Supporting InformationDenby et al. 10.1073/pnas.1116360109SI Materials and MethodsRepressor of HypOXia (ROX1) KO strains were generated byreplacing the endogenous ROX1 gene and the upstream 467 bpwith a URAcil requiring (URA)3 (1) or kanamycin resistancecassette (2). To generate constructs for transcriptional reporters,a portion of the ROX1 promoter was amplified from yeast ge-nomic DNA and fused to CFP (Addgene) flanked by a kanamy-cin resistance cassette (2) by ligation-independent cloning (LIC)(3). To generate constructs for protein fusion reporters with WTcoding sequences, the ROX1 promoter and coding sequencewere amplified from yeast genomic DNA and fused to GFP(Addgene) flanked by a kanamycin resistance cassette (2) byLIC. To generate constructs for protein fusion reporters withsuboptimized codons, we first generated each coding sequenceseparately; sequences are provided in Fig. S6. A suboptimizedROX1 sequence was synthesized and incorporated into a cloningvector (GeneOracle). We also generated by PCR a fusion ofcommercial oligonucleotides (IDT) corresponding to sub-optimized GFP regions with overlapping ends and ends con-taining LIC sites; the fused product was introduced into an LICvector. For reporter fusions bearing WT ROX1 and sub-optimized GFP, we then used primers containing LIC sites toamplify two regions with overlapping ends, one containing theWT ROX1 promoter and ROX1 coding sequence and the othercontaining suboptimized GFP; we generated a fusion of theseregions by PCR and introduced the resulting construct into anLIC vector. For reporter fusions bearing suboptimized ROX1and suboptimized GFP, we used primers containing LIC sites to

amplify three regions with overlapping ends: one containing theWT ROX1 promoter, a second containing the ROX1 codingsequence, and a third containing suboptimized GFP. We fusedthese regions into one combined fragment and cloned as above.Yeast strains bearing each reporter were ultimately generated bytransforming ROX1::URA3 strains with PCR-amplified con-struct products. Integrants were selected by growing on YeastPeptone Dextrose media supplemented with 300 μg/mL G418(Cellgro).To generate feedback mutant strains, Rox1 binding sites in the

ROX1 promoter were identified as perfect matches to the Rox1consensus sequence (4). Site-directed mutagenesis of these siteson cloned promoter constructs was performed with the Quik-Change II XL Site-Directed Mutagenesis Kit (Stratagene).Sanger sequencing confirmed DNA sequences of the resultingconstructs, and ROX1::URA3 strains were transformed withPCR products and selected for integration as above.ROX1::CFP strains were generated in the S288c background.

The background of ROX1-GFP strains was as follows. The W303derivative JRY2334 (5) was mated to a HEM1Δ DY150 de-rivative (6) [gifts from J. Rine (University of California, Berke-ley, CA) and J. Kaplan (University of Utah, Salt Lake City, UT),respectively], and a haploid recombinant WT for HEM1 wasisolated from the progeny. This recombinant was then crossed toBY4742 (Open Biosystems), and a haploid recombinant bearingthe W303 allele of HAP1 rather than the S288C allele (7) wasisolated.

1. Goldstein AL, Pan X, McCusker JH (1999) Heterologous URA3MX cassettes for genereplacement in Saccharomyces cerevisiae. Yeast 15:507–511.

2. Sheff MA, Thorn KS (2004) Optimized cassettes for fluorescent protein tagging inSaccharomyces cerevisiae. Yeast 21:661–670.

3. Haun RS, Serventi IM, Moss J (1992) Rapid, reliable ligation-independent cloning of PCRproducts using modified plasmid vectors. Biotechniques 13:515–518.

4. Badis G, et al. (2008) A library of yeast transcription factor motifs reveals a widespreadfunction for Rsc3 in targeting nucleosome exclusion at promoters.Mol Cell 32:878–887.

5. Thomas BJ, Rothstein R (1989) Elevated recombination rates in transcriptionally activeDNA. Cell 56:619–630.

6. Crisp RJ, et al. (2003) Inhibition of heme biosynthesis prevents transcription of ironuptake genes in yeast. J Biol Chem 278:45499–45506.

7. Gaisne M, Bécam AM, Verdière J, Herbert CJ (1999) A ‘natural’ mutation inSaccharomyces cerevisiae strains derived from S288c affects the complex regulatorygene HAP1 (CYP1). Curr Genet 36:195–200.

Fig. S1. Predicted binding sites for the Rox1 protein in the ROX1 promoter. Shown is the yeast genome sequence between the ROX1 start codon (bold) and600 bp upstream; predicted Rox1 binding sites are in yellow.

Denby et al. www.pnas.org/cgi/content/short/1116360109 1 of 5

Fig. S2. Suboptimized ROX1 expression reporters. Each bar reports expression from an ROX1-GFP fusion reporter in a haploid yeast strain. Each bar labeledwith so corresponds to a reporter construct with the indicated sequence component encoded with suboptimized codons; bars labeled with 4s correspond toreporters where the four Rox1 binding sites in the ROX1 promoter are mutagenized as in Fig. 2. 4s-soROX1-soGFP corresponds to the suboptimized feedbackmutant in Figs. 3 and 4, and 4s-ROX1-GFP corresponds to the feedback mutant with WT ROX1 and GFP sequences in Figs. 3 and S4.

Fig. S3. Perturbing ROX1 expression affects yeast growth. Each panel reports a comparison of growth across yeast strains in one environmental medium. Eachrow represents a haploid strain with an ROX1-GFP fusion bearing the indicated modifications diluted to varying densities; naming is as in Fig. S2. Δ, whole-genedeletion.

Fig. S4. Eliminating feedback in anotherwiseWTRox1-GFP reporter compromisesmutational robustness.Data are as in Fig. 4, except that no feedback indicates anROX1-GFP reporter construct inwhich thepromoters in all four Rox1 binding sites aremutagenized and coding sequences bearWT codons (Figs. 2 and3 andFig. S2).

Denby et al. www.pnas.org/cgi/content/short/1116360109 2 of 5

Fig. S5. Complete dataset from the screen of regulatory feedback across transcription factors. All symbols and values are as in Fig. 1.

Fig. S6. Suboptimized ROX1-GFP sequence. Lowercase letters represent codons that have been modified from the WT sequence to a codon corresponding toa lower codon index value (1). Uppercase letters represent unchanged codons. Uppercase italicized letters represent optimal non-WT codons. Blue-shadedcodons were modified from suboptimal codons to prevent synthesis errors caused by repeat sequence. The yellow-shaded sequence represents the linkerbetween ROX1 and GFP incorporated to improve expression level (2). The green-shaded sequence represents the GFP coding sequence. Bold text represents thestop codon.

1. Jansen R, Bussemaker HJ, Gerstein M (2003) Revisiting the codon adaptation index from a whole-genome perspective: Analyzing the relationship between gene expression and codonoccurrence in yeast using a variety of models. Nucleic Acids Res 31:2242–2251.

2. Sheff MA, Thorn KS (2004) Optimized cassettes for fluorescent protein tagging in Saccharomyces cerevisiae. Yeast 21:661–670.

Denby et al. www.pnas.org/cgi/content/short/1116360109 3 of 5

Table S1. Abundances of yeast transcription factors

Gene Abundance

SWI4 35,186.53ZAP1 26,578.11HIR1 26,386.28GLN3 26,353.92STP2 27,919.37YAP3 26,555.68AZF1 27,732.27RPH1 26,549.68GZF3 26,801.2TEC1 26,060.8OPI1 28,371.46YML081W 26,105.32HAL9 27,385.85ARG81 26,149.43BAS1 26,061.96ASH1 26,845.78MOT3 31,713.58PHD1 25,435.17YFL044C 26,348.15PUT3 24,942.47UPC2 31,527.14YDR520C 24,605.21IXR1 41,224.61UGA3 24,337.57WTM2 27,119.07ROX1 34,123.42INO2 29,926.03YAP5 24,166.04ZMS1 24,004.79PDR3 23,735.8UME6 27,712.91RCO1 46,841.26WTM1 135,981.4MAC1 44,675.11KSS1 51,426.07SUT1 46,944.43DAT1 49,860.42DAL81 54,671.8ECM22 49,117.57XBP1 53,446.87MIG1 54,576.54DIG1 52,827.92HMS2 48,718.59CUP9 48,009.46YAP7 45,950.59MBP1 55,450.88SKO1 48,905.9YDR026C 23,436.41GAT1 24,449.59HOG1 29,015.13CBF1 34,552.59FZF1 24,368.94HAP5 25,887.41TBS1 22,157.88LEU3 23,521.49YRR1 22,629.15SKN7 23,984.04GAL80 28,860.63YHP1 49,703.81CIN5 20,915.41RGM1 20,419.31NRG1 21,071.44MET28 20,420.98

Denby et al. www.pnas.org/cgi/content/short/1116360109 4 of 5

Table S1. Cont.

Gene Abundance

ARG80 20,768.52RTG1 26,675.7INO4 20,302.5HAP3 21,632.88

Each row reports the total cellular fluorescence in arbitrary units of theindicated transcription factor as a GFP protein fusion in a diploid hemizy-gote measured by quantitative microscopy.

Denby et al. www.pnas.org/cgi/content/short/1116360109 5 of 5