Functional Characterization of the S. cerevisiae Genome by Gene Deletion and Parallel Analysis

Bassett D. E., et al., Nature Genet. 15, 339 (1997);

Foury F., Gene 195, 1 (1997).

Smith V., Botstein D., Brown P. O., Proc. Natl. Acad. Sci. U.S.A. 92, 6479 (1995);

Burns N., et al., Genes Dev. 8, 1087 (1994);

Ross-Macdonald P., Sheehan A., Roeder G. S., Snyder M., Proc. Natl. Acad. Sci. U.S.A. 94, 190 (1997).

Smith V., Chou K. N., Lashkari D., Botstein D., Brown P. O., Science 274, 2069 (1996).

Baudin A., Ozier-Kalogeropoulos O., Denouel A., Lacroute F., Cullin C., Nucleic Acids Res. 21, 3329 (1993);

Wach A., Brachat A., Pohlmann R., Philippsen P., Yeast 10, 1793 (1994);

Lorenz M. C., et al., Gene 158, 113 (1995).

Wach A., Brachat A., Alberti-Segui C., Rebischung C., Philippsen P., Yeast 13, 1065 (1997).

Shoemaker D. D., Lashkari D. A., Morris D., Mittmann M., Davis R. W., Nature Genet. 14, 450 (1996).

Hensel M., et al., Science 269, 400 (1995).

To construct deletion strains, two long oligonucleotide primers are synthesized, each containing (3′ to 5′) 18 or 19 bases of homology to the antibiotic resistance cassette, KanMX4 (U1, D1), a unique 20-bp tag sequence, an 18-bp tag priming site (U2 or D2), and 18 bases of sequence complementary to the region upstream or downstream of the yeast ORF being targeted (including the start codon or stop codon; see ). These 74-mers are used to amplify the heterologous KanMX4 module, which contains a constitutive, efficient promoter from a related yeast strain, Ashbya gosspii, fused to the kanamycin resistance gene, nptI (5). Because oligonucleotide synthesis is 3′ to 5′ and the fraction of full-size molecules decreases with increasing length, improved targeting is achieved by performing a second round of PCR using primers bearing 45 bases of homology to the region upstream and downstream of a particular ORF. Transformation with the PCR product results in replacement of the targeted gene upon selection for G418 resistance. The unique 20-mer tag sequences are covalently linked to the sequence that targets them to the yeast genome, creating a permanent association and genetic linkage between a particular deletion strain and the tag sequence. The use of two tags increases confidence in the analyses, and the redundancy is useful in case one of the tags carries a mutation or performs poorly in hybridization assays. To verify correct integration of the deletion cassette, genomic DNA was prepared from the resistant strains and used as template in PCR reactions using two primers common to the KanMX4 module (KanB (5′-CTGCAGCGAGGAGCCGTAAT-3′) and KanC (5′-TGATTTTGATGACGAGCGTAAT-3′) and four ORF-specific primers (A, B, C, and D). A and D are from regions 200 to 400 bases upstream or downstream of the start codon, whereas B and C are from within the ORF (see ). For verification, both the A-KanB and the D-KanC PCR reactions were required to give the correct size product when analyzed by gel electrophoresis. If one of either the A-KanB or D-KanC reactions failed to yield a product, the identification of the correctly-sized AD product could suffice. In addition, haploid deletion strains were tested for the disappearance of the wild-type AB and CD products. All ORFs encoding proteins greater than 100 amino acids in size were initially selected for deletion. The deletion cassettes were designed to remove the entire coding sequence for a given ORF but to leave the start and stop codon intact. Although ∼10% of ORFs in S. cerevisiae overlap one another, the positions of the deletions were not adjusted, nor was any attempt made to avoid essential genes, genes in which a previous deletion had been constructed, or genes with a well-defined function. Genes represented multiple times in the genome (telomeric ORF, Ty-elements) were usually not deleted as their targeted disruption would pose a challenge due to the conservation of upstream and downstream regions. Some smaller nonannotated ORFs (NORFs) will be deleted in the future. Transcripts from many of these NORFs have been detected in SAGE analysis, warranting their inclusion in the study [

Velculescu V. E., et al., Cell 88, 243 (1997);

]. All oligonucleotide primers (5 nmol scale) were synthesized on an automated multiplex oligonucleotide synthesizer [

Lashkari D. A., Hunicke-Smith S., Norgren R. M., Davis R. W., Brennan T., Proc. Natl. Acad. Sci. U.S.A. 92, 7912 (1995);

] in batches of 96 primers. Scripts were written to automate the selection of primers. Primer sequences and ORF locations were chosen from the Stanford Genome Database (http: //genome-www.stanford.edu/Saccharomyces/) at various times over a 2-year period. The KanMX4 cassette was PCR-amplified, and the resulting PCR products were sent to participating laboratories where 1 μg of PCR product was used to transform yeast by means of a variation on the standard lithium acetate procedure (. edu/group/yeast_deletion_project/protocols.html) in a 96-well format. Electronic records, accessible over the World-Wide Web, were kept for every strain constructed. MATahaploid strains were given record numbers of less than 10,000, MATα haploid strains were given record numbers between 10,000 and 20,000, the heterozygous diploid, between 20,000 and 30,000, and the homozygous diploid, greater than 30,000. Each record consists of primer sequence information, the results of the different diploid tests that were performed, and notes about the phenotype. Data for completed strains are accessible from www-sequence.stanford.edu/group/deletion/index. html. These strains, frozen in 15% glycerol, can be obtained from Research Genetics (Huntsville, AL) or EUROSCARF (Frankfurt, Germany).

Four different types of strains, containing several genetic markers (Table 1), were generated for each ORF—two haploid strains, one of each mating type, and two diploid strains, heterozygous and homozygous for the deletion loci and kanamycin marker. The homozygous diploid was constructed by mating the two haploid strains, obtained from independent transformations. Typically, the heterozygous diploid and one of the haploid strains were obtained by direct transformation, while the other haploid strain and the homozygous diploid were obtained by sporulation and mating, respectively. Essential genes were identified by 2:2 segregation of viability in tetrads derived from the heterozygous diploid. Strains were sporulated by patching them on a fresh GNA plate (5% D-glucose, 3% Difco nutrient broth, 1% Difco yeast extract, 2% Difco Bacto agar) for 1 day at 30°C before transfer to liquid sporulation medium (1% potassium acetate and 0.005% zinc acetate supplemented with 0.1 mM uracil, 0.15 mM histidine-HCL, or 1.0 mM leucine as necessary). Sporulation cultures were incubated on a rollerwheel for 4 to 5 days at 25°C. If 2:2 segregation of viability was consistently observed in two independently-transformed heterozygous deletion strains, the gene was designated essential. All confirmed diploid strains obtained through mating were required to pass two of three tests: a diploid budding pattern, the ability to sporulate, and the inability to mate. In a few cases, essential genes overlapped other essential genes (20 pairs) or a gene whose viability status was unknown (four pairs), making it difficult to determine the cause of lethality.

Mewes H. W., Albermann K., Heumann K., Liebl S., Pfeiffer F., Nucleic Acids Res. 25, 28 (1997).

Cho R. J., et al., Mol. Cell 2, 65 (1998).

To construct the pools, each deletion strain was patched on YPD plates in the presence of 150 mg/liter G418. Approximately equal numbers of cells were harvested from the plate for each strain and combined. Aliquots of the pools were stored in the presence of 15% glycerol at–80°C.

The UPTAG and DOWNTAG sequences were separately amplified from genomic DNA by means of PCR using primers B-U2-comp (5′ biotin-GTCGACCTGCAGCGTACG-3′) and U1 (5′-GATGTCCACGAGGTCTCT-3′), or primers B-D2-comp (5′-biotin-CGAGCTCGAATTCATCG-3′) and D1 (5′-CGGTGTCGGTCTCGTAG-3′). In both cases, a twofold molar excess of biotinylated primers was used in the reactions. The amplified UPTAG and DOWNTAG sequences were combined and hybridized to high-density oligonucleotide arrays in 200 μl of 6×SSPE [1 M NaCl, 66 mM NaH₂PO^–4, 6.6 mM EDTA, (pH 7.4)], containing 0.005% Triton X-100 (SSPE-T), 200 pmol U1, 200 pmol U2 (5′-CGTACGCTGCAGGTCGAC-3′), 200 pmol D1, and 200 pmol primer D2 (5′-CGATGAATTCGAGCTCG-3′). The addition of complementary primers to the hybridization mix was shown to improve the signal-to-noise ratio [D. D. Shoemaker, thesis, Stanford University, Stanford, CA (1998)]. Samples were heated to 100°C for two min, and then cooled on ice before being applied to the array. Samples were hybridized for 1 hour at 42°C. The arrays were washed two times with six changes of 6×SSPE-T. The arrays were then stained at 42°C for 10 min with 6×SSPE-T containing 2 μg/ml phycoerythrin-streptavidin (Molecular Probes) and 1 μg/ml acetylated bovine serum albumin, washed two times with five changes 6×SSPE-T, and scanned at an emission wavelength of 560 nm using an Affymetrix GeneChip Scanner. Of the strains analyzed, 157 contained only a single tag sequence (UPTAG). Six of these strains were not detected in the hybridization mix. Of the strains represented with two-tag sequences, 98.5% exhibited either an UPTAG or DOWNTAG bar code hybridization signal that was greater than threefold over background. Sequencing of 186 deletion regions tags showed that 25% of mutations in the tags or tag priming sites resulted in a nonfunctional tag that could not be amplified or detected by hybridization, or both. In only 1.1% of cases was a complete lack of hybridization signal not associated with a mutation in the tag or tag priming site. Mutations were most often found in the tags or tag priming sites (0.85% per base) and were less frequent in the regions of yeast homology (0.25% per base), most likely due to selection against the mutated PCR products during the recombination event or to the two-step PCR strategy.

Grids were aligned to the scanned images using the known feature dimensions of the array. The hybridization intensities for each of the elements in the grid were determined using the 75th percentile method in the Affymetrix GeneChip software package. Subsequent analysis of the hybridization intensities consisted of two steps: adjustment of data to achieve approximate equality of background and maximal signals on each array, and analysis of the decrease (or increase) of the UPTAG and DOWNTAG signal strength over time. Equalization of signal strength relied on the fact that for most array sites, the amount of tag DNA present did not vary over time. A consensus score for these sites was obtained from the first principal component of the logarithms of the signals. The logarithms of the hybridization intensity for each element on the array were linearly transformed to make each array's overall signal approximately equal to the consensus. The growth rate for each strain was determined by using the model log₂(signal) = max(a + b t, 0) + e, where a and b are model parameters, t represents the number of population doublings, and e is a random error term. The growth rate is calculated as 1 + b. For strains that have dropped out of the pool (a + bt < 0), the model describes the statistical distribution of background signals. When e is normally distributed, background signals have a lognormal distribution. This appeared to be roughly true in our data. However, we found that a small fraction of signals on each array are liable to have very high values, much larger than can be accounted for by a purely lognormal background model. To obtain a degree of robustness against these occasional outliers, we assumed that the error term e had a scaled t distribution with one degree of freedom. The use of this heavy-tailed distribution reduced the likelihood of false positive identifications of deficient strains due to occasional high signal levels at t = 0, at the expense of a possible reduction in the ability to detect marginally deficient strains.

The analysis was not as accurate for strains with growth rates of less than 0.5 that of the wild type, because generally only three data points were above background for these.

Du L. L., Collins R. N., Novick P. J., J. Biol. Chem. 273, 3253 (1998).

A comprehensive study of chromosome V genes using genetic footprinting (3) provided an opportunity to validate the data: the results generally agreed, with a few exceptions. For example, of the 52 genes whose disruption had no effect on strain fitness under all conditions tested by genomic footprinting, we detected a growth defect in deletants yel033w (0.68, R; 0.83, M), yel050c (0.73, R; 0.68, M) and yer028c (0.79, R; 0.69, M). The observed phenotype for yel033w probably results from interference with a neighboring gene (HYP2, encoding translation initiation factor eIF-5A). In addition, the hem14 deletant showed a strong growth defect in rich medium, while genomic footprinting revealed a salt-specific defect, but no defect in rich medium. Of the 11 genes that had been shown by genomic footprinting to have a severe growth defect, in three cases our deletants appeared to have no discernible phenotype (nrf1, gda1, pcl6). These differences could be due to our use of diploid versus haploid strains, to the auxotrophies carried by our deletants, or to our using 30°C versus 25°C as the growth temperature. The effect of temperature is probably the cause of discrepancies for the nrf1 strain, which grows slowly at 25°C but grows faster than normal at 36.5°C. Among other disparities, two (YER082C and YEL026W) of the 22 genes on chromosome V determined to be essential by tetrad analysis were wild-type by footprinting (perhaps because of cross-feeding in the Ty pool or because the products are required for germination) and the minimal medium–specific growth defects we detected for hom3 and ilv1 mutants were not detected by genomic footprinting.

Cultures of BY4743 ho::KanMX4 were grown at 30°C in YPD or minimal media supplemented with histidine, uracil, lysine, methionine, and adenine, in the presence of G418, and were harvested at mid-log phase. cDNA was prepared from 20 μg of polyadenylated RNA from each sample, using a dT₂₁ primer and Superscript II reverse transcriptase (GibcoBRL), according to the manufacturer's recommendation. cDNA was fragmented using DNaseI (GibcoBRL), biotinylated using ddATP (NEN) and Terminal Transferase (Boehringer), and hybridized to yeast full-genome arrays (Affymetrix) as described in L. Wodicka et al. [Nature Biotechnol.15, 1359 (1997)]. After scanning, the average signal from each array was normalized to the average signal strength of all eight chips.

Baganz F., et al., Yeast 14, 1417 (1998).

Contributing groups include all authors, G. Valle, S. Kelley, J. Strathern, and D. Garfinkel.

EUROFAN projects B0 and B9 at www.mips.biochem.mpg.de/proj/eurofan/index.html; R. Niedenthal et al., Yeast, in press.

Giaever G., et al., Nature Genet. 21, 278 (1999).

Wolfe K. H., Shields D. C., Nature 387, 708 (1997).

Bickle M., Delley P. A., Schmidt A., Hall M. N., EMBO J. 17, 2235 (1998);

Corominas J., et al., FEBS Lett. 310, 182 (1992);

Santos M. A., Garcia-Ramirez J. J., Revuelta J. L., J. Biol. Chem. 270, 437 (1995).

Chow T. Y., Ash J. J., Dignard D., Thomas D. Y., J. Cell Sci. 101, 709 (1992);

Rad M. R., et al., Yeast 13, 281 (1997);

Brown J. D., et al., EMBO J. 13, 4390 (1994);

Machado A. K., Morgan B. A., Merrill G. F., J. Biol. Chem. 272, 17045 (1997) .

Yamamoto A., Guacci V., Koshland D., J. Cell Biol. 133, 85 (1996);

Leidich S. D., Orlean P., J. Biol. Chem. 271, 27829 (1996).

Storms R. K., et al., Genome 40, 151 (1997);

Elledge S. J., Davis R. W., Genes Dev. 4, 740 (1990);

Kasahara S., et al., J. Bacteriol. 176, 1488 (1994).

Fodor S. P. A., et al., Science 251, 767 (1991);

Pease A. C., et al., Proc. Natl. Acad. Sci. U.S.A. 91, 5022 (1994).

Brachmann C. B., et al., Yeast 14, 115 (1998).

We thank D. Lashkari for establishing the oligonucleotide synthesis facility, T. Nguyen, M. Sigrist, and K. Tanner for help in tetrad analysis, S. Voegeli for DNA sequence analyses, P. Koetter for distribution of deletion strains, J. Rine for helpful advice, and M.Cherry and K. Wolfe for files. E.A.W. is supported by the John Wasmuth fellowship in Genomic Analysis (HG00185-02). Supported by NIH grants HG01633, HG01627, HG00198, by an operating grant from the Medical Research Council of Canada, by grants from the European Commission (BIO4-CT97-2294), by the Swiss Federal Office for Education and Science, and by the region de Bruxelles-Capital, Belgium.