. 13: 1519–1534 (1997)

Two-dimensional Electrophoretic Separation of YeastProteins using a Non-linear Wide Range (pH 3–10)Immobilized pH Gradient in the First Dimension;Reproducibility and Evidence for Isoelectric Focusing ofAlkaline (pI>7) Proteins

JOAKIM NORBECK AND ANDERS BLOMBERG*

General and Marine Microbiology, Goteborg University, Medicinaregatan 9C, 413 90 Goteborg, Sweden

Received 24 June 1996; accepted 8 July 1997

The proteome of the yeast Saccharomyces cerevisiae was analysed by two-dimensional (2D) polyacrylamide gelelectrophoresis utilizing a non-linear immobilized pH gradient (3–10) in the first-dimensional separation. Cells werelabelled by [35S]methionine incorporation in the respiro-fermentative phase during exponential growth on glucose.Gels were run, visualized with phosphoimager technology and all resolved proteins automatically quantified.Proteins were well resolved over the whole pH interval, and evidence for isoelectric focusing on the basic side of thepattern was generated by sequencing of some spots, revealing the 2D positions of Tef1p, Pgk1p, Gpm1p, Tdh1p andShm2p. Roughly 25% of the spots were resolved at the alkaline side of the pattern (pI>7). The positionreproducibility was high and in the range 1–2 mm in the x-and y-dimension, respectively. No quantitative variationwas linked to a certain size or charge class of resolved proteins, and the average quantitative standard deviation was17&11%. The obtained immobilized pH gradient based pattern could easily be compared to the old ampholine-based 2D pattern, and the previously reported identifications could thus be transferred. Our yeast pattern currentlycontains 43 known proteins, all identified by protein sequencing. Utilizing these identified proteins, relevant pI andMr scales in the pattern were constructed. Normalization of the expression of identified spots by compensating forthe number of methionine residues a protein contains allowed stoichiometric comparisons. The most dominantproteins under these growth conditions were Tdh3p, Fba1p, Eno2p and Tef1p/Tef2p, all being expressed at morethan 500 000 copies per cell. The differential carbon source response during exponential growth on either glucose,galactose or ethanol was examined for the alkaline proteins identified by micro-sequencing in this study. ? 1997John Wiley & Sons, Ltd.

Yeast 13: 1519–1534, 1997.

— proteome analysis; 2D-PAGE; protein identification; protein expression; S. cerevisiae

INTRODUCTION

The 24th of April 1996 demarcates a new era instudies on the molecular physiology of eukaryotic

cells; it was the date of the public release ofthe fully sequenced 13 Mb genome of the yeastSaccharomyces cerevisiae. Thus, for the first timethe complete protein repertoire of a eukaryotic cellwas accessible, information which surely will haveimportant implications on general cell biology,biotechnology as well as medicine. In order toevaluate this impressive amount of DNA/protein

*Correspondence to: Anders Blomberg; E-mail:anders.blomberg@gmm.gu.seContract grant sponsor: Swedish National Board for NaturalScience Contract grant sponsor: Swedish National Board forTechnical Development.

CCC 0749–503X/96/161519–16 $17.50? 1997 John Wiley & Sons Ltd

sequence data, softwares have been developedthat combine an array of search algorithms withthe information in a number of databases (Casariet al., 1995). One of the most striking outcomes ofthese analyses is that roughly 35% of the yeastgenome encodes proteins of currently unknownfunction, i.e. FUN-genes (Casari et al., 1996).These tentative open reading frames code for pro-teins which display no homology to presentlyidentified proteins, thus, they highlight the vastamount of biological information that awaits to beunravelled even in a ‘simple’ unicellular organismlike yeast. For proteins where functional in-formation is available, about half are linked tofunction by direct experimental evidence in yeast,while the data about the remaining half stem frominter-species sequence homology and transfer offunctional information. The most recent up-dated figure on the proportion of FUN-genes canbe found at the www-site ‘gene-crunch’ (http://www.sander.ebi.ac.uk/genequiz).The completion and release of the yeast ge-

nome sequence will also make a drastic impacton the technical side of molecular biology onyeast, since it partly ends the era of cloning andsequencing of novel genes in this organism.Surely, cloning by complementation or multicopysuppression of interesting mutants will persist;however, sequencing of DNA will be restricted tosequence tags used in database searches or confir-mation of DNA constructs. Instead we will entera phase of research, the focus of which will bethe functionality of gene products, especiallythose without known function or those whosesequence homology only provides vague hints totheir functionality (Oliver, 1996). Protein-directedstudies involving enzyme purification, character-ization and subsequent sequencing will make asubstantial contribution in deciphering the linkbetween activities and genes, and by this ap-proach the functionality of many FUN-genes willbe revealed (e.g. Norbeck et al., 1996). However,not only information regarding the catalytic func-tion of a protein will be of interest, but also thephysical or metabolic link between a particularprotein and other cellular functions. One type ofexperimental data offering clues to functionalityis the impact a disruption or overexpression of aFUN-gene will have on the production of otherproteins. The logic behind this is that mutationsin functionally related proteins will affect theexpression of each other. This has been tested fora number of genes and in about 50% of cases an

expression effect on the global protein patternhas been observed (Kahn, 1995).A technology well suited for these types of

global analysis of proteins is two-dimensionalpolyacrylamide gel electrophoresis (2D-PAGE),described initially in the mid-seventies (Klose,1975; O’Farrell, 1975), which separates proteinsaccording to their pI and size. Originally thefirst-dimensional charge separation was achievedin a pH gradient generated by ampholines, atechnology that in the eighties was refined to utilizestable and easily designed immobilized pH gradi-ents (IPGs; Gorg et al., 1988). Studies on theexpression of proteins under certain environmentalconditions (Bataille et al., 1988; Blomberg, 1995;Boucherie, 1985; Boucherie et al., 1995a; Milleret al., 1982; Norbeck and Blomberg, 1997) or theinterlink between metabolic pathways (Norbecket al., 1997) are neatly conducted with the aid of2D-PAGE, since this technique not only resolves avast proportion of all tentative gene products, butalso via computerized image analysis will yieldinformation about the level of expression of allresolved proteins (Garrels, 1989). In a total yeastextract roughly 2000 proteins can be resolved,which will constitute a good representation of thegenes expressed. In a gel of a total cell extractnormally only the most abundant proteins will bedetected with a codon bias in the range 0·4–1·0.However, many more proteins can be detected ifpre-fractionation techniques are employed. Identi-fication of 2D-PAGE resolved yeast proteins tobiochemical function or linkage to their corre-sponding genes has been conducted utilizing anumber of methodologies (Bataille et al., 1987;Boucherie et al., 1995b; Garrels et al., 1994;Ludwig et al., 1982; Norbeck and Blomberg, 1995,1996; Norbeck et al., 1996). The ampholine-based2D technology still struggles with two problems: (i)the non-standard pattern and (ii) the lack ofresolution on the basic side of the pH gradient.The first point will be accurately addressed withthe IPG technology, and its applicability has alsobeen tested in an interlaboratory study concerningthe pattern reproducibility (Blomberg et al., 1995).It is also clear that proteins with pI>7 will bemissed in the ampholine pattern, and that theseproteins constitute a high proportion of all tenta-tive yeast gene products (Boucherie et al., 1995b).We report here on the separation of yeast pro-

teins in an immobilized wide range non-linearpH gradient (pH 3–10). The pattern obtainedreveals good resolution over the entire range and

1520 . .

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (1997)

micro-sequencing identified 2D resolved proteinswith theoretical pIs between 7 and 9.1. The ob-tained pattern also exhibited strong similarities tothe previously used ampholine-driven 2D systems,as well as good position and quantitative repro-ducibility. We thus believe that this system, orslight variations to it, could be a long-lasting 2Dsystem for the construction of yeast proteomeexpression databases, of which one currentlyis in operation at our www-server (http://yeast-2dpage.gmm.gu.se).

MATERIALS AND METHODS

Strain, growth conditions, radioactive labelling andpreparation of cell extract

The diploid strain SKQ2n of S. cerevisiae (a/á;ade1/+; +/ade2; +/his1) was grown at 30)C insynthetic medium containing 2% (w/v) glucose, 2%(w/v) galactose or 2% (w/v) ethanol as carbon andenergy source. Growth and labelling in the mid-exponential phase (about 5#106 cells/ml) as wellas the subsequent preparation of cell extract was aspreviously reported (Blomberg et al., 1995).

High-resolution 2D-PAGE with IPG in the firstdimensionAll chemicals used were of electrophoresis grade

and supplied by Oxford Glycosystem, exceptwhere indicated. Isoelectric focusing on IPGs wasessentially performed as described earlier (Gorg,1994) with the following modifications. The IPGstrips were non-linear pH 3–10 (0·5 mm thin) caston GelBond PAGfilm and 3 mm wide (Pharmacia,nr. 17-1235-01). Before use the IPG strips wererehydrated overnight at room temperature in 8 -urea, 1% (v/v) NP-40, 13 m-DTT, and 1% (v/v)Pharmalyte (pH range 3–10; Pharmacia, nr. 17-0456-01) and a trace of Bromophenol Blue,~0·01% (w/v). The protein extract (20 ìl persample containing 1#106 DPM and about 3 ìgprotein) was applied to the IPG strips using thesample cups and gels were run at 20)C under oilaccording to the manufacturer’s recommendationfor about 40 000 volt hours with the followingprogram: linear gradient 0–500 V, 5 h; 500 V, 5 h;linear gradient 500–3500 V, 9.5 h, 3500 V, 5 h.Extending or decreasing the number of volt hoursby altering the time at the final 3500 V setting didnot change the obtained 2D pattern. For prepara-tive gels 3500 V were prolonged to 48 h. The strips

were frozen at "80)C prior to application to thesecond dimension.The second-dimension acrylamide gels were 10

or 12·5% T, 2·1% C (Duracryl 0·65% Bis; ELCR2DC 070) containing 0·1% (w/v) SDS, 0·37 -Tris-base and 0·27 -Tris–HCl. Gels were cast at roomtemperature and samples run on a vertical systemat 20)C (Investigator; Oxford Glycosystems)with gels fully submerged in running buffer forefficient cooling. First-dimensional IPG strips wereequilibrated for 30 min in 10 ml of equilibrationsolution containing 3% (w/v) SDS, 50 m-DTT,0·3 -Tris-base, 0·075 -Tris–HCL and 0·01%(w/v) Bromophenol Blue. The strips wereloaded onto the second dimension through a 65)Csolution containing 0·5% (w/v) agarose (Sea-Plaque; FMC) dissolved in gel buffer. The runningbuffer was 25 m-Tris-base, 0·1% (w/v) SDS and192 m-glycine with the upper tank containingapproximately 2 litres of 2# concentrated runningbuffer. Electrophoresis in the second dimensionwas performed at limiting power of 16 000 mW(max voltage 500 V) per gel for about 5 h until thedye front reached the bottom of the gel. Immedi-ately after the run the gels were dried on filter-paper without any previous fixation. The gelsurface was covered by plastic film during thedrying process.

Visualization and image analysisDried gels were exposed to image plates for

roughly 72 h and subsequently scanned in a phos-phoimager (Molecular Dynamics) at a resolutionof 176#176 ìm. Image files were transferred to asparc 1 station (SUNMicrosystems) and raw scanswere processed by the 2D software PDQuest (PDI)according to the following procedure: make ODimage, average smooth, Fourier smooth (4#4)and ball background subtraction (wheel 200),spots detected and fitted to Gaussian volumes.Spots allocated in the front or at the very top ofthe gels were erased manually. For some of themore dominant spots in some gels multiple spotcentres had been positioned and these were manu-ally corrected into one single spot centre. Nofurther processing of the gels was performed. Thedifferent gel patterns were by visual comparisonmanually matched to each other (landmark func-tion). These matched spots were compared as totheir x- and y-position and their quantities. Datawere exported as spreadsheet data (export datafunction) and imported to Excel v. 3·0 (Microsoft)

1521-

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (1997)

for further comparisons. Individual quantificationof proteins resolved were normalized to the totalamount of radioactivity in all quantified spots inthe gel. Systematic errors in quantification betweengels were compensated for by log-normalization(Garrels, 1989), where the normalization factors indifferent gels varied between 1·05 and 0·93. A pIand Mr scale for the whole gel was constructedwith the PDQuest software utilizing the selectedstandard proteins (Table 2, Figure 5; see Resultssection for details on the selection). Figure presen-tations were made from raw scans via Tiff-imagesproduced by the PDQuest software, as previouslydescribed (Blomberg et al., 1995).

Calculation of number of actins per cellCalculation of number of actin molecules per

cell was based on the data provided from actinpurification (Nefsky and Bretscher, 1992). From100 g packed cells (equivalent to 8#1011 cells) ayield of 1·2 mg of pure actin was obtained. Therecovery in the purification process was 36%,which, together with a molecular weight of41·7 kDa, resulted in an estimate of 60 000 actinmolecules per cell.

Microsequencing of resolved spotsPrior to sequencing, 2D-resolved protein were

in-gel digested with trypsin (Promega SDS) andgenerated peptides automatically fractionated onreversed-phase HPLC using a SMART system(Pharmacia, Uppsala; Norbeck and Blomberg,1995). The N-terminal sequence analysis on thegenerated peptides was performed on a pulsed-liquid sequencer 473A (Applied Biosystems, NewYork).

RESULTS

Resolving potential of alkaline proteins (pI>7) inthe pH 3–10 immobilized gradientWhole-cell protein extracts of exponentially

growing cultures of S. cerevisiae, labelled by incor-poration of [35S]methionine, were subjected to2D-PAGE. Proteins were first resolved by isoelec-tric focusing (IEF) on a non-linear IPG, spanninga pH range of 3–10, and subsequently applied tosize separation on SDS gels. Protein extract corre-sponding to roughly 1#106 DPM was loaded atthe acidic/anodic end of the IPG-strips, and thedried 2D gels were routinely exposed to imageplates for 3 days. The gel scans were processed and

quantified automatically by the image analysissoftware PDQuest. Prior to gel comparison, spotsallocated in the front or at the top of the gels wereerased. No major manual editing of the gel imageswas performed for the low quantity spots or spotsin streaks, even though for these spot-categoriesthe computer algorithms differently detected spotsin different gels. Manual editing of these spots,which would have been rather time consuming,would have greatly improved the spot detection forthese spot-categories. The generated gel imageswere compiled into a matchset (one gel imagechosen as the reference), which was the unit for thesubsequent analyses.The total number of spots resolved and quanti-

fied in each gel by this procedure, when the sameprotein extract was applied to five gels, was onaverage 1088&72 (SD; n=5), covering a 1000-foldexpression range encompassing 50–50 000 ppm. Byapplying higher amounts of radioactivity, or pro-longing the exposure time, about 2000 spots werelocated in the gels and the quantitative rangecovered 5–50 000 ppm (Andlid et al., 1997; anddata not shown). The obtained 2D pattern re-vealed great similarities to previously published 2Dmaps of yeast proteins separated by ampholine-driven IEF; however, at the alkaline side of theelectrogram a substantial number of ‘new’ proteinswere resolved.The broken line in Figure 1 roughly indicates the

analytical limit in the ampholine gels (Boucherieet al., 1995b; Garrels et al., 1994; Norbeck andBlomberg, 1995, 1996), a border usually set at theposition of the most dominant isoform of theglycolytic enzyme glyceraldehyde 3-phosphate de-hydrogenase (Tdh3p). In addition, on the acidicside of the Tdh3 protein (pH~6·0–6·5) the resolu-tion in the ampholine gels is usually of less qualitycompared to the rest of the gel, exhibiting a highproportion of streaky spots, thus even furtherlimiting the effective range resolved. Using theIPG-based 2D technology good resolution wasobtained on the acidic as well as the alkaline sideof Tdh3p (Figure 1). The number of proteinsresolved with an apparent pI more basic thanTdh3p was on average 246&46 (SD; n=5), whichindicated that about 25% of the proteins in thewhole cell extract focused in this part of thepattern, outside the range of the ampholine IEFgels.Isoelectric focusing of proteins with alkaline

theoretical pI values in the obtained IPG-2D-PAGE pattern was substantiated by sequencing of

1522 . .

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (1997)

3

152-

tee,s-ee-ae,).

t-eer-cne-rsics.elyts-fsss,d.frlyllela.-lyrs.lyt

esde-

7)

five of the resolved proteins on the basic side(indicated in Figure 1). In order to obtain enoughprotein material for sequencing, preparative gelswere run with a load of roughly 1 mg protein pergel in the first dimension. The second-dimensionalslab gels were stained with Coomassie blue andspots of interest cut out and subjected to in-geltrypsination (Norbeck and Blomberg, 1995). Anumber of the peptides fractionated on RP-HPLCwere sequenced and these sequences unequivocallyidentified the spots as the products of the TEF1/TEF2, SHM2, PGK1, TDH1 and GPM1 genes(Table 1). All these gene products have alkaline(>7) theoretical pI values (Table 2), thus providing

evidence that the immobilized pH 3–10 gradienneatly resolved alkaline proteins by IEF. Threglycolytic enzymes resolved on the alkaline sidnamely phosphoglycerate kinase (Pgk1p), phophoglycerate mutase (Gpm1p) and one of thminor isoforms of glyceraldehyde 3-phosphate dhydrogenase (Tdh1p). This Tdh1p isoform hastheoretical pI of 8·32 and is thus supposed to blocated on the alkaline side of Tdh2p and Tdh3pboth having calculated pI values of 6·5 (Table 2

Qualitative and quantitative reproducibility of theyeast 2D patternThe position reproducibility of the protein pa

tern is of great importance in 2D-PAGE to ensurcomparison of old and new data locally in onlaboratory, but, even more so, if the data geneated is to be of any use to the whole scientificommunity. The high position reproducibility ithese IPG-2D-PAGE-generated images was rvealed in the matching process, since the numbeof landmarks which had to be added manually astarting points for the computer-based automatmatching was much less than for ampholine gelFollowing the introduction of 27 landmarks (threlandmark spots in each of nine gel portions evendistributed over the gel), almost no further spowere automatically matched when additional landmarks were introduced (Figure 2). The placing othese 27 landmarks was a fairly quick proce(roughly 15 min) and all together the analysiincluding editing, introduction of landmarks anautomatic matching, was finished within 30 minVisual inspection of the gel images after a total o53 landmarks had been introduced gave no furthematches, thus the matching process was apparentcompleted. In total, 680 spots were matched to afive gels, leaving on average 150 proteins per gunmatched to the reference gel. In no case wasmore dominant and well-focused spot mismatchedExamination of hundreds of the unmatched proteins revealed that they were almost exclusivelow abundance spots not detected by the computealgorithms or, less frequently, spots in streakThus, extensive manual editing would have easibrought the number of fully matched proteins to aleast 1000.The average position standard deviation for th

680 spots automatically matched to all gels wafound to be 1·04%&0·28 (SD; n=5) an0·81%&0·27 (SD; n=5) for the x- and thy-position, respectively. Since the physical dimen

Figure 1. Enlargements of a portion of the IPG-2D-PAGEpattern of protein extract from Saccharomyces cerevisiae, util-izing a non-linear immobilized pH (3–10) gradient in the firstdimension and a 10% acrylamide SDS gel in the seconddimension. Cells were labelled with [35S] methionine in therespiro-fermentative phase during exponential growth on mini-mal medium with glucose as carbon and energy source.Roughly 1#106 cpm (~10 ìg protein) was applied to the firstdimension and the final 2D gel was exposed to a Phospho-Imager plate for 72 h. The broken line indicates roughly theanalytical limit when using broad-range ampholytes for thefirst-dimensional isoelectric focusing (Norbeck and Blomberg,1995, 1996). Encircled proteins on the basic side have beensubjected to sequencing (Table 1) and identified by comparisonto sequences in the database: spot 1, TEF1; spot 2, SHM2; spot3, PGK1; spot 4, TDH1; spot 5, GPM1. The TDH3 spot isindicated in the figure for reference. For information about theprotein abbreviations see Table 2.

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

ee-dei-ei-ds.d-

et.,7;er-r

7)

sion of the generated pattern is about 16#20 cm,the absolute standard deviation would be in the1–2 mm range. The quantitative variation wasgreater and amounted to 17%&11 (SD; n=5);however, only 8 of the 680 proteins analysedexhibited a standard deviation greater than 50%.This quantitative variability was more or lessevenly distributed over the gel and no biases wereseen in either the pI or Mr dimension (data notshown). However, low quantity spots close to thebackground (50–150 ppm) revealed a slightlyhigher standard deviation (data not shown).

IPG pattern comparison to 2D patterns generatedusing the ampholine systemIn order to compare the 2D pattern obtained in

our ampholine-generated system (Norbeck andBlomberg, 1995, 1996, 1997) with the one resultingfrom IPG-based separation, the same protein ex-tracts were applied and analysed in both 2D sys-tems. In addition to glucose-grown cells, proteinextracts were made from cells exponentially grow-ing on ethanol. Since the expression of a large

number of proteins is affected by the carbon sourc(Bataille et al., 1988), we could in this way utilizcarbon source responders as tie-points in our comparison. A small portion of the gels (same pI anMr range) is seen in Figure 3, where some of thcarbon source responders are indicated. It is evdent from the comparison of these gels that thpatterns generated were rather similar. The domnant spots are easily recognized in the pattern, anso are the indicated carbon source responderThus, data could in this way be transferrebetween the old ampholines and the new IPGsystems.

Identified proteins and their use in defining a pIand Mr scale in the patternThe ease of pattern recognition between th

ampholine and the IPG systems made clear thaprevious spot identifications (Blomberg et al1997; Norbeck and Blomberg, 1995, 1996, 199Norbeck et al., 1996) could be transferred to thIPG system (Figure 4). These identifications curently include proteins involved in diverse cellula

Table 1. Peptide sequences obtained by micro-sequencing for some of the IPG-2D-PAGEresolved proteins on the alkaline side (pH>7) of the yeast 2D pattern.

Proteinnumbera

Amino acid sequence in one-letternotationb

Residues in identifiedproteinc Gene assigned

1 YA(W)VLD 56–61 TEF1/TEF2d

VETGVIX(P)GM 265–274XETGVIXPGMVVXFAP 265–280

2 ISAVSTYFESF 151–161 SHM2e

EVLYDLEN 274–2813 HELXXLADVYIN 149–160 PGK1

ALENPTRPFLAI 197–208VLENTEIXXXIF 244–255

4 HIIIDGV 73–79 TDH1f

VLPELQGK 217–2245 FNTYR 108–112 GPM1g

XNIPTGIPLVF(E)LD 203–216

aSee Figure 1.bLetters in brackets indicate the most likely amino acid designation from micro-sequencing (Xindicates a missing identification). Underlined amino acids correspond to differences between isogeneproducts.cData taken from the Swiss-Prot database.dThe gene products from TEF1 and TEF2 are identical in amino acid sequence and thus cannot bedistinguished.eShm1p is mitochondrial and contains signal sequence. Underlined amino acids correspond todifferences found at position 168–176 and at position 290–297 in the SHM1 product.fI76 is V in both Tdh2p and Tdh3p. V79 is K in Tdh3p and H in Tdh2p. Sequence 217–224 is also foundin Tdh2p and Tdh3p.gNo sequence similarities found to proteins designated Gpm2p (TDL021w) and Gpm3p (YOL056w).

1524 . .

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

ge

e7

th

7)

Table 2. 2D resolved proteins identified by in-gel trypsination and sequencing of generated peptides (Blomberet al., 1997; Norbeck and Blomberg, 1995, 1996, 1997; Norbeck et al. 1996). Data on the encoded proteins comfrom the Swiss-Prot database (http://expasy.hcuge.ch). Genes are presented in alphabetical order.

Gene numbera DescriptionNumberof Metb

TheoreticalpI/Mrb

ExperimentalpI/Mrc

ACT1 Actin 16 5·44/41·7 5·3/41·9ADH1 Alcohol dehydrogenase I 6 6·26/36·7 6·2/40·1ALD6 Putative aldehyde dehydrogenase 11 5·31/54·4 5·4/58·4ATP2 ATP synthase beta chain, mitochondrial 7 5·36/52·8d 5·1/47·3BGL2 Glucan 1,3-beta glucosidase 4 4·28/31·7d 4·3/29·0CTT1 Catalase T (cytoplasmic) 6 6·17/64·5 6·0/59·5DAK1 Putative dihydroxyacetone kinase 10 5·25/62·2 5·2/58·1EFB1* Elongation factor 1-beta 3 4·30/22·7 4·3/30·4ENO1*# Enolase 1 5 6·17/46·7 6·2/46·7ENO2 Enolase 2 9 5·67/46·8 5·8/45·2FBA1 Fructose-bisphosphate aldolase 7 5·51/39·5 5·7/38·4GDH1 NADP-specific glutamate dehydrogenase 8 5·68/49·6 5·4/50·7GPD1 Glycerol-3-phosphate dehydrogenase 6 5·30/42·9 5·3/43·8GPM1* Phosphoglycerate mutase 1 8·86/27·5 8·9/25·5GPP1 -glycerol 3-phosphatase 1 1 5·35/27·8 5·4/30·2GPP2 -glycerol 3-phosphatase 2 2 5·80/27·8 5·9/27·7HSP26 Heat shock protein 26 0 5·31/23·7 5·2/25·3HXK2 Hexokinase PII 13 5·16/53·9 5·2/54·6ILV5 Ketol-acid reductoisomerase 7 6·31/39·2d 6·3/38·0IPP1*# Inorganic pyrophosphatase 2 5·36/32·2 5·4/32·2LYS9* Saccharopine dehydrogenase (NADP+) 9 5·10/48·9 5·1/45·7MET6 Methionine synthase 8 6·07/85·7 6·1/79·3MET17* o-Acetylhomoserine sulfhydrylase 1 5·98/48·5 6·0/46·0PDC1 Pyruvate decarboxylase 12 5·99/61·5 5·7/55·3PGK1 Phosphoglycerate kinase 3 7·09/44·6 6·8/44·1SAM1 S-adenosylmethionine synthase 1 7 5·04/41·8 5·0/46·7SAM2 S-adenosylmethionine synthase 2 6 5·18/42·1 5·1/43·3SSA1 Heat shock homolog ssa1 7 5·00/69·5 4·9/68·7SSA2 Heat shock homolog ssa2 7 4·95/69·3 4·8/66·5SSB1# Heat shock homolog ssb1 9 5·32/66·5 5·2/66·5SSB2 Heat shock homolog ssb2 8 5·37/66·5 5·3/66·5SSE1 Heat shock homolog sse1 10 5·12/77·3 5·1/78·0SHM2* Serine hydroxymethyltransferase 11 6·98/52·2 7·0/46·8TDH1* Glyceraldehyde 3-phosphate dehydrogenase 1 6 8·32/35·6 8·3/33·0TDH2 Glyceraldehyde 3-phosphate dehydrogenase 2 7 6·5/35·7 6·5/32·8TDH3* Glyceraldehyde 3-phosphate dehydrogenase 3 6 6·5/35·6 6·5/33·6TEF1/TEF2 Elongation factor 1-alpha 8 9·14/50·0 9·1/46·9TPI1 Triosephosphate isomerase 0 5·75/26·7 5·8/23·3VMA1 Vacuolar H+-ATPase (subunit A) 15 5·21/66·7e 5·1/69·1YKL056c Translationally controlled tumor protein

homolog7 4·41/18·7 4·4/24·6

YMR116c Putative nucleotide binding protein 7 5·8/34·8 5·6/29·0YST1* Nucleic acid-binding protein 1 4·65/27·9 4·7/29·5YST2 Nucleic acid-binding protein 1 4·69/27·8 4·7/29·5

a *Indicates that protein was selected and used as pI standard. #Indicates that protein was selected and used as Mr standard.bValues taken from Swiss-Prot.cCalculated by the PDQuest software utilizing the selected pI and Mr standard proteins.dpI/Mr given for the processed form. These proteins contain signal sequences which are deleted after proper localization. ThpI/Mr for the full-length forms is estimated to: ATP2 5·71/54·9 (1–19 is signal sequence for the mitochondria); ILV5 9·10/44·4 (1–4is signal sequence for the mitochondria) and BGL2 4·32/34·1 (1–23 is signal sequence for the cell wall).epI/Mr given for the processed form. An internal peptide of 454 amino acids is post-translationally cleaved out. The full lengtranslation product of VMA1 has a theoretical value of 5·83/118·6.

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

ald.,alesttse-

.-);et-drisdonfe-aln–).rirr;

,,e-e-nr-,nreg.);ly-s

7)

functions, like glycolysis (e.g. ENO2, TDH1,PGK1), heat stress response (e.g. SSA1, SSE1),osmoregulatory response (e.g. GPD1, GPP2),translation (e.g. TEF1, EFB1), biosynthesis(e.g. ILV5, SAM1, SHM2), ATP generation(ATP2), nitrogen assimilation (GDH1) structuralorganization (ACT1), as well as some FUN(function unknown) gene products (YMR116c,YKL056c).

calculations (Table 2). For example, (i) Ilv5pinvolved in the biosynthesis of lysin. It is localizeto the mitochondria and an N-terminal 47 aminacid signal peptide is cleaved off in the localizatioprocess (Kassow, 1992). (ii) Vma1p is a subunit othe vacuolar ATPase complex, and its final squence is determined by an autocatalytic internsplicing mechanism bringing about removal of ainternal domain of 454 amino acids (residues 284737) named the intein (Hirata and Anruka, 1992In all these cases the theoretically calculated pI/Mof the processed products agreed well with theposition in the 2D map (experimental pI/MFigure 5; Table 2).

Expression-stoichiometry during exponentialgrowth on glucose

The global separation of proteins by 2D-PAGEand their subsequent biochemical identificationwould be of little scientific value if the quantitativaspect of the technology was neglected. The potential to build expression databases is great sincrather accurate quantitation can be obtained simultaneously on roughly 1000 proteins. Expressiowill mostly be recorded as the relative incorpoation of a labelled amino acid into a proteinproviding valuable information on the regulatioof a specific gene product during certain growth ostress conditions. However, between proteins, thnumber of a particular amino acid residue (e.methionine) can vary considerably (Table 2Tpi1p and Hsp26p lack methionine, Met17p onharbours one residue (0·2 mole%), while the FUNgene product, Yk1056c, exhibits seven methionine

Figure 3. Position comparison of two 2D-PAGE systems; ourpreviously used ampholine-driven system (effective pH range4–7) in the first dimension and 10% acrylamide in the SDS-dimension, and the presented system using IPG (pH 3–10) and12·5% acrylamide in the SDS-dimension. Identical sampleshave been analysed in the two systems. Cells were growingexponentially in minimal medium, either with glucose orethanol as energy and carbon source, when labelled with[35S]methionine. Arrows represent proteins that were eitherinduced (arrows upwards) or repressed (arrows downwards)during growth on ethanol compared to glucose. The position ofactin (Act1p) is indicated by a white circle in the different gelsfor reference.

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

These identified proteins can be used as internstandards for the generation of a relevant pI anMr scale for the whole pattern (Bjellqvist et al1993, 1994). To evaluate how well the theoreticvalues adhered to the recorded 2D position, threlative x- and y-migration was plotted againthese calculated values (Figure 5). Most spoadhere to a general trend in the x-dimension. Somof these were selected as standards, and the obtained pI plot closely resembles that expected fromcalculations concerning the casting process (BBjellqvist, personal communication). The calculated Mr values scattered somewhat (Figure 5Bhowever, most spots adhered to the straight linin this log-plot. For four of the spots postranslational modifications are reported which hato be taken into account in the theoretical pI/M

Figure 2. The number of spots automatically matched in allgels by the PDQuest software as a function of number ofmanually added landmarks.

1526 . .

7

7)

Figure 4. Protein identification of 2D-resolved proteins via sequencing in our laboratory (Blomberg et al.,1997; Norbeck and Blomberg, 1995, 1996, 1997; Norbeck et al., 1996). IPG-2D-PAGE pattern of proteinextract from Saccharomyces cerevisiae, utilizing a non-linear immobilized pH (3–10) gradient in the firstdimension and a 12·5% acrylamide SDS gel in the second dimension. Yeast protein extract grown and labelledas in Figure 1. Black circles indicate the position of spots which are close to the level of detection in thispresentation; white circles indicate spots which are partly covered by more dominant spots, and white spotsindicate the position of proteins which do not contain any methionine and are thus not visualized in these gels.For explanation to the biochemical identity of the abbreviations, see Table 2. Our sequences have notdistinguished between the isoproteins SSA1 and SSA2 (Norbeck and Blomberg, 1996) as well as YST1 andYST2 (Norbeck and Blomberg, 1997); the positions of these taken from Garrels et al. (1994). The Mr and pIscales are automatically constructed by the PDQuest software utilizing the selected standard proteins (Table2). An alternative presentation of the gel with the acidic side to the right is provided at our www-server(http://yeast-2dpage.gmm.gu.se).

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

152-

-ts-e-n,axe-te

rs-0ee

ealrg,sp-d-hl,s-falnol-dfssherre-e0

stn

7)

(4·3 mole%). Thus, the apparent expression-stoichiometry between different proteins under onecondition can clearly be misleading if ppm valuesare compared. The identification and gene linkageof 2D spots will, however, make detailed stoichio-metric comparisons possible if ppm values arenormalized to the number of methionines in theproteins (Figure 6). The methionine-normalizedppm value for every protein was compared to themethionine-normalized ppm value for Act1p, pro-viding a means of transforming obtained values tonumber of copies per cell, since the number ofactin molecules per yeast cell can be estimated atroughly 60 000.The glyceraldehyde 3-phosphate dehydrogenase

isoform Tdh3p was about 17-fold more abundantthan actin in these exponentially growing cells, andappeared under these conditions to be the mostabundant protein, amounting to more than 1 mil-lion copies per cell. Some of the other glycolyticproteins were also found at high levels, and Fba1p,Eno2p and Gpm1p were all expressed to at least

300 000 copies per cell. A comparison of the expression of the three TDH isoforms indicated thaTdh3p constituted 92% of the total TDH expresion of this abundant gene family. A lower rported value of Tdh3p (50–60%) was based oactivity measurements (McAlister and Holland1985). However, it was also reported that the Vmof Tdh3p homotetramers was about two- to threfold lower compared to Tdh2p, indicating thaactivity measurements will underestimate thamount of Tdh3 proteins. The minor isoformTdh1p only amounted to about 25 000 copies pecell during these growth conditions, and was thuless abundant than actin, while Tdh2p was equivalent to actin in expression and present at 64 00copies per cell. The present study constitutes thfirst report where all three TDH isoforms arsimultaneously resolved and properly quantified.Proteins involved in the translation process wer

also rather abundant, especially the translationelongation factor 1-alpha (Tef1p). The cells undestudy are growing rapidly, probably explainintheir high translational capacity. In additionmany proteins involved in assimilatory functionwere rather highly expressed, e.g. Met17(sulphur-assimilation) and Gdh1p (nitrogenassimilation) were expressed to 190 000 an140 000 copies per cell, respectively. The constitutive form of glycerol 3-phosphatase, Gpp1p, whicparticipates in the cellular production of glycerowas expressed to 160 000 copies per cell, surpriingly high, and in parity with the amount opyruvate decarboxylase, Pdc1p. Under these basgrowth conditions, the stoichiometric expressioof Gpp1p was at least 10-fold higher compared tthe osmo-stress induced and also glyceroproducing enzyme Gpd1p (Norbeck anBlomberg, 1997). The steady-state expression oother stress-induced proteins, like the heat streresponder Ssa1p, were also high and in parity witAct1p. The FUN-gene products identified in th2D pattern, Yk1056c and Ymr116c, were ratheabundant proteins and expressed to levels highethan actin under glucose growth. Thus, thstoichiometric range covered for the identified proteins encompassed a more than 100-fold differencin number, from 7500 (Dak1p) to 1 050 00(Tdh3p) molecules per cell.

Carbon source-dependent differential expressionThe quantitative utility of an annotated yea

2D map was evident from studies on the carbo

Figure 5. Theoretical values, obtained from Swiss-Prot, for pIand Mr of identified proteins in the 2D pattern as a function oftheir relative x- and y-migration. (A) Theoretical pI. Allproteins indicated in Table 2. Filled symbols indicate theproteins used as standards in generating an appropriate pI scalein our gels: EFB1, NAB1A, LYS9, IPP1, TPI1, MET17,ENO1, TDH3, SHM2, TDH1, GPM1 and TEF1. (B) Theoreti-cal Mr. All proteins indicated in Table 2. Spots used forconstructing our Mr scale is shown by filled symbols: IPP1,ENO1 and SSB1.

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

1528 . .

pnlyehoae,e,-ttsohnn

9

7)

source dependent expression during exponentialgrowth on either glucose, galactose or ethanol. Theexpression pattern of the five proteins identified bymicro-sequencing in this study were compared tothe carbon source-dependent differential expres-sion of some of the previously identified proteins(Figure 7). A full record of the global systemapproach to protein responders during exponentialgrowth on different carbon sources will be pre-sented elsewhere (J. Norbeck and A. Blomberg,manuscript in preparation).When cells are shifted from a fermentable to a

non-fermentable carbon source, the glycolytic pathis converted from a mainly catabolic route to ananabolic one. This changed carbon flux is reflectedin the drastic reduction in expression of many ofthe members of this metabolic path; during expo-

nential growth on ethanol Pgk1p and Gpm1constituted only about 40% of their glucose-growvalues (Figure 7). Even a shift to another, slightless favourable, fermentable carbon source likgalactose (generation time 3·5 h compared to 2·2for glucose), reduced the expression of these twenzymes by 50%. A similar expression pattern ofgradually reduced expression in the series glucosgalactose and ethanol, could be seen for somother glycolytic proteins like Hxk2p, Fba1pTdh3p and Eno2p. The enolase 2 isoform exhibited the most drastic carbon source-dependenchanges in expression of these proteins, and ilevel during growth on ethanol was decreased tonly 10% of its glucose value, consistent witprevious findings (Bataille et al., 1988; Entiaet al., 1984). The Eno2 protein had this expressio

Figure 6. Stoichiometric comparison of protein expression duringexponential growth in the respiro-fermentative phase with glucose ascarbon source. The expression of identified proteins (ppm) was nor-malized to the number of methionines the protein contains, and relatedto the value obtained for Act1p (Act1p constituted 8228 ppm ofradioactivity in all resolved protein spots). For calculations onthe amount of Act1p per cell (60 000) see Materials and Methods.Abbreviations are explained in Table 2.

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

152-

r;poledt

rei-ylnf/als-ee-sddc-d-hn.,

renise-netsgisa-oA

ctels1x,h

7)

pattern in common with two enzymes clearly notof significance during respiratory metabolism,but crucial during fermentative ethanol produc-tion, i.e. Pdc1p and Adh1p. Glyceraldehyde3-phosphate dehydrogenase I (Tdh1p), which inthis strain under glucose growth only constitutedabout 2% of the total amount of this enzyme(Figure 6), was also reduced in expression duringgrowth on galactose; however, during solely respir-atory growth on ethanol Tdh1p increased expres-sion compared to during galactose fermentation.This pattern of regulation was similar to thatfound for Eno1p, which displayed enhanced ex-pression during ethanol growth with values even

exceeding the expression on glucose (50% higheFigure 7). A slightly higher expression of Eno1under respiratory conditions (acetate compared tglucose growth) has been reported earlier (Batailet al., 1988). It was also apparent that Tdh1p anTdh3p were not regulated co-ordinately, consistenwith earlier data (Boucherie et al., 1995a).The abundant translational elongation facto

1alpha (Tef1p/Tef2p) was strongly affected by thcarbon source. The carbon source not only instgates changes in catabolism, but cells also displadifferential rates of biomass formation; ethanogrowth being the slowest of the three carbosources investigated with a generation time o6·0 h. Probably the decreased expression of Tef1pTef2p is partly a reflection of the differentigrowth rate, reflecting a decreased need for tranlational capacity in the series glucose, galactosand ethanol. A very similar carbon sourcdependent expression profile to Tef1p/Tef2p wafound for Ssb1p, which is known to be associatewith functional ribosomes (Nelson et al., 1992) anhas been implicated as having a chaperone funtion in the initial folding of newly synthesizepolypeptides. Its expression is also strongly downregulated during conditions of sudden growtarrest, as under heat shock, and has thus beedesignated a heat stroke protein (Ludwig et al1982).The slower the cells grow, the lower the need fo

biosynthetic capacity, and this was reflected in thexpression pattern for some enzymes involved iamino acid synthesis, i.e. Ilv5p and Lys9p. Thpattern of regulation was also found for thGdh1p enzyme involved in ammonium assimilation (Figure 7). However, the biosynthetic proteiShm2p involved in the interconversion of glycinand serine clearly differed in its regulation, since iexpression was almost two-fold higher duringrowth on ethanol compared to galactose. Thdifference in regulatory pattern might partly bereflection of its dual cellular functions; it is involved not only in amino acid synthesis but alsin the generation of 1-carbon units for DNsynthesis (McNeil et al., 1994).The FUN-gene products Yk1056c and Ymr116

displayed rather similar carbon source-dependenprofiles; almost invariant expression on glucosand galactose but reduced expression on ethano(Figure 7). None of the other identified proteindisplayed a comparative expression profile. The Fsubunit of the mitochondrial ATPase compleAtp2p, was invariantly expressed during growt

Figure 7. Differential protein expression of the identifiedproteins with alkaline pI values, as well as for some otherpreviously identified spots, during exponential growth on eitherglucose (left bar; white), galactose (middle bar; grey) or ethanol(right bar; black). Values are expressed relative to the highestexpression value (set to one).

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

1530 . .

-ytsf.,alrresrr,tsiceyeenoee-f6;

olefestaew

)te);hne-heneddd

1

7)

on the two fermentable carbon sources; however, itdiffered in response compared to the FUN geneproducts during solely respiratory growth whereAtp2p displayed almost two-fold enhanced expres-sion. In addition, the Ipp1p protein (inorganicpyrophosphatase) was invariantly expressed dur-ing growth on not only glucose and galactose, buton either of the three carbon sources. Actin wasslightly regulated since its expression decreased inthe series glucose, galactose and ethanol, makingIpp1p potentially a better candidate for an internalcontrol in quantitative protein analysis under thesedifferent growth conditions.

DISCUSSION

Large-scale global analysis of ‘all’ cellularproteins—the proteomeWe report here on the use of an immobilized

wide range pH gradient in the first-dimensionalseparation of whole cell extracts of yeast by the2D-PAGE technology, an analytical tool of highpotential in global studies concerning changes inprotein expression. The presented procedure willcover proteins with isoelectric points between 3and 10 and molecular weights in the range 10–150 kDa. Proteins with pI/Mr values outside thisrange seem to amount to about 15–20% of thetotal open reading frames identified in the yeastgenome (Boucherie et al., 1995b). The yeast ge-nome codes for an estimated 6021 proteins, ofwhich roughly 5000 seem to be expressed underexponential growth in basal medium. Thus, the 2Danalysis which resolves and quantifies roughly1–2000 proteins easily and with good precision,misses out on more than 60% of the expressedproteins. So, why can all proteins not be resolved?Some proteins are not visualized even though theirpredicted pI and Mr value would indicate this.These are mainly proteins with a low codon biasvalue, expressed at low levels and not detectedunder these analysis conditions. Some of theseminor proteins could be studied by 2D-PAGEutilizing pre-electrophoretic fractionation tech-niques to enrich for subgroups of proteins, eithervia isolation of certain organelles or through chro-matographic fractionation of proteins with specificfeatures, e.g. DNA binding, tubulin binding etc.There are also difficult proteins which tend toaggregate and precipitate, either during the samplepreparation or during focusing at their respectivepI values. Efforts to circumvent these problems

have been addressed, and recent progress in increasing the denaturing conditions during IEF benhancing the concentration of chaotrophic agenshows great promise even in the resolution ointegral membrane proteins (Rabilloud et al1997). In addition, despite the resolving potentiof the technology, dominant proteins will coveminor neighbours preventing detection or propequantification. This problem can effectively bsolved by utilizing a number of pI gradients, thuextending the pattern for a certain pI range; fothis the immobilines are aptly designed. Howevethe main problem in large-scale proteome projecwill be to obtain a high throughput of genetvariants or growth conditions in the analysis; thpace of the analysis will be strongly hampered bthe use of multiple gel systems. We thus proposthese wide pH range gradients as well suited for thinitial screening of a large set of samples. Iaddition, even though it might be impossible tresolve all cellular proteins simultaneously in onsingle 2D gel, it is well documented that th2D-resolved proteins from a total cell extract constitute a good representation of a vast array ometabolic features (Sagliocco et al., 199Shevchenko et al., 1996; this study).

Standardization in proteome analysis; aprerequisite for general expression databasesIt is important that a protein analysis system t

be utilized for the construction of easily accessibimage databases should be robust in a number orespects. First of all, it should resolve a wide rangof pI and Mr values, to allow for the analysis of amany proteins as possible in the most time-efficienway. Use of alternative gel-systems to obtaincomplete coverage of all proteins could of coursbe used, but that will, as mentioned above, slodown the proteome analysis. Non-equilibrium 2Dtechnologies (NEPHGE; O’Farrell et al., 1977have been explored to cover a more complerange of yeast proteins (Bataille et al., 1987however, the interlaboratory comparison of sucpatterns is difficult. Furthermore, the spot positioin the pattern should ideally reflect some sequencbased feature of the separated proteins, whicmake IEF in the first dimension a more attractivtechnology. Secondly, the generated 2D pattershould also be easily reproduced in more than onlaboratory. This is a major drawback of the olampholine-based technology, since the producebatch quality of ampholines varies over time an

153-

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

fe-e-s.n-dnn-).lle2t-e-e3),esectdy,e-snryg)oalrlea

tsaldi-lyalls

7)

among different companies. Studies on the inter-laboratory reproducibility of patterns utilizingIPG-2D-PAGE, however, hold good promise forthe future development of useful and standardizedimage-based expression databases (Blomberget al., 1995; Corbett et al., 1994).However, it appears that one aspect of 2D-based

proteome analyses is still in its infancy: the stand-ardization of growth conditions, labelling regimesand extract procedures. The 2D patterns generatedin either our old ampholine system or in our newimmobiline-based system displayed a high degreeof similarities in the region of the 2D pattern thatoverlapped. It was concluded in a study on thereproducibility of 2D patterns between differentlaboratories that a number of factors (e.g. chemi-cals, physical dimensions of tanks, etc) apparentlydid not have a major impact on the resultingpattern (Blomberg et al., 1995). Thus, in manycases where interlaboratory image comparison isdifficult, we propose that the protein extractsthemselves are actually different. To our knowl-edge no study on the standardized production ofyeast extracts has been conducted, which seemsurgently needed in the approach to standardizedimage-linked expression databases on the yeastproteome.The identification of 2D-resolved proteins has

taken a leap forward with the advent of massspectrometry (Larsson et al., 1997; Sagliocco et al.,1996; Schevchenko et al., 1996). It appears that therapid and accurate identification of most proteinsin a gel will soon be within reach. Thus, the nextchallenge will be the construction of extensive2D-PAGE expression databases containing infor-mation about protein expression responses to alarge number of growth conditions or geneticalterations. Certainly, these databases should besearchable by more sophisticated tools than justthe identification of a particular spot—one shouldalso be able to screen the database for proteinswith a specific expression profile.

Alternative global approaches to the analyses ofexpression of genesCurrently, no serious analytical competitor to

the 2D-PAGE technology has been presented, letalone applied, to proteome analysis. However,several alternative approaches to the high-throughput evaluation of gene expression on thetranscript level have recently been described(Nguyen et al., 1995; Valculescu et al., 1997). One

of these novel technologies, called serial analysis ogene expression, (SAGE), provides quantitativexpression data without the prerequisite of a hybridization probe for each transcript, via largscale sequencing of concatenated sequence tagComparison of our 2D-PAGE-generated proteidata to the reported transcript levels during exponential growth in rich medium (YPD) revealesimilarities but also discrepancies (proteins knowto be under the regulation of medium compositio[YPD or YNB; Norbeck and Blomberg; unpublished data] were left out of the comparisonAmong the most abundant transcripts in the ceduring exponential growth were found to be thosfrom TDH2/TDH3, TEF1/TEF2, FBA1, ENOand PGK1 (Valculescu et al., 1997), consistenwith the data presented here on intracellular protein levels. However, the relation between thnumber of proteins and transcripts differed substantially between different genes; e.g. for thFBA1 gene there were 660 000 proteins and 18transcripts per cell (3600 proteins per transcriptwhile for the ACT1 gene 60 000 proteins werencoded by 60 transcripts per cell (1000 proteinper transcript). When the comparisons were madfor all identified proteins it was clear that no strirelation between the amount of transcript anprotein applied (data not shown). Apparentlyeast cells utilize in many instances alternativlevels of control of gene expression besides transcriptional regulation. A similar conclusion warecently presented for human cells from studies omRNA abundance and protein levels in the live(Anderson and Seilhamer, 1997). In addition, anpost-translational modification (protein-trimminor chemical modification, e.g. phosphorylationwould be missed if the analysis of proteins was tbe neglected. Ideally, a combination of globsystem approaches, focusing on either mRNA oproteins, should be applied to the large-scaanalysis of gene expression in order to obtaincomplete picture of the flow of information fromthe genome.

ACKNOWLEDGEMENTS

This study was financially supported by granfrom the Swedish National Board for NaturScience (NFR) and the Swedish National Boarfor Technical Development (NUTEK). The prmary data from the SAGE analysis was kindprovided by Dr V. E. Valculescu. The technicexpertise of Lena Blomberg in running the 2D ge

1532 . .

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

st

l-8..e

.inisinn.

).-es

a-.

).e

c-i-k.

eh6.ee)-g.

at

ees

c-as

.--n

n,-hll

al3-0,

3

7)

as well as the constructive criticism by StefanHohmann on the manuscript are acknowledged.

REFERENCES

Anderson, L. and Seilhamer, J. (1997). A comparison ofselected mRNA and protein abundances in humanliver. Electrophoresis 18, 533–537.

Andlid, T., Blomberg, L., Gustafsson, L. and Blomberg,A. (1997). Global system approach to the analysis of awild-type Saccharomyces cerevisiae, isolated from fishgut. Submitted.

Bataille, N., Peypouquet, M.-F. and Boucherie, H.(1987). Identification of polypeptides of the carbonmetabolism machinery on the two-dimensionalmap of Saccharomyces cerevisiae. Location of 23additional polypeptides. Yeast 3, 11–21.

Bataille, N., Thoraval, D. and Boucherie, H. (1988).Two-dimensional gel analysis of yeast proteins: Ap-plication to the study of changes in the levels of majorpolypeptides of Saccharomyces cerevisiae dependingon the fermentable or nonfermentable nature ofcarbon source. Electrophoresis 9, 774–780.

Bjellqvist, B., Basse, B., Olsen, E. and Celis, J. E. (1994).Reference points for comparisons of two-dimensionalmaps of proteins from different human cell typesdefined in a pH scale where isoelectric points correlatewith polypeptide compositions. Electrophoresis 15,529–539.

Bjellqvist, B., Hughes, G. J., Pasquali, C., et al. (1993).The focusing position of polypeptides in immobilizedpH gradients can be predicted from their amino acidsequences. Electrophoresis 14, 1023–1031.

Blomberg, A. (1995). Global changes in protein syn-thesis during adaptation to 0·7 -NaCl medium ofSaccharomyces cerevisiae. J. Bacteriol. 177, 3563–3572.

Blomberg, A., Blomberg, L., Norbeck, J., et al. (1995).Interlaboratory reproducibility of yeast proteinpatterns analyzed by immobilised pH gradienttwo-dimensional electrophoresis. Electrophoresis 16,1935–1945.

Blomberg, A., Nilsson, A., Norbeck, J., O} lz, R. andGustafsson, L. (1997). Fermentative capacity andprotein content in Saccharomyces cerevisiae duringgrowth, starvation and storage. Manuscript.

Boucherie, H. (1985). Protein synthesis during transitionand stationary phases under glucose limitation inSaccharomyes cerevisiae. J. Bacteriol. 161, 385–392.

Boucherie, H., Bataille, N., Fitch, I. T., Perrot, M. andTuite, M. F. (1995a). Differential synthesis ofglyceraldehyde-3-phosphate dehydrogenase poly-peptides in stressed yeast cells. FEMS Microbiol. Lett.125, 127–134.

Boucherie, H., Dujardin, G., Kermorgant, M.,Monribot, C., Slonimski, P. and Perrot, M. (1995b).Two-dimensional protein map of Saccharomyces

cerevisiae: construction of a gene-protein index. Yea11, 601–613.

Casari, G., Andrade, M., Bork, P., et al. (1995). Chalenging times for bioinformatics. Nature 376, 647–64

Casari, G., Daruvar, A., Sanders, C. and Schnider, R(1996). Bioinformatics and the discovery of genfunction. TIGS, 12, 244–245.

Corbett, J. M., Dunn, M. J., Posch, A. and Gorg, A(1994). Positional reproducibility of protein spotstwo-dimensional polyacrylamide gel electrophoresusing immobilised pH gradient isoelectric focusingthe first dimension: An interlaboratory comparisoElectrophoresis 15, 1205–1211.

Entian, K. D., Frohlish, K. U. and Mecke, D. (1984Regulation of enzymes and isoenzymes of carbohydrate metabolism in the yeast Saccharomyccerevisiae. Biochim. Biophys. Acta 799, 181–186.

Garrels, J. I. (1989). The QUEST system for quantittive analysis of two-dimensional gels. J. Biol. Chem264, 5269–5282.

Garrels, J. I., Futcher, B., Kobayashi, R., et al. (1994Protein identifications for a Saccharomyces cerevisiaprotein database. Electrophoresis 15, 1–21.

Gorg, A. (1994). High-resolution two-dimensional eletrophoresis of proteins using immobilized pH gradents. In (eds), Cell Biology: A Laboratory HandbooAcademic Press, New York, pp. 231–242.

Gorg, A., Postel, W. and Gunther, S. (1988). Thcurrent state of two-dimensional electrophoresis witimmobilized pH gradients. Electrophoresis 9, 531–54

Hirata, R. and Anruku, Y. (1992). Mutations at thputative junction sites of the yeast VMA1 protein, thcatalytic subunit of the vacuolar membrane H(+ATPase, inhibits its processing by protein splicinBiochem. Biophys. Res. Comm. 188, 40–47.

Kahn, P. (1995). From genome to proteome: lookinga cell’s proteins. Science 270, 369–370.

Kassow, A. (1992). Metabolic effects of deleting thregion encoding the transit peptide in Saccharomyccerevisiae ILV5.

Klose, J. (1975). Protein mapping by combined isoeletric focusing and electrophoresis in mouse tissues:novel approach to testing for induced point mutationin mammals. Human Genetics 26, 231–243.

Larsson, T., Norbeck, J., Karlsson, H., Karlsson, K.-Aand Blomberg, A. (1997). Identifications of twodimensional gel electrophoresis resolved yeast proteins by matrix-assisted laser desorption ionisatiomass spectrometry. Electrophoresis 18, 418–423.

Ludwig, J. R., Foy, J. J., Elliot, S. G. and McLaughliC. S. (1982). Synthesis of specific identified, phosphorylated, heat shock, and heat stroke proteins througthe cell cycle of Saccharomyces cerevisiae. Mol. CeBiol. 2, 117–126.

McAlister, L. and Holland, M. J. (1985). Differentiexpression of the three yeast glyceraldehyde-phosphate dehydrogenase genes. J. Biol. Chem. 2615019–15027.

153-

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (199

ia2,

g,r-oli-ess-in.

-l.

.-2,

al

J.sd

l.ind

.,sso-i.

).8,

McNeil, J. B., McIntosh, E. M., Taylor, B. V., Zhang, F.R., Tang, S. and Bognar, A. L. (1994). Cloning andmolecular characterisation of three genes, includingtwo genes encoding serine hydroxymethyltransferases,whose inactivation is required to render yeast auxo-trophic for glycine. Mol. Biol. Cell 3, 29–47.

Miller, M. J., Xuong, N.-H. and Geiduschek, E. P.(1982). Quantitative analysis of heat shock responseof Saccharomyces cerevisiae. J. Bacteriol. 151, 311–327.

Nefsky, B. and Bretscher, A. (1992). Yeast actinis relatively well behaved. Eu. J. Biochem. 206, 949–955.

Nelson, R. J., Ziegelhoffer, T., Nicolet, C., Werner-Washburne, M. and Craig, E. A. (1992). Thetranslational machinery and 70 kd heat shock proteincooperate in protein synthesis. Cell 71, 97–105.

Nguyen, C., Rocha, D., Grunjeaud, S., et al. (1995).Differential gene expression in the murine thymusassayed by quantitative hybridisation of arrayedcDNA clones. Genomics 29, 207–216.

Norbeck, J., Ansell, R., Adler, L. and Blomberg, A.(1997). Deletion of the redox regulated GPD2 gene inSaccharomyces cerevisiae strongly alters protein ex-pression during anaerobic growth. Manuscript.

Norbeck, J. and Blomberg, A. (1995). Gene linkage oftwo-dimensional polyacrylamide gel electrophoresisresolved proteins from isogene families in Saccharo-myes cerevisiae by microsequencing of in-gel trypsingenerated peptides. Electrophoresis 16, 149–156.

Norbeck, J. and Blomberg, A. (1996). Protein expressionduring exponential growth in 0·7 -NaCl medium ofSaccharomyces cerevisiae. FEMSMicrobiol. Lett. 137,1–8.

Norbeck, J. and Blomberg, A. (1997). Metabolic andregulatory changes associated with growth of Saccha-romyces cerevisiae in 1·4 -sodium chloride: Evidence

for osmotic induction of glycerol dissimilation vthe dihydroxyacetone pathway. J. Biol. Chem. 275544–5554.

Norbeck, J., Pahlmann, A.-K., Akhtar, N., BlomberA. and Adler, L. (1996). Purification and characteization of two isoenzymes of -glycer3-phosphatase from Saccharomyces cerevisiae. Identfication of the corresponding GPP1 and GPP2 genand evidence for osmotic regulation of Gpp2p expresion by the osmosensing mitogen-activated protekinase signal transduction pathway. J. Biol. Chem271, 13843–13875.

O’Farrell, P. H. (1975). High resolution twodimensional electrophoresis of proteins. J. BioChem. 250, 4007–4021.

O’Farrell, P. Z., Goodman, H. M. and O’Farrell, P. H(1977). High resolution two-dimensional electrophoresis of basic as well as acidic proteins. Cell 11133–1142.

Oliver, G. S. (1996). From DNA sequence to biologicfunction. Nature 379, 597–600.

Rabilloud, T., Adessi, C., Giraudel, A. and Lunardi,(1997). Improvements of the solubilisation of proteinin two-dimensional electrophoresis with immobilisepH gradients. Electrophoresis 18, 307–316.

Sagliocco, F., Guillemot, J.-C., Monribot, C., et a(1996). Identifications of proteins of the yeast protemap using genetically manipulated strains anpeptide-mass fingerprinting. Yeast 12, 1519–1533.

Shevchenko, A., Jensen, O. N., Podtelejnikov, A. Vet al. (1996). Linking genome and proteome by maspectrometry: large-scale identification of yeast prteins from two dimensional gels. Proc. Natl. Acad. ScUSA. 93, 14440–14445.

Valculescu, V. E., Zhang, L., Zhou, W., et al. (1997Characterization of the yeast transcriptome. Cell 8243–251.

1534 . .

? 1997 John Wiley & Sons, Ltd. . 13: 1519–1534 (1997)