3008.full

7/28/2019 3008.full

1/17

3008 | A. A. Petti et al. Molecular Biology o the Cell

MBoC |ARTICLE

Combinatorial control of diverse metabolic andphysiological functions by transcriptionalregulators of the yeast sulfur assimilationpathwayAllegra A. Pettia,*, R. Scott McIsaaca,b,, Olivia Ho-Shinga,,, Harmen J. Bussemakerc,and David Botsteina,daThe Lewis-Sigler Institute or Integrative Genomics, bGraduate Program in Quantitative and Computational Biology,and dDepartment o Molecular Biology, Princeton University, Princeton, NJ 08544; cDepartment o Biological Sciences,Columbia University, New York, NY 10027

ABSTRACT Methionine abundance aects diverse cellular unctions, including cell division,

redox homeostasis, survival under starvation, and oxidative stress response. Regulation o

the methionine biosynthetic pathway involves three DNA-binding proteinsMet31p,

Met32p, and Cb1p. We hypothesized that there exists a division o labor among these

proteins that acilitates coordination o methionine biosynthesis with diverse biological pro-

cesses. To explore combinatorial control in this regulatory circuit, we deleted CBF1, MET31,

and MET32 individually and in combination in a strain lacking methionine synthase. We ol-

lowed genome-wide gene expression as these strains were starved or methionine. Using a

combination o bioinormatic methods, we ound that these regulators control genes in-

volved in biological processes downstream o sulur assimilation; many o these processes

had not previously been documented as methionine dependent. We also ound that the di-

erent actors have overlapping but distinct unctions. In particular, Met31p and Met32p areimportant in regulating methionine metabolism, whereas p unctions as a generalist tran-

scription actor that is not specic to methionine metabolism. In addition, Met31p and

Met32p appear to regulate ironsulur cluster biogenesis through direct and indirect mecha-

nisms and have distinguishable target specicities. Finally, CBF1 deletion sometimes has the

opposite eect on gene expression rom MET31 and MET32deletion.

INTRODUCTIONCells must adapt to a great variety o internal and external signals.One way they accomplish this is by using condition-specic,sequence-specic transcription actors (TFs) to make compensatory

changes in gene expression. Instead o using a distinct TF or everypossible combination o environmental perturbations, the cell usesa parsimonious strategy o transcriptional control in which TFs can

be used in combinations that elicit distinct, stimulus-specic, tran-scriptional responses rom their target genes.

The existence o genome-wide expression and TF-binding datain Saccharomyces cerevisiae and other organisms has enabled re-searchers to systematically catalogue many such combinations, in-cluding pairwise combinations (Pilpel et al., 2001), higher-ordercombinations such as the multi-input moti, and more complex TFcircuits such as the regulator chain, multicomponent loop,eedorward loop, and dense-overlapping regulon (Lee et al.,2002; Milo et al., 2002; Shen-Orret al., 2002). However, compara-tively ew examples o combinatorial control have been studied indetail experimentally. In yeast, well-studied examples o combinato-rial control include the response to atty acids by Oa1p, Oa3p,

Pip2p, and Adr1p (Smith et al., 2007); combinatorial use o Hog1p

Monitoring EditorCharles BooneUniversity o Toronto

Received: Mar 23, 2012Revised: Jun 4, 2012Accepted: Jun 6, 2012

This article was published online ahead of print in MBoC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E12-03-0233) on June 13, 2012.These authors contributed equally.

Present addresses: *Department of Human Genetics, University of Chicago,Chicago, IL 60637; Department of Molecular and Cellular Biology, HarvardUniversity, Cambridge, MA 02138.

Address correspondence to: Allegra A. Petti ([email protected]),David Botstein ([email protected]).

2012 Petti et al. This article is distributed by The American Society for Cell Biol-ogy under license from the author(s). Two months after publication it is availableto the public under an AttributionNoncommercialShare Alike 3.0 UnportedCreative Commons License (http://creativecommons.org/licenses/by-nc-sa/3.0).

ASCB, The American Society for Cell Biology, and Molecular Biology of

the Cell are registered trademarks of The American Society of Cell Biology.

Abbreviations used: Fe-S, ironsulfur; Met, methionine; TF, transcription factor;TFBM, transcription factorbinding motif.

7/28/2019 3008.full

2/17

Volume 23 August 1, 2012 Coordination by methionine regulators | 3009

The more widespread physiological infuence o the Met path-way was rst observed when Unger and Hartwell (1976) ound thatmethionine starvation causes cell cycle arrest. According to an earlymicroarray-based cell cycle analysis, the Met biosynthetic genesbut not the biosynthetic genes o any other amino acidare ex-pressed periodically throughout the cell cycle (Spellman et al.,1998). It was subsequently shown that constitutive activation oMet4p (such as by deletion oMET30) causes cell cycle arrest at theG1/S transition and that this arrest depends on the transcriptionalactivation domain o Met4p (Patton et al., 2000). Methionine starva-tion, unlike starvation or other amino acids, results in cell cycle ar-rest and survival; survival during methionine starvation is abolishedby double deletion oMET31 and MET32and is correlated with theability o cells to mount an eective oxidative stress response (Pettiet al., 2011). It is notable in this context that activity o the Met tran-scription actors is more tightly correlated with the yeast metaboliccycle than any other TF (Tu et al., 2005; Murray et al., 2007). Finally,a detailed study o the slow-growth phenotype oMET4 deletionmutants showed a direct regulatory connection between the sulurassimilation pathway and the phosphatidylcholine biosyntheticpathway (Hickman et al., 2011), in which the SAM2gene (encodingone o two SAM synthetases) is regulated by both MET4 and OPI1,

which regulates lipid metabolism.Two recent articles examined in detail the binding o Cb1p and

Met31p/Met32p to promoters in the Met biosynthetic pathway (Leeet al., 2010; Siggers et al., 2011). Nevertheless, neither study lookedin depth beyond the sulur assimilation pathway, and the biologicalrationale or the complexity o this transcriptional regulatory net-work remains unclear. We hypothesized that there is some divisiono labor among these TFs that acilitates the coordination o methi-onine metabolism with the great variety o cellular processes thatdepend on sulur metabolism. We reasoned that by comparing phe-notype and gene expression across a panel o isogenic TF-deletionmutants undergoing methionine starvation, we could understandthis division o labor and learn how the multiplicity o actors coordi-

nates the many biological unctions that depend, directly or indi-rectly, upon the sulur assimilation pathway.

We began by surveying the genome-wide eects o methioninestarvation in methionine auxotrophs lacking MET6 (which encodesmethionine synthase) orMET13 (which encodes methylenetetrahy-droolate reductase). In a preliminary bioinormatic analysis o thisdata, we ound evidence that the Met TFs Met31p/Met32p andCb1p are not redundant, lending support to our division o laborhypothesis. We also used these data to systematically identiy cel-lular processes that depend on methionine and to generate TF-binding matrices or Met31p/Met32p and Cb1p, which we use inmost subsequent analyses.

To expand upon the division o labor result, we designed a set

o ollow-up experiments and computational analyses to investigatedistinctions between these TFs and the potential role o combinato-rial regulation in coordinating methionine-dependent processeswith methionine abundance. In these experiments, we deleted thegenes encoding the three DNA-binding proteins in this systemCBF1, MET31, and MET32individually and in pairwise combina-tions. To assure that all the mutants were starving equally or methi-onine, we made these deletions in the met6 background. Wemeasured and compared the gene expression patterns and cell cy-cle progression o these strains during methionine starvation tothose o theirmet6 parent. (We also deleted MET4, but not in themet6 background, because met4 is growth impaired even whensupplemented with excess methionine [Hickman et al., 2011]; weused results with this strain or qualitative comparisons only.)

and Msn2p/Msn4p (Capaldi et al., 2008); and, on a smaller scale,signal integration at the FLO11 promoter through the use o multi-ple TF-binding sites (TFBS; Rupp et al., 1999).

We developed an integrated experimental and computationalramework to systematically investigate combinatorial transcrip-tional regulation in response to methionine abundance in S. cerevi-siae. Many previous studies o combinatorial regulation ocused onthe use o combinatorial control to integrate multiple environmentalsignals. Methionine provides an example o special interest becausea single signal (methionine abundance) controls a wide variety ointracellular responses in a way that has not been studied. Here weinvestigate the mechanisms o this control.

Methionine is synthesized by the sulur assimilation pathway, alsoknown as the methionine (Met) pathway, which occupies a centralrole in metabolism and growth control in yeast. The pathway synthe-sizes cysteine, methionine, and S-adenosyl methionine (SAM) rominorganic sulate and leads to a host o other essential metabolites.Some o these contain sulur atoms (e.g., glutathione and acetyl-CoA), but others are connected to the pathway because they con-tain methyl groups derived rom SAM (e.g., phosphatidyl choline) oraminopropyl groups rom the SAM derivative S-adenosyl methioni-namine (e.g., polyamines). Essential macromolecular products o

the sulur assimilation pathway include proteins in general; mem-branes, which contain phospholipids whose biosynthesis dependson methyl groups derived rom SAM; and many specic proteins,notably the diverse ironsulur proteins that carry out electron trans-er reactions. Methionine and SAM are well situated to be controlpoints or protein synthesis (every polypeptide chain is initiated withmethionyl-tRNA), membrane biosynthesis (SAM is required at sev-eral steps o the pathways leading not only to phospholipids butalso sterols), redox balance (via glutathione), and methylation o his-tones and DNA itsel (although yeast DNA is not known to bemethylated).

Much o what is known about the biological connections amongmethionine biosynthesis, central metabolism, and growth control

has been learned through studies o mutants (mostly obtained asmethionine auxotrophs) that either abrogate or deregulate the Metpathway. These studies have revealed a complex transcriptionalregulatory network (Figure 1A; reviewed in Thomas and Surdin-Ker-jan, 1997). At the metabolic level, this circuit is governed by SAM(Kuras and Thomas, 1995a), which acts through an ensemble o TFs.In response to high SAM levels, the SCFMet30ubiquitin ligase ubiquit-inylates the transcriptional activator Met4p, resulting in Met4p inac-tivation, but not degradation (Cai and Davis, 1990; Blaiseau et al.,1997; Rouillon et al., 2000; Kaiseret al., 2006). Met4p binds DNAwith the assistance o three DNA-binding proteins (Met31p, Met32p,and Cb1p) and one protein (Met28p) that acilitates the interactionbetween Met4p and the DNA-binding proteins. Met31p and

Met32p, which are largely overlapping in unction, bind to the DNAsequence moti AAACTGTGG, whereas Cb1p binds to the motiCACGTG (Cai and Davis, 1990; Blaiseau et al., 1997; Blaiseau andThomas, 1998). Cb1p, variously known as CP1, CEP1, and CPF1,also binds centromeres through the same CACGTG sequence and isrequired or proper chromosome segregation (Bram and Kornberg,1987; Baker et al., 1989; Baker and Masison, 1990; Thomas andSurdin-Kerjan, 1997). Met4p acts with dierent DNA-binding pro-teins to regulate dierent Met pathway biosynthetic genes (reviewedin Thomas and Surdin-Kerjan, 1997), and Cb1p was shown early onto be partly dispensable or the expression o most o the Metpathway genes (Kuras and Thomas, 1995b). Evidence in the litera-ture, although sparse, suggests that MET31 and MET32might notbe perectly redundant (Su et al., 2008; Cormieret al., 2010).

7/28/2019 3008.full

3/17


FIGURE 1: Overview o the background and the computational methods. (A) The transcriptional circuit governing sulurassimilation and methionine biosynthesis. (B) Overview o the preliminary bioinormatic analysis using met6 expression datato identiy TFBMs or Met31p/Met32p and Cb1p and to distinguish the unctional roles o these TFs using a Web-basedtool called Gene Ontology Term Finder (GOTF). (C) Overview o the use o TF-dependency analysis to compare TF-deletionmutants in order to identiy and characterize genes that depend specifcally on Met31p/Met32p, Cb1p, or both.

(D) Overview o the use o multiple regression and Studentst-test to identiy dierences between

MET31and

MET32.

7/28/2019 3008.full

4/17


Beore choosing met6 as the background and control strain orthe regulatory studies described later, we compared the met6 andmet13 data in order to veriy that most gene expression is ex-tremely similar in both strains. This allowed us to conclude that theeects we see are indeed the result o methionine depletion and arenot specic to the loss o methionine synthase.

Multiple regression was used to identiy 466 genes whose ex-pression depends signicantly on time, diers signicantly betweenmet6 and met13, and changes by twoold or more (see Materialsand Methods). Minor dierences were ound in the expression kinet-ics or amplitude o genes involved in sulur metabolism (more highlyinduced in met6), cell cycle and nucleotide metabolism (repressedearlier in met13), and regulation o translation (repressed earlier inmet6).

We then characterized the general transcriptional response toMet depletion using the remaining 2669 genes that change by two-old or more, depend signicantly on time, and behave similarly inmet6 and met13 (Supplemental Figure S1). We clustered thesegenes using the k-means algorithm, identied clusters with distinc-tive expression proles, and analyzed their Gene Ontology (GO)term enrichment using an online tool called the Generic Gene On-tology Term Finder (GOTF; Supplemental Data Set 1; see Materials

and Methods). The most rapidly and highly induced cluster (122genes, cluster 9 in Supplemental Figure S1) is enriched or pro-cesses related to metabolism o sulur, methionine, and metaboli-cally related amino acids; oxidative stress response; response toheavy metals; and targets o the Met transcription actors Met4p,Cb1p, Met31p, and Met32p. However, we also observed rapid,highly coordinated induction o genes involved in nonglycolytic en-ergy generation, including storage compound metabolism (cluster16) and electron transport (cluster 11), implying increased relianceon mitochondrial energy generation during Met starvation. Consis-tent with the progressive cell cycle arrest observed during methion-ine starvation (Unger and Hartwell, 1976; Petti et al., 2011), all geneclusters enriched or biological processes related to cell division,

the chromosome cycle, and DNA replication are gradually repressedduring the starvation (clusters 3, 4, 10, and 19). A small set o mark-edly coregulated genes with a distinctive expression prole (cluster4) is enriched or a variety o seemingly disparate unctions, includ-ing DNA packaging (e.g., the key histone-encoding genes HTA1,HTA2, HTB1, HHF1, HHF2, HHT1 HHT2), purine metabolism, sin-gle-carbon metabolism, and amino acid biosynthesis. This is in-triguing, given the putative role o methionine in regulating the cellcycle. Both HHF2and HHT2contain binding sites or Met4p, sug-gesting that the transcriptional regulation o these histone-encod-ing genes may depend directly on Met abundance (MacIsaac et al.,2006). All o the clusters enriched or ribosomal biogenesis genesare repressed, consistent with numerous previous ndings that en-

vironmental stress represses ribosome biogenesis.

Using gene expression patterns in mutants to dissect thecomplex combinatorial regulation o the pathwayTo understand how the Met pathway TFs might regulate Met-infuenced processes that go beyond the sulur assimilation path-ways per se, we rst undertook a simple bioinormatic analysis otranscriptional regulation, as outlined in Figure 1B. Here we usedthe gene expression data we collected or the met6 strain toderive key transcription actorbinding motis (TFBMs), and char-acterized the biological unction o moti-containing genes thatchange substantially during methionine starvation (Figure 2). Wehierarchically clustered genes that change by twoold or more inthe met6 time course and identied a set o 45 genes that

We designed these experiments with particular computationalmethods in mind. Using these methods, we analyzed the expressiondata to identiy dierences in specicity among the TFs. The analyti-cal pipeline or several key methods is summarized in Figure 1, BD;another method, transcription actor activity analysis, is summarizedin Lee and Bussemaker (2010). Our analysis ocused on two mainquestions: rst, whether the dierent TFs regulate dierent genes;second, whether there are unctional dierences between the genesregulated by each TF, which we would take as evidence that thereexists a division o labor that enables dierent TFs to coordinateMet metabolism with dierent cellular processes. We ocused ondierences between the two major TF classesCb1p and Met31p/Met32pbut also examined dierences between Met31p andMet32p in two genetic backgrounds (with and without CBF1).

We ound abundant evidence that the Met TFs regulate genesinvolved in biological processes that lie downstream o sulur as-similation. Some o these were previously known to depend onsulur or methionine, but many (e.g., copper transport, iron assimi-lation, ironsulur cluster biogenesis, maltose metabolism, andmicroautophagy) were not. Met4p is involved in virtually all othese; there is a striking similarity between the patterns o geneexpression in the met4 strain and the double met31met32

strains. Consistent with a regulatory circuit that serves to coordi-nate multiple cellular processes, we ound that the dierent ac-tors regulate overlapping but distinct groups o genes character-ized by overlapping but distinct biological unctions. We nd thatMet31p and Met32p have distinguishable target specicities, withsignicant gene expression dierences among the mutant pheno-types under methionine starvation. It appears rom our resultsthat, depending on circumstances, all three o the DNA-bindingproteins can act as repressors and activators through the same(or very similar) sequence motis. In many cases, we ound that agiven target gene is regulated dierently by dierent TFs. Finally,we note that the regulatory consequences oCBF1 deletion some-times oppose those oMET31 and MET32deletion, suggesting

that Cb1p and Met31p/Met32p can infuence gene expression inopposing ways and that the infuence o one is unaected by dele-tion o the other.

The experiments in this study are designed to perturb the Metpathway by starving auxotrophic TF-deletion mutants. By compar-ing phenotype and gene expression across a variety o deletion mu-tants, one can detect unctional distinctions among dierent TFs(and TF combinations). An alternative approach, based on sepa-rately inducing each TF-encoding gene and measuring the eectson gene expression, provides both support and additional inorma-tion. Those results are presented in an accompanying article (McIsaacet al., 2012).

RESULTSMethionine starvation infuences many diversecellular processesTo gain a preliminary overview o the cellular response to methion-ine starvation, we used gene expression data previously collectedduring starvation o strains that cannot make methionine becausethey lack a biosynthetic enzyme, as opposed to strains that have aregulatory deect. We had studied two such Met biosynthetic aux-otrophs, met6 and met13 (Petti et al., 2011). Here we used thisdata in several ways. First, we used it to gain an overview o methi-onine-dependent genes and processes. Second, we used met6 asa control against which to compare the transcription actordeletionmutants. Third, we used the met6 data to identiy transcriptionactorbinding motis or Met31p/Met32p and Cb1p.

7/28/2019 3008.full

5/17


the metabolism o related amino acids andoxidationreduction processes. In contrast,targets o only Cb1p or only Met31p/Met32p are involved in a wider variety oprocesses including, but not limited to,methionine and sulur metabolism.

Met31p/Met32p and Cb1p regulateoverlapping but distinct gene setsenriched or dierent unctionsTo investigate experimentally the special-ization between Met31p/Met32p andCb1p, we deleted MET31, MET32, andCBF1 individually and in all pairwise combi-nations in a met6 background. We mea-sured gene expression during the early ex-ponential phase o growth and duringmethionine depletion in these strains andselected genes that a) depend signicantlyon time and change by twoold or morein at least one strain, or b) are constitutivelyexpressed twoold relative to met6 in

at least one strain (see Materials andMethods). Based on the met4 expressiondata, this panel o single- and double-TFdeletions captures the ull range o expres-sion proles available to Met TF-deletionmutants: compared with the other TF-dele-tion mutants, we ound expression prolesunique to met4 in only 14 o 6256 genesassayed (notable examples include HXT1,SNQ2, CAT8, ESF1, SSA2, SPE2, DDI2, andDDI3).

We rst examined the specialization between Cb1p andMet31p/Met32p without distinguishing between Met31p and

Met32p. In an initial, coarse-grained analysis, we surveyed all GOcategories with at least 10 members to identiy those in whichmet31met32met6 and cb1met6 dier most. Specically,or each GO category i, we computed a distance dened asd n c mi

it j t j t j

ni

==

1

1max( )

, , . Here ni is the number o genes in category

i, cj,tis the expression level o genejat time point tin cb1met6,and mj,t is the expression level o gene j at time point t inmet31met32met6. As shown in Figure 3, the dierences be-tween the strains are roughly normally distributed across the GOcategories. Gene sets that are dierently aected by deletion oMET31/MET32and CBF1 are located in the tails o the distribu-tion. The right tail o the distribution, where expression is higherin cb1met6 than in met31met32met6, is dominated by

cysteine and methionine biosynthesis, sulur metabolism, andgrowth-related terms such as ribosome biogenesis and assem-bly, rRNA processing, and translation. In the let tail, we ndterms that are mainly related to redox homeostasis, includingiron transport and homeostasis, copper transport, metal iontransport, atty acid metabolism, and glutathione metabolism.This means that Met31p/Met32p and Cb1p dierentially regu-late genes related to methionine metabolism, as well as genesrelated to growth and cellular redox. We iner, as indicated inFigure 3, that the categories on the let are induced by Cb1pand/or repressed by Met31p/Met32p, whereas the categories onthe right are induced by Met31p/Met32p and/or repressed byCb1p. Independent evidence or this view can be ound inMcIsaac et al. (2012).

respond early and dramatically to methionine starvation. Most (43o 45) o these genes are induced within 10 min o methionine

removal (Figure 2A), and most change by a actor o 20 or more;a similar result was ound by Lee et al. (2010). Primarily located incluster 9 discussed earlier, these genes unction in methionine,cysteine, and S-adenosyl methionine biosynthesis, sulur metabo-lism, single-carbon metabolism (CHA1), DNA repair (RAD59), andbiosynthesis o the antioxidants glutathione (GTO1) and NADPH(BNA3). Using the moti identication algorithm MEME (seeMaterials and Methods), we identied two sequence motis in thepromoters o these 45 genes (Figure 2B). The motis match previ-ously derived motis or Met31p and Met32p, which are believedto bind the same moti, and Cb1p (Blaiseau et al., 1997; Leeet al., 2010; Siggers et al., 2011).

These sequence motis have been well studied in the context o

methionine metabolism, but the wide variety o methionine-infu-enced processes suggests that they also infuence non-Met pathwaygenes. We used MAST to search the genome or additional in-stances o each moti (MAST score

7/28/2019 3008.full

6/17


synthetic enzymes. The others (e.g., MHT1,MMP1, MET2, MUP3, CYS3, MUP1, SAM1,SAM2, SAM3) are involved in ancillary unc-tions such as methionine transport and theSAM cycle. This supports and renes ourprevious result: many Met pathway genes,particularly methionine and sulur transport-ers, depend on Met31p/Met32p but not onCb1p. The second striking subset o genesappears, surprisingly, to be repressed spe-cically by Met31p/Met32p. These genesare strongly enriched or iron assimilationand iron transport (e.g., SIT1, ENB1, FIT3,ARN2, ISU2) and are dramatically induced inthe met31met32met6 strain. We alsoound one Met31p/Met32p target, ISU2,which, by virtue o its role in the synthesis oironsulur (Fe-S) proteins, suggests a com-pelling connection between sulur assimila-tion and iron homeostasis.

These results are urther supported bythe additional, nondirect target genes in-

volved in the homeostasis o iron (e.g.,FET3, TIS11, FTR1, FRE1, FRE2, FRE3, FIT2,AFT1, SCS3, VHT1) and other metals, suchas copper (CUP1-1, CUP1-2, ARN1; Figure

4B). We noticed in previous work that deletion oMET31 and MET32(Petti et al., 2011) or depletion oMET4 (Hickman et al., 2011) in-duces iron-related genes, but the TF binding site (TFBS) data usedin those analyses (MacIsaac et al., 2006) indicated that these genesare not direct targets o any o the Met TFs. In those studies, there-ore, we concluded there that these genes are indirectly induced bymethionine depletion. In the target genes listed in the precedingparagraph, however, we have evidence that Met31p/Met32p di-rectly regulate iron assimilation.

The infuence oMET31 and MET32is not limited to the overrep-resented processes discussed earlier. Key regulatory genes or pro-cesses such as carbohydrate metabolism (HXT1, RIM1) and phos-phate metabolism (PHO11,PHO12, PHO81, PHO84,PHO86, PHO89)are dramatically aected by MET31/MET32 deletion. The glucosetransporter gene HXT1 appears to be strongly repressed inmet31met32met6 and more weakly repressed in cb1met6.

Cb1 regulates glutamate, meiotic, and carbohydrate genesFewer genes are aected by CBF1 deletion than by MET31/MET32deletion, but they span at least as wide a variety o processes,among which methionine metabolism is weakly represented. An

Transcription actor-dependency analysis dierentiatesbetween the Met31p/Met32p and Cb1p regulonsOur analysis so ar paints a broad picture o unctional specializationamong the Met pathway TFs. However, we wanted a more detailedunderstanding o the combinatorial regulation among these TFsand o how each TF aected the response to Met starvation relativeto the control strain, met6. To answer these questions, we devel-oped a nonparametric statistical method, which we reer to asTF-dependency analysis, to identiy genes whose expression de-

pends on Cb1p, Met31p/Met32p, or both. As outlined in Figure1C, we dened expression signatures corresponding to each sce-nario and ound genes whose expression proles were signicantlycorrelated with these signatures (see Materials and Methods). At aalse discovery rate (FDR) o 5%, we identied 89 genes that de-pend on both Cb1p and Met31p/Met32p, 455 that depend primar-ily on Met31p/Met32p alone, and 88 that depend primarily onCb1p alone. Only 15% o the classied genes contain a motior either TF, but moti prevalence accurately refects expression de-pendence within each class (Table 1).

Met31p/Met32p induces Met transporters and repressesiron homeostasis genes

In the Supplement, we provide an overview o the 455 genes thatdepend more strongly on Met31p/Met32p than on Cb1p, whichare enriched or a variety o processes ranging rom sulur and me-thionine metabolism to carbohydrate metabolism and iron homeo-stasis (Supplemental Figure S2 and Supplemental Data Set 1). Herewe ocus on the 13% that contain a moti or Cb1p and/or Met31p/Met32p (the target genes in Figure 4A), and the nontargetgenes that lack a moti but show particularly strong, specic depen-dence on Met31p/Met32p (Figure 4B). There are two striking sub-sets o target genes. One requires Met31p/Met32p or inductionand is enriched or a small number o processessulur assimilationand methionine, aspartate, cysteine, and glutathione metabolism.O note, only three o the genes involved in the sulur and methion-ine pathwayMET3, MET13, and MET17encode methionine bio-

FIGURE 3: Transcription actor specifcity among biological processes. Histogram o GObiological processes, showing the average maximum expression dierence betweencbf1met6 and met31met32met6 or each GO category with at least 10 members.

Percent containing moti(s)Met31p/

Met32p

moti only

Cb1p

moti only

Both

motis

Any

moti

Genes dependenton Met31p/Met32ponly

7.3 3.5 1.8 13

Genes dependenton Cb1p only

1.1 14 2.3 14

Genes dependenton both

6.7 9 14 29

TABLE 1. Motis ound in each TF-dependency class.

7/28/2019 3008.full

7/17


2007; http://avis.princeton.edu/pixie/index.php), a program that analyzes the unc-tional enrichment o genes that interactphysically or genetically with a query gene.bioPIXIE indicates that additional Cb1p-dependent genes interact with carbohy-drate metabolism genes, including YPR1(pentose metabolism), JID1 (respiration),and APA2 (regulation o gluconeogenesis).Most o these lack a moti or either Cb1por Met31p/Met32p, raising the possibilitythat Cb1p aects these genes indirectlyor that they represent a general stressresponse. However, the cb1met6strain grows to a higher cell density dur-ing methionine starvation than themet31met32met6 strain, and these car-bohydrate genes are not perturbed in thelatter strain. This suggests that the eect oCBF1 deletion on these genes is specic,not a general stress or slow-growth re-sponse. The infuence o Cb1p on TYE7

could be particularly consequential becauseTYE7is a transcriptional activator that bindsthe E-boxes o glycolytic genes.

The dearth o Cb1p-dependent methi-onine biosynthetic genes is striking andsupports previous evidence here and else-where (Lee et al., 2010) that Cb1p is notthe primary regulator o sulur assimilationand Met metabolism. MET4 and MET22arethe only specifcally Cb1p-dependentgenes with Met-related annotations. How-ever, we ound one gene that potentiallyconnects Met abundance with the cell cycle:

RMD6, which is required or meiosis, re-quires CBF1 or induction, contains theCb1p TFBM, and interacts physically and/or genetically with 20 genes involved in themetabolism, transport, and utilization o sul-ur and Met (bioPIXIE; see Materials and

Methods). Given that a number o other genes involved in sporula-tion or cell division are Cb1p dependent (RME1, BNS1, SPR6, KEL2,SCM4, GSC2, SPO22, MCM6, YKL053W [bioPIXIE], YLR042C[bioPIXIE], and YLR463C[bioPIXIE]), these data suggest that Cb1pmay help couple methionine abundance with cell division throughthe regulation o these genes.

Met31p/Met32p and Cb1p coregulate most methioninebiosynthetic genesDeleting eitherCBF1 or MET31/MET32perturbs the expression o 89genes. These jointly regulated genes are enriched or sulur assimi-lation and the metabolism o Met and related amino acids, as well asor cell wall organization and lipid metabolism (notable examples in-clude OYE2, IZH4, and YEH1; Supplemental Figure S4). Twenty-ninepercent o the genes in this class contain motis or Cb1p and/or Met31p/Met32p, and 14% contain motis or both (Table 1 andFigure 6). This group, which contains most o the enzyme-encodinggenes in the Met pathway, is also enriched or siroheme and hememetabolism, oxidationreduction, coactor metabolism, and the me-tabolism o related amino acids (serine, cysteine, aspartate). Otherjointly regulated target genes represent a variety o other processes:

overview o Cb1p-dependent genes (Supplemental Figure S3)shows weak enrichment or deoxyribonucleotide biosynthesis(RNR2, RNR4) and several genes associated with meiosis (RME1,BNS1, SPR6, RMD6, SPO22). In the more ocused view o putativelydirect Cb1p targets (Figure 5), it is notable that Cb1p-dependentgeneseven those containing TFBMs are very weakly enrichedor Met metabolism (Figure 5A), although they include MET4, the

master transcriptional activator o the Met pathway. Instead, Cb1-dependent target genes are enriched or glutamate and glutaminemetabolism and or nitrogen compound metabolism, and containdiverse metabolic genes such as isocitrate dehydrogenase (IDH1),NADP-dependent glutamate dehydrogenase (GDH3), the ATPaseATP7, and LPP1, which controls levels o the phospholipid-regulatorphosphatidic acid.

Consistent with the coarse-grained GO term analysis in Figure 3,we ound a number o Cb1p-dependent carbohydrate metabolismgenes, including TYE7, RIM15, SOK2, RIP1, VAN1, IDH1, ATP7,and GDH3, most o which appear to be repressed by Cb1p (Figure5). Because there are comparatively ew Cb1p-dependent genesand many o them are poorly annotated, we obtained more inorma-tion about these genes using bioPIXIE (Myers and Troyanskaya,

FIGURE 4: Genes whose expression depends on MET31/MET32. (A) Genes that depend

statistically signifcantly on MET31/MET32(but not CBF1) and contain a TFBM or Cb1p (graydot under column C) and/or Met31p/Met32p (green dot under column M). Strain order or bothpanels is met6, cbf1met6, met31met32met6, met4. The colored triangles are colorcoded or each strain and represent the decrease in methionine over each time course.(B) Selected genes that depend statistically signifcantly on MET31/MET32and lack a TFBM.

7/28/2019 3008.full

8/17


miss many synergistic interactions by ignoring the quadruple mu-tant met31met32cb1met6.) We used linear regression tocompare each TF mutant to met6 (see Materials and Methods) andchose genes whose expression in cb1met31met6 andcb1met32met6 is signicantly dierent rom that in met6(Fstatistic q 0.05, old-change dierence 1.5). We also requiredthat the signicance o this dierence be at least 10 times greaterthan that or the dierence between met6 and eithercb1met6ormet31met32met6.

Based on these criteria, 114 genes exhibit subtle synergy be-tween the two TF classes (Supplemental Data Set 1). The vast major-ity o these genes show diminished response to methionine starva-tion, although several (PHO5, HHF1, HHT2) show more extremeinduction. Most are involved in ribosome biogenesis and transla-tion, with a ew notable exceptions such as the carbon metabolismgenes TKL1, ALD6, PDC1, ADH1, GDH1, GPM1, and TDH1, theglutathione S-transerase GTT2, the transcription actorYAP5, andthe phosphatase PHO5, which is the most dierentially regulatedgene in met31cb1met6 and met32cb1met6. O note,PHO5, HHT2, and HHF1 interact, according to bioPIXIE.

Very ew o these genes have binding sites or either Cb1p(GAR1, ASP3-1, GTT2, MBF1, UTP21, PIR1, RPP2B, BNS1, GUA1) orMet31p/Met32p (RPL4B, ATG29, RPS7A, NMD3, ECM33), and only

FIGURE 5: Genes whose expression depends on CBF1. (A) Genesthat depend statistically signifcantly on CBF1 (but not MET31/MET32) and contain a TFBM or Cb1p (gray dot under column C)and/or Met31p/Met32p (green dot under column M). Strain order or

both panels is met6, cbf1met6, met31met32met6, met4.(B) Selected genes that depend statistically signifcantly on CBF1 andlack a TFBM.

FIGURE 6: Genes whose expression depends on MET31/MET32andCBF1. (A) Genes that depend statistically signifcantly on both CBF1and MET31/MET32and contain a TFBM or Cb1p (gray dot undercolumn C) and/or Met31p/Met32p (green dot under column M). Strain

order or both panels is met6, cbf1met6, met31met32met6,met4. (B) Selected genes that depend statistically signifcantly onCBF1 and MET31/MET32and lack a TFBM.

ADH3 (mitochondrial alcohol dehydrogenase), GIT1 (uptake oglycerol and choline), MNN4 (glycosylation o oligosaccharides),BNA3 (biosynthesis o NADH and cell cycle [bioPIXIE]), ATM1 (Fe-Scluster biosynthesis and respiration [bioPIXIE]), RAD59(DNA repair,also ound in other studies),VHS3 (G1/S cell cycle transition [bioPIXIE]and ion homeostasis), and JIP5(ribosome biogenesis). These resultssuggest that Cb1p and Met31p/Met32p must both be present orull induction o methionine biosynthesis genes and additional genesthat may be particularly important in Met metabolism.

Synergy between Met31p/Met32p and Cb1pThus ar, we have shown that Cb1p and Met31p/Met32p regulateoverlapping but distinct gene sets. In particular, genes that are co-regulated by Cb1p and Met31p/Met32p deviate rom wild-typeexpression roughly equally i eitherCBF1 orMET31/MET32is de-leted. We next looked or additive eects o CBF1 and MET31/MET32by asking whether there are genes whose expression is al-tered more substantially in mutants where both TF classes (CBF1and eitherMET31orMET32) are deleted. To answer this, we searchedor genes whose expression is perturbed relative to the control strain(met6) in cb1met31met6 and cb1met32met6, but not incb1met6 and met31met32met6. (As discussed earlier, ewgenes are uniquely expressed in met4, suggesting that we did not

7/28/2019 3008.full

9/17


genes involved in other metabolic pathways, including NAD (POS5),NADPH (ZWF1), phospholipid biosynthesis (CHO2), and atty acidmetabolism (LEE1, based on bioPIXIE). The second cluster is verystrongly enriched or genes involved in pheromone response, includ-

ing, or example, STE2, FAR1, FUS2, BAR1, MFA1, MFA2, and so on.

Variants o the canonical Met31p/Met32p DNA-bindingmoti are associated with dierent unctional responsesAs described, we ound that many genes depend on Met31p/Met32p,including some that require Met31p and Met32p or ull inductionand some that are only induced in the absence o Met31p andMet32p. Furthermore, we identied some genes that are regulateddierently by Met31p and Met32p. Reasoning that there might bevariants o the Met31p/Met32p TFBM that account or this wide vari-ety o expression proles, we looked or moti variants that distinguishbetween induced or repressed genes in met31met32met6. Usinggene sets described earlier, we identied genes that depend on

Met31p/Met32p but exhibit no signicant dierences between

one gene, YAP5, has motis or both TFs. On one hand, the lowincidence o TFBMs and the prevalence o ribosomal genes in thislist suggest that most o these genes are responding to an indirect,additive eect o methionine starvation in the TF mutants, such aslow growth rate due to depletion o sulur-containing compoundsand insucient methylation. On the other hand, many ribosomalgenes display minor expression deects in cb1met6, suggestingthat cb1 may indeed regulate ribosomal genes. This is consistentwith our initial, coarse analysis, which showed that numerous ribo-some biogenesis GO categories are less repressed in cb1met6than in met31met32met6.

Distinguishing between MET31 and MET32To pursue possible dierences between the unctions oMET31 andMET32(Su et al., 2008; Cormieret al., 2010), we used multiple re-gression as outlined in Figure 1D to compare met31met6 tomet32met6 (see Materials and Methods). Briefy, we requiredthat the strain pairs dier signicantly according to multiple regres-sion (F-test, q 0.1) or average expression level (Students t-test,q 0.1). To remove results that are statistically signicant but stillmodest in magnitude, we also required that the two strains dier byat least 1.5-old in at least one time point.

By these criteria, 35 genes dier in expression betweenmet31met6 and met32met6. The 15 most striking examplesare shown in Figure 7A. (Three additional genes, MXR1, VHT1, andSAM1, also appear to depend dierently on Met31p and Met32pbut did not meet our strict selection criteria.) For many o thesegenes, deleting MET31 appears to have the opposite eect o de-leting MET32: the genes are constitutively overexpressed inmet31met6 (note the strong expression at time zero) comparedwith met6, but underexpressed in met32met6 relative to met6.This subset is heavily enriched or genes involved in transport (VBA2,YCT1, YIL166C, AGP3, YOL162W, YOL163W). Although YOL162Wand YOL163W are ocially unannotated, bioPIXIE indicates thatthey interact physically and genetically with a group o genes that isheavily enriched or Met biosynthesis (and associated unctions) andthiamine metabolism. CRF1 is a TOR-activated transcriptional re-pressor o the ribosomal genes. Two genes, GRX8and PDC6, arehighly induced (8- and 64-old, respectively) in met31met6 butvirtually uninduced in met32met6. GRX8 is one o several glu-taredoxins whose expression is regulated by Met abundance. PDC6is an isoorm o pyruvate decarboxylase whose induction during sul-ur limitation has also been observed by others (Boeret al., 2003).More recently, PDC6 was shown to be regulated by Met32p througha noncanonical binding moti (Cormieret al., 2010).

We next examined the Met31p/Met32p-dependent genes de-scribed earlier (including genes that also depend on Cb1p) to de-termine the extent to which Met31p and Met32p can substitute oreach other. With the exception o the genes mentioned previously,

Met31p and Met32p appear to be able to substitute or each otherin all Met31p/Met32p-dependent genes.

We suspected, on the basis o preliminary data, that the dier-ences between met31 and met32 might be more pronounced in acb1 background. To narrow the long list o genes that dier be-tween cb1met31met6 and cb1met32met6, we increasedthe FDR stringency to 1%. Even then, 166 genes pass our criteria, thestrongest o which are shown in Figure 7B. The top cluster is stronglyenriched or Met and sulur metabolism, again indicating that Met31and Met32 play dierent roles even in the standard pathways withwhich they are associated. However, only two o the Met/sulur-asso-ciated genes (MET2and MET17) encode enzymes o the sulur as-similation pathway. Five regulate glutathione abundance and metab-

olism (GRX8, DUG2, DUG3, GLO4, OPT1). This cluster also contains

FIGURE 7: Genes that are regulated dierently by Met31p andMet32p. (A) Select genes that dier signifcantly betweenmet31met6 and met32met6. Strain order is met6,met31met6, met32met6, met4. (B) Selected genes that diersignifcantly between cbf1met31met6 and cbf1met32met6.Strain order is met6, cbf1met31met6, cbf1met32met6,met4.

7/28/2019 3008.full

10/17


characterize the three moti variants, we used MAST to search thegenome or additional examples o each variant (using a more strin-gent MAST score o 50), ltered the MAST results or genes thatchange at least twoold in our data, and measured the unctionalenrichment o the resulting genes. Based on this analysis, the posi-tive variant is ound preerentially in genes involved in Met and sul-ur metabolism, chromatin silencing, mitochondrial degradationglutathione metabolism, and spindle pole separation, whereas thecore variant is more specic to Met and sulur metabolism genes.(One complicating actor is that the core moti is a ragment o thepositive variant, so some promoters match both the positive and thecore motis. However, the unctional enrichment results hold evenwhen such promoters are eliminated rom the unctional enrichmentanalysis.) The negative variant is ound almost exclusively in genesinvolved in iron and ion homeostasis and transport, consistent withits close match to At1p, a major transcriptional regulator o ironmetabolism (Yamaguchi-Iwai et al., 1995). As discussed later, thissuggests that Met31p and Met32p regulate iron metabolismthrough direct and indirect mechanisms.

Transcription actor activity analysisThe previous analyses were designed to be stringent and restrictive.

To obtain a broader overview o the physiological dierences be-tween the Met pathway regulatory mutants we studied, we calcu-lated the transcription actor activity o each o 124 nonMet-path-way TFs (see Materials and Methods). Briefy, this metric quantiesthe activity o each TF at each time point by perorming linearregression o the genome-wide mRNA expression levels on the a-nity with which the TF is predicted to bind to each genes promoterregion (Lee and Bussemaker, 2010). It provides inormation abouthow each o our deletions aects each regulon in early log phase(the initial time point), as well as during Met depletion. This ap-proach is particularly useul because the biological roles o manyyeast regulons have already been airly well described. Thus we cansummarize the unctional eects o the TF deletions in terms o pre-

viously characterized regulons.This approach is complementary to the analyses given earlier

and diers rom them in several key ways. First, a given regulon canspan multiple cellular processes. Second, this approach does nottell us whether a Met pathway TF directly or indirectly infuenceseach regulon. Third, this method captures more subtle interstraindierences because we do not impose any thresholds on the extentto which a regulon must dier among strains. Moreover, the analysiswas perormed on all genes in the genome, including those withsubtle changes in gene expression.

The activity o all 124 non-Met TFs is represented as a hierarchi-cally clustered heatmap in Supplemental Figure S5. The gure illus-trates the number and diversity o cellular processes infuenced by

the Met pathway TFs, including Met biosynthesis and sulur metabo-lism (MET4, MET31, MET32, MET28, CBF1; cluster 1), carbohydratemetabolism (TYE7[cluster 1], MIG1), response to metals and metalion homeostasis (RCS1, AFT2, YAP1, CAD1, YAP7; cluster 2), pH reg-ulation (RIM101), phosphate metabolism (PHO4), ribosome biogen-esis (RAP1, FHL1, SFP1; cluster 3), mitochondrial respiration (HAP2,HAP3, HAP5; cluster 4), and the cell cycle (STB1, SWI4, SWI6, MBP1;cluster 5). Importantly, this analysis does not indicate whether theregulation is direct or which genes are perturbed by TF deletionthisinormation is better supplied by the preceding analyses.

This analysis reveals marked dierences among the Met path-way TFs, with striking examples highlighted in Figure 9. We seesubstantial dierences between Cb1p and Met31p/Met32p,Met31p and Met32p, and Cb1p/Met31p and Cb1p/Met32p.

FIGURE 8: Met31p/Met32p moti variants. (A) Expression levels ogenes that are induced early during Met starvation o met6, shownor all strains (let), and the canonical Met31p/Met32p TFBM derivedrom the promoters o these genes (right). Strain order or all panels ismet6, met31met32met6, met31met6, met32met6,cbf1met31met6, cbf1met32met6, met4. (B) Genes inducedby Met31p/Met32p (top) and the positive moti variant derivedrom their promoters (bottom). (C) Genes repressed by Met31p/Met32p (top) and the negative moti variant derived rom theirpromoters (bottom). (D) Genes that dier between met31met6 andmet32met6 (top) and the core moti variant derived rom theirpromoters (bottom).

met31met6 and met32met6 (in either the CBF1 or the cb1background). We divided the most convincing genes in this list (Figure8A) into two groups, based on whether they are positively regulated(i.e., repressed in the deletion mutant; Figure 8B) or negatively regu-lated (i.e. induced in the deletion mutant; Figure 8C) by Met31p/Met32p. Using MEME, we derived dierent versions o the Met31p/Met32p TFBM rom the promoters o each group, which we call the

positive variant (Figure 8B) and the negative variant (Figure 8C).The positive variant is almost identical to the canonical moti shown inFigure 8A, with the exception that its upstream adenine residues aremuch more highly conserved. In contrast, the negative variant is sub-stantially dierent rom the canonical moti and contains an exactmatch to the At1p moti (Zhu et al., 2009).

To search or motis in genes that respond dierently inmet31met6 and met32met6, we applied MEME to the promot-ers o the genes in Figure 8D. The result was a third variant, the corevariant, consisting o the central, conserved TGTGG o the canonicalmoti. We also searched the promoters o the genes in Figure 7B andderived a moti very similar to the canonical Met31/Met32 moti.

One measure o the validity o a computationally derived TFBMis the unctional specicity o the genes in which it occurs. To urther

7/28/2019 3008.full

11/17


FIGURE 9: Transcription actor activities. Calculated activity profles or selected TFs across all strains. Each panel showsTF activity (vertical axis) or the indicated TFs throughout the time courses or all eight strains. The strains arerepresented in the ollowing order along the horizontal axis: met6, cbf1met6, met31met32met6,cbf1met31met6, cbf1met32met6, met31met6, met32met6, met4. The triangles are color coded or eachstrain and represent the decrease in methionine over each time course. TF activity corresponds to the tscore in Leeand Bussemaker (2010), which is the regression coefcient obtained rom regression o the expression data on theposition-specifc afnity matrix or the given TF (see Materials and Methods).

7/28/2019 3008.full

12/17


Cb1p and Met31p/Met32p might regulate carbohydrate metabo-lism dierently. We used a 96-well plate-reader to measure in tripli-cate the growth ocb1met6 and met31met32met6 in richmedia containing 16 dierent carbon sources (see Materials andMethods). We ound small but highly reproducible dierences ingrowth rate and/or steady-state cell density between the two strains.In particular, cb1met6 was growth impaired relative tomet31met32met6 in glucose, inositol, ructose, and mannosebut grew better than met31met32met6 in maltose and galac-tose (Supplemental Figure S6A). In additional experiments using thesame panel o TF mutants in a MET6 background, we ound thatcb1, cb1met31, and cb1met32 cells survive ar longer thanmet6 and met31met32 during Met starvation when the carbonsource is galactose (unpublished data).

Cb1p may cooperate with Met31p and Met32p to regulatecell cycle progressionThe literature is rich in evidence that methionine abundance regu-lates cell cycle progression. Unlike other auxotrophs, methionineauxotrophs arrest in the G0/G1 phase o the cell cycle during methi-onine starvation. To understand whether any particular Met TF or TFcombination is required or arrest, we measured bud index (the per-

centage o cells lacking a bud due to arrest in G0/G1) and cell den-sity concurrently with mRNA expression in each o the methionine-depletion experiments described earlier. Bud index diers amongstrains across the length o the time course, even though each timecourse was started at the same cell density (an early exponential-phaseKlett o22). Consistent with previous ndings (Unger and Hartwell,1976; Petti et al., 2011), met6 cells arrest eciently by 6 h (average6-h bud index is 96%), whereas met4 cells do not (6-h bud index,82%; Supplemental Figure S6B). Note that the 6-h bud index ormet6 is comparable to the nal, steady-state bud index ormet6measured over 27 h in (Petti et al., 2011). The consistently lower budindex o met4 was recapitulated in cb1met31met6 andcb1met32met6, whose 6-h bud index is signicantly lower that

omet6 (p= 2.5 10

5, Students t-test) but not in cb1met6,met31met6, met32met6, ormet31met32met6. Thus cellcycle arrest is impaired at all methionine concentrations when mem-bers o both classes o TFCb1p and Met31p/Met32paredeleted.

This result is consistent with the TF activity analysis, whichshows that Cb1p cooperates with Met31p and Met32p to regu-late cell cycle genes (Figure 9, N and O). To understand in greaterdetail which cell cycle genes are regulated by Cb1p/Met31p/Met32p, we examined the expression o genes associated withthe cell cycle GO term, paying particular attention to those thatare repressed during methionine starvation o met6 (Supple-mental Figure S1, clusters 3, 4, and 19). Consistent with the

TF activity analysis, many cell cycle genes repressed in met6

are somewhat less repressed in cb1met31met6 andcb1met32met6. However, it was dicult to identiy particulargenes that might contribute to the arrest deect, so we undertooka broader search or genes that are 1) potentially involved in cellcycle regulation and 2) expressed dierently incb1met31met6and cb1met32met6 than in the other strains. We ound 17genes that meet these criteria (Supplemental Figure S6C), includ-ing nine that are less repressed (or more activated) and 8 that areless activated (or more repressed) in cb1met31met6 andcb1met32met6. The ormer includes HHT1, HTA2, HTB1,HTB2, HHF2, YOX1, SCW10, GIC1, and PCL1, and the latter in-cludes YLR112W, FAR1, PRM1, GIP1, BMH1, GLC8, CDC53, andMBF1. Two genesBMH1 and MBF1contain Cb1p-binding

Consistent with the TF-dependency classications given earlier, TFactivity reveals that Cb1p is negligible compared with Met31p/Met32p in methionine and sulur metabolism: whereas deletion oCBF1 barely aects the methionine regulon, represented by theTFs Met4p, Met31p, Met32p, Cb1p, and Met28p in Figure 9A,double deletion oMET31 and MET32nearly obliterates the regu-lons response to methionine depletion. met31 and met32 alsodier in their eects on this regulon: met31 appears to increaseand met32 to decrease the steady-state expression level o thisregulon, consistent with the dierences between Met31p andMet32p identied earlier using regression. In contrast, the activityo the ribosomal regulon TFs appears to depend almost entirelyon CBF1 (Figure 9I and Supplemental Figure S5, cluster 3). Al-though these activities are all negative during the starvation, de-leting CBF1 eliminates the time dependence seen in a CBF1 back-ground. Striking dierences between Met31p and Met32p canalso be seen in the regulons controlled by Cb1p, Tye7p, Mig1p,and Pho4p. Subtler dierences in the respiration regulons (Hap2p,Hap3p, and Hap5p), and the Ino2p and Ino4p regulons are alsoseen (Figure 9, J and M).

Consistent with Figure 3, a striking number o regulons dependin opposite ways on Cb1p and Met31p/Met32p (Figure 9, B, D, E,

H, and M). Another unexpected result is the requency with whichMet31p/Met32p behaves as a transcriptional repressor, as indicatedby high time-zero expression levels in met31met32met6. Basedon exponential-phase TF activities (time 0), Met31p/Met32prepresses, and Cb1p activates, the iron regulons represented byRcs1p (At1p) and At2p, the copper regulon represented by Mac1p,the carbohydrate metabolism regulons represented by Tye7p andMig1p, the Pho4p regulon, and the phospholipid biosynthesis regu-lons represented by Ino2p and Ino4p. Although our detailed analy-ses identied individual Cb1p-dependent and Met31p/Met32p-dependent genes associated with these processes, ew genes wereregulated by both Cb1p and Met31p/Met32p. This suggests thatCb1p and Met31p/Met32p exert opposing eects on dierent

genes that participate in these processes.Synergy between Cb1p and either Met31p or Met32p is re-

fected in regulons that are perturbed (relative to met6) only incb1met31met6 and cb1met32met6. Specically, Met31p/Met32p and Cb1p cooperate to regulate respiratory gene expres-sion (Hap2p, Hap3p, and Hap5p), response to oxidative stress(Stb5p), and the cell cycle (Stb1p, Swi4p, Swi6p, and Mbp1p), aswell as Dig1p, Ste12p, and Mcm1p (Figure 9, J, K, N, and O).

The TF activity proles highlight two interesting aspects o TF-binding specicity. First, the Cb1p moti is almost identical to theTye7p moti, as indicated by the strong correlation o Cb1p andTye7p activity in Figure 9B. This could be a biologically irrelevantcoincidence, or, as discussed later, it could support the previously

observed connections between sulur/methionine metabolism andcarbohydrate metabolism. Second, the activity proles in Figure9A show that the binding proles o Met31p and Met32p dierslightly. The moti source used in this analysis (MacIsaac et al.,2006) reports the Met31p and Met32p motis as GTGTGG andAACTGTGGC, respectively. As discussed later, the Met31p motiin (MacIsaac et al., 2006) is similar to the core variant that wederived rom genes that dier in expression betweenmet31met6and met32met6.

Cb1 and Met31/Met32 have distinct roles in carbohydratemetabolismMultiple lines o evidence, as described earlier, suggest that Metabundance regulates carbohydrate metabolism and, moreover, that

7/28/2019 3008.full

13/17


motis and none contains the Met31p/Met32p moti. However,given that cb1met31met6 and cb1met32met6 strainsdier rom met6 strains in their ailure to arrest the cell cycle, wecannot rule out the possibility that the dierences we detect maybe the eect, rather than the cause, o the arrest.

DISCUSSIONThe motivation or this study was to understand whether and howcombinatorial transcriptional regulation enables the coordination odiverse cellular processes with the sulur assimilation pathway, whichsynthesizes methionine. Methionine abundance infuences manybiological processes and is regulated by a relatively complex circuitinvolving three DNA-binding proteinsMet31p, Met32p, andCb1p. We hypothesized that there exists among these TFs a divi-sion o labor that acilitates the coordination o diverse processes.We collected a large set o data in which we measured genome-wide gene expression proles under conditions that perturb theMet pathway: namely, starvation o a methionine auxotroph. Weused a MET6 (methionine synthetase) deletion mutant as our con-trol strain and compared this to a series o double- and triple-dele-tion mutants in which MET31, MET32, and/orCBF1 were also de-leted. Data were obtained over a time series beginning in exponential

growth phase and continuing beyond the cessation o growthcaused by methionine starvation. We used a variety o new and es-tablished computational methods to investigate whether and howthe TFs dier rom each other, including two main complementaryapproaches. One, TF-dependency analysis, quanties thedependence o each gene on each TF and characterizes the result-ing gene classes according to biological unction and DNA-bindingmotis; the second, TF activity analysis, measures the aggregate be-havior o genes in known regulons. As described, both approachesyield similar general conclusions.

First, we identied biological processes that were not previouslyknown to be regulated by the Met TFs and are not typically associ-ated with sulur or Met metabolism. Second, we ound clear unc-

tional distinctions between Met31p/Met32p and Cb1p that sup-port our hypothesis that there exists a division o labor among theseTFs, including a clear biological distinction between the numerousprocesses activated by Met31p/Met32p and/or repressed by Cb1p,on one hand, and activated by Cb1p and/or repressed by Met31p/Met32p, on the other (Figure 2). Third, we ound strong evidence orboth positive (activator) and negative (repressor) activities or theMet TFs. Finally, we were able to distinguish the activities oMET31and MET32rom one another and identied a variant o the Met31p/Met32p TFBM that appears to be specic to Met32p. The arrange-ment we observedin which dierent transcription actors regulateoverlapping but distinct genes and processesis analogous to thedense overlapping regulon (DOR) described in Shen-Orr et al.

(2002). The DOR has been hypothesized to unction in the coordina-tion o diverse cellular processes.

How is combinatorial regulation accomplished?Our results suggest that methionine biosynthesis is coordinatedwith other biological processes through the use o two compara-tively methionine-specic transcription actorsMet31p andMet32pand a generalist transcription actorCb1p. Althoughall three TFs regulate a wide variety o cellular processes, Met31pand Met32p are more important in methionine metabolism thanCb1p, whose regulatory roles appear to be more evenly distributedacross cellular metabolism and other processes such as sporulationand translation. The metabolism-related regulation we ound issummarized in Figure 10, in which direct and indirect TF targets are

superimposed on a skeletal backbone o metabolism. Clearly,Met31p and Met32p regulate every gene in the methionine biosyn-thetic pathway, whereas Cb1 only regulatesjointly with Met31p/Met32pthe sulate importers, MET14, and the two steps that de-pend on NADPH (MET16, MET10, and ECM17).

This diagram also highlights the broad infuence o the Met reg-ulators over central metabolism, including phosphatidylcholinebiosynthesis (PSD1, CHO1); cardiolipin biosynthesis (PGS), pyrimi-dine biosynthesis (URA5); tryptophan (TRP3) and NAD biosynthesis(BNA2,3,4, 5, and 6); central metabolism (PDC6, ADH3, PYK2,IDH1, SDH1, SDH2, DAL7); atty acid biosynthesis (FAS2, MCT1,HFA1); glutathione metabolism (GDH3); lipid metabolism leadingto phosphatidyl choline (SER33); nitrogen metabolism (IDP1, GDH3,GLT1); allantoin degradation (DAL1, DAL2, DAL3); purine biosyn-thesis (IMD2); and the production o deoxyribonucleotides (RNR2,RNR4).

Dierent ways o implementing combinatorial regulation are evi-dent rom Figure 10. Many metabolic pathways are jointly regulatedby Cb1p and Met31p/Met32p. Sometimes this is accomplished bya joint target (BNA3, SER33, PDC6, and many genes in methioninebiosynthesis itsel) and sometimes by two dierent targets in thesame pathway (PSD1 and CHO1; BNA2and BNA4, BNA5, BNA6).

Clearly, there must be other ways to accomplish joint regulation thatwere missed by our experimental design. For instance, SAM2 isjointly regulated by Opi1p and Met4p, and its promoter containsbinding sites or Met31p/Met32p, Cb1p, and Opi1p (Hickmanet al., 2011). However, we ound no dependence oSAM2on Cb1p.An article by McIsaac et al. (2011) explores an independent ap-proach to regulation by inducing the Met pathway TFs one at a timeThese experiments show that CBF1 induction causes SAM2repres-sion, suggesting that many (i not all) o the regulatory eects wemay have missed in the starvation experiments can be ound with analternative approach.

Methionine-dependent processes refect the role o sulur in

diverse electron transer reactionsRedox reactions are central to the mitochondrial electron transerchain, which contains proteins rich in Fe-S clusters (Voet and Voet,2010). Our results show that Met31p/Met32p represses genes in-volved in iron homeostasis (particularly iron transport) and Fe-S clus-ter biogenesis through both direct and indirect mechanisms. In sup-port o direct repression, several iron-homeostasis genes and onegene involved in Fe-S biogenesis depend on MET/MET32 and con-tain the Met31p/Met32p TFBM. In support o indirect repression,we derived the At1p TFBM rom genes that are signicantly in-duced in met31met32met6. At1p induces iron-related geneexpression specically in response to deects in Fe-S cluster biogen-esis (Chen et al., 2004). Thus, deletion oMET31 and MET32ap-

pears to cause deects in Fe-S biogenesis that ultimately activateAt1p and the iron-homeostasis genes under its control. Taken to-gether, these results suggest that methionine (or sulur) deprivation,which activates Met31p and Met32p (via Met4p), may concomi-tantly repress iron import in order to refect the lack o sulur atomsavailable or Fe-S biogenesis. The electron transport chain also con-tains two copper atoms in complex IV. Consistent with this, Met31p/Met32p-dependent genes are enriched or copper ion homeostasis,suggesting that Met31p/Met32p more generally regulates theabundance o electron transport chain metals in response to suluravailability.

Sulur is also critical to the redox reactions central to the activityo antioxidants such as glutathione and glutaredoxin (SupplementalFigure S7). We ound that many antioxidant biosynthetic genes are

7/28/2019 3008.full

14/17

7/28/2019 3008.full

15/17


For analysis o the met6data in isolation, we used the 2669 genesthat change by twoold or more in the met6 time course, dependsignicantly on time (q < 0.01, F-test), and behave similarly in met6and met13. For analysis o the total data set, we used the 3671genes that change by twoold or more in at least one strain and either1) depend signicantly on time (p 0.05, F-test) or 2) are constitutivelyexpressed twoold relative to met6 (p 0.05, t-test). All regression,correlation, and randomization analyses used in this work were codedin Matlab, version R2011a (The MathWorks, Natick, MA).

We developed a correlation-based method that we reer to as TF-dependency analysis to identiy genes whose wild-type expression,as measured in met6, depends on Cb1 only (case 1), Met31/Met32only (case 2), or both (case 3). First, an articial gene expression tem-plate was constructed to match each case. For instance, the templateor case 1 refects that, or a gene that depends on Cb1 but notMet31/Met32, expression levels in cb1 dier rom those in met6and met31met32met6. The three templates are as ollows:

Case 1: met6 = 0 (at all time points), cb1met6 = 1,met31met32met6= 0.



Second, the Pearson correlation between each gene and eachtemplate was calculated. A bootstrapped p-value or each correla-tion coecient was calculated by repeating the correlation calcula-tion on 1 104 randomly permuted sets o expression data anddetermining the raction o random correlation coecients thatexceeded the real r in absolute value. Q-VALUE sotware wasused to calculate q-values rom the p-values (Storey and Tibshirani,2003). Finally, a gene was assigned to a particular TF-dependencyclass i the q-value or the correlation between the gene and thetemplate was at most 0.05 (corresponding to p 0.007).

Pairwise comparison o strains (comparison omet31met6 tomet32met6 and comparison o cb1met31met6 tocb1met32met6) was perormed using the regression modelspecied in Eqs. 1 and 2, as well as a two-tailed Students t-test. Inthe ull regression model (Eq. 1, Dclassies the genotype with re-spect to met31 and met32 (orcb1met31 and cb1met32),such that D= 1 ormet32 (orcb1met32) and D= 0 ormet31(orcb1met31). Regression signicance was calculated rom the Fstatistic comparing the t o the ull model (Eq. 1) with the t o thereduced model (Eq. 2). Pairwise comparison o strains or the syn-ergy analysis was perormed analogously, using a regression modelthat allowed or simultaneous pairwise comparisons between morethan two strains. Nonparametric p-values were calculated or the F

statistic and the tstatistic by randomly permuting the expressiondata 1 105 times. Q-VALUE sotware was used to calculate q-valuesrom the p-values (Storey and Tibshirani, 2003). A gene was consid-ered to dier between two strains i the q-value or the Fstatistic orthe tstatistic was

7/28/2019 3008.full

16/17


desired carbon source in a fat-bottomed, polystyrene 96-well platecovered with a Breathe-Easy membrane (Sigma-Aldrich, St. Louis,MO). Carbon sources were supplied at a nal concentration o 2%.Each strain was grown in triplicate in each carbon source with con-tinuous shaking at 30C, and the OD600 was recorded every 30 minor 70 h.

o values ork. The value giving the largest raction o unctionallyenriched clusters and best discrimination between distinct expres-sion proles was used in subsequent analyses.

Functional enrichment o gene clusters was measured with respectto the Biological Process classications specied in the GO(Ashburneret al., 2000). Enrichment or GO terms was measured us-ing the GOTF (Boyle et al., 2004; available at http://go.princeton.edu/cgi-bin/GOTermFinder), using the FDR option or multiple hypothesiscorrection. Here we report enrichments with FDR at most 0.1.

DNA sequence motis were identied using the Web-based mo-ti-detection algorithm MEME (http://meme.sdsc.edu/meme/intro.html; Bailey et al., 2009). Beore using MEME, Regulatory SequenceAnalysis Tools (http://rsat.ulb.ac.be) were used to obtain the pro-moter sequence, ranging rom positions 800 to 1 and excludingoverlapping open reading rames, upstream o every gene (Thomas-Chollieret al., 2011). This genome-wide set o promoter sequenceswas used to generate the nucleotide-requency background modelused by MEME. In MEME, moti width was allowed to range rom 6to 12 nucleotides. All other MEME parameters were set to deaultvalues. To identiy additional instances o each moti throughout thegenome, the motis identied by MEME were used as input or thecompanion program MAST (http://meme.sdsc.edu/meme/cgi-bin/

mast.cgi). MAST takes as input a list o motis represented asposition-specic scoring matrices (such as those produced byMEME) and searches a user-specied sequence database (here theS. cerevisiae upstream sequence database supplied with the MEMEsuite) or instances o those motis. MAST assigns each input se-quence an Evalue, which refects the expected number o se-quences in a random sequence database that match the input mo-tis at least as well as the actual sequence. Although the Evalue isnot a p-value, it is calculated rom the p-values o the moti matchesin the input sequence. We applied two dierent Evalue cutos,depending on the nature o the analysis: or MAST hits that weresubsequently ltered using additional criteria (e.g., those in Figure1), we used hits with an Evalue o

7/28/2019 3008.full

17/17

Rupp S, Summers E, Lo H-J, Madhani H, Fink G (1999). MAP kinase andcAMP lamentation signaling pathways converge on the unusually largepromoter o the yeast FLO11 gene. EMBO J 18, 12571269.

Saeed AI, Bhagabati NK, Braisted JC, Liang W, Sharov V, Howe EA, Li J,Thiagarajan M, White JA, Quackenbush J (2006). TM4 microarray sot-ware suite. Methods Enzymol 411, 134193.

Shen-Orr SS, Milo R, Mangan S, Alon U (2002). Network motis in thetranscriptional regulation network oEscherichia coli. Nat Genet 31,6468.

Siggers T, Duyzend MH, Reddy J, Khan S, Bulyk ML (2011). Non-DNA-binding coactors enhance DNA-binding specicity o a transcriptionalregulatory complex. Mol Syst Biol 7, 114.

Smith JJ, Ramsey SA, Marelli M, Marzol B, Hwang D, Saleem RA,Rachubinski RA, Aitchison JD (2007). Transcriptional responses to attyacid are coordinated by combinatorial control. Mol Syst Biol 3, 115.

Spellman PT, Sherlock G, Zhang MQ, Iyer VR, Anders K, Eisen MB, BrownPO, Botstein D, Futcher B (1998). Comprehensive identication ocell cycle-regulated genes o the yeast Saccharomyces cerevisiae bymicroarray hybridization. Mol Biol Cell 9, 32733297.

Storey JD, Tibshirani R (2003). Statistical signicance or genomewide stud-ies. Proc Natl Acad Sci USA 100, 94409445.

Su N-Y, Ouni I, Papagiannis CV, Kaiser P (2008). A dominant suppressormutation o the met30 cell cycle deect suggests regulation o the Sac-charomyces cerevisiae Met4-Cb1 transcription complex by Met32. JBiol Chem 283, 1161511624.

Thomas D, Surdin-Kerjan Y (1997). Metabolism o sulur amino acids inSaccharomyces cerevisiae. Microbiol Mol Biol Rev 61, 503532.

Thomas-Chollier M, Derance M, Medina-Rivera A, Sand O, Herrmann C,Thiery D, van Helden J (2011). RSAT 2011: regulatory sequence analy-sis tools. Nucleic Acids Res 39, W86W91.

Tu BP, Kudlicki A, Rowicka M, McKnight SL (2005). Logic o the yeastmetabolic cycle: temporal compartmentalization o cellular processes.Science 310, 11521158.

Unger MW, Hartwell LH (1976). Control o cell division in Saccharomycescerevisiae by methionyl-tRNA. Proc Natl Acad Sci USA 73, 16641668.

Voet D, Voet JG (2010). Biochemistry, New York: John Wiley and Sons.Yamaguchi-Iwai Y, Dancis A, Klausner RD (1995). AFT1: a mediator o iron

regulated transcriptional control in Saccharomyces cerevisiae. EMBO J14, 12311239.

Zhu C et al. (2009). High-resolution DNA binding specicity analysis o yeasttranscription actors. Genome Res 19, 556566.

Kuras L, Thomas D (1995b). Identication o the yeast methionine bio-synthetic genes that require the centromere binding actor 1 or theirtranscriptional activation. FEBS Lett 367, 1518.

Lavoie H, Hogues H, Mallick J, Sellam A, Nantel A, Whiteway M (2010).Evolutionary tinkering with conserved components o a transcriptionalregulatory network. PLoS Biol 8, e1000329.

Lee E, Bussemaker HJ (2010). Identiying the genetic determinants o tran-scription actor activity. Mol Syst Biol 6, 412.

Lee TA, Jorgensen P, Bognar AL, Peyraud C, Thomas D, Tyers M (2010).Dissection o combinatorial control by the Met4 transcriptional complex.Mol Biol Cell 21, 456469.

Lee TI et al. (2002). Transcriptional regulatory networks in Saccharomycescerevisiae. Science 298, 799804.

MacIsaac KD, Wang T, Gordon DB, Giord DK, Stormo GD, Fraenkel E(2006). An improved map o conserved regulatory sites orSaccharomy-ces cerevisiae. BMC Bioinormatics 7, 113.

McIsaac RS, Petti AA, Bussemaker HJ, Botstein D (2012). Perturbation-based analysis and modeling o combinatorial regulation in the yeastsulur assimilation pathway. Mol Biol Cell 23, 29933007.

McIsaac RS, Silverman SJ, Mcclean MN, Gibney PA, Macinskas J, HickmanMJ, Petti AA, Botstein D (2011). Fast-acting and nearly gratuitousinduction o gene expression and protein depletion in Saccharomycescerevisiae. Mol Biol Cell 22, 44474459.

Milo R, Shen-Orr S, Itzkovitz S, Kashtan N, Chklovskii D, Alon U (2002).Network motis: simple building blocks o complex networks. Science298, 824827.

Murray DB, Beckmann M, Kitano H (2007). Regulation o yeast oscillatory

dynamics. Proc Natl Acad Sci USA 104, 22412246.Myers CL, Troyanskaya OG (2007). Context-sensitive data integration and

prediction o biological networks. Bioinormatics 23, 23222330.Patton EE, Peyraud C, Rouillon A, Surdin-Kerjan Y, Tyers M, Thomas D

(2000). SCF(Met30)-mediated control o the transcriptional activatorMet4 is required or the G(1)-S transition. EMBO J 19, 16131624.

Petti AA, Crutcheld CA, Rabinowitz JD, Botstein D (2011). Survival o starv-ing yeast is correlated with oxidative stress response and nonrespiratorymitochondrial unction. Proc Natl Acad Sci USA 108, E1089E1098.

Pilpel Y, Sudarsanam P, Church GM (2001). Identiying regulatory networksby combinatorial analysis o promoter elements. Nat Genet 29, 153159.

Rouillon A, Barbey R, Patton EE, Tyers M, Thomas D (2000). Feedback-regulated degradation o the transcriptional activator Met4 is triggeredby the SCFMet30 complex. EMBO J 19, 282294.

3008.full

Documents

Transcript of 3008.full