Plasticity of genetic interactions in metabolic networks of yeast

Harrison et al. 10.1073/pnas.0607153104.

Supporting Information

Files in this Data Supplement:

SI Table 2
SI Materials and Methods
SI Table 3
SI Table 4
SI Table 5
SI Figure 5
SI Table 6
SI Table 7
SI Figure 6
SI Table 8
SI Table 9

Fig. 5. Synthetic genetic interactions between gene triplets. Three-gene interactions were revealed, in all three cases, by severe growth defects of triple-gene-deletant strains compared with the corresponding double- and single-gene deletants. In addition to the double-mutant strains listed here, we note that double deletants cho2D/cpt1D, opi3D/pct1D, and opi3D/cpt1D also exhibit growth phenotypes indistinguishable from the corresponding single-gene deletants (see also Table 1).

Fig. 6. Interdependence between the extent of compensatory mechanisms and environmental specificity. Systematic mapping of genetic interactions by the SGA method (1) provides information on the number of SSL interactions in which a given gene participates. Such data were available for 147 genes (2), including those for which no interaction was detected. Single-gene-deletion strains that exhibit a pronounced growth defect (<80% of wild-type growth rate) on SD medium (3) were excluded (see also Materials and Methods in the main text), leaving 132 genes for the analysis. Altogether, we were able to find evidence for an environment-specific fitness effect for 30 of these genes. Those with environment-specific phenotypes had almost twice as many reported SSL interactions as those without evidence of environmental specificity (Mann-Whitney U test, P = 0.002).

Tong AH, Evangelista M, Parsons AB, Xu H, Bader GD, Page N, Robinson M, Raghibizadeh S, Hogue CW, Bussey H et al. (2001) Science 294:2364-2368.
Tong AH, Lesage G, Bader GD, Ding H, Xu H, Xin X, Young J, Berriz GF, Brost RL, Chang M, et al. (2004) Science 303:808-813.

3. Warringer J, Ericson E, Fernandez L, Nerman O, Blomberg A (2003) Proc Natl Acad Sci USA 100:15724-15729.

SI Materials and Methods

Computational Procedures. Definition of simulated environmental conditions. To find synthetic lethal interactions in the S. cerevisiae metabolic network under different growth conditions, we defined a large set of environmental conditions following a protocol described in ref. 1. All external metabolites were tested in silico for their ability to support aerobic growth in minimal medium: 50 external metabolites supported aerobic growth as the sole carbon source. We also defined a medium mimicking YPD, a medium where all possible external metabolites were allowed to enter the cell, and a minimal vitamin medium [lacking pantothenate, because yeast is capable of de novo pantothenate biosynthesis (2)] . This produced a total of 53 nutrient conditions.

List of 50 carbon sources: acetate, acetaldehyde, adenosine, 2-oxoglutarate, L-alanine, L-arginine, L-asparagine, L-aspartate, allantoin, allantoate, myristic acid, palmitate, stearate, citrate, cytidine, ethanol, formaldehyde, formate, D-fructose, fumarate, 4-aminobutanoate, glycerol, D-galactose, a-D-glucose, L-glutamine, L-glutamate, glycine, guanosine, inosine, (R)-lactate, D-lactaldehyde, (S)-lactate, malate, a-D-mannose, melibiose, maltose, L-ornithine, adenosine 3',5'-bisphosphate, L-proline, pyruvate, D-ribose, S-adenosyl-L-methionine, L-serine, sucrose, succinate, L-threonine, a,a-trehalose, uridine, xanthosine, and D-xylulose.

In addition to a principal carbon source, all simulated minimal media (except for the one lacking pantothenate) included the following metabolites to mimic the in vivo growth conditions: biotin, H⁺, L-histidine, L-leucine, L-lysine, L-methionine, NH₄⁺, O₂, inorganic phosphate, (R)-pantothenate, riboflavin, SO₄^2-, thiamine, and uracil.

Simulated composition of YPD (based on refs. 3-5): L-alanine, L-arginine, L-asparagine, L-aspartate, biotin, myristic acid, palmitate, stearate, choline, L-cysteine, deoxycytidine, thymidine, ergosterol, a-D-glucose, L-glutamate, glycine, guanine, H⁺, L-histidine, L-leucine, L-lysine, L-methionine, NH₄⁺, O₂, L-1-pyrroline-3-hydroxy-5-carboxylate, inorganic phosphate, (R)-pantothenate, L-proline, putrescine, riboflavin, L-serine, SO₄^2-, spermine, spermidine, thiamine, L-threonine, L-tryptophan, L-tyrosine, and uracil

A detailed list of exchange flux constraints for each simulated environment can be found in SI Table 8.

Validation of predicted synthetic lethal interactions. To validate our computationally predicted SL interactions and their environment dependence, we examined experimental data for the list of SL pairs generated under two well studied growth conditions. Our simulations identified 59 gene pairs that show SL on either nutrient-rich (YPD) or glucose minimal (SD) medium, with 24 gene pairs predicted to interact under both conditions, resulting in a list of 83 cases of synthetic lethal interactions. First, we cross-checked the viability of the single deletants included in our list with published large-scale data on single-gene-deletion phenotypes (6, 7). In some cases, we found that one of the genes of the pair was essential in the predicted environment, indicating that the prediction was false (13 cases). Second, we created double-mutant strains to verify synthetic genetic interactions in a subset of cases where viability of single deletants was correctly predicted (another 17 cases, see Materials and Methods of the main text for details of the selection criteria). Finally, we augmented our data set of validated interactions with literature information on double-gene-deletion strains (19 cases). Altogether, we were able to test 49 of 83 predicted interactions, corresponding to 38 unique gene pairs (see SI Table 5).

Twenty four of the 49 predictions proved to be correct, corresponding to a success rate of 49%. A systematic SL mapping strategy previously demonstrated that only 0.56% of randomly chosen gene pairs show interactions (8, 9). Because the few metabolic genes involved in this genome-wide screen had similar interaction densities to other genes (Mann-Whitney U test, P = 0.245), we conclude that the hit rate of the modeling framework outperforms chance expectations by almost two orders of magnitude [P < 10^-287 Yates-corrected c-square test based on frequencies 24 of 49 versus 3,868 of 692,865 (9)].

To assess the fraction of previously described metabolic SL gene pairs that can be recovered by FBA, we extracted a list of published SL interactions from a comprehensive literature-curated data set of protein and genetic interactions (10). We retained only genes present in the metabolic network and recurated the original publications to ensure that only those reporting pair-wise interactions were included in our analysis (i.e., nonviable double mutants). The above procedure resulted in 29 SL interactions, of which 7 are predicted by the model under at least one environmental condition. This 24.14% success rate is >400 times higher than expected by chance (P < 10^-300, Yates-corrected c-square test based on the frequency that the model predicts interaction between randomly chosen genes: 97 of 179,700 versus the frequency that it predicts interaction between true SL pairs: 7 of 29). True interactions missed by the model are listed in SI Table 6.

Phylogenetic cooccurrence of SSL pairs in eukaryotic genomes. We quantified the cooccurrence of gene pairs in genomes in terms of the mutual information between genes, which indicates what extra information we gain about the probability that gene i is present in a genome, from the knowledge that another gene j is also present (11). Occurrence of orthologous groups across 16 sequenced eukaryotic genomes was taken from the STRING database (12). We were able to assign phylogenetic profiles to 1,850 of the 2,666 SSL gene pairs investigated (see above). SSL pairs belonging to the same orthologous group were excluded from the analysis. A two-sided P value was calculated by randomizing gene pairings within the set of 1,850 SSL pairs 10⁵ times. Very similar results were obtained when genes occurring in all 16 species were excluded from the analysis (P = 0.0894, n = 1018 pairs). We performed pair-wise BLASTP (13) similarity searches to find SSL pairs showing significant sequence similarity (at the permissive E value cut-off of 0.01) and identified 70 such pairs (of 1,850).

Phylogenetic cooccurrence of protein-complex subunits was calculated in a similar manner by using a catalog of literature-curated protein complexes as extracted from the MIPS CYGD database (14).

Experimental Procedures. Double mutant construction. We constructed double mutants by crossing haploid yeast strains containing single-gene deletions in BY4742 (MATa his3D 1 leu2D 0 lys2D 0 ura3D 0) and BY4741 (MATa his3D 1 leu2D 0 met15D 0 ura3D 0) backgrounds onto YPD and replica-plated onto selective medium lacking methionine and lysine. The surviving diploids were streaked out for single colonies and incubated overnight in presporulation medium. Cells were harvested, washed, and sporulated on plates at 20°C for 1 week. Tetrads were dissected on YPD (10 tetrads were selected per strain) and incubated at 30°C until colonies were of a reasonable size to determine phenotype. Sporulation and tetrad dissection were carried out according to ref. 15. The genotypes of germinated spores were verified by colony PCR (see below). Colonies were also replica-plated onto SD to assess phenotypes on glucose minimal medium (15).

Triple-mutant construction. Single mutants were transformed with the plasmid pUG-KlURA3 in a one-step gene-deletion strategy as described in ref. 15, and the marker was not excised from the genome. These newly created double mutants were then crossed to create a triple-mutant homozygous at the new deletion locus and heterozygous at the other two loci. Diploids were then sporulated and dissected in an identical manner to the double mutants.

Primers: 40-mers diagnostic for triple mutants.

Forward EPT1, 5'-ATTAGAAGTGTAGAATAAGAAAAACAAGCTAAGGTATAAAcagctgaagcttcgtacgc-3';

reverse EPT1, 5'-ccttgttttcgaataaaaaaaagtagatacaaagtgcgatCATAGGCCACTAGTGGATCTG-3';

forward YGR007w, 5'-AAATGCTTTACAGGATCGGGACTTGAAATATACTGACTGGcagctgaagcttcgtacgc-3';

reverse YGR007w, 5'-atccatttaatttacgttcgaagaagttttcaacatttgtCATAGGCCACTAGTGGATCTG-3'.

List of yeast strains constructed in this study. All yeast strains are derivatives of:

S. cerevisiae BY4742 MATa his3 D leu2 D lys2 D ura3 D; S. cerevisiae BY4741 MATa his3 D leu2 D met15 D ura3 D. See SI Table 9 for the complete list of strains.

Verification by colony PCR. Germinated spores from the dissected plates were streaked out into patches on YPD and colony PCR was performed to verify the visually observed phenotype. Colonies from the germinated spores were grown on YPD medium overnight at 30°C. Small scrapings were placed in 200 ml Eppendorf tubes for colony PCR. These were heated at full power for 1 min 15 s in the microwave, and the PCR mastermix (see below) was added. The tubes were then centrifuged and a colony PCR program (see below) was run on a Techne FTC41H2D machine. When completed, 20 ml of product was run for 45 min at 120 V on a 1% agarose gel. The presence or absence of the kanamycin marker was verified for each gene deleted in each strain with a primer internal to the Kan gene and an external primer homologus to the chromosome sequence flanking the deleted gene. The gel was photographed by using a Bio-Rad Gel dock with 1D PD-Quest software

Reaction mix for colony PCR. 25 ml of 10× PCR buffer with Mg²⁺, 20 ml of (2 mM) dNTP mix, 1 ml of Primer A, 1 ml of Primer B, 0.8 ml of Taq Polymerase, and water to 50 ml.

PCR program for colony PCR. Five minutes at 94°C,

Drop tests.Cultures were grown overnight to an OD of 2-3. Dilutions were made of 1/20, 1/50, and 1/100, and the dilution corresponding an optimum cell density of 60 on a hemocytometer was selected. This was then diluted to 20 ´10⁶ cells per ml, and 5 ml was spotted onto a YPD plate. Serial dilutions of 10⁴, 10³, 10², and 10 cells were made and spotted in an identical manner.

1. Price,ND, Reed JL, Palsson BO (2004) Nat Rev Microbiol 2:886-897.

2. White WH, Gunyuzlu PL, Toyn JH (2001) J Biol Chem 276:10794-10800.

3. Forster J, Famili I. Palsson BO, Nielsen J (2003) Omics 7:193-202.

4. Kaouass M, Audette M, Ramotar D, Verma S, De Montigny D, Gamache I, Torossian K, Poulin R (1997) Mol Cell Biol 17:2994-3004.

5. Kneifel H, Schoberth SM (1981) FEMS Microbiol Lett 11:59-61.

6. Giaever G, Chu AM, Ni L, Connelly C, Riles L, Veronneau S, Dow S, Lucau-Danila A, Anderson K, Andre B, et al.. (2002) Nature 418:387-391.

7. Kuepfer L, Sauer U, Blank LM (2005) Genome Res 15:1421-430.

8. Tong AH, Lesage G, Bader GD, Ding H, Xu H, Xin X, Young J, Berriz GF, Brost RL, Chang,M, et al.. (2004) Science 303:808-13.

9. Wong SL, Zhang LV, Tong AH, Li Z, Goldberg DS, King OD, Lesage G, Vidal M, Andrews B, Bussey H, et al.. (2004) Proc Natl Acad Sci USA 101:15682-15687.

10. Reguly T, Breitkreutz A, Boucher L, Breitkreutz BJ, Hon GC, Myers CL, Parsons A, Friesen H, Oughtred R, Tong A, et al.. (2006) J Biol 5:11.

11. Huynen M, Snel B, Lathe W, III, Bork P (2000) Genome Res 10:1204-1210.

12. von Mering C, Jensen LJ, Snel B, Hooper SD, Krupp M, Foglierini,M, Jouffre N, Huynen MA, Bork P (2005) Nucleic Acids Res 33:D433-D437.

13. Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ (1997) Nucleic Acids Res 25:3389-3402.

14. Güldener U, Münsterkötter M, Kastenmüller G, Strack N, van Helden J, Lemer C, Richelles J, Wodak SJ, García-Martínez J, Pérez-Ortín JE, et al.. (2005) Nucleic Acids Res 33:D364-D368.

15. Amberg,DC, Burke DJ, Strathern JN (2005) Methods in Yeast Genetics: A Cold Spring Harbor Laboratory Course Manual (Cold Spring Harbor Lab Press, Cold Spring Harbor, New York).