Integrated Genomic and Proteomic Analyses of a Systematically Perturbed Metabolic Network T. Ideker, V. Thorsson, J. A. Ranish, R. Christmas, J. Buhler, J. K. Eng, R. Bumgarner, D. R. Goodlett, R. Aebersold, L. Hood

Integrated physical-interaction network (PDF format, enlarged from Fig. 4A). Gene expression data from this study are shown in the context of a network of previously observed physical interactions. Nodes represent genes and are labeled with their corresponding gene names. Connections between nodes display physical interactions as recorded in the public databases, where a yellow arrow directed from one node to another represents a protein --> DNA interaction, and a blue line between nodes represents a protein-protein interaction. Global changes in mRNA expression (in this case, in response to a deletion of GAL4 in the presence of galactose) are visually superimposed on the network. The grayscale intensity of each node indicates the change in mRNA expression of the corresponding gene, where medium gray represents no change, darker or lighter shades represent an increase or decrease in expression, respectively (as in Fig. 2), and node diameter scales with the overall magnitude of change. GAL4 is colored in red to signify that its expression level has been perturbed by external means. When viewing this PDF in Acrobat Reader, use the magnifying glass to obtain more detail on network regions of interest and to enlarge the names of individual genes. Highly interconnected groups of genes tend to have common biological function and are annotated accordingly (rectangular labels).

Download supplemental figure 1

Supplemental Figure 2. Comparison of Northern vs. microarray analysis. Total RNA from yeast growing in each of eight perturbation conditions (column headings) was probed with radiolabeled GAL1 or GAL80 cDNA using a Northern blot assay. The actin gene (ACT1) was also labeled as a control expected to be expressed at constant levels over all conditions. Corresponding changes in gene expression measured with the yeast-genome microarray are depicted graphically beneath each blot. Microarray data are shown relative to wt yeast growing in galactose (column 5), with medium-gray representing no change, darker or lighter shades representing increasing or decreasing amounts of expression respectively, and spot size scaling with the magnitude of change. The quantitative log₁₀ change in expression level is annotated in each spot center.

Medium version | Full size version

Supplemental Table 1. mRNA- and protein-expression responses to galactose pathway perturbations (Microsoft Excel format). This table reports mRNA-expression ratios measured across all 23 experimental perturbations (20 initial and three follow-up perturbations) using yeast-genome microarrays, as well as protein-abundance ratios for each of 289 proteins measured between wt+gal and wt-gal conditions using ICAT technology.

To view Supplemental Table 1 in excel format click here

Supplemental Table 2. Physical interactions linking genes with strong expression-profile correlations [accompanies Step (iii) of the text]. Pearson correlations rAB were computed between the expression profiles of pairs of genes A and B, where each profile consists of log10 mRNA-expression ratios over our galactose-pathway perturbations (all 20 except the wt+gal versus wt+gal control). Strong correlations (|rAB| > 0.4) are reported for all cases in which [a] protein A is involved in a known proteinDNA interaction with gene B (AB); [b] protein A affects B through a signaling pathway involving a protein-protein interaction between A and a third protein C (A-CB); [c] genes A and B are regulated by proteinDNA interactions from a common transcription factor C (CA,B). These interactions are hypothesized to have transmitted expression changes from one gene to another in one or more perturbations. When protein C regulates more than two genes (A, B, ...), we list the range of all pairwise correlations among regulated genes.
[a] AB
A	B	rAB
Alpha2	STE2	-0.545
	BAR1	-0.469
Cbf1	PGK1	-0.574
	MET16	0.471
Cup2	CUP1B	0.631
	CUP1A	0.573
Gcn4	ARG1	0.422
Gcr1	PGK1	0.579
	TPI1	0.575
	CDC19	0.544
	ENO1	0.535
	ENO2	0.493
Hap1	CTT1	-0.45
Hsf1	SSA4	-0.409
Mac1	FRE1	0.476
Mcm1	FAR1	0.817
	MFA2	-0.675
	MFA1	-0.526
	PIS1	-0.445
Met28	MET16	0.471
Met4	MET16	0.52
Mig1	FBP1	0.633
	HAP4	0.601
	GAL1	-0.549
	GAL3	-0.477
	GAL10	-0.465
Msn2	SUC2	-0.416
Pdr1	HXT9	-0.427
Put3	PUT1	0.535
Rox1	HEM13	-0.431
Sip4	PCK1	0.513
	FBP1	0.5
	ICL1	0.48
	MLS1	0.429
Sko1	SUC2	0.433
Ste12	MFA2	-0.456

[b] A-CB
A	C	B	rAB
Gcr2	Gcr1	TPI1	-0.859
		PGK1	-0.675
		CDC19	-0.463
		ENO2	-0.443
Hsp42	Rap1	PGK1	0.529
		TPI1	0.48
		CDC19	0.441
Mcm1	Alpha2	MFA1	-0.526
		MFA2	-0.675
	Ste12	MFA1	-0.526
		MFA2	-0.675
Rad52	Rfa1	CAR1	0.697
	Rfa2	CAR1	0.697
Tpo1	Cup2	CUP1A	0.469
Ydr412W	Pdr1	HXT9	-0.451

[c] C(A, B, ...)
C	(A, B, ...)	rAB
Alpha2	BAR1, MFA1, MFA2, STE2	0.427 to .772
Gal11	GAL1, GAL2, GAL7, GAL10	0.807 to 0.919
Gal4	GAL1, GAL2, GAL7, GAL10, GCY1	0.787 to 0.919
Gcn4	ADE4, HIS3, HIS4, HIS7	0.489 to 0.893
Gcr1	ENO1, ENO2, CDC19, PGK1, TPI1	0.563 to 0.891
Hap1	CYB2, CYC1, CYC7	0.388 to 0.753
Hsf1	HSP26, CUP1	0.555
Mcm1	BAR1, MFA1, MFA2, PIS1, STE2	0.427 to 0.854
	CLB1, FAR1	0.579
Mig1	GAL1, GAL3, GAL10	0.726 to 0.898
	FBP1, HAP4	0.665
Rap1	CDC19, ENO1, ENO2, PDC1, PGK1, TPI1	0.502 to 0.891
	HIS4, RPL16A, RPL18A, RPL18B, RPL25, RPS17A, RPS24A, RPS24B	0.503 to 0.969
Sip4	FBP1, ICL1, MLS1, PCK1	0.669 to 0.882
Ste12	MFA1, MFA2, STE2	0.679 to 0.772

Supplemental Table 3. Gal4p binding-site predictions [accompanies Step (iii) of the text]. We looked for the well-characterized Gal4p-binding site (UASGAL) upstream of the 997 genes whose mRNA levels were significantly affected by our 20 GAL-pathway perturbations. Of the 41 genes in this set with UASGAL predictions, Gal4p-binding had been confirmed experimentally for seven of these (shown in bold) [R. J. Bram, N. F. Lue, R. D. Kornberg, EMBO J. 5, 603-608 (1986); M. Angermayr, W. Bandlow, J. Biol. Chem. 272, 31630-31635 (1997); W. Zheng, H. E. Xu, S. A. Johnston, J. Biol. Chem. 272, 30350-30355 (1997)]. Since all experimentally-confirmed genes fell into expression clusters 1, 2, or 3, we suggest that other genes in these clusters with predicted UASGAL sites may be directly regulated by Gal4p. Note that the average expression profiles of clusters 1, 2, and 3 are also highly correlated with one another (r12 = 0.8; r13 = 0.8; r13 = 0.7).
Cluster	Gene	Upstream position (+/- strand)	Core similarity score	Matrix similarity score	Sequence
1	GAL1 (GAL10 shares regul. region)	456 (+)	1.000	0.840	GTACGGATTAGAAGCCGCCGAGC
		437 (+)	1.000	0.880	GAGCGGGCGACAGCCCTCCGACG
		419 (+)	0.875	0.920	CGACGGAAGACTCTCCTCCGTGC
		355 (+)	1.000	0.860	CCTCGCGCCGCACTGCTCCGAAC
1	GAL7	336 (-)	1.000	0.860	CCTCGCGCCGCACTGCTCCGAAC
		272 (-)	0.875	0.920	CGACGGAAGACTCTCCTCCGTGC
		254 (-)	1.000	0.880	GAGCGGGCGACAGCCCTCCGACG
2	GCY1	372 (-)	1.000	0.840	CCCCGGAATAGTCTGCCCCGATT
2	GAL6 (LAP3)	67 (+)	1.000	0.870	CGCCGGCTGACAAGTCGCCGACG
2	MLF3	169 (-)	1.000	0.924	CCGCGGAGTGCTCTTCGCCGAGA
2	PCL10	235 (-)	1.000	0.848	GATCGGTGCAATATACTCCGAGC
2	YEL057C	417 (+)	1.000	0.828	GGACGGGCGGCTGCCGTCCGGGG
2	YPL066W	101 (+)	1.000	0.827	TCACGGTCATCACTGCTCCGACA
2	YPS3	211 (+)	1.000	0.876	GATCGGATTACTATTCGCGGAAA
3	GAL2	533 (-)	1.000	0.905	TTCCGGAAGGAAGCTTTCCGAAT
		419 (+)	0.800	0.829	CACCGGCGGTCTTTCGTCCGTGC
		400 (-)	0.875	0.811	GAACGGCGCAGATATCTCCGCAC
		336 (+)	1.000	0.856	TATCGGGGCGGATCACTCCGAAC
		331 (+)	0.725	0.806	GGGCGGATCACTCCGAACCGAGA
3	GAL3	291 (-)	1.000	0.818	GTTCGGCACACAGTGGACCGAAC
3	GAL80	175 (+)	0.875	0.903	TACCGGCGCACTCTCGCCCGAAC
3	RPA49	249 (+)	0.800	0.825	GACCGGACACCTAATCACCGACG
3	YMR318C	239 (+)	1.000	0.867	GTCCGGTCCGTCCTTGACCGAAG
3	YPR194C	624 (-)	1.000	0.855	CGTCGGACAGCAACCCCCCGATT
5	YLR201C	143 (-)	1.000	0.938	CTTCCGCCTAATATAGTCCGAAA
7	ADK1	226 (-)	1.000	0.952	CTGCTGCGGACAGTTCTCCGTGA
7	RPL27A	175 (+)	1.000	0.804	TTGCTGCAGAGATTCGCCCGAAG
7	RPL34B	389 (+)	1.000	0.830	TTTCGGAGGTCCCGCTTCCGACA
7	YGR090W	381 (-)	1.000	0.814	CAACGGCATGCAGCGAGCCGTAG
8	BAP3	750 (+)	1.000	0.812	GTGCGAAGTAGTATGATCCGAAG
8	YLR042C	276 (+)	0.875	0.813	ATTTGGCCAAGAATGCCCCGAAC
10	MTH1	472 (-)	1.000	0.801	GCACGGACTCCATTTCCCCGGAC
10	NAR1	460 (-)	1.000	0.823	CGCCGGCTGACAAGTCGCCGACG
10	YJL217W	259 (-)	1.000	0.819	GAACGGTACTTATTTCCCCGAAA
10	YLR352W	298 (-)	0.875	0.819	GCGCGGGTAACATACCTCCGTGA
11	ISU1	306 (-)	1.000	0.805	AAGCGGAGATAAAGCCTCCGAAC
12	MRK1	268 (-)	1.000	0.800	CATCGGACGACTTTGCTCCCAGG
12	POR1	281 (+)	0.875	0.825	GATCGGGGTTCAATTCCCCGTCG
12	USV1	251 (+)	0.800	0.834	AACCGTTCAACAGTCTTCCGTAT
12	YJL112W	526 (+)	1.000	0.827	GATCGGGGTTCAATTCCCCGTCG
12	YMR031C	305 (-)	1.000	0.840	TTTTGGGTAACAGCGGACCGAAG
12	YMR044W	23 (+)	0.875	0.836	GAACGGCGTGTCATTCTCCGATA
12	YMR098C	299 (+)	1.000	0.853	AATGGGGTCACAATCATCCGAAC
13	GAL5 (PGM2)	337 (-)	0.725	0.828	CTCCGCGCTTCTCTTCACCGAGC
		175 (+)	0.875	0.817	ATCTGGATGACTGCCGCCCGAAC
14	MBR1	313 (+)	1.000	0.804	GAGCGGCTCCCCTTTCCCCGGAA
14	YJL045W	524 (+)	0.875	0.802	GATCGGGGTTCAATTCCCCGTCG
14	YLR164W	243 (+)	0.875	0.817	GATTGGAGTACCCTTATCCGAAG
15	ICL1	407 (+)	1.000	0.804	CCCAGGTTTCCATTCATCCGAGC
16	YIL057C	192 (+)	1.000	0.801	CGGCGGTTGGCAATCGTCCGTAT

Supplemental Table 4. New observations, hypotheses, and possible tests
	OBSERVATIONS	EXAMPLE HYPOTHESES	SYSTEMS-LEVEL TESTS
[1]	Effect of gal80D-gal: slow growth and widespread changes in metabolic-gene expression	Stress-related, caused by derepression of either the GAL enzymes or transporter	Examine EP of a gal4Dgal80D-gal double deletion, in which the GAL enzymes and transporter are not expressed
[2]	Decrease in GAL-gene expression in response to gal7D+gal or gal10D+gal	Dependent on levels of Gal-1-P or a derivative metabolite	Examine EP of a gal1Dgal10D+gal double deletion strain, in which Gal-1-P levels are reduced
[3]	GAL5 and GAL6 mRNA levels are unaffected by galactose addition or by deletion of GAL3, 4, or 80	Caused by differences in strains and/or media between this and previous studies	Obtain EPs for identical strains and media as in previous studies
[4]	GAL6 deletion does not affect mRNA levels of GAL enzymes
[5]	Expression levels of genes in many other metabolic pathways respond to perturbations of the GAL pathway	Each affected pathway depends on galactose, specific GAL genes, or on the total amount of available energy	Examine EPs of yeast growing in carbon sources other than galactose, e.g. 2% glucose
[6]	In wt+gal vs. wt-gal, approx. 15 genes change in protein but not mRNA abundance (Fig. 3)	These genes are regulated at the level of protein translation or degradation	Compare global translation state of proteins between + vs. - gal, using method of [Q. Zong et al. Proc Natl Acad Sci U.S.A. 96, 10632-6. (1999)]
[7]	In wt+gal vs. wt-gal, most ribosomal subunits increase in mRNA but not protein abundance (Fig. 3)
[8]	Nine genes w/ predicted Gal4p-binding sites have EPs that are similar to those of known GAL genes	Gal4p regulates transcription of these genes via protein-DNA interactions	Verify predicted interactions by global chromatin immuno-precipitation experiments [B. Ren et al., Science 290, 2306-9. (2000)]
Possible model refinements shown in Fig. 1.
Tests explored in Step (iv) of the paper.
EP = Expression profile.