Quantitative analysis of the microbial metabolome by isotope dilution mass spectrometry using...

ANALYTICALBIOCHEMISTRY

Analytical Biochemistry 336 (2005) 164–171

www.elsevier.com/locate/yabio

Quantitative analysis of the microbial metabolome by isotope dilution mass spectrometry using uniformly 13C-labeled cell extracts

as internal standards

Liang Wu¤, Mlawule R. Mashego, Jan C. van Dam, Angela M. Proell, Jacobus L. Vinke, Cor Ras, Wouter A. van Winden, Walter M. van Gulik, Joseph J. Heijnen

Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC, Delft, The Netherlands

Received 28 April 2004Available online 23 November 2004

Abstract

A novel method was developed for the quantitative analysis of the microbial metabolome using a mixture of fully uniformly (U)13C-labeled metabolites as internal standard (IS) in the metabolite extraction procedure the subsequent liquid chromatography–elec-trospray ionization-tandem mass spectrometry (LC–ESI-MS/MS) analysis. This mixture of fully U 13C-labeled metabolites wasextracted from biomass of Saccharomyces cerevisiae cultivated in a fed-batch fermentation on fully U 13C-labeled substrates. Theobtained labeled cell extract contained, in principle, the whole yeast metabolome, allowing the quantiWcation of any intracellularmetabolite of interest in S. cerevisiae. We have applied the labeled cell extract as IS in the analysis of glycolytic and tricarboxylic acid(TCA) cycle intermediates in S. cerevisiae sampled in both steady-state and transient conditions following a glucose pulse. The use oflabeled IS eVectively reduced errors due to variations occurring in the analysis and sample processing. As a result, the linearity of cal-ibration lines and the precision of measurements were signiWcantly improved. Coextraction of the labeled cell extract with the sam-ples also eliminates the need to perform elaborate recovery checks for each metabolite to be analyzed. In conclusion, the methodpresented leads to less workload, more robustness, and a higher precision in metabolome analysis. 2004 Elsevier Inc. All rights reserved.

Keywords: Metabolomics; Isotope dilution; LC-MS; Intracellular metabolites; 13C labeling; Fed-batch cultivation; Saccharomyces cerevisiae

The integrated approach for the study of microbialphysiology requires comprehensive analysis of diVerentfunctional entities within the cell such as the levels ofmRNAs (transcriptome), proteins (proteome), and low-molecular weight metabolites (metabolome) [1]. Recently,the importance of metabolomic data has been increas-ingly recognized in various application Welds [2] such asthe quantiWcation of in vivo kinetics using stimulus–response experiments [3–5] and qualitative studies infunctional genomics [6] and physiology [7]. Despite the

* Corresponding author. Fax: +31 15 2782355.E-mail address: [email protected] (L. Wu).

0003-2697/$ - see front matter 2004 Elsevier Inc. All rights reserved.doi:10.1016/j.ab.2004.09.001

advent of metabolomics, eYcient and reliable quantiWca-tion of (intracellular) metabolite concentrations is stillimpeded by a number of experimental drawbacks inboth sample preparation and sample analysis.

In practice, sample preparation commences with thefast arrest of metabolism by, for example, quenching ina cold methanol–water solution [8–10] followed by anextraction step. DiVerent extraction procedures thatseek to optimize between complete extraction and lim-ited degradation of metabolites have been developed.Established protocols include, for example, boiling eth-anol treatment [9,10] and acid/alkaline extraction withfreeze–thaw cycles [11]. Because of the severe conditionsapplied (i.e., high temperature or acidity/alkalinity) and

mailto: [email protected]

mailto: [email protected]

Microbial metabolome analysis using uniformly 13C-labeled cell extracts / L. Wu et al. / Anal. Biochem. 336 (2005) 164–171 165

the instability of a large variety of intracellular species,degradation of metabolites is to be expected and recov-ery of each metabolite of interest during the sampleextraction procedure must be veriWed experimentally[9,10]. Therefore, the development of proper sample pro-cessing procedures is very time-consuming. Moreover,the overall sample preparation process is elaborate andinvolves multiple manual handlings that are prone tooperator-to-operator and run-to-run variations.

Intracellular metabolite concentrations in the obtainedcell extracts can be measured with a range of analyticaltechniques. The traditional enzymatic and chromato-graphic methods are being increasingly replaced by high-throughput mass spectrometric techniques that allow thesimultaneous identiWcation and quantiWcation of alarger number of metabolites with high selectivity, ade-quate sensitivity, and minimal sample use (e.g., 5–10 �lvs. 100 �l or more for enzyme assays). Various studieshave shown that liquid chromatography–electrosprayionization-mass spectrometry (LC–ESI-MS)1 is particu-larly appropriate for the quantiWcation of phosphory-lated carbon compounds such as those that areencountered in the study of primary metabolic pathways[12,13]. However, the performance of LC–ESI-MS canbe compromised by ion suppression eVects [14] that ren-der a diVerent signal response of the analyte due tochanges in the sample matrix. The composition of thesample matrix is inXuenced by, for example, samplepreparation procedures and the species of microorgan-ism. Thus, the evaluation of ion suppression eVectsencompasses extensive standard addition protocols foreach metabolite and each microorganism of interest,thereby seriously impairing the Xexibility of LC–ESI-MSin its implementation and strongly increasing therequired developmental eVorts.

The above-mentioned drawbacks in sample prepara-tion and LC–ESI-MS analysis can be tackled by the iso-tope dilution technique, where a stable isotope analog(i.e., isotopologue [15]) of the analyte compound is usedas an internal standard (IS) in the MS analysis. Thistechnique has found a large variety of applications, asreviewed in [16]. Because of the high physical–chemicalsimilarities between the labeled IS and the analyte,degradation during sample preparation, variations ininstrumental response, and ion suppression eVects inLC–ESI-MS can be compensated [17–20].

1 Abbreviations used: LC–ESI-MS, liquid chromatography–electro-spray ionization-mass spectrometry; IS, internal standard; U, uniform-ly; MS/MS, tandem mass spectrometry; TCA, tricarboxylic acid; G6P,glucose 6-phosphate; F6P, fructose 6-phosphate; F1,6bP, fructose 1,6-bisphosphate; G1P, glucose 1-phosphate; T6P, trehalose 6-phosphate;2PG, 2-phosphoglycerate; 3PG, 3-phosphoglycerate; PYR, pyruvate;CIT, citrate; OGL, oxoglutarate; FUM, fumarate; SUC, succinate;MAL, malate; iCIT, isocitrate; IDMS, isotope dilution mass spectrom-etry; PEP, phosphoenol pyruvate; RSE, relative standard error.

However, the availability of labeled intracellularmetabolites is scarce, and de novo synthesis of the isoto-pologue of interest is often required. Biosynthesis is thepreferred method for the preparation of biomolecules ofcomplex structure [21]; for example, 13C-labeled sugarswere obtained by means of photosynthesis from 13CO2[22]. In a number of publications [19,23], small-scalemicrobial fermentations were used to yield desiredlabeled compounds from labeled precursors. For metab-olome wide analysis, it is desirable to produce all labeledintracellular metabolites in one stroke. This can beachieved by cultivating organisms exclusively on sub-strates containing a speciWc isotope, such as deuteratedalgae grown on heavy water, where all of the hydrogenpresent is in the form of the heavy isotope [24]. Recently,Mashego et al. [25] cultivated 13C-labeled biomass byswitching the unlabeled feed of a chemostat culture to100% uniformly (U) 13C-labeled feed. They found that ashort labeling period (4 h) led to an incomplete replace-ment of the unlabeled metabolites in the primary metab-olism by U 13C-labeled ones, possibly due to therelatively slow turnover of unlabeled storage pools.Coextraction of the partially labeled biomass and unla-beled biomass samples and subsequent LC–ESI-tandemmass spectrometry (MS/MS) analysis provided the ratiosof each unlabeled metabolite and its U 13C-labeled isoto-pologue (with the necessary correction of the partiallabeling in the IS), whereby an improved precision wasobtained compared with conventional analysis withoutIS. However, the adopted methodology still has a num-ber of disadvantages. The production of labeled biomassby chemostat cultivation is rather laborious, and muchof the labeled material is lost in the eZuent from the fer-mentor. Also, the direct use of labeled biomass as IS pro-vides only the relative ratios between unlabeled and U13C-labeled metabolites and does not allow determina-tion of absolute metabolite concentrations in samples.The incomplete labeling of the metabolites in the pri-mary metabolism must be corrected, thereby increasingthe computational complexity and the error in theresults. More importantly, certain metabolite pools,especially the conserved moieties such as the adenineand pyridine nucleotides, have a relatively slow turnoverrate and, therefore, will be poorly labeled after a shortlabeling period, thereby impeding the isotopic dilution-based analysis of these physiologically important metab-olites.

We have circumvented the above-mentioned draw-backs by cultivating biomass in a small scale fed-batchfermentation on 100% U 13C-labeled substrates, fromwhich the labeled metabolites were extracted usingproper rapid sampling and quenching procedures [10].By using a fed-batch culture, all labeled precursors areeVectively incorporated into the biomass, resulting in ametabolome-wide labeling that, in theory, provides fullyU 13C-labeled IS for each intracellular metabolite of

166 Microbial metabolome analysis using uniformly 13C-labeled cell extracts / L. Wu et al. / Anal. Biochem. 336 (2005) 164–171

interest. Moreover, by adding the labeled cell extract toboth the unlabeled calibration standards and the sam-ples prior to extraction, absolute metabolite concentra-tions can be obtained instead of isotopologue ratios. Inthis article, we demonstrate the applicability of theselabeled cell extracts as IS in LC–ESI-MS/MS analysisand its advantages, as compared with conventional 12C-based methods, for the analysis of glycolytic andtricarboxylic acid (TCA) cycle intermediates of Saccharo-myces cerevisiae in both steady-state and transientconditions.

Materials and methods

Materials

U 13C6 glucose (>99%) and U 13C2 ethanol (>99%)were purchased from Campo ScientiWc (Veenendaal,The Netherlands). The naturally labeled metabolitesglucose 6-phosphate (G6P), fructose 6-phosphate (F6P),fructose 1,6-bisphosphate (F1,6bP), glucose 1-phos-phate (G1P), trehalose 6-phosphate (T6P), 2-phospho-glycerate (2PG), 3-phosphoglycerate (3PG), phosphoenolpyruvate (PEP), pyruvate (PYR), citrate (CIT), oxoglu-tarate (OGL), fumarate (FUM), succinate (SUC), andmalate (MAL) were obtained from Sigma (St. Louis,MO, USA). All chemicals used were of analyticalgrade.

Cultivation of labeled biomass

Saccharomyces cerevisiae (CEN.PK 113-7D) was cul-tivated in a 1-L fermentor (Applikon, Schiedam, TheNetherlands) equipped with a sampling port at the bot-tom of the vessel. Throughout the cultivation, the pHwas kept at 5, the temperature was kept at 30 °C, and theoverpressure was kept at 0.1 bar. The composition of themineral medium (0.2 L) used for the batch phase wasbased on Verduyn et al. [26], with 14.1 g/L U 13C6 glucoseas the sole carbon source. On depletion of the carbonsource, an aerobic fed-batch phase was started with anexponential feed proWle supporting a speciWc growthrate of 0.05 h¡1. The fed-batch medium contained 56.8 g/L U 13C6 glucose, 3.0 g/L U 13C2 ethanol, 16.7 g/L(NH4)2SO4, 10 g/L KH2PO4, 1.7 g/L MgSO4 · 7H2O,3.3 ml/L vitamin solution, and 3.3 ml/L solution withtrace elements. The vitamin solution contained 0.05 g/Lunlabeled D-biotin and 1 g/L unlabeled vitamin B1. Thecomposition of the trace element solution was accordingto Verduyn et al. CO2 in the air supply was completelyremoved during the entire cultivation, except for the ini-tial batch phase, by passing the air through two 1-L bot-tles containing 4 M KOH solution connected in series.The stirrer speed and the gas Xow rate were adjusted tokeep the dissolved oxygen tension above 20%.

Preparation of labeled cell extracts

The labeled biomass was sampled three times with a2-h interval between each sampling. To compensate forthe volume changes, the feed proWle was adjusted afterthe sampling to ensure a constant growth rate of0.05 h¡1. During each sampling, 100 ml of broth waswithdrawn from the bottom sampling port of the fer-mentor into 500 ml of continuously stirred 60% metha-nol kept at ¡40 °C. The 600-ml broth–methanol mixtureobtained from each sampling was divided into 100 testtubes in aliquots of 6 ml and kept at ¡40 °C. Centrifuga-tion and washing of the cells in each test tube were car-ried out according to Mashego et al. [25]. Thus, the threesampling runs gave 300 test tubes containing cell pelletsthat were stored at ¡80 °C no longer than 100 h beforebeing extracted in boiling ethanol (75%) as described byMashego et al. The 300 cell extracts obtained werepooled, well mixed, and subsequently redistributed in2-ml vials and then stored at ¡80 °C for later use as IS.

Preparation of unlabeled biomass samples

An aerobic carbon-limited chemostat culture of S.cerevisiae was grown at pH 5, 30 °C, and a dilution rateof 0.05 h¡1 in a 7-L fermentor (Applikon) with a workingvolume of 4 L. The feed contained 27.1 g/L glucose and1.4 g/L ethanol, which supports a biomass concentrationof approximately 15 g dry weight/L. After approximately170 h, steady-state samples were taken within a period of300 s. The chemostat culture was subsequently perturbedby a sudden increase of the extracellular glucose concen-tration by 1 g/L from the steady-state level of 20 mg/L.During the transient condition after the perturbation,samples were taken within a time window of 300 s. Thesampling, quenching, and extraction procedures fol-lowed essentially those described by Mashego et al. [25]except that 100 �l of labeled cell extract was added toeach cell pellet prior to the ethanol boiling step. Thisresulted in a 5£ dilution of the added labeled cell extractin the Wnal sample of 500 �l.

LC–ESI-MS/MS analysis

The LC–ESI-MS/MS method for quantiWcation ofthe glycolytic intermediates in S. cerevisiae has beendescribed previously [13]. The citric acid cycle com-pounds were analyzed with the same analysis methodand conditions. The metabolites 2PG and 3PG cannotbe resolved with the applied analytical procedure, andthe same held for CIT and isocitrate (iCIT). Conse-quently, only the sums of these compounds (i.e.,2PG + 3PG and CIT + iCIT) can be determined.

The labeling content of the obtained cell extract wasquantiWed by measuring the concentration of U 13C-labeled and unlabeled metabolite with a conventional


calibration line of unlabeled standards, of which thepeak areas corresponding to the mass M + 0 were deter-mined (where M is the molecular mass of the metaboliteor its fragment). With this procedure, we assume that thesensitivity of the analysis is the same for the U 13C-labeled and 12C-labeled compounds. We have no purelabeled standards, and this assumption cannot bechecked; however, it is plausible considering the similar-ity of physical and chemical behaviors of labeled andunlabeled compounds.

For isotope dilution mass spectrometry (IDMS) anal-ysis, labeled cell extracts were added to the naturallylabeled calibration standards in a 1:4 ratio. The ratio ofthe unlabeled and U 13C-labeled metabolites in the sam-ples and the calibration standards was subsequentlydetermined by measuring peak areas corresponding tothe masses M + 0 and M + N (where N is the number ofcarbon atoms of the metabolite or its fragment).

Results and discussion

Production of labeled cell extract

With the aim to obtain fully U 13C-labeled biomass, S.cerevisiae was grown on fully U 13C-labeled substrates(glucose and ethanol) in an aerobic fed-batch culture andthe preceding batch phase. Incorporation of unlabeledcarbon was prevented by removing unlabeled CO2 fromthe air supply and omitting unlabeled nonessential vita-mins from the medium. Therefore, the only sources of12C were the inoculum of the batch phase (»1 ml on atotal end volume of 500 ml), the 12C contamination of thelabeled substrates (<1%), the essential vitamins (unla-beled D-biotin and vitamin B1) present in trace amountsin the medium, and the 12CO2 incorporated into the bio-mass during growth in the initial batch phase when CO2in the air supply was not removed to prevent retardationof growth. Thus, extraction of the cultivated and U 13C-labeled biomass yields essentially the whole yeast metab-olome and so can, in principle, provide U 13C-labeled ISfor each intracellular metabolite of interest.

To assess the extent to which the obtained cell extractwas 13C labeled, the cell extract was analyzed by LC–ESI-MS/MS. The results, summarized in Table 1, show thatthe U 13C enrichments of the metabolites measured areclose to 1, indicating a negligible presence of unlabeledfractions that might interfere with the IDMS analysis.

Calibration with labeled cell extract as IS

Labeled cell extracts were added as IS in equalamounts to the unlabeled calibration standards, and thepeak areas of the U 13C and 12C metabolites were mea-sured by LC–ESI-MS/MS. Linear calibration lines wereobtained when the area ratios between the U 13C and 12C

metabolites are plotted against the known concentra-tions of 12C metabolites in the calibration standard.Because the peak areas of unlabeled metabolites weremeasured as well in the IDMS analysis, it was also possi-ble to obtain a conventional 12C calibration line. Thecalculated coeYcients of determination (R2) given inTable 2 show an improved linearity of the IS-based cali-bration lines compared with 12C-based calibration linesfor all metabolites analyzed. Fig. 1 shows several parityplots where the known concentrations in the standardsare plotted against the calculated concentrations accord-ing to the two diVerent calibration lines. It can be seen inthe Wgure that the outliers (e.g., SUC, CIT + iCIT) andnonlinearity [e.g., MAL, PEP] observed with the 12C cali-bration lines, which may be caused by instrumental vari-ations and/or ion suppression eVects, are eVectivelycorrected by the use of U 13C-labeled IS. This demon-strates the principle of isotope dilution, i.e., the ratiobetween the U 13C-labeled IS and the unlabeled analyte

Table 1Concentration ratios of U 13C-labeled and the sum of U 13C- and 12C-labeled intracellular metabolites in the labeled cell extract

Note. Concentration ratios and 95% conWdence intervals were deter-mined from quadruple measurements of a sample of the labeled cellextract.

Metabolite [U 13C]/([U 13C] + [12C])

G6P 0.99 § 0.01F6P 0.99 § 0.00F1,6bP 0.99 § 0.032PG + 3PG 1.00 § 0.03PEP 0.99 § 0.00PYR 0.97 § 0.00CIT + iCIT 1.00 § 0.00OGL 0.99 § 0.03SUC 1.00 § 0.00FUM 1.00 § 0.00MAL 1.00 § 0.00

Table 2CoeYcients of determination (R2 values) obtained from IS- and12C-based calibration lines

Note. For a proper comparison, linear calibration lines were usedwithout forced zero intercept for the same set of 12C-standard concen-trations.

Metabolite IS-basedcalibration line

12C-basedcalibration line

G6P 0.9972 0.9928F6P 0.9896 0.9724F1,6bP 0.9986 0.99272PG + 3PG 0.9919 0.9695PEP 0.9963 0.9340PYR 0.9755 0.8592CIT + iCIT 0.9927 0.9703OGL 0.9937 0.9903SUC 0.9978 0.9385FUM 0.9933 0.9700MAL 0.9900 0.9641


is much less sensitive to adverse experimental and instru-mental conditions than are the absolute peak areas.

The slope of each IS-based calibration line equals thereciprocal of the corresponding U 13C-labeled metabo-lite concentration in the labeled cell extracts and should,in principle, be independent of possible instrumentalvariations and show no variation between diVerent anal-ysis runs. To conWrm this, a single set of calibration stan-dards, to which labeled extracts were added, wassubjected to four independent calibration runs (Table 3).In the Wnal two calibration runs, a decrease of both the12C and U 13C peak areas was observed for all metabo-lites, indicating either loss of sensitivity of the LC–ESI-MS/MS or degradation of metabolites. This leads tolarge changes in the slopes of 12C-based calibration lines(e.g., for F6P, the estimated slopes in the four runs, in thesame sequence as in Table 3, are 207.6, 120.6, 11.0, and

20.4) but hardly aVects the IS-based calibration, demon-strating the robustness of the applied isotope dilutionmethodology.

Metabolite analysis in steady-state and transient samples

IDMS analysis was applied to the steady-state samplesfrom a chemostat culture of S. cerevisiae and transientsamples when the chemostat was perturbed by a pulse ofconcentrated glucose. Absolute concentrations of intra-cellular metabolites were obtained from the peak arearatios of 12C- and U 13C-labeled metabolites in the sam-ples and the IS-based calibration lines. This was possiblebecause the labeled cell extract was added to the calibra-tion standards and the samples to the same Wnalconcentration (i.e., diluted 5£). Because of the high physi-cal–chemical resemblance of the U 13C-labeled IS and

Fig. 1. Parity plots for selected metabolites: (A) PEP, (B) CIT + iCIT, (C) SUC, and (D) MAL. Known concentration in the standards (x axis) againstcalculated concentrations (y axis) according to the IS-based calibration line (�) and 12C-based calibration line (+).

Table 3Slopes and their 95% conWdence intervals of selected IS-based calibration lines in four LC–ESI-MS/MS runs

G6P F6P F1,6bP PEP OGL FUM

Run 1 0.042 § 0.002 0.30 § 0.02 0.24 § 0.01 0.59 § 0.03 0.15 § 0.01 0.32 § 0.02Run 2 0.042 § 0.002 0.32 § 0.02 0.24 § 0.01 0.60 § 0.05 0.16 § 0.01 0.31 § 0.01Run 3 0.042 § 0.002 0.28 § 0.01 0.25 § 0.00 0.59 § 0.03 0.16 § 0.01 0.32 § 0.02Run 4 0.042 § 0.002 0.29 § 0.03 0.24 § 0.01 0.63 § 0.02 0.15 § 0.01 0.33 § 0.02


the unlabeled analyte, metabolite degradation duringsample preparation, as well as ion suppression eVects andvariations in the sample analysis, were corrected. Itshould be mentioned, however, that the addition of IScannot correct for cell leakage during quenching and cen-trifugation or for incomplete extraction of metabolitesfrom the cells. For S. cerevisiae, it has been shown thatthe sample preparation procedure, as applied in the cur-rent study, eVectively preserves cell integrity by quench-ing with cold methanol and leads to complete release ofthe metabolites by extraction with boiling ethanol [9,10].

Table 4Steady-state intracellular metabolite concentrations (�mol/g dryweight) in a chemostat culture of S. cerevisiae

Note. Concentrations and 95% conWdence intervals were determinedfrom duplicate measurements of 16 steady-state samples.

Metabolite Concentration

G6P 2.11 § 0.05F6P 0.38 § 0.02F1,6bP 0.20 § 0.032PG + 3PG 1.33 § 0.02PEP 1.05 § 0.05CIT + iCIT 5.57 § 0.11OGL 0.09 § 0.00SUC 0.05 § 0.00FUM 0.06 § 0.00MAL 0.28 § 0.02

A number of metabolites (e.g., G1P, T6P) are presentin the labeled extracts in extremely low concentrationsand, therefore, cannot be properly detected in thecurrent LC–ESI-MS/MS setup. Therefore, the concen-tration of these metabolites in samples cannot be deter-mined by means of the isotope dilution method.However, this might be circumvented by, for example,adding a larger amount of labeled extract to samples andstandards, increasing sample injection volumes, orapplying diVerent fed-batch cultivation strategies for theproduction of labeled extract, thereby leading to diVer-ent concentrations of intracellular metabolites.

The average steady-state concentrations, determinedfrom 16 steady-state samples, are given for each metabo-lite in Table 4. The absolute concentrations of glycolyticintermediates correspond very well with those reportedin [25], as determined by conventional LC–MS/MS anal-ysis. The temporal proWles of several representativeintracellular metabolites during the transient conditionafter a glucose pulse are plotted in Fig. 2. The dynamicsof the glycolytic intermediates is comparable to pub-lished data [4,25], showing an increase in the concentra-tion of hexose phosphates and decreases in theconcentrations of 2PG, 3PG, and PEP. The concentra-tions of TCA cycle metabolites are much less perturbedthan those of glycolytic intermediates because a major

Fig. 2. Intracellular metabolite concentrations following a glucose pulse to a chemostat culture of S. cerevisiae: (A) G6P, (B) 2PG + 3PG, (C) OGL,and (D) MAL.


part of the increased carbon inXux is redirected to etha-nol instead of entering the TCA cycle.

Precision of isotope dilution LC–ESI-MS/MS

To evaluate the precision of IDMS analysis, the rela-tive standard errors (RSEs) of metabolites using IDMSand conventional 12C-based LC–ESI-MS/MS analysiswere compared. For each metabolite, the RSE of the twoanalytical methods was evaluated by calculating the RSEof each duplicate measurement of all 42 samples (steady-state and transient) from the S. cerevisiae fermentation.For IDMS, 42 RSEs of peak area ratios were calculated,and for the conventional 12C-based analysis, 84 RSEs ofpeak areas of both the 12C- and U 13C-labeled metabo-lites were obtained. By using duplicate measurements ofthe same sample, error sources other than the MS analy-sis are excluded. Moreover, the fairly large number of cal-culated RSEs ensures that the obtained distribution isrepresentative for both analytical methods. The averageRSEs for metabolites obtained from the two analyticalmethods are given in Table 5, which demonstrates thatisotope dilution leads to much smaller RSEs comparedwith the 12C-based method and, therefore, provides meta-bolomic data with a higher precision.

The decrease in the average RSE of duplicate mea-surements indicates that the inXuence of variations dur-ing LC–ESI-MS/MS analysis is eVectively reduced byisotope dilution. A number of error sources during sam-ple preparation (e.g., metabolite degradation) can beeVectively eliminated by isotope dilution as well. How-ever, inaccuracies in sample weight determination, aswell as errors in added volumes of labeled extracts,could still inXuence the accuracy of the Wnal result. Toevaluate the possible contribution of these remainingerror sources, we calculated the RSEs of 16 individually

Table 5Precision of the IDMS analysis evaluated by RSEs (percentages)

a RSE of the intracellular concentration measured in 16 steady-state samples.

b Average of RSEs from 42 duplicate area ratio measurements ofthe same steady-state samples and transient samples.

c Average of RSEs from 84 duplicate 12C and U 13C area measure-ments of the same steady-state samples and transient samples.

Metabolite RSE

Steady-state samples by IDMSa

IDMSb 12C-basedLC–ESI-MS/MSc

G6P 4.9 3.7 13.3F6P 7.8 11.3 15.7F1,6bP 10.4 3.7 9.62PG + 3PG 4.2 6.0 10.2PEP 9.5 7.5 13.5CIT + iCIT 3.7 3.4 5.4OGL 11.1 6.0 6.1SUC 20.1 12.3 14.8FUM 13.1 8.2 13.7MAL 10.3 7.4 14.5

processed steady-state samples of S. cerevisiae (i.e., con-taining all experimental error sources) and comparedthose in Table 5 with the average RSEs of the duplicatemeasurements of single samples by IDMS. It should bementioned that because of the limited population ofsteady-state samples, the calculated RSEs of steady-stateconcentrations might not be precise. Table 5 shows thatmore variations are present in steady-state concentra-tions than could be accounted for by errors in IDMSanalysis alone (except F6P and 2 + 3PG), indicating thatadditional errors might be involved in the sample prepa-ration. However, it must be noted that variation insteady-state metabolite concentrations partially reXectsthe true Xuctuations of intracellular metabolites in cellspresent in the fermentor occurring during the timeperiod when the samples were taken.

In conclusion, fully labeled cell extracts, obtained byfed-batch cultivation of S. cerevisiae on fully U 13C-labeled substrates, were used as IS in the IDMS analysisof low-molecular weight metabolites typically present inthe micromolar range in samples. We have demon-strated, with the analysis of glycolytic and TCA cycleintermediates in yeast, that the IDMS method presentedhere signiWcantly improves the precision of metaboliteconcentration measurements [by roughly a factor of two(see the Wnal two columns of Table 5)], as well as the lin-earity of calibration lines, compared with conventionalLC–MS/MS analysis. Method development eVort can belargely reduced as coextraction of the labeled extracttogether with the samples eliminates the need for labori-ous recovery checks. Extension of the presented methodto other mass spectrometric analyses can be readilymade because any possible metabolite degradation (e.g.,during derivatization steps needed for gas chromatogra-phy–MS) can be accounted for by labeled IS. Thus,IDMS with labeled extracts as IS enables a robust quan-tiWcation of intracellular metabolite concentrations withhigh precision and holds great potential for future meta-bolomics research.

Acknowledgment

This work was supported by the Netherlands Organi-zation for ScientiWc Research (NWO) and DSM.

References

[1] D. Delneri, F.L. Brancia, S.G. Oliver, Towards a truly integrativebiology through the functional genomics of yeast, Curr. Opin. Bio-technol. 12 (2001) 87–91.

[2] O. Fiehn, Metabolomics: the link between genotypes and pheno-types, Plant Mol. Biol. 48 (2002) 155–171.

[3] M. Rizzi, M. Baltes, U. Theobald, M. Reuss, In vivo analysis ofmetabolic dynamics in Saccharomyces cerevisiae: II. Mathematicalmodel, Biotechnol. Bioeng. 55 (1997) 592–608.


[4] D. Visser, Measuring and modeling in vivo kinetics of primarymetabolism, Ph.D. thesis, Delft University of Technology, 2002.

[5] A. Buchholz, J. Hurlebaus, C. Wandrey, R. Takors, Metabolo-mics: quantiWcation of intracellular metabolite dynamics, Biomol.Eng. 19 (2002) 5–15.

[6] L.M. Raamsdonk, B. Teusink, D. Broadhurst, N. Zhang, A. Hayes,M.C. Walsh, J.A. Berden, K.M. Brindle, D.B. Kell, J.J. Rowland,H.V. WesterhoV, K. van Dam, S.G. Oliver, A functional genomicsstrategy that uses metabolome data to reveal the phenotype ofsilent mutations, Nat. Biotechnol. 19 (2001) 45–50.

[7] H. Tweeddale, L. Notley-McRobb, T. Ferenci, EVect of slowgrowth on metabolism of Escherichia coli, as revealed by globalmetabolite pool (metabolome) analysis, J. Bacteriol. 180 (1998)5109–5116.

[8] W. de Koning, K. van Dam, A method for the determination ofchanges of glycolytic metabolites in yeast on a subsecond timescale using extraction at neutral pH, Anal. Biochem. 204 (1992)118–123.

[9] B. Gonzalez, J. Francois, M. Renaud, A rapid and reliable methodfor metabolite extraction in yeast using boiling buVered ethanol,Yeast 13 (1997) 1347–1355.

[10] H.C. Lange, M. Eman, G. van Zuijlen, D. Visser, J.C. van Dam, J.Frank, M.J. de Mattos, J.J. Heijnen, Improved rapid sampling forin vivo kinetics of intracellular metabolites in Saccharomyces cere-visiae, Biotechnol. Bioeng. 75 (2001) 406–415.

[11] U. Theobald, W. Mailinger, M. Baltes, M. Reuss, M. Rizzi, In vivoanalysis of metabolic dynamics in Saccharomyces cerevisiae: I.Experimental observations, Biotechnol. Bioeng. 55 (1997) 305–316.

[12] A. Buchholz, R. Takors, C. Wandrey, QuantiWcation of intracellu-lar metabolites in Escherichia coli K12 using liquid chromato-graphic-electrospray ionization tandem mass spectrometrictechniques, Anal. Biochem. 295 (2001) 129–137.

[13] J.C. van Dam, M. Eman, J. Frank, H.C. Lange, G.W.K. vanDedem, J.J. Heijnen, Analysis of glycolytic intermediates in Sac-charomyces cerevisiae using anion exchange chromatography andelectrospray ionization with tandem mass spectrometric detection,Anal. Chim. Acta 460 (2002) 209–218.

[14] T.M. Annesley, Ion suppression in mass spectrometry, Clin. Chem.49 (2003) 1041–1044.

[15] M.K. Hellerstein, R.A. Neese, Mass isotopomer distribution anal-ysis at eight years: theoretical, analytic, and experimental consid-erations, Am. J. Physiol. 276 (1999) E1146–E1170.

[16] T.A. Baillie, The use of stable isotopes in pharmacologicalresearch, Pharmacol. Rev. 33 (1981) 81–132.

[17] L.E. Sojo, G. Lum, P. Chee, Internal standard signal suppressionby co-eluting analyte in isotope dilution LC–ESI-MS, Analyst 128(2003) 51–54.

[18] L. Grey, B. Nguyen, P. Yang, Liquid chromatography-electro-spray ionization isotope dilution mass spectrometry analysis ofparaquat and diquat using conventional and multilayer solid-phase extraction cartridges, J. Chromatogr. A 958 (2002) 25–33.

[19] F. Magni, S. Pereira, M. Leoni, G. Grisenti, M. Galli Kienle,Quantitation of cyclosporin A in whole blood by liquid chroma-tography/stable isotope dilution electrospray ionization massspectrometry, J. Mass Spectrosc. 36 (2001) 670–676.

[20] Z. Kuklenyik, D.L. Ashley, A.M. Calafat, Quantitative detectionof trichloroacetic acid in human urine using isotope dilution high-performance liquid chromatography-electrospray ionization tan-dem mass spectrometry, Anal. Chem. 74 (2002) 2058–2063.

[21] N.A. MatwiyoV, D.G. Ott, Stable isotope tracers in the life sci-ences and medicine, Science 181 (1973) 1125–1133.

[22] V.H. Kollman, J.L. Hanners, J.Y. Hutson, T.W. Whaley, D.G. Ott,C.T. Gregg, Large-scale photosynthetic production of carbon-13-labeled sugars: the tobacco leaf system, Biochem. Biophys. Res.Commun. 50 (1973) 826–831.

[23] J. Evans, T.C. Wang, M.P. Heyes, S.P. Markey, LC/MS analysis ofNAD biosynthesis using stable isotope pyridine precursors, Anal.Biochem. 306 (2002) 197–203.

[24] J.J. Katz, H.L. Crespi, Deuterated organisms: cultivation and uses,Science 151 (1966) 1187–1194.

[25] M.R. Mashego, L. Wu, J.C. Van Dam, C. Ras, J.L. Vinke, W.A.Van Winden, W.M. Van Gulik, J.J. Heijnen, MIRACLE: mass iso-topomer ratio analysis of U-13C-labeled extracts—A new methodfor accurate quantiWcation of changes in concentrations of intra-cellular metabolites, Biotechnol. Bioeng. 85 (2004) 620–628.

[26] C. Verduyn, E. Postma, W.A. ScheVers, J.P. Van Dijken, EVect ofbenzoic acid on metabolic Xuxes in yeasts: a continuous-culturestudy on the regulation of respiration and alcoholic fermentation,Yeast 8 (1992) 501–517.

Quantitative analysis of the microbial metabolome by isotope dilution mass spectrometry using...

Documents

Transcript of Quantitative analysis of the microbial metabolome by isotope dilution mass spectrometry using...