Victoria Smith, Karen N. Chou, Deval Lashkari, David Botstein, Patrick O. Brown ^*

Genetic footprinting was used to assess the phenotypic effects of Ty1 transposon insertions in 268 predicted genes of chromosome V of Saccharomyces cerevisiae. When seven selection protocols were used, Ty1 insertions in more than half the genes tested (157 of 268) were found to result in a detectable reduction in fitness. Results could not be obtained for fewer than 3 percent of the genes tested (7 of 268). Previously known mutant phenotypes were confirmed, and, for about 30 percent of the genes, new mutant phenotypes were identified.

V. Smith and K. N. Chou, Department of Biochemistry, Stanford University School of Medicine, Stanford, CA 94305, USA.
D. Lashkari and D. Botstein, Department of Genetics, Stanford University School of Medicine, Stanford, CA 94305, USA.
P. O. Brown, Howard Hughes Medical Institute and Department of Biochemistry, Stanford University School of Medicine, Stanford, CA 94305, USA.
^*   To whom correspondence should be addressed. E-mail: pbrown{at}cmgm.stanford.edu

We subjected a large population of haploid yeast cells (10¹¹ cells) to mutagenesis by transiently inducing transposition of a marked Ty1 transposable element. DNA was extracted from a portion of this culture (the "time-zero" DNA). Other representative samples of this population were subjected to one of several selections (Table 1). DNA was extracted from the cells recovered after each selection. The presence and relative abundance of cells carrying Ty1 insertions within a gene of interest was assessed for each of these samples by means of a polymerase chain reaction (PCR) (2). In general, for each gene, we surveyed a minimum of 500 to 900 base pairs (bp) of coding sequence, along with 400 to 600 bp of upstream sequence. Smaller genes (<700 bp) were analyzed in their entirety, along with several hundred base pairs of sequence flanking the start and stop codons. A growth disadvantage to cells carrying insertions in a gene, under a particular selection, was reflected by the loss or depletion of the PCR products representing those insertions (the "genetic footprint") when DNA samples from the selected population were compared with the time-zero DNA samples. The method not only detects severe growth disadvantages, but also sensitively measures moderate reductions in fitness.

Table 1. Selections used in genetic footprinting analysis. The time points at which DNA samples were isolated correspond to the given numbers of population doublings in each selection after Ty1 mutagenesis. Each gene was analyzed using DNA (1 µg) isolated at a primary time point for each selection; the pattern of PCR products was compared with that obtained with the time-zero DNA sample. Secondary time points were used to confirm potential growth defects or to resolve ambiguities identified by the primary analysis. Analysis of both primary and secondary time points was useful for confirmation of general quantitative growth defects as well as for identification of growth defects in specific selections. As a control to identify any nonspecific PCR products, each gene-specific primer was also used with the Ty1-specific primer for a PCR that used DNA isolated from cells in which Ty1 transposition had not been induced. Ty1 transposition mutagenesis was performed as described (2). At least 4 × 108 cells were transferred to the appropriate medium for each selection, and cell density was maintained at 1 × 105 to 3 × 107 cells per milliliter (or less) over the course of each selection. In some cases, cell density was allowed to reach 5 × 107 to 1 × 108 cells per milliliter for harvesting at the final time point. All selections were performed at 25°C, except high-temperature growth (36.5°C) and mating (30°C). The media for all selections contained 2% glucose, except rich-lactate medium (2% lactate). The pH of the media was 5.5. Standard rich medium, with supplements for auxotrophies of the host strain, was used for the rich-medium, lactate, high-temperature, caffeine, and high-salt selections (1% yeast extract, 2% bactopeptone, 0.008% tryptophan, 0.0026% adenine, 0.0022% uracil, and 0.0046% histidine). The medium for the caffeine selection contained, in addition, 6 mM caffeine, and the high-salt selection medium included 0.9 M NaCl. The minimal medium used for selection lacked amino acids and nucleotides, except as required by the auxotrophy of the strain (0.67% yeast nitrogen base, 0.0022% uracil, and 0.0046% histidine). For the mating selection, mutagenized cells (his3 HIS4) were grown in rich-glucose medium for 4 hours, then mixed with a threefold excess of cells of opposite mating type (HIS3 his4), collected on a filter, and incubated on rich-glucose medium at 30°C for 7 hours. Cells were then washed into SC-histidine liquid medium and incubated at 30°C for 16 hours (10 population doublings) to select the HIS+ diploid products of successful mating. To compensate for the presence of the extra haploid genome, we used 2 µg of DNA from the selected diploids for each PCR. Selection Primary time point (population doublings) Secondary time points (population doublings) Rich medium 18 5, 12, 15, 23 Rich medium 51 56, 60 Minimal medium 18 10, 15 Rich-lactate medium 18 11 Caffeine 18 12 High temperature 18 10 High salt 10 Mating 15

For each predicted protein-coding sequence on chromosome V, a color was assigned on the basis of whether a particular selection protocol resulted in a perceptible depletion of the PCR products representing insertions in that coding sequence (Fig. 1). Overall, we were able to obtain satisfactory genetic footprinting data for 261 (97%) of the predicted protein-coding genes (4, 5). This total includes six putative genes contained in repeated telomeric sequences (boxed in Fig. 1): cells with mutations in any of these six genes appeared wild type for growth in all selections. However, because a unique priming site was found for only one of these genes, they are excluded from the following discussion. Of the remaining 255 genes, we could detect a phenotype that distinguished the corresponding mutants from wild-type cells for 157 genes (61.6%). For 98 genes (38.4%), Ty1 insertions resulted in a reduction in growth of less than 5 to 10%, or a decrease in mating proficiency (as an ) of less than 50%; that is, the mutants were indistinguishable from the wild type for growth or mating, within the sensitivity of our assays.

YEL Series: 3w: small gene, low number of insertions detected, appeared WT. 7w: WT, possibly Q3. 8w: Q3/Q2 specifically in S; this assessment was based on depletion of ~6 detected insertions in this small gene, but quantitative depletion in high salt was reproducible for several independent PCRs. 9c: Q2/Q3 specifically in M. 10w: WT, good target for Ty1 transposition. 12w: WT, good target for Ty1 transposition. 13w: Q3, possibly WT for growth in rich-glucose medium at 25^oC, with Q3 specifically in M, L, H, and C. 15w: WT, X data were weak. 16c: Q3, possibly more severe Q in X, NR for S. 17w: WT based on result obtained with two independent primers. However, an intense cluster of peaks was detected in the time-zero DNA by one primer, which allowed an extra 110 bp of 3´ coding sequence to be examined. This cluster was not present in the selected DNAs. This could represent selective depletion of cells carrying mutations in just this portion of the gene, or in an alternative (unrecognized) gene, or it could be an artifact of the PCR. 18w: WT, possibly Q3 in M only. 18ca: NR, insertions were not reproducibly detected within this small coding sequence (159 bp). 22w: WT, X data were weak. 24w: NR, insertions were detected upstream of the ATG, but insertions in the coding sequence were not detected reproducibly. 25c: WT, X data were weak. 26w: NR for X. 27w: Q2, very strong upstream site preference for Ty1 insertion. The Q2 classification is based on a small number of insertions detected in the coding sequence, but the apparent depletion could reflect sampling variability. Thus, possibly Q1 or even WT. 29c: Q2, possibly Q1. 30w: WT, possibly Q3 in H only. 35c: Q2, possibly Q1. 36c: Q1 (possibly Q2) specifically in M, H, and S. Q3 in all other selections, ~95 to 98%. 37c: Q1, possibly Q2. 38w: Q3, ~95 to 98%. 40w: X, C, and rich-medium 51-generation data were slightly weak but probably WT. 43w: Q3, ~95 to 98%. 44w: There was a conflict between the results obtained with two independent gene-specific primers: data obtained with one primer indicated a Q1 growth defect in M only; results with the second primer indicated a Q1 (possibly Q2) growth defect in all selections. Both primed uniquely at YEL044w. 46c: Q1 specifically in M, Q2 (possibly Q1) in all other selections. 49w: WT, time-zero data were weak. The signal throughout the coding sequence was of low intensity: a very strong upstream site preference for Ty1 insertion could cause PCR artifacts. 50c: Q2, possibly Q1 in M and C, specifically. Time-zero data were weak. 51w: Q1, but insertions in the 3´ ~150 bp of coding sequence gave a WT phenotype (tolerated insertions). 52w: Q3, possibly WT. 53c: Q1, possibly Q2. 57c: Q2; however, three or four peaks corresponding to insertions in the coding sequence remained after 51 population doublings. Thus, possibly Q1, with artifact peaks or tolerated insertions. 59w: Q3, possibly Q2. 61c: Q2, possibly Q1. 62w: Q1, possibly Q2, ~70 to 80%. 63c: WT, possibly Q3 or Q2 in rich-media selections. 64c: WT, time-zero data were weak. 65w: NR for H-18 but WT for H-11. 66w: Q2 (possibly WT) specifically in L, with some tolerated 5´ insertions. NR for X, NR for H-18 but WT for H-11. 68c: WT, X data were weak. 71w: WT, X data were weak. 72w: WT, time-zero data were weak, NR for X. 73c: In telomeric repeat. The primer primes at a unique sequence 398 bp upstream of the start ATG. WT, but NR for X. 74w: in telomeric repeat. One primer had a unique nucleotide at the 3´-most position, but did not detect many insertions in the coding sequence. Another primer yielded WT results for all selections, but could prime at homologous telomeric sites in other chromosomes. 75c: In telomeric repeat. The primer could prime at homologous telomeric sites in other chromosomes. NR for X. 76c: In telomeric repeat. The primer could prime at homologous telomeric sites in other chromosomes.

CEN5: No Ty1 insertions were detected in the 154-bp region of conserved centromere sequences. Insertions were detected immediately adjacent to this region, and these insertions did not result in any apparent growth defect.

YER Series: 5w: Q2 interpretation reflects the average of the data, which varied by >10%. 6w: Q1 specifically in M, Q3 in other selections. Possibly Q1/Q2 in H. 7w: WT, possibly Q3. 7ca: Q3, possibly Q2. 12w: Q1, signal corresponding to insertions in the coding sequence was very weak but was detected in the time-zero DNA sample; some insertions were also detected in the rich-medium 5-generation population doubling sample. 13w: Q1, the signal throughout coding sequence was low but was reproducibly detected in the time-zero DNA sample. There was potential for generation of PCR artifacts, due to the strong signal from insertions upstream of the coding sequence. 14w: Q1 specifically in S in one high-salt selection, but Q2 or possibly even WT in a second high-salt selection. Possibly Q3 in H. 15w: S data were weak, but probably WT. 17c: There was a conflict between the results obtained with two independent gene-specific primers. Data obtained using one primer indicated Q1 specifically in L, and Q2 in all other selections except possible Q1 in H; data obtained using the other primer indicated Q3 or WT in all selections. Both primers primed uniquely at this gene. A third primer was uninformative (it generated Ty-independent background). The data obtained with the first primer are consistent with the published data for AFG3 [E. Guelin, M. Rep, L. A. Grivell, Yeast 10, 1389 (1994); M.-F. Paul and A. Tzagoloff, FEBS Lett. 373, 66 (1995)]. 18c: Q1, but several peaks corresponding to insertions in the coding sequence were still observed in the analysis of the rich-medium 51-generation population doubling sample. 19w: WT, but only a few insertions were detected in the coding sequence. 20w: Q1 specifically in C; Q2 in all other selections. Possibly Q1 in S. NR for X. 26c: Q1, but WT specifically in S. There was a strong 5´ site preference for Ty1 insertions, but insertions in the coding sequence were reproducibly detected with two independent primers in time-zero DNA. Both primers confirmed that mutants were resistant to high salt. This phenomenon was also confirmed in two independent high-salt selections: in one selection, the mutants appeared to be growing at the population rate (PCR gave a pattern similar to that observed with the time-zero sample); in the other, the mutants appeared to have a growth advantage (a richer pattern of insertions was detected after selection). 27c: WT, the time-zero PCR gave a richer pattern of PCR products through the coding sequence: analysis of all selected DNAs gave a consistent pattern that supports WT, but the pattern of peaks was less rich. This is consistent with some selection against gal83 mutants during mutagenesis; gal83 mutants would be expected to have a growth disadvantage during the mutagenesis (as galactose is the carbon source) but would be expected to be able to grow as wild-type on the carbon sources used in subsequent selections. 28c: X data were variable, but probably WT. 30w: Q3, possibly Q2. 31c: Q1, possibly Q2, ~70 to 80%. 32w: Q3, possibly Q2. 35w: WT, possibly Q3. 38c: Q3, possibly WT. Time-zero data were weak, NR for X. 45c: WT, S data were weak, NR for C. 47c: Q3, possibly Q2 with tolerated insertions in ~200 bp of coding sequence adjacent to the start ATG. Possibly Q2/Q1 in S. 49w: Q1, possibly Q2. Strong 5´ site preference for Ty1 insertion, but insertions in the coding sequence were reproducibly detected in the time-zero DNA. 50c: WT, X data were weak. 51w: Q3, possibly Q2, NR for X. 52c: Q3, Q2 supported by H, C and S but with weaker signal overall. NR for X. 54c: Q2, possibly Q3. 55c: Q3, possibly Q2. As the yeast strain used in this study is a his3 mutant, histidine was added to all media. 56ca: Q2, the predicted gene product has similarity to ribosomal protein L34e. 59w: Q1, possibly Q2. 60w: Q2, cells not frozen following mutagenesis may be at less of a growth disadvantage, although a growth disadvantage was still apparent. 61c: Q1 specifically in L, Q2 in other selections. 62c: Q3, possibly WT, ~95 to 100%. 63w: Q3, possibly WT, ~95 to 100%. 64c: Q2, possibly Q1. 69w: Q1 specifically in M, Q3 in other selections, ~95 to 97%. 70w: Q1, but insertions appear to be tolerated in the 100 to 200 bp of coding sequence adjacent to the start ATG. 72w: Q1, but Q2 specifically in H and Q3/Q2 specifically in X. 73w: Q3, ~95 to 99%. 74w: Q1, despite the presence of a close homolog on chromosome IX. The primer used primes uniquely at YER074w. 75c: Q3, ~95 to 98%, very strong 5´ site preference for Ty1 insertion. 79w: WT, possibly Q3. 87w: Q1 specifically in L, Q2 in other selections. Possibly Q1 in all selections, but mutants were nonetheless at a greater apparent growth disadvantage in lactate medium. The predicted gene product has similarity to E. coli prolyl tRNA synthetase. 88c: Q3, possibly WT, ~95 to 100%. 89c: Q2, possibly Q1. 90w and 91c: Q1 specifically in M, probably Q3 in other selections, ~95 to 100%. 92w: WT, possibly Q3, rich-51 data were weak. 93c: Q1 specifically in C, Q3 in other selections. 93ca: Q2, but WT specifically in H. This gene is described as being spliced: an alternative possibility is a sequencing or cloning error that introduces a single stop codon at position 178 of an otherwise continuous open reading frame. 95w: Q2, possibly Q1, ~70 to 80%. 97w: WT, unusually good target for Ty1 transposition. 98w: WT, strong 5´ site preferences for Ty1 insertion. 101c: WT, possibly Q3, ~97 to 100%. 102w: Q1, possibly Q2. 107c: Q1 specifically in X, Q3 (possibly Q2) in other selections. The predicted gene product has similarity to BUB3 protein and other GTP-binding regulatory proteins. 108c: Q3, possibly WT. Cells not frozen and thawed following mutagenesis appeared to be at less of a growth disadvantage, possibly WT. 111c: NR, peaks representing Ty1 insertions were not reproducibly detected in the coding sequence. 112w: Q1, possibly Q2 (results using one primer supported Q1, results with a second primer were consistent with Q2). 113c: Q3, depletion was more severe (Q1/Q2) for mutants with insertions in a ~400-bp region adjacent to the primer, in all selected DNAs, relative to the time zero DNA (indicated by a small red box within the box representing this gene). Less depletion was apparent in the next ~300 bp. Analysis with a second primer that primed at a site 133 bp further from the start codon produced weaker data, but confirmed the absence of tolerated insertions in the 3´ portion of the gene after selection. 114c: Q3, possibly Q2. 116c: Q2, possibly Q1. 123w: NR, peaks representing Ty1 insertions were not reproducibly detected in coding sequence. 126c: Q1, numerous peaks corresponding to insertions were detected in the time-zero DNA, but the region of depletion was extensive for all selected DNAs, such that peaks were not always detectable upstream of the coding sequence. Analysis using the Sau3A library DNA as an internal standard confirmed this result was not due to PCR failure. This extensive depletion may reflect the proximity of this gene to an adjacent Q1 gene, YER127w. 128w: WT, possibly Q3. 129w: Q1 specifically in C, Q2 in other selections. 131w: Q2, possibly Q1, Q2 represents the average of the data for all selections. The growth disadvantage of mutant cells was possibly less severe in lactate medium (Q3/WT). For most selections, insertions well upstream of the designated start codon, ~1000 bp, showed depletion, although perhaps to a lesser extent (Q2/Q3), than those in the predicted coding region. As noted by MIPS, YER131w has a potential upstream intron in the 5´ untranslated region. This result may this reflect a growth disadvantage of mutants bearing insertions in the upstream intron, assuming the gene is spliced. In addition, there is a potential (unannotated) gene 544 bp upstream, of length 213 bp, which could account for the observed phenotype of upstream insertions. A third possibility is that important promoter elements might extend unusually far upstream of the coding sequence. This gene has close homolog on chromosome VII, but the primer used here primes uniquely to YER131w. 132c: Q1, but WT/Q3 specifically in X. Strong signal corresponding to Ty1 insertions was detected in the first ~280 bp of coding sequence adjacent to the ATG start codon in all selected DNAs (tolerated insertions). Signal was weak in the time-zero PCR, but peaks corresponding to insertions in the coding sequence were detected across the entire 719-bp region of coding sequence analyzed. A similar signal (eight or nine independent peaks) was detected in the coding sequence on analysis of DNA from the diploid products of successful mating. These data suggest that mutations in this gene (beyond the first 280 bp) confer a Q1 growth disadvantage but allow mating. This gene has strong 5´ site preferences for Ty1 insertion, creating some potential for PCR artifact. R133w: Q1, but the strong 5´ preference for Ty1 insertion could have caused PCR artifacts. Using independent primers, the data obtained for all selected DNAs indicated that insertions in exon1, the intron, and the first ~150 bp of exon 2 were tolerated. PCR products corresponding to insertions in the 3´ ~550 bp of coding sequence in exon 2 were detected with one of the primers in the time-zero DNA sample, but were undetectable in DNA from any of the selected cell populations. In all selected DNAs, insertions were not detected in this region of exon 2. 134c and 135c: WT, very good targets for Ty1 transposition. 136w: NR, very strong 5´ preference for Ty1 insertion, artifact peaks were generated by standard PCR analysis (revealed by 23 cycle and forward primer analysis). Insertions were not reproducibly detected in the coding sequence. However, peaks corresponding to insertions were detected in the 5´ 200 to 300 bp of the coding sequence, adjacent to the stop codon, in the time-zero and selected DNAs. 137c: WT, very good target for Ty1 transposition. 139c: Q1 (possibly Q2) specifically in H, Q3 in other selections, ~ 95 - 98%. 141w: Q1 specifically in L, Q3 in other selections, ~ 95-99%. 142c: WT, possibly Q3 (data from one primer support WT, the other suggest possible Q3). 143w: Q3, ~95 to 100%. 144c: WT, possibly Q3, ~97 to 100%. 145c: Q1, possibly Q2. Moderately strong 5´ site preference for Ty1 insertion. 146w: Q3 specifically in M, ~85 to 90%, possibly Q2 in M with insertions tolerated near the stop codon. The predicted gene product has similarity to small nuclear ribonuclear protein SNP2. 147c: Q2, possibly Q1. 149c: WT, strong 5´ site preference for Ty1 insertion created potential for PCR artifacts. A relatively low number of insertions was detected in the coding sequence. 150w: WT, in a region of strong preference for Ty1 insertion. 151c: Q2 (the average of data from two independent primers), possibly more severe Q2 or Q1 specifically in H. 152c: Q3, a greater than expected number of peaks remained at 18 and 51 population doublings. An alternative interpretation of these data is that mutant cells were at a disadvantage during mutagenesis, but grew at the same rate as the general population in subsequent selections. 153c: Q1 specifically in L, Q3 in other selections, ~85 to 90%. 154w: Q1 specifically in L, possibly Q3 in other selections. 155c: BEM2, Q2, possibly Q1, ~70 to 80%. Probably Q1 in H. 156c: WT, but the very strong 5´ preference for Ty1 insertion created the potential for PCR artifacts. Possible Q3 in all selections, or possible Q3 in C only. 157w: WT, strong 5´ preference for Ty1 insertion. 159c: WT, good target for Ty1 transposition. 161c: WT, X data were variable but probably WT. 163c: Q3, a greater than expected number of peaks remained at 18 and 51 population doublings. An alternative interpretation of these data is that mutant cells were at a disadvantage during mutagenesis, but grew at the same rate as the general population in subsequent selections. 167w: Q2, possibly Q3. 169w: WT, possibly Q3, or possible selection against mutants during mutagenesis followed by growth at the population rate in subsequent selections. 172c: NR, insertions were not reproducibly detected in the ~800 bp of coding sequence analyzed. However, PCR products corresponding to insertions were detected by an independent primer scanning a region ~2100 to 2800 bp from the start ATG. These insertions were detected in the selected DNAs, suggesting that insertions in this portion of the gene may not result in a growth defect, but the signal was low. 174c: Q3, possibly Q2, ~85 to 88%. 175c: Q3, ~95 to 98%. 176w: Q2, possibly Q3, ~80 to 90%. 180c: WT, very strong 5´ preference for Ty1 insertion. 182w: NR, very strong 5´ preference for Ty1 insertion created PCR artifacts. The results from PCR performed using only 23 cycles to minimize artifacts indicated a possible general quantitative growth defect for mutant cells. 181c: WT, very good target for Ty1 transposition. 184c: WT, possibly Q3. 186c: Q3, ~95 to 98%, the mutant growth defect was possibly more severe in C. 189w and 190w: in telomeric repeat. The primers could prime at homologous telomeric sites in other chromosomes.

[View Larger Version of this Image (36K GIF file)]

The most common mutant phenotype we found among all the genes analyzed was a general growth disadvantage for mutant cells under all selections. By analyzing multiple time points using the least stringent selection for growth--growth in rich-glucose medium at 25°C (Table 1)--we could estimate the growth rates of each mutant. We divided the results into three categories: Q1, Q2, and Q3. Q1 indicates that mutant cells grew at no more than 75% of the growth rate of the overall population (for example, after 15 population doublings the mutants had doubled no more than 12 times, and thus were represented at no more than 1/2³ = 1/8 of their abundance in the time-zero population). This category includes genes whose product is absolutely required for vegetative cell growth ("essential" genes). The Q2 and Q3 categories include genes for which mutant cells were at more subtle growth disadvantages, growing at apparent rates of 75 to 85% and 85 to <100% of the population growth rate, respectively (Fig. 1 and Table 2). None of the phenotypes in these classes was a consequence of the freezing of cells in glycerol after mutagenesis, or of their subsequent resuscitation (6). These data were confirmed by analysis of DNA isolated from three independent rich-medium selections. In addition, the data obtained from the other selections corroborated the general growth defects observed in rich-glucose medium. For >90% of the genes in the Q2 and Q3 categories, growth deficits were reproducibly detected to within ±5% variability in independent samples.

Table 2. Numbers of genes giving mutant phenotypes in each selection. The percentage of genes associated with each phenotype was calculated with respect to the 255 nontelomeric genes on chromosome V for which reliable data were obtained. Some genes were associated with more than one phenotype. For 98 genes (38.4%), Ty1 insertion mutants were indistinguishable from wild-type cells in any of the selections we used. R refers to genes for which mutants with general severe growth defects (Q1 or Q2) were able to grow at notably improved rates under particular selections. Many of the phenotypes we detected for previously characterized genes are, in fact, novel phenotypes (not listed). These new phenotypes were typically Q2 or Q3 (sometimes Q1) growth disadvantages for mutants of genes previously described as wild type for vegetative growth. Selection Number of genes (%) Number of novel genes Rich: Q1 (<75%) 51  (20.0) 14 Rich: Q2 (75 to 85%) 44  (17.3) 22 Rich: Q3 (85 to 100%) 55  (21.6) 35 Minimal 9  (3.5) 3 Lactate 8  (3.1) 1 Caffeine 3  (1.2) 1 High temperature 2  (0.8) 1 High temperature R 1  (0.4) 1 High salt 3  (1.2) 2 High salt R 1  (0.4) 0 Mating 1  (0.4) 1 Mating R 2  (0.8) 2

Fifty-one genes (20%) fell into the Q1 category; 14 of these were previously undescribed (Table 2 and Figs. 1 and 2). In addition to several known essential genes, this category included ILV1 and PRO3, mutations of which are known to result in auxotrophy. As described previously for ade2 insertions (2), genetic footprinting revealed considerable growth defects for ilv1 and pro3 mutant cells even in rich medium thought to contain adequate quantities of the relevant nutrients [the inability of pro3 mutants to grow in rich medium was reported previously (7)]. MOT2, which encodes a transcriptional repressor, also fell in this class; although null mutants were viable at 25°C, their growth defect was clearly exposed by genetic footprinting analysis (8). Other genes previously reported to be nonessential, such as VMA8 (encoding the vacuolar H-adenosine triphosphatase subunit), GDA1 (encoding guanosine diphosphatase of the Golgi membrane), and MMS21 (encoding a DNA repair protein) (9), were found to fall in the Q1 category, which indicated a substantial growth disadvantage of the corresponding mutant strains.

Some selection against cells with mutations in essential genes may occur during the 4-day Ty1 mutagenesis. It is possible that this factor accounted for some cases in which we failed to obtain an interpretable result (3% of the genes analyzed) (10). Most essential genes were not excluded from analysis in this way, as we were reliably able to obtain PCR products representing Ty1 insertions in almost all of the previously identified essential genes on chromosome V with the use of DNA from the time-zero cells.

Insertion mutations in 99 genes (38.8%) resulted in more subtle quantitative growth defects in all selections. These genes were divided into two classes (Q2 and Q3, Fig. 1 and Table 2) on the basis of the estimated growth rate disadvantage of mutants. Most of the genes in these groups (58%) had not been previously characterized. Moreover, for many of the previously characterized genes in these categories, the growth disadvantage we observed had not previously been reported. An example of such a gene is SSA4, for which we found a Q3 mutant growth defect. This gene is a member of a large family of genes that encode apparently interchangeable cytoplasmic heat shock proteins (11). We also observed a Q2 growth disadvantage for mutants of RAD51 (encoding a RecA-like DNA repair protein), PRB1 (encoding vacuolar protease B), and GLN3 (encoding a positive nitrogen-regulatory protein) (12).

Spurious PCR products (those unrelated to real Ty1 insertions in the gene of interest) would not be expected to be depleted under a selection that requires the gene's activity, leading to underestimation of the selective disadvantage of the mutants. A similar behavior would be expected for PCR products that represent actual insertions within a gene that do not impair its function. PCR products corresponding to tolerated Ty1 insertions were frequently observed immediately adjacent to the 5 and 3 boundaries of coding sequences and were even observed within coding sequences (13). Because we analyzed multiple time points, we were usually able to determine whether these anomalous products were consistently present and omitted them from the quantitative analysis. The 50- to 60-generation time points were particularly useful for this purpose (14).

When a gene is important for cell growth, the dilution of the corresponding mutant cells as the general cell population expands leaves progressively fewer targets for PCR amplification. Reduced competition for PCR reagents could then lead to more efficient amplification of these remaining DNA targets. For example, PCR products representing insertions upstream of an important coding region were often more abundant among the products amplified with the use of DNA samples from the corresponding selection than among the products of the time-zero PCR. We tried to take this effect into account by normalizing the signal produced in different PCRs relative to insertions upstream of the coding regions, under the assumption that these upstream insertions had no effect on fitness. To provide a more robust normalization, we developed an independent internal standard for the PCR reactions. A library of Sau 3A restriction fragments of yeast genomic DNA was cloned into a vector carrying the marked Ty1 primer sequence, such that PCR amplification of this library with the Ty1-specific primer and any labeled gene-specific primer resulted in a predictable pattern of products (15). DNA from the library was mixed at a fixed concentration with DNA samples from selected cells and time-zero cells, respectively. The mixtures were subjected to a second set of PCR reactions for each of 63 genes that were quantitatively important for growth (16). The intensities of the PCR products from the different selected DNAs were then normalized using the library-specific peaks (Fig. 3). The growth rate defect of the mutant cells was estimated by comparing the normalized signals from the different time points. These data facilitated the grouping of genes into the Q classifications and were also useful in resolving ambiguous results concerning general or specific growth defects.

Ty1 insertion mutations in a few genes resulted in growth defects that were discernible only in one of the selection protocols. Mutations in several other genes resulted in general growth defects as well as more severe specific ones. Genes required for growth in minimal medium include four previously characterized genes involved in amino acid biosynthesis [GLY1, TRP2, MET6, and ARG5,6 (17)] and two novel genes, YER006w and YEL044w (18). (Novel genes are genes that have not been previously characterized and have not had a putative function assigned on the basis of compelling homology to a characterized gene.) Mutations in GLY1, ARG5,6, MET6, TRP2, and YER006w also produced more subtle growth defects in rich medium (Q2 for GLY1, Q3 for the others). GCN4, which encodes a transcriptional activator of amino acid biosynthetic genes, displayed a Q2 phenotype only in minimal medium (19). Mutations in YER146w produced a Q3 growth defect only in minimal medium. Mutations in ANP1, which encodes a protein involved in retention of glycosyltransferases in the Golgi apparatus (20), produced numerous phenotypes: Q1 (possibly Q2) in minimal medium, in high-salt medium, and at high temperature, and Q3 in all other selections.

Genes known to be required for respiration (PET117, PET122, CEM1, OXA1, and AFG3) behaved as expected, showing at least a Q1 defect in rich-lactate medium (21, 22). YER141w was also important for growth on lactate, consistent with its identification as COX15 (23). Mutations in most of these genes resulted in less severe general growth defects in all other selections as well. One novel gene, YER087w, was important for growth in lactate medium (Q1 in lactate) and was quantitatively important for growth in rich medium (Q2). Mutants of YEL066w had a moderate growth defect (Q2) only in lactate medium.

Mutants of three genes were found to have growth defects (Q1) in medium containing 6 mM caffeine. Two of these, PAK1 and GPA2 (24), encode a protein kinase and a guanosine triphosphate-binding regulatory protein, respectively; the third, YER093c, was a novel gene (Fig. 4). Mutations in all three of these genes also resulted in a less severe general growth disadvantage (Q2 or Q3) (Fig. 4). One novel gene, YER139c, was identified as important for growth at high temperature. Two novel genes, YEL008w and YER014w, were found to be important for growth in high-salt medium.

There were some examples of unexpected survival in particular selections. CHO1, which encodes a phosphatidylserine synthetase, is important for phosphatidylcholine synthesis (25). By genetic footprinting analysis, cho1 mutants were at a severe growth disadvantage in rich medium (Q1) and all other selections, except rich medium containing high salt (0.9 M NaCl). Cells containing insertions in the CHO1 coding sequence were abundantly represented after selection in the high-salt medium, indicating that the cho1 mutants were not only viable, but grew at least as well as the general population of cells in this medium. This unexpected result has no obvious explanation (26), but it highlights the value of applying genetic footprinting analysis to all genes, even the apparently well-characterized ones. Similarly, mutants in the novel genes YER093ca (Q2) and YER072w (Q1) showed general growth defects at 25°C but grew at improved rates at 36.5°C (Fig. 1).

The ability to mate was also used as a selection protocol. Selection for diploids was done after Ty1-mutagenized MAT cells were mated with excess MATa partners. One novel gene on chromosome V, YER107c, was required for mating. Mutations in this gene also resulted in a Q3 growth defect in all selections. Conservative interpretation of mating data was necessary in cases of mutants with more severe general growth defects (Q1 and Q2), as these mutants were more susceptible to inadvertent selection during the mating procedure (27). PCR products representing detrimental insertions in mutants with the most severe growth defects (Q1) were frequently (but not always) absent upon analysis of the DNA from the selected diploid products of a successful mating. There were two notable exceptions. Peaks corresponding to insertions in the analyzed coding sequence of YER132c were barely detectable after PCR analysis of the time-zero DNA and were depleted in all other selections. However, these peaks were detected upon PCR analysis of DNA isolated from the diploid products of mating, which suggested that these severely growth-impaired mutants were nonetheless able to mate. Similarly, peaks corresponding to insertions in the coding sequence of YER072w (Q1) were detected upon PCR analysis of the successful maters. Mutants with less severe growth defects (Q2 and Q3) were generally able to mate (28).

Genetic footprinting has allowed us to determine that an unexpectedly large fraction of the genes of S. cerevisiae make detectable contributions to fitness under standard laboratory conditions. This finding may be attributable in part to the essentially quantitative nature of the data obtained with this method. Unlike standard gene disruption methods, genetic footprinting provides an estimate of the fitness of mutants relative to the population as a whole. The estimates obtained for the fraction of genes that are "essential" is governed by this feature: the Q1 category includes not only essential genes but genes for which mutants have severe growth defects but might still manage a visible colony. The number observed here (20%) is indeed higher than that reported in other studies estimating the proportion of "essential" genes (2). Mutations in another 39% of the nontelomeric genes on chromosome V resulted in general growth defects. These results suggest that despite the apparent redundancy in the yeast genome, more than half of all yeast genes contribute detectably to competitive fitness. Even redundant genes can be important for vegetative growth. For example, on chromosome V, YER074w (RP50A) and YER131w (RPS26b), which encode predicted ribosomal proteins, have nearly perfect homologs on other chromosomes. Mutants of YER074w and YER131w were nonetheless at substantial growth disadvantages (Q1 and Q2) in rich medium (29).

We anticipate that simply extending the methods described above to the whole genome will enable us to assign mutant phenotypes to more than half of the genes. In our work to date, the more specialized selections yielded fewer mutant phenotypes, but the number obtained is of consequence nonetheless. Mutants of 11 novel genes were found to have specific growth defects or growth advantages in specialized selections. Extrapolating this result to the rest of the genome predicts that these specialized selections alone would identify specific mutant phenotypes for another 250 to 300 novel genes, in addition to any new discoveries for previously characterized genes. Incorporation of additional selection protocols should increase the fraction of genes for which a specific role can be inferred. However, even when mutations in a gene produced no discernible phenotype under any of the selections we used, the result was still informative: we learned that the gene under study was dispensable for growth in each of the media tested, to a resolution of 5 to 10% of the population growth rate (30).

Genetic footprinting provides an efficient and economical way to take advantage of the genomic sequences of microorganisms so as to learn about the functions of the newly identified genes. Thanks to the high degree of functional conservation among eukaryotic proteins, we can expect that the information gathered in this way will also, in many cases, provide important clues to the functions of the cognate human genes.

REFERENCES AND NOTES

1. M. G. Goebl and T. D. Petes, Cell 46, 983 (1986) [Medline]; S. G. Oliver et al., Nature 357, 38 (1992) [Medline].
2. V. Smith, D. Botstein, P. O. Brown, Proc. Natl. Acad. Sci. U.S.A. 92, 6479 (1995) [Medline].
3. F. Dietrich et al., Nature, in press. In subsequent releases, additional telomeric sequences were added, making the full length of chromosome V 574,860 bp [updated in the Martinsried Institute for Protein Sequences (MIPS) Yeast Genome Database]. The number of predicted genes is derived from Dietrich et al. and from the MIPS database. Most of the possible additional and alternative open reading frames noted in the MIPS database were also covered by the primers used in this study. One to three independent gene-specific primers were used to analyze each gene. Primers were designed with the use of the program PRIMER (Whitehead Institute for Biomedical Research, Cambridge, MA) with a specified melting temperature of 69° to 73°C. Primers were synthesized by means of a 96-well array synthesizer [ D. A. Lashkari, S. P. Hunicke-Smith, R. Norgren, R. W. Davis, T. Brennan, Proc. Natl. Acad. Sci. U.S.A. 92, 7912 (1995) [Medline]] and labeled at their 5 end with 5-carboxyfluorescein (Applied Biosystems or Pharmacia). Approximately 85% of primers produced usable data. Alternative primers were synthesized for any gene for which the first primer failed to produce usable data. Each labeled gene-specific primer was used in a PCR with an unlabeled Ty1-specific primer, as described (2).
4. Although genetic footprinting analysis of chromosome V had a high success rate, there were some failures. For seven genes (RIP1, PMP2, SWI4, CLK3, GDI1, BRR2, and YER182w), PCR reactions that used two different gene-specific primers failed to produce enough amplified products to allow meaningful data interpretation. In some cases, products corresponding to insertions upstream of the start ATG codon were readily detected, but products representing insertions in the coding sequence were not reproducibly detected. This may reflect a low frequency of Ty1 insertion in the coding sequences of these genes. It is also possible that Ty1 insertions in essential genes may not be detected, even in the time-zero sample, because of selection during the mutagenesis (10). However, at least some of the genes for which no insertions could be detected are known from previous work not to be essential for vegetative growth, so the latter hypothetical explanation cannot account for all failures resulting from insufficient signal. In almost all cases, when two independent primers for any particular gene both produced Ty1-dependent signal upon PCR analysis, a very similar distribution of PCR products was obtained. In three cases, however, two independent primers for the same gene produced apparently credible data yet gave different results by genetic footprinting analysis (YEL044w, AFG3, and RSP5; Fig. 1). On the basis of the relative frequency of these cases of discordant data produced by two gene-specific primers relative to the frequency of concordant result, we estimate that each analysis using a single primer has a 2 to 3% chance of producing unreliable data.
5. Caution was necessary in interpreting data for genes that had very strong local preferences for Ty1 insertion upstream of the start ATG codon. On chromosome V, these genes were generally located in the immediate vicinity of tRNA genes. Preferred upstream sites for Ty1 transposition frequently generated a cluster of PCR products with 100 to 1000 times the signal typically observed. This raises the possibility of exhaustion of PCR reagents at a stage earlier than 30 cycles, because of the increased number of initial template molecules. It is thus possible that some of the less abundant, smaller DNA products observed in these cases, which would ordinarily be assumed to represent insertions in the coding sequence of the gene, could be artifactual. In cases of this kind, PCR was repeated with only 23 cycles in an attempt to minimize this possible artifact. In addition, new primers that did not encompass the preferred site were also used; typically, these primers were located in the first 100 bp of the gene and were directed downstream, toward the stop codon. These "reverse" primers typically allowed a more reliable survey of products corresponding to insertions in coding sequences of genes with very strong upstream site preferences for Ty1 insertion.
6. Typically, after mutagenesis, aliquots of 2 × 10⁸ cells were stored in 25% glycerol at 80°C. After one mutagenesis, cells were immediately transferred to rich medium for growth, without intervening storage. The DNA isolated from these cells was analyzed with primers specific to every gene for which a quantitative growth defect was detected. No cases were found in which storage in glycerol, freezing and resuscitation, or both accounted for the apparent growth defect, although possible contributions of these procedures to some phenotypes are noted in Fig. 1.
7. M. C. Brandriss, J. Bacteriol. 138, 816 (1979) [Medline].
8. K. Irie, K. Yamaguchi, K. Kawase, K. Matsumoto, Mol. Cell. Biol. 14, 3150 (1994) [Medline].
9. L. A. Graham, K. J. Hill, T. H. Stevens, J. Biol. Chem. 270, 15037 (1995) [Medline]; C. Abeijon et al., J. Cell Biol. 122, 307 (1993) [Medline]; S. Prakash and L. Prakash, Genetics 87, 229 (1977) [Medline].
10. Selection against the mutants during the mutagenesis procedure itself may account for our failure to detect insertions in most of the coding sequence of GDI1, an essential gene [ M. D. Garrett, J. E. Zahner, C. M. Cheney, P. L. Novick, EMBO J. 13, 1718 (1994) [Medline]]. Factors that may affect the degree to which mutants are lost during the period of mutagenesis, and thus our ability to detect insertions in an essential gene, include the integrity of the mutant cell (rapid cell lysis would limit recovery of the cell's DNA) and the stability and turnover of residual RNA and protein produced before Ty1 insertion.
11. M. Werner-Washburne, D. E. Stone, E. A. Craig, Mol. Cell. Biol. 7, 2568 (1987) [Medline].
12. A. Shinohara, H. Ogawa, T. Ogawa, Cell 69, 457 (1992) [Medline]; C. M. Moehle, M. W. Aynardi, M. R. Kolodny, F. J. Park, E. W. Jones, Genetics 115, 255 (1987) [Medline]; A. P. Mitchell and B. Magasanik, Mol. Cell. Biol. 4, 2758 (1984) [Medline].
13. These observations presumably reflect the fact that the Ty1 element could supply promoter or terminator functions and could even provide an ATG start codon [ J. D. Boeke, D. J. Garfinkel, C. A. Styles, G. R. Fink, Cell 40, 491 (1985) [Medline]].
14. For example, for a gene in which mutations reduce the growth rate to 90% of the population rate, 29% of the signal corresponding to insertions affecting gene function would remain after 18 population doublings, but only 3% would remain after 50 population doublings.
15. There are 1628 Sau 3A sites in the 569,202-bp sequence of chromosome V investigated in this analysis. Thus, a Sau 3A site occurs, on average, every 350 bp. In this analysis, for 600 to 900 bp of coding sequence, two or three peaks resulting from Sau 3A sites are expected, assuming a random distribution.
16. We analyzed 63 genes for which quantitative mutant growth deficits were detected in rich medium (42%) independently by library DNA normalization and normalization to upstream peaks; the remainder were analyzed only by normalization to upstream peaks. Variations in intensities of library peaks were observed. In some cases, library peaks could not be identified because of the density of Ty1-specific peaks. The library DNA was most useful as a normalization standard for genes that contained Sau 3A sites within the region undergoing depletion. Most of the genes analyzed (94%) satisfied this criterion.
17. J. B. McNeil et al., J. Biol. Chem. 269, 9155 (1994) [Medline]; G. Nass and K. Poralla, Mol. Gen. Genet. 147, 39 (1976) [Medline]; M. Masselot and H. de Robichon-Szulmajster, ibid. 139, 121 (1975) [Medline]; C. Boonchird, F. Messenguy, E. Dubois, ibid. 226, 154 (1991) [Medline].
18. YEL044w may be important for growth under all selections (Q1). The two primers used to analyze this gene gave conflicting data.
19. A. P. Mitchell and B. Magasanik, Mol. Cell. Biol. 4, 2767 (1984) [Medline].
20. R. E. Chapman and S. Munro, EMBO J. 13, 4896 (1994) [Medline].
21. J. E. McEwen, K. H. Hong, S. Park, G. T. Preciado, Curr. Genet. 23, 9 (1993) [Medline]; J. D. Ohmen, B. Kloeckener-Gruissem, J. E. McEwen, Nucleic Acids Res. 16, 10783 (1988) [Medline]; A. Harrington, C. J. Herbert, B. Tung, G. S. Getz, P. P. Slonimski, Mol. Microbiol. 9, 545 (1993) ; N. Bonnefoy, F. Chalvet, P. Hamel, P. P. Slonimski, G. Dujardin, J. Mol. Biol. 239, 201 (1994) [Medline].
22. E. Guelin, M. Rep, L. A. Grivell, Yeast 10, 1389 (1994) [Medline].
23. GenBank accession numbers YSCCOX15A and L38643.
24. S. Thiagalingam, K. W. Kinzler, B. Vogelstein, Proc. Natl. Acad. Sci. U.S.A. 92, 6062 (1995) [Medline]; M. Nakafuku et al., ibid. 85, 1374 (1988) [Medline].
25. K. D. Atkinson et al., J. Bacteriol. 141, 558 (1980) [Medline].
26. Perhaps an alternative synthesis or salvage pathway is activated when cells are exposed to 0.9 M NaCl, possibly as part of a general process in which cell wall structure is altered to enhance viability in the high-osmolarity medium.
27. Approximately four to five population doublings elapsed during the mixing and plating of cells to allow mating before selection for diploid (mated) cells.
28. Depending on the severity of the growth defect, the number of peaks corresponding to detrimental insertions detected in the PCR analysis of DNA from successful maters typically ranged between the number observed in the time-zero analysis and the number observed at 18 population doublings.
29. Explanations for this type of phenomenon have been reported previously. For example, RNR1 and RNR3 are closely related in sequence, and both encode large subunits of ribonucleotide reductase. However, RNR1 is constitutively expressed and is essential for vegetative growth, whereas RNR3 is induced in response to exposure to DNA-damaging agents [ S. J. Elledge and R. W. Davis, Genes Dev. 4, 4740 (1990) ].
30. After 60 population doublings, mutants with a 5% growth rate disadvantage will have doubled only 57 times, and thus will be underrepresented by a factor of 2³ = 8. Thus, by examining cells grown for 60 population doublings in rich-glucose medium, we could readily detect a 5% deficit in growth rate in this medium. Because selections for growth in other media involved fewer population doublings (typically 15 to 18), we may only have been able to recognize growth deficits of 10% in those selections (15).
31. Y-J. Kim, L. Francisco, G.-C. Chen, E. Marcotte, C. S. M. Chan, J. Cell Biol. 127, 1381 (1994) [Medline].
32. We thank R. Norgren for assistance with data analysis and the generation of Fig. 1, and L. McAllister and K. Davis for discussions. Supported by NIH grant 1PO1 HG00205-05 (to R. Davis, P.O.B., and D.B.) and by the Howard Hughes Medical Institute (P.O.B.). Supported in part by grant LT-141/93 from the Human Frontier Science Program Organization to V.S. P.O.B. is an assistant investigator of the Howard Hughes Medical Institute.

31 July 1996; accepted 7 November 1996