 contains the interactions observed for BNI1 and those for seven other query genes,BBC1 (MTI1), ARC40, ARP2,BIM1, NBP2, SGS1, andRAD27, as described below. The network contains 204 genes, represented as nodes on the graph, and 291 genetic interactions, represented as edges connecting the genes. To visualize subsets of functionally related genes, we color-coded the genes according to their YPD cellular roles and aligned them with one another on the basis of their roles and connectivity (16). The function of the genes with unknown cellular roles (colored black) is predicted by the roles of surrounding genes that show a similar connectivity.

Figure 3

Genetic interaction network representing the synthetic lethal/sick interactions determined by SGA analysis. Genes are represented as nodes, and interactions are represented as edges that connect the nodes; 291 interactions and 204 genes are shown. All of the interactions were confirmed by tetrad analysis, with 8 to 14 tetrads examined in each case. The genes are colored according to their YPD cellular roles (18). For genes assigned multiple cellular roles, we chose one that we considered the most probable on the basis of a review of published abstracts for studies concerning the gene. Synthetic lethal relationships can also be represented by two-dimensional hierarchical clustering, as is used for analysis of DNA microarray experiments (16).

If these interactions identify functionally related genes, then some of the uncharacterized genes from the bni1Δ screen should also participate in cortical actin assembly or spindle orientation. To test this, we conducted an SGA screen using a strain deleted for a previously uncharacterized gene, BBC1, which leads to a synthetic sick phenotype in combination withbni1Δ. We scored 17 potential synthetic lethal/sick interactions for bbc1Δ, most of which have YPD-classified cell polarity or cell structure (cytoskeletal) roles (Fig. 3). In particular, bbc1Δ showed interactions with several genes whose products control actin polymerization and localize to cortical actin patches (CAP1, CAP2, SAC6, and SLA1), suggesting that BBC1 may be involved in assembly of actin patches or their dependent processes. Further experiments demonstrated that Bbc1 localized predominantly to cortical actin patches and binds to Las17 (Bee1), a member of the WASp (Wiskott-Aldrich Syndrome protein) family proteins that controls the assembly of cortical actin patches through regulation of the Arp2/3 actin nucleation complex (21, 22).

We next focused on ARC40 and ARP2, both of which encode subunits of the Arp2/3 complex (23), a major regulator of actin nucleation, the rate-limiting step for actin polymerization. Because ARC40 is an essential gene, we first isolated a temperature-sensitive conditional lethal allele,arc40-40, by polymerase chain reaction (PCR) mutagenesis and then conducted the screen at a temperature that was semipermissive for growth. We scored 40 synthetic lethal/sick interactions forarc40-40 (Fig. 3), most of which have cell polarity or cell wall maintenance roles, including two genes involved in Arp2/3 activation (VRP1 and MYO5). From a similar screen with a temperature-sensitive ARP2 allele,arp2-33, we scored a total of 44 synthetic lethal/sick interactions. The ARP2 screen identified 31 interactions that were shared with ARC40 (Fig. 3). Both screens identified a relatively small number of unique interactions, which included genes implicated in actin patch assembly (MYO5 andARC18 with arc40-40 but not arp2-33;PRK1 with arp2-33 but not arc40-40), and may reflect roles specific to each Arp2/3 subunit (23, 24).

Bim1 determines spindle orientation through physical interactions with cytoplasmic microtubules and with Kar9, which marks the microtubule capture site at the bud cortex (25). With a BIM1gene deletion as the query, we scored a total of 56 genetic interactions by SGA synthetic lethal analysis (Fig. 3). Many of theBIM1 interactions occurred with mitotic genes, including multiple genes involved in nuclear migration and spindle orientation during mitosis (BIK1, SLK19, KIP3,PAC11, PAC1, NUM1, DYN1,JNM1, ARP1, ASE1), and with several genes encoding components of the Bub2- and Mad2-dependent spindle-assembly checkpoints (BUB1, BUB2,BUB3, BFA1, MAD1, MAD2, andMAD3). BIM1 also interacted with a group of genes assigned a chromatin/chromosome structure cellular role, many of which have been implicated in kinetochore function (CTF8,CTF19, MCM21, MCM22,CHL4), and with ARP6, a gene encoding a nuclear-localized actin-related protein of unknown function. The interactions with spindle-assembly checkpoint and kinetochore genes suggests that Bim1 may participate in the attachment of microtubules to the kinetochore, a possibility supported by a two-hybrid interaction with the kinetochore component Duo1 (18). A total of 16 genes with unclassified cellular roles interacted with BIM1. To examine one of the uncharacterized genes in more detail, we conducted an SGA analysis of a deletion of NBP2, which revealed interactions with several genes involved in nuclear migration and spindle function (KAR9, CIN2,KIP3), actin assembly (CAP1, CAP2), and the de novo folding of actin and tubulin (PAC10,GIM5) (Fig. 3), suggesting a general role for Nbp2 in cytoskeletal organization.

To explore the potential for SGA methodology to identify interactions for genes with roles distinct from cytoskeletal organization and cell polarity, we undertook screens with two nonessential genes, SGS1 and RAD27.SGS1 encodes the yeast homolog of the human Werner's syndrome protein, WRN, a member of the RecQ family of DNA helicases (9), whereas RAD27 encodes an enzyme that processes Okazaki fragments during DNA synthesis and repair (26). SGA analysis identified 24 synthetic interactions for SGS1 and 36 synthetic interactions forRAD27, most of which involve genes with cellular roles in DNA synthesis and repair (Fig. 3).

One of the caveats of the SGA method is the absence of essential genes from the genomewide arrays. This drawback could be solved by construction of a complete set of marked conditional mutations in essential genes. It is straightforward to extend SGA analysis to multiple gene interactions, by beginning with a multiple mutant starting strain; for example, by beginning with a double-mutant query strain, arrays of triple mutants can be selected with the SGA protocol. Arrays of yeast cells each carrying an open reading frame on a high-copy plasmid or under high-level conditional expression will allow comprehensive dosage suppression and synthetic dosage lethality screens (27), which could be performed with yeast genes or heterologous genes. To facilitate conventional forward genetic approaches, the marked gene deletions can be used to systematically map both recessive and dominant mutations, and potentially multigenic traits, that lead to a gain-of-function phenotype (28). In theory, SGA methodology should also allow us to backcross the entire set of deletion mutations into another genetic background to analyze strain-specific traits (28). The high-density arrays in SGA screens should also permit efficient detection of chemical-genetic interactions on medium containing a compound of interest (29). The unique oligonucleotide “bar code” tags built into each deletion strain may be exploited for quantitative analysis of growth phenotypes by hybridization of pooled SGA colonies to DNA microarrays of bar code sequences (30). Each of these variations can be tailored to high-throughput analysis of specific pathways by use of mini-arrays of defined composition, as derived from initial genomewide surveys.

By extrapolation of the results presented here, we estimate that on the order of 300 SGA screens covering judiciously selected query genes will provide an effective working genetic scaffold, which should reveal many of the molecular mechanisms behind genetic robustness and buffering (2, 3). Because gene function is often highly conserved, a comprehensive functional genetic map of S. cerevisiae will provide a template to understand the relationships among analogous pathways in metazoans. With the advent of systematic genetic-perturbation methodologies, such as large-scale RNA interference analysis of gene function in Caenorhabditis elegans (31), the SGA approach is in principle applicable to metazoan systems.

  • * To whom correspondence should be addressed. E-mail: brenda.andrews{at}, tyers{at}, and charlie.boone{at}


Stay Connected to Science

Navigate This Article