destabilized green fluorescent protein for monitoring dynamic changes in yeast gene expression with...

Research Article

Destabilized green ¯uorescent protein for monitoringdynamic changes in yeast gene expression with ¯owcytometry

Carolina Mateus1 and Simon V. Avery2*1 Department of Biology, Georgia State University, University Plaza, Atlanta, GA 30303, USA2 School of Life and Environmental Sciences, University of Nottingham, University Park, Nottingham NG7 2RD, UK

*Correspondence to:S. V. Avery, School of Life andEnvironmental Sciences,University of Nottingham,University Park,Nottingham NG7 2RD, UK.E-mail: [email protected]

Received: 16 April 2000

Accepted: 31 May 2000

Abstract

Green ¯uorescent protein (GFP) has many advantages as a reporter molecule, but its

stability makes it unsuitable for monitoring dynamic changes in gene expression, among

other applications. Destabilized GFPs have been developed for bacterial and mammalian

systems to counter this problem. Here, we extend such advances to the yeast model. We

fused the PEST-rich 178 carboxyl-terminal residues of the G1 cyclin Cln2 to the C

terminus of yEGFP3 (a yeast- and FACS-optimized GFP variant), creating yEGFP3-

Cln2PEST. We tested the hybrid protein after integrating modules harbouring the yEGFP3or yEGFP3±CLN2PEST ORFs into the Saccharomyces cerevisiae genome. yEGFP3±

Cln2PEST had a markedly shorter half-life (tK) than yEGFP3; inhibition of protein

synthesis with cycloheximide lead to a rapid decline in GFP content and ¯uorescence

(tK y30 min) in cells expressing yEGFP3±Cln2PEST, whereas these parameters were quite

stable in yEGFP3-expressing cells (tK y7 h). We placed yEGFP3±CLN2PEST under the

control of the CUP1 promoter, which is induced only transiently by copper. This transience

was readily discernible with yEGFP3±Cln2PEST, whereas yEGFP3 reported only on CUP1switch-on, albeit more slowly than yEGFP3±Cln2PEST. Cell cycle-regulated transcriptional

activation/inactivation of the CLN2 promoter was also discernible with yEGFP3±

Cln2PEST, using cultures that were previously synchronized with nocodazole. In

comparison to CLN2, expression from the ACT1 promoter was stable after release from

nocodazole. We also applied a novel ¯ow-cytometric technique for cell cycle analysis with

asynchronous cultures. The marked periodicities of CLN2 and CLB2 (mitotic cyclin)

transcription were readily evident from cellular yEGFP3±Cln2PEST levels with this non-

perturbing approach. The results represent the ®rst reported successful destabilization of a

yeast±GFP. This new construct expands the range of GFP applications open to yeast

workers. Copyright # 2000 John Wiley & Sons, Ltd.

Keywords: Saccharomyces cerevisiae; green ¯uorescent protein; reporter molecule; half-

life; cell cycle; transcription; cyclin; PEST sequence

Introduction

In recent years, the green ¯uorescent protein (GFP)from Aequorea victoria has become one of the mostwidely used reporters of gene expression andsubcellular protein localization (Chal®e et al.,1994; Cormack, 1998; Tsien, 1998). One principaladvantage of GFP over conventional reporters suchas LacZ is that detectable GFP ¯uorescence devel-ops spontaneously in cells without the need foraddition of exogenous substrates. Thus, with GFP,

gene induction or changes in protein localizationcan be monitored readily in live cells with ¯uores-cence microscopy, confocal microscopy or ¯owcytometry.

GFP is very stable, which can be advantageousfor studies where genes are expressed weakly andsensitivity may be a problem (Mistelli and Spector,1997). However, the stability of GFP also imposessigni®cant limitations on its use. For example,whereas native GFP is an ideal reporter of geneswitch-on, the protein's stability masks reliable

YeastYeast 2000; 16: 1313±1323.

Copyright # 2000 John Wiley & Sons, Ltd.

detection of gene switch-off. As a consequence,events such as transient gene induction, a commonincident during the cell division cycle, tend to bemissed by GFP. To circumvent the drawbacksassociated with GFP stability, destabilized GFPconstructs have recently been developed for usewith mammalian cells and bacteria (Andersen et al.,1998; Li et al., 1998; Corish and Tyler-Smith, 1999;Weber-Ban et al., 1999). These modi®ed proteinswere constructed by fusing GFP to sequences thatare subject to rapid turnover, such as the degrada-tion domain of mouse ornithine decarboxylase (Liet al., 1998; Corish and Tyler-Smith, 1999). Theestimated half-lives of the resultant proteins rangedfrom 40 min upwards in bacteria and 2 h upwardsin mammalian cells. These unstable GFPs havefound multiple applications, including dynamicmonitoring of bacterial growth rate (Sternberget al., 1999) and direct correlation of gene inductionwith protein translocation (Li et al., 1998).

Despite such advances in other systems, thesuccessful destabilization of a GFP that is codon-optimized for yeast has yet to be reported. This isan important gap in the array of tools availablewith S. cerevisiae, particularly since this organismhas already proved to be an excellent system forGFP applications tested to date, such as geneinduction (Niedenthal et al., 1996; Avery et al.,2000), protein localization (Niedenthal et al., 1996;Reiser et al., 1999; Fikus et al., 2000), mRNAlocalization (Beach et al., 1999), biosensing (Billin-ton et al., 1998) and genome-wide expressionanalyses (Bell et al., 1999). In this study we soughtto redress the balance with other model systems bydeveloping a destabilized GFP for use with yeast.Our approach was to create a hybrid proteincomprising the yeast-optimized GFP (yEGFP3)constructed by Cormack et al. (1997) fused to theC-terminal residues of the yeast G1 cyclin, Cln2p.The latter residues encompass the PEST motifs ofCln2 and are thought to target the protein forubiquitin(Ub)-dependent degradation. An impor-tant premise for this work was that such degrada-tion is constitutive (Salama et al., 1994; Schneideret al., 1998); the cell cycle dependence of yeast Cln2levels is regulated transcriptionally. Previously, acarboxyl-terminus sequence from Cln2 was usedsuccessfully to destabilize human thymidine kinaselacking its native carboxyl terminus, in yeast(Salama et al., 1994). The chimeric protein createdhad a half-life of approximately 12 min, compared

with the 2 h half-life of wild-type thymidine kinasein yeast. Here we accomplished a comparativedegree of yeast±GFP destabilization with a similarapproach. Moreover, we show that, in contrast towild-type GFP, the modi®ed protein is suf®cientlyunstable to enable ready detection of dynamicchanges in gene expression, e.g. transient geneinduction or cell cycle-dependent gene expression.In addition, we apply a novel ¯ow cytometricapproach with this new construct to detect cellcycle-dependent gene expression using asynchro-nous yeast populations.

Materials and methods

Strains and plasmids

A DNA fragment comprising the 3k±terminal 534nucleotides of CLN2 was ampli®ed by PCR withVent DNA polymerase (New England Biolabs)from yeast genomic DNA (Ausubel et al., 2000)using the primers 5k-GAATTGTACAAAGCATCCAACTTGAACATTTCG-3k and 5k-GAAGTGGCGCGCCCTATATTACTTGGGTATTGCC-3k.After digestion with BsrGI and AscI (sites under-lined), the fragment was inserted into BsrGI/AscI-digested pSVA12, creating pSVA13; we generatedpSVA12 previously (Avery et al., 2000) by sub-stituting the FACS- and yeast-optimized GFP openreading frame (yEGFP3) from pYGFP3 (Cormacket al., 1997) in place of the wild-type GFP sequencein pFA6a-GFPMT-His3MX6 (Wach et al., 1997).BsrGI provided a convenient restriction site atthe 3k terminus of yEGFP3 for translational fusionwith the CLN2 fragment, creating yEGFP3±CLN2PEST. Sequence ®delity in all PCR productswas routinely con®rmed by automated DNAsequencing. Promoter sequences (encompassingnucleotides up to ATG) for CUP1 (450 bp), CLN2(614 bp), ACT1 (479 bp), SOD1 (600 bp) and CLB2(600 bp) were tagged with appropriate restrictionsites by PCR, using yeast genomic DNA astemplate, and inserted into the multiple cloningsite immediately upstream of the GFP open readingframe in pSVA12 or pSVA13 (primer sequences areavailable on request). To integrate the resultingtranscriptional fusions into the yeast genome at thenon-essential HO locus, modules (encompassingpromoter, GFP and HIS3MX6 marker) wereampli®ed with Vent DNA polymerase by short¯anking homology (SFH) PCR, as described pre-

1314 C. Mateus and S. V. Avery

Copyright # 2000 John Wiley & Sons, Ltd. Yeast 2000; 16: 1313±1323.

viously (Wach et al., 1997; Avery et al., 2000).Flanking sequences targeted the PCR product toHO. After transformation (Gietz and Woods, 1998)into S. cerevisiae S150-2B, His+ colonies wereselected and appropriate integration of the moduleswas con®rmed by PCR. The strains created arelisted in Table 1.

Culture and experimental conditions

Strains were routinely maintained on YNB agarwithout amino acids (Difco) minus histidine, or onYEPD agar as described (Avery et al., 1996).Starter cultures were inoculated from plates intoYEPD broth in Erlenmeyer ¯asks and incubated at30uC with rotary shaking (120 rpm). Experimentalcultures in YEPD were inoculated from 48 h startercultures and grown overnight to OD600 y2.0(y2r107 cells/ml) before analysis. Where speci®ed,cultures were synchronized by incubation with50 mM nocodazole for 2.5 h. Release from arrestwas accomplished by washing twice with YEPDlacking nocodazole and ®nal suspension in the samemedium.

Western blotting

Preparation of crude lysates from S. cerevisiae andWestern blotting were performed using standardprocedures (Ausubel et al., 2000). Proteins(5±50 mg) were separated by SDS±PAGE (10%)using a Biorad Mini-PROTEAN II electrophoresissystem, and transferred to nitrocellulose membrane(Biorad). The blots were probed with mousemonoclonal anti-GFP (Clontech) (1 : 500 dilution),and alkaline phosphatase-conjugated anti-mouseIgG (Promega) (1 : 3000) antibodies. GFP wasdetected with BCIP-NBT (Promega). GFP was

quantitated by densitometry with a Quantity Onesystem (Biorad).

Microscopy

Microscopic observations were with a Nikon E600¯uorescence microscope equipped with NomarskiDIC. A r10 eyepiece and a r40 objective wereused for examination of cells. The excitation sourcefor ¯uorescence was a 100 W Hg vapour arc lamp.Fluorescence from GFP was visualized using a B-2E/C FITC ®lter block (excitation 465±495 nm,dichroic mirror 505 nm, barrier ®lter 515±555 nm).Images were captured with a DAGE DC330 three-chip colour camera and Flashpoint video framegrabber.

Flow cytometry

Cellular ¯uorescence from GFP was determinedquantitatively with a FACSCalibur ¯ow cytometer(Becton Dickinson, CA) equipped with a 15 mW,488 nm argon ion laser. All samples were suspendedin PBS and sonicated for 10 s to disperse cell-clumps just prior to analysis. Voltage and gainsettings, respectively, were 582 and 1.00 in log modefor FL1 (green ¯uorescence) readings, and E00 and1.00 in linear mode for FSC (forward scatter)readings. Typically, 10 000 cells were analysed persample. Data acquisition and analysis were per-formed using CELLQuest software (BD).

Results

Destabilization of yEGFP3

The carboxyl terminal 178 residues of Cln2 wereshown previously to promote for rapid proteinturnover in S. cerevisiae (Salama et al., 1994). Totest whether we could produce a destabilizedversion of GFP in yeast, we fused these 178 residuesto the C terminus of a modi®ed GFP (yEGFP3)that was previously optimized for FACS analysisand yeast codon usage (Cormack et al., 1997), thuscreating yEGFP3±Cln2PEST. We placed yEGFP3and yEGFP3±CLN2PEST under the control ofPSOD1 (the Cu±Zn superoxide dismutase promoter),which gives high constitutive transcriptional activity(Avery et al., 2000), and integrated the modules intothe S. cerevisiae genome, yielding strains SVY31and SVY32, respectively. Whole-cell extracts fromSVY31 and SVY32 were prepared and analysed by

Table 1. Strains used in this study

Strain Genotype*

S150-2B MATa leu2-3, 112 ura3-52 trp1-289 his 3-D1

SVY14 HO::PCUP1±yEGFP3±HIS3MX6

SVY15 HO::PCUP1±yEGFP3±CLN2PEST±HIS3MX6

SVY17 HO::PCLN2±yEGFP3±CLN2PEST±HIS3MX6SVY18 HO::PACT1±yEGFP3±CLN2PEST±HIS3MX6

SVY31 HO::PSOD1±yEGFP3±HIS3MX6

SVY32 HO::PSOD1±yEGFP3±CLN2PEST±HIS3MX6

SVY34 HO::PCLB2±yEGFP3±CLN2PEST±HIS3MX6

*All modi®ed strains were derived from S. cerevisiae S150-2B.

Destabilized GFP for yeast gene expression 1315


immunoblotting with anti-GFP antibody, asdescribed in Materials and methods. A proteinband was observed in SVY31 extracts at thepredicted molecular weight for yEGFP3(y27 kDa; arrow a, Figure 1A) and in SVY32 foryEGFP3±Cln2PEST (y48 kDa; arrow b, Figure 1A).To test whether yEGFP3±Cln2PEST was degradedmore rapidly than yEGFP3, we added cyclohex-imide, a protein synthesis inhibitor, to the culturesand monitored the abundance of these proteins byimmunoblotting. yEGFP3 was highly stable, show-ing only an approximate 10% decline during 2 hincubation of S. cerevisiae SVY31 with cyclohex-imide (Figure 1A, B). In contrast, there was amarked reduction in the level of yEGFP3±Cln2PEST

in SVY32 extracts during the same period. Within

20 min of cycloheximide addition, the yEGFP3±Cln2PEST content was diminished by approximately50%, and after 2 h the level of detected protein wasonly approximately 5% of that prior to cyclo-heximide addition. The graphs were transformedto approximate linearity when data were plotted aslogarithms, indicating that GFP is degraded with®rst-order kinetics.

Since the most useful property of GFP is its¯uorescence, we tested whether the apparent suc-cessful destabilization of the protein in S. cerevisiaedescribed above was also measurable with ¯uores-cence. We examined cellular green ¯uorescence qual-itatively by ¯uorescence microscopy (Figure 1C)and quantitatively by ¯ow cytometry (Figure 1D).S. cerevisiae SVY31 expressing yEGFP3 from PSODI

Figure 1. Destabilization of yEGFP3. Exponential phase S. cerevisiae SVY31 and SVY32 cultures in YEPD medium expressingyEGFP3 and yEGFP3±CLN2PEST, respectively, under the control of PSOD1, were exposed to 10 mg/ml cycloheximide. GFPcontent was monitored by protein content and ¯uorescence. (A) GFP in protein extracts obtained at intervals aftercycloheximide addition, detected with anti-GFP antibody; 5 mg and 50 mg protein were loaded for SVY31 and SVY32 extracts,respectively. (B) Quantitative analysis of results shown in (A): (O), SVY31; ($), SVY32. (C) Fluorescence images obtained atintervals after cycloheximide addition. Each ®eld of view contains four to eight cells. (D) Flow cytometric determination ofcellular green ¯orescence after cycloheximide addition, symbols as in (B). Points represent the mean ¯uorescence from10 000 cells, after subtraction of auto¯uorescence determined with S. cerevisiae S150-2B lacking GFP. Typical results areshown from one of three independent experiments conducted on different days. The trends shown were reproduced in eachexperiment. Absolute measurements of ¯uorescence showed <20% variation between experiments. (E) Green ¯uorescence(FL1) histograms obtained for S. cerevisiae S150-2B (Ð), SVY31 (. . .) and SVY32 (Ð) before cycloheximide addition; note that¯uorescence is presented on a logarithmic scale



displayed bright green ¯uorescence that was notvisibly diminished after cycloheximide addition(Figure 1C). This was borne out by ¯ow cytometricanalyses showing that the average level of cellulargreen ¯uorescence (corrected for auto¯uorescence,see Figure 1 legend) declined only slightly overthe 2 h incubation period with cycloheximide(Figure 1D). Cellular green ¯uorescence was lowerin cells expressing yEGFP3±CLN2PEST compared tothose expressing yEGFP3, but was higher thanauto¯uorescence in both cases (Figure 1C, E).Moreover, the ¯uorescence of S. cerevisiae SVY32was reduced further during incubation with cyclo-heximide. Analysis by ¯ow cytometry revealed thatthese cells' mean green ¯uorescence declined byapproximately 52% and 91% after 30 min and2 h incubation with cycloheximide, respectively(Figure 1D). The ¯uorescence half-lives (tK) of theproteins were estimated from the slopes of linear-ized plots ®tted by linear regression; tK values wereapproximately 7 h 23 min for yEGFP3 and 34 minfor yEGFP3±Cln2PEST. The former value wasconsistent with the previously reported half-life ofy7 h for wild-type GFP in S. cerevisiae (Natarajanet al., 1998). The value for yEGFP3±Cln2PEST wascomparable to that derived from the linearizedprotein data (Figure 1B), which yielded a tK ofapproximately 26 min. The results indicate thatfusion of the stated C-terminal Cln2 residues toyEGFP3 shortened the latter protein's longevity inS. cerevisiae by approximately 15-fold.

Detection of transient induction of CUP1transcription using yEGFP3±Cln2PEST

The transcription of CUP1, encoding yeast copper-binding metallothionein, has previously been shownto be induced only transiently following cellular Cuexposure (Pena et al., 1998). To test whether therapid turnover of yEGFP3±Cln2PEST made it asuitable reporter of such dynamic gene regulation,we monitored the ¯uorescence from GFP in cellsexpressing yEGFP3 (SVY14) or yEGFP3±CLN2-

PEST (SVY15) under the control of PCUP1. The pre-induction level of GFP ¯uorescence in cells expres-sing stable GFP was considerably higher than thatof cells expressing destabilized GFP (Figure 2).Induction of PCUPI with 0.25 mM Cu(NO3)2 yieldeda sharp rise in GFP ¯uorescence of S. cerevisiaeSVY15 between 15 and 45 min, which contrastedwith a more gradual increase in SVY14, mostly

between 30 and 75 min. Moreover, after peaking,the mean green ¯uorescence of SVY15 diminishedrapidly (by almost 75%) between 45 and 90 min(Figure 2). Expression of yEGFP3±CLN2PEST sub-sequently rose again by 1.5-fold to 2-fold but didnot attain the maximal level evident 45 min afterinduction. In contrast, the GFP ¯uorescence ofSVY14 remained high after peaking, and 4 h afterinduction was still >80% of the maximal valueevident at 75 min. Control cultures lacking GFPconstructs showed low (auto)¯uorescence that wasnot affected by incubation with Cu (not shown).The results obtained with yEGFP3±Cln2PEST wereconsistent with those obtained elsewhere by North-ern blotting for CUP1 transcripts (Pena et al.,

Figure 2. Transient induction of CUP1 transcription.Cu(NO3)2 was added to exponential phase S. cerevisiaecultures in YEPD medium to a ®nal concentration of0.25 mM. The green ¯uorescence of strains SVY14 (O) andSVY15 ($) expressing yEGFP3 and yEGFP3±CLN2PEST, respec-tively, under the control of PCUP1, was monitored with ¯owcytometry. Points represent the mean ¯uorescence from10 000 cells, after subtraction of auto¯uorescence deter-mined with S. cerevisiae S150-2B lacking GFP. Typical resultsare shown from one of three independent experimentsconducted on different days. The trends shown werereproduced in each experiment. Absolute measurements of¯uorescence showed <20% variation between experiments



1998), indicating that this novel protein is a suitablereporter of dynamic gene expression in S. cerevisiae.

Detection of cell cycle-dependent genetranscription in synchronous cultures

To determine whether yEGFP3±Cln2PEST was asuitable reporter of cell cycle-dependent geneexpression, we compared its production underPCLN2 and PACT1 control in nocodazole-synchronized cultures; CLN2, but not ACT1, issubject to cell cycle regulation at the transcriptionallevel (Spellman et al., 1998). Successful nocodazolearrest at G2/M was con®rmed microscopically,whereby >90% of cells exhibited a large buddedphenotype. Cell morphology was monitored follow-ing release from arrest, and initiation of synchro-nous budding occurred at around 70 min. We used¯ow cytometry to monitor the ¯uorescence of S.cerevisiae SVY17 and SVY18 (PCLN2± and PACT1±yEGFP3±CLN2PEST, respectively). Since cellvolume can in¯uence cellular ¯uorescence non-speci®cally, ¯uorescence data for samples fromsynchronous cultures were normalized against thecorresponding forward scatter (cell volume) data, asdescribed elsewhere (Williams et al., 1999). The dataare plotted as logarithm values to enable directcomparison of relative changes in expression ofCLN2 and ACT1 (Figure 3). The GFP ¯uorescenceof SVY18 was subject to relatively minor detectablevariation (<30%) during the 160 min experimentaltime course and no signi®cant trend was evident atany stage. In contrast, whereas the GFP ¯uores-cence of SVY17 remained quite low initially, after60 min there was a marked y5.5-fold increase inGFP arising from PCLN2 that peaked at around120 min (Figure 3). This increase coincided with theonset of budding. Subsequently, the GFP ¯uores-cence of SVY17 declined, as cells proceededtowards the next division. The results were consis-tent with the accepted regulatory patterns of CLN2and ACT1 during the S. cerevisiae cell cycle (seeDiscussion).

Detection of cell cycle-dependent genetranscription in asynchronous cultures, byforward scatter gating with ¯ow cytometry

Forward scatter is a ¯ow cytometric parameter thatis widely used as an indicator of particle/cell size(Lloyd, 1993). We recently showed that by appro-priate gating of forward scatter histograms, sub-

populations in asynchronous yeast cultures thatcorrespond to cells at differing cell cycle stage canbe readily differentiated. Thus, cells with progres-sively increasing forward scatter show the expectedshift from a 1C to a 2C DNA content (Figure 4A,B). We sought to test whether such an approachcould be adapted to provide a novel means ofdetecting cell cycle dependent transcription, usingasynchronous cultures. Thus, subpopulations ofexponential-phase cultures were gated according tocell volume (forward scatter). Gates were adjustedso that each of 10 fractions represented y10% ofthe cells analysed in each sample (y1000 cells perfraction). Average GFP ¯uorescence (from FL1histograms) was determined in each gated popula-tion. To eliminate non-speci®c effects on ¯uores-cence in these experiments, we normalized¯uorescence data against those obtained with

Figure 3. Cell cycle-dependent CLN2 transcription insynchronous cultures. Exponential phase cultures weresynchronized with nocodazole, and then released fromsynchrony. The mean green ¯uorescence of S. cerevisiaeSVY17 ($) and SVY18 (O) expressing yEGFP3±CLN2PEST

under the control of PCLN2 and PACT1, respectively, wasmonitored by ¯ow cytometry. To correct for non-speci®ceffects of cell volume, ¯uorescence measurements weredivided by the corresponding mean forward scatter value foreach sample, as described by Williams et al. (1999).Auto¯uorescence determinations for synchronous S. cerevi-siae S150-2B cultures were subtracted. Points representmean data collected for 10 000 cells. Typical results areshown from one of two independent experiments conductedon different days. The trends shown were reproduced ineach experiment. Absolute measurements of ¯uorescenceshowed <20% variation between experiments



PACT1±yEGFP3±CLN2PEST (see legend, Figure 4C);this approach should be less empirical than normal-ization against forward scatter (as used above) andby this stage we had con®rmed that the constitutivenature of PACT1 activity was detectable with ¯owcytometry (see Figure 3). To provide an additionaltest to PCLN2 we examined PCLB2, which is inducedlater in the cell cycle as cells enter G2. The GFP¯uorescence of S. cerevisiae SVY17 showed acyclical variation with progressive increases in cellvolume (Figure 4C). Peak expression from PCLN2

was evident in cells in the fraction at 30±40% ofmaximal cell volume. There was a decline in GFP

production with further increases in cell volume.However, in cells larger than y80% of maximal cellvolume, GFP ¯uorescence rose again and this trendwas continued in the smallest (newly-divided) cells(Figure 4C). Note that data for 0±10% fractionswere very variable and consequently are not shown;this variability may be related to the unequal sizesof newly budded yeast cells.

Expression of yEGFP3±CLN2PEST from PCLB2

(in S. cerevisiae SVY34) also ¯uctuated in cellpopulations differentiated by forward scatter(Figure 4C). In this case, marked increases in GFP¯uorescence were evident with progressive increases

Figure 4. Cell cycle-dependent gene transcription in asynchronous cultures. A, B adapted from Howlett and Avery (1999) todemonstrate the correlation between forward scatter and cell cycle stage. (A) Forward scatter (FSC) histogram ofexponential phase S. cerevisiae, gated into 10 regions (R2-R11), each comprising 10% of all 100 000 cells in the sample. (B)Cellular DNA content, as assessed by propidium iodide (PI) staining of regions cells in R2-R11. (C) Exponential phase S.cerevisiae SVY17 (O) and SVY34 ($) cultures expressing yEGFP3±CLN2PEST under the control of PCLN2 and PCLB2, respectively,were analysed by ¯ow cytometry. Points represent data collected for cells in fractions gated by forward scatter, each fractionrepresenting 10% of the total population. Anto¯uorescence determinations for corresponding fractions with S. cerevisiaeS150-2B were subtracted. To correct for non-speci®c volume effects, the resultant values were normalized (by division)against those obtained for S. cerevisiae SVY18, expressing yEGFP3±CLN2PEST under the control of the constitutive ACT1promoter. Each point represents corrected mean data collected for a fraction of 3000 cells in the case of SVY17 (3r1000from independent cultures), and 10 000 cells in the case of SVY34 (10r1000 from independent cultures). Typical results areshown from one of three independent experiments conducted on different days. The trends shown were reproduced in eachexperiment. Absolute measurements of ¯uorescence showed <20% variation between experiments



in cell volume above 45% of maximal volume. Thistrend continued until cells were at the apparentpoint of division (90±100% fraction). Small (newly-divided) cells showed lower GFP levels and furtherreductions in transcription from PCLB2 were evidentin cell populations progressing to y45% of max-imal volume. The patterns of yEGFP3±CLN2PEST

expression evident in Figure 4 are fully consistentwith the cyclical ¯uctuations in PCLN2 and PCLB2

transcription that occur during the S. cerevisiae cellcycle (see Discussion).

Discussion

This paper describes the ®rst successful constructionof a destabilized GFP for use with yeast. The GFPvariant that we destabilized was previously FACS-optimized (Cormack et al., 1996) and codon-optimized for yeast (Cormack et al., 1997). Theestimated 40-fold signal enhancement that the latterfeatures impart when expressed in S. cerevisiae(Cormack et al., 1997) evidently compensated forthe signal reduction associated with low steady-statelevels of destabilized GFP. For example, auto¯uor-escence accounted for only approximately 30% ofthe total green ¯uorescence of cells expressingyEGFP3±CLN2PEST constitutively from the SOD1promoter, and considerably less than this in thecase of expression from the CUP1 promoter afterCu induction. Moreover, it is stressed that thesensitivity of ¯ow cytometric detection ofyEGFP3±Cln2PEST that we report here correspondsto expression from single genome-integratedcopies of constructs. Multicopy expression couldreadily be used to enhance the signal from weakerpromoters. Whereas ¯ow cytometric analysis is idealfor quantitative purposes, yEGFP3±CLN2PEST-expressing strains (e.g. SVY32) could also bediscriminated qualitatively from non-GFP-contain-ing cells by simple observation with ¯uorescencemicroscopy. As expected, the contrast was greaterwith strains expressing stable GFP.

Our experiments indicate a half-life (tK) foryEGFP3±Cln2PEST of approximately 30 min. Thedegree of GFP destabilization was of the sameorder as that attained previously (tK y12 min from2 h) with a thymidine kinase derivative followingtranslational fusion to the same C-terminal Cln2fragment that we used (Salama et al., 1994). Thus,our results corroborate the notion that this PEST-

rich Cln2 sequence can serve as a universalubiquitin-targeting sequence. The slightly fasterdecline in yEGFP3±Cln2PEST that was detectablewith immunoblotting than with ¯ow cytometryafter cycloheximide addition, was similar to resultsobtained with mammalian GFP (Li et al., 1998),and may be attributable to the existence of non-¯uorescent premature GFP (Cormack et al., 1996;Li et al., 1998). For the purposes of this study, itsuf®ced that 50% turnover of yEGFP3±Cln2PEST

clearly could occur within the y2 h cell cycleexhibited by S. cerevisiae under the present experi-mental conditions.

The lower steady-state ¯uorescence (e.g. duringconstitutive expression from PSOD1) of cells expres-sing yEGFP3±Cln2PEST compared to yEGFP3 wasconsistent with the higher turnover rate of theformer. Indeed, from the ¯uorescence half-lives ofthe two proteins, we calculated that, at steady state(i.e. when synthesis rate matches degradation rate,during constitutive expression from the samepromoter), cellular levels of stable yEGFP3 shouldbe approximately 13-fold higher than those ofyEGFP3±Cln2PEST. This corresponds very wellwith the approximate 16-fold difference measuredexperimentally. The closeness of this correlationeffectively discounts the possibility that the lower¯uorescence of yEGFP3±CLN2PEST-expressing cellsresults from impaired ¯uorescence formation in thehybrid protein, as similarly concluded for otherhybrid GFPs (Weber-Ban et al., 1999).

The transient nature of copper-induced PCUP1

transcription (Pena et al., 1998) was readily detectedusing the yEGFP3±Cln2PEST reporter. Our inciden-tal observation that yEGFP3±Cln2PEST also reportsmore rapidly on initial induction than stable GFP isconsistent with results obtained for mammaliandestabilized GFP (Li et al., 1998). This observationmay re¯ect the higher basal accumulation of stableGFP, which could mask the extent of subsequentinduction. It should also be noted that, because ofthe short half-life of yEGFP3±Cln2PEST, the fullextent of induction is underestimated; the meantime from GFP translation to ¯uorophore forma-tion (y1 h) suggests that much of the GFP will bedegraded before reaching full ¯uorescence.

The results obtained here correlated well withthose obtained previously for CUP1 transcripts(Pena et al., 1998). As in the study of Pena et al.(1998), CUP1 expression eventually stabilized at alevel considerably higher than the basal level



observed prior to induction. The reproducibleincrease in yEGFP3±Cln2PEST production thatoccurred between 90 and 105 min after initialinduction, suggested a possible cyclical relationshipbetween copper availability and CUP1 expression.Such a scenario could arise if, for example, thedecline in Cup1p production predicted between 45and 90 min coincided with some Cup1p turnover,and if copper released from degraded Cup1 inducedfurther transcription from PCUP1. Further workwould be required to test this possibility. It shouldbe noted that the copper concentration (0.25 mM)used in our experiments was non-toxic to cellsincubated in YEPD medium. At toxic Cu concen-trations (i5 mM in YEPD), we observed reductionsin GFP ¯uorescence (not shown), which we attrib-uted to leakage of the protein across permeabilizedplasma membranes (Avery et al., 1996). This resultunderscores the need to adopt caution wheninterpreting reporter-gene data under conditionsthat are potentially lethal.

The cyclical change in PCLN2 activity detectedwith yEGFP3±Cln2PEST in synchronous culturescontrasted markedly with the stable expressionevident from PACT1. The constitutive nature ofACT1 expression during the yeast cell cycle isre¯ected in the widespread use of ACT1 mRNA asa loading control for Northern blots. In contrast,CLN2 is activated during the G1 stage of the yeastcell cycle (Wittenberg et al., 1990), as a prelude toinduction of B-type cyclins at S phase. Thus,induction of Cln2 synthesis normally precedes theonset of budding. The almost-coincident timings ofyEGFP3±Cln2PEST production from PCLN2 and ofbudding evident in our experiments probably re¯ectthe folding time required for GFP to reach full¯uorescence (Cormack et al., 1996). The generationof fast-folding GFP mutants has been described(Cormack et al., 1996) and similar variants in S.cerevisiae could, if necessary, help to improvefurther the coordination between GFP expressionand other events. It should also be borne in mindthat Cln2p has a shorter half-life thanyEGFP3±Cln2PEST (Lanker et al., 1996), so thedegree to which Cln2p oscillates during the cellcycle is greater than as portrayed with GFP inFigure 3.

The ¯uctuations in CLN2 and CLB2 expressionevident with our recently developed ¯ow cytometricfractionation approach (Howlett and Avery, 1999),were based on the correlation between cell size

(forward scatter) and cell cycle stage. The principaladvantage of this technique over conventionalapproaches for cell cycle analysis is that it circum-vents the need to generate synchronous cultures andis, therefore, truly non-perturbing. In theory, themethod is applicable to any cellular parameter thatcan be probed with ¯uorescence, and has been usedto study the cell cycle-dependence of coppersensitivity (Howlett and Avery, 1999) and phagocy-tosis (Avery et al., 1995). Here we extended theapproach to gene expression for the ®rst time.

The cyclical y1.4 to 1.7-fold changes in PCLN2

and PCLB2 activity evident here were not as markedas seen with certain other approaches (e.g. seeFigure 3). Several factors can account for this,principally that the asymmetry of S. cerevisiae celldivision means that cells of the same size are notstrictly of the same cell cycle stage (Alberghina et al.,1998). This contributes some heterogeneity tosubpopulations gated by forward scatter, as evi-denced by several fractions containing both 1C- and2C-DNA cells (see Figure 4B). In addition, the ¯owcytometric method does not readily distinguish G2/M-phase cells from any G1 `doublets', and someoverlap between fractions comprising cells at thesestages would be expected. This is borne out by ourobservation that transcription from PCLN2 hadcommenced in cells within the largest-volumefraction (Figure 4C). Nonetheless, our resultsshowed clear ¯uctuations in CLN2 and CLB2transcription that correlated closely with thechanges anticipated during the S. cerevisiae cellcycle. Thus, by assuming a relatively constantincrease in cell size during the cell cycle and acon®rmed doubling time of y2 h in these experi-ments, we can estimate that the peak in PCLN2-generated GFP ¯uorescence evident at y35%maximal cell volume corresponded to cells atbetween y40 and 45 min of their division cycle.This slight delay compared with the Cln2 protein,which peaks at y30±40 min (late G1/S phase)(Wittenberg et al., 1990; Nasmyth, 1996), mayagain be attributable to the folding time of GFP,as well as to the longer half-life of yEGFP3±Cln2PEST than Cln2p (see above). Since degradationof the mitotic cyclin Clb2p is subject to differentregulatory mechanisms than the constitutive degra-dation conferred by the Cln2p PEST sequences(Amon et al., 1994; Barral et al., 1995; Lanker et al.,1996; Schneider et al., 1998), it is more appropriateto interpret our PCLB2 data in relation to CLB2



transcript levels than protein levels. Thus, the peakin PCLB2-generated GFP evident in cells with thelargest volumes was consistent, allowing for delay,with the normal induction of CLB2 transcriptionduring entry to G2, and with peak CLB2 transcrip-tion occurring just prior to anaphase (Spellmanet al., 1998).

In conclusion, yEGFP3±Cln2PEST now introducessimilar bene®ts to the yeast system that haverecently been afforded to human and bacterialsystems by GFP destabilization (Li et al., 1998;Sternberg et al., 1999). Furthermore, since there isfar lower net accumulation of destabilized thanstable GFP, potential concerns of GFP toxicity areeffectively redundant. We have validated the use ofthis GFP variant as a reporter of dynamic geneexpression that is detectable either microscopically,by Western blotting, or most convenientlyÐsince itallows precise quantitation in real time without theneed for cell disruptionÐby ¯ow cytometry. Thegeneration of destabilized colour variants of GFP,by the same approach used here, could allowdynamic changes in the expression of multiplegenes to be monitored simultaneously. yEGFP3±Cln2PEST provides a powerful new means ofexploring cell cycle-dependent gene expression inyeast, which could be integrated with the genomicscreening technologies that have to date relied onmRNA transcripts to detect cell cycle-dependentgene expression (Cho et al., 1998; Spellman et al.,1998). Moreover, our novel ¯ow cytometric fractio-nation approach with yEGFP3±Cln2PEST couldestablish a new dimension in such studies bycircumventing the need for synchronous culturesor cell disruption.

Acknowledgements

This work was funded by an award to SVA from the

National Institutes of Health (ROI GM57945). Plasmid

pYGFP3 was a gift from Brendon Cormack (Johns Hopkins

Med. School) and pFA6a-GFPMT-His3MX6 was from

Peter Philippsen (University of Basel). John E. Houghton

and Angela M. Avery are thanked for helpful discussions.

References

Alberghina L, Smeraldi C, Ranzi BM, Porro D. 1998. Control by

nutrients of growth and cell cycle progression in budding yeast,

analyzed by double-tag ¯ow cytometry. J Bacteriol 180:

3864±3872.

Amon A, Irniger S, Nasmyth K. 1994. Closing the cell cycle

circle in yeast: G2 cyclin proteolysis initiated at mitosis persists

until the activation of G1 cyclins in the next cycle. Cell 77:

1037±1050.

Andersen JB, Sternberg C, Kongsbak-Poulsen L, Petersen-Bjorn

S, Givskov M, Molin S. 1998. New unstable variants of green

¯uorescent protein for studies of transient gene expression in

bacteria. Appl Environ Microbiol 46: 2240±2246.

Ausubel FM, Brent R, Kingston RE, Moore DD, Seidman JG,

Smith JA, Struhl K (eds). 2000. Current Protocols in Molecular

Biology. Wiley: New York.

Avery SV, Harwood JL, Lloyd D. 1995. Quanti®cation and

characterization of phagocytosis in the soil amoeba Acantha-

moeba castellanii by ¯ow cytometry. Appl Environ Microbiol

61: 1124±1132.

Avery SV, Howlett NG, Radice S. 1996. Copper toxicity towards

Saccharomyces cerevisiae: dependence on plasma-membrane

fatty acid compositon. Appl Environ Microbiol 62: 3960±3966.

Avery SV, Malkapuram S, Mateus C, Babb KS. 2000. Copper/

zinc-superoxide dismutase is required for oxytetracycline

resistance of Saccharomyces cerevisiae. J Bacteriol 182: 76±80.

Barral Y, Jentsch S, Mann C. 1995. G1 cyclin turnover and

nutrient uptake are controlled by a common pathway in yeast.

Genes Devel 9: 399±409.

Beach DL, Salmon ED, Bloom K. 1999. Localization and

anchoring of mRNA in budding yeast. Curr Biol 9: 569±578.

Bell PJL, Davies IW, Att®eld PV. 1999. Facilitating functional

analysis of the Saccharomyces cerevisiae genome using an

EGFP-based promoter library and ¯ow cytometry. Yeast 15:

1747±1759.

Billinton N, Barker MG, Michel CE, Knight AW, Heyer W,

Goddard NJ, Fielden PR, Walmsley RM. 1998. Development

of a green ¯uorescent protein reporter for a yeast genotoxicity

biosensor. Biosens Bioelectron 13: 831±838.

Chal®e M, Tu Y, Euskirchen G, Ward WW, Prasher DC. 1994.

Green ¯uorescent protein as a marker for gene expression.

Science 263: 802±805.

Cho RJ, Campbell MJ, Winzeler EA, Steinmetz L, Conway A,

Wodicka L, Wolfsberg TG, Gabrielian AE, Landsman D,

Lockhart DJ, Davis RW. 1998. A genome-wide transcriptional

analysis of the mitotic cell cycle. Mol Cell 2: 65±73.

Corish P, Tyler-Smith C. 1999. Attenuation of green ¯uorescent

protein half-life in mammalian cells. Prot Eng 12: 1035±1040.

Cormack BP, Valdivia RH, Falkow S. 1996. FACS-optimized

mutants of the green ¯uorescent protein (GFP). Gene 173:

33±38.

Cormack BP, Bertram G, Egerton M, Gow NAR, Falkow S,

Brown AJP. 1997. Yeast-enhanced green ¯uorescent protein

(yEGFP): a reporter of gene expression in Candida albicans.

Microbiology 143: 303±311.

Cormack BP. 1998. Green ¯uorescent protein as a reporter of

transcription and protein localization in fungi. Curr Opin

Microbiol 1: 406±410.

Fikus MU, Mieczkowski PA, Koprowski P, Rytka J, Sled-

ziewska-Gojska E, Ciesla Z. 2000. The product of the DNA

damage-inducible gene of Saccharomyces cerevisiae, DIN7,

speci®cally functions in mitochondria. Genetics 154: 73±81.

Gietz RD, Woods RA. 1998. Transformation of yeast by the

lithium acetate single-stranded carrier DNA/PEG method. In

Methods in Microbiology, Vol. 26, Yeast Gene Analysis, Brown

AJP, Tuite MF (eds). Academic Press: London; 53±66.

Howlett NG, Avery SV. 1999. Flow cytometric investigation of



heterogeneous copper-sensitivity in asynchronously grown

Saccharomyces cerevisiae. FEMS Microbiol Lett 176: 379±386.

Lanker S, Valdivieso MH, Wittenburg C. 1996. Rapid degrada-

tion of the G1 cyclin Cln2 induced by CDK-dependent

phosphorylation. Science 271: 1597±1601.

Li X, Zhao X, Fang Y, Jiang X, Duong T, Fan C, Huang C-C,

Kain SR. 1998. Generation of destabilized green ¯uorescent

protein as a transcription reporter. J Biol Chem 273:

34970±34975.

Lloyd D (ed.) 1993. Flow Cytometry in Microbiology. Springer-

Verlag: Berlin.

Misteli T, Spector DL. 1997. Applications of the green

¯uorescent protein in cell biology and biotechnology. Nature

Biotechnol 15: 961±964.

Nasmyth K. 1996. At the heart of the budding yeast cell cycle.

Trends Genet 12: 405±412.

Natarajan A, Subramanian S, Srienc F. 1998. Comparison of

mutant forms of the green ¯uorescent protein as expression

markers in Chinese hamster ovary (CHO) and Saccharomyces

cerevisiae cells. J Biotechnol 62: 29±45.

Niedenthal RK, Riles L, Johnston M, Hegemann JH. 1996.

Green ¯uorescent protein as a marker for gene expression and

subcellular localization in budding yeast. Yeast 12: 773±786.

Pena MMO, Koch KA, Thiele DJ. 1998. Dynamic regulation of

copper uptake and detoxi®cation genes in Saccharomyces

cerevisiae. Mol Cell Biol 18: 2514±2523.

Reiser V, Ruis H, Ammerer G. 1999. Kinase activity-dependent

nuclear export opposes stress-induced nuclear accumulation

and retention of Hogl mitogen-activated protein kinase in the

budding yeast Saccharomyces cerevisiae. Mol Biol Cell 10:

1147±1161.

Salama SR, Hendricks KB, Thorner J. 1994. G1 cyclin degrada-

tion: the PEST motif of yeast Cln2 is necessary, but not

suf®cient, for rapid protein turnover. Mol Cell Biol 14:

7953±7966.

Schneider BL, Patton EE, Lanker S, Mendenhall MD, Witten-

burg C, Futcher B, Tyers M. 1998. Yeast G1 cyclins are

unstable in G1 phase. Nature 395: 86±89.

Spellman PT, Sherlock G, Zhang MQ, Iyer VR, Anders K, Eisen

MB, Brown PO, Botstein D, Futcher B. 1998. Comprehensive

identi®cation of cell cycle-regulated genes of the yeast

Saccharomyces cerevisiae by microarray hybridization. Mol

Biol Cell 9: 3273±3297.

Sternberg C, Christensen BB, Johansen T, Toftgaard NA,

Andersen JB, Givskov M, Molin S. 1999. Distribution of

bacterial growth activity in ¯ow-chamber bio®lms. Appl

Environ Microbiol 65: 4108±4117.

Tsien RY. 1998. The green ¯uorescent protein. Ann Rev Biochem

67: 509±544.

Wach A, Brachat A, Alberti-Segui C, Rebischung C, Philippsen

P. 1997. Heterologous HIS3 marker and GFP reporter

modules for PCR-targeting in Saccharomyces cerevisiae. Yeast

13: 1065±1075.

Weber-Ban EU, Reid BG, Miranker AD, Horwich AL. 1999.

Global unfolding of a substrate protein by the Hsp100

chaperone ClpA. Nature 401: 90±93.

Williams I, Paul F, Lloyd D, Jepras R, Critchley I, Newman M,

Warrack J, Giokarini T, Hayes AJ, Randerson PF, Venables

WA. 1999. Flow cytometry and other techniques show that

Staphylococcus aureus undergoes signi®cant physiological

changes in the early stages of surface-attached culture.

Microbiology 145: 1325±1333.

Wittenberg C, Sugimoto K, Reed SI. 1990. G1-speci®c cyclins of

S. cerevisiae: cell cycle periodicity, regulation by mating

pheromone, and association with the p34CDC28 protein

kinase. Cell 62: 225±237.



destabilized green fluorescent protein for monitoring dynamic changes in yeast gene expression with...

Documents