Metabolomics

, Volume 4, Issue 3, pp 240–247

Leakage of adenylates during cold methanol/glycerol quenching of Escherichia coli

Authors

  • Hannes Link
    • Lehrstuhl für BioverfahrenstechnikTechnische Universität München
  • Bernd Anselment
    • Lehrstuhl für BioverfahrenstechnikTechnische Universität München
    • Lehrstuhl für BioverfahrenstechnikTechnische Universität München
Original Article

DOI: 10.1007/s11306-008-0114-6

Cite this article as:
Link, H., Anselment, B. & Weuster-Botz, D. Metabolomics (2008) 4: 240. doi:10.1007/s11306-008-0114-6

Abstract

Effective and rapid inactivation of cellular metabolism is a prerequisite for accurate metabolome analysis. Cold methanol quenching is commonly applied to stop any metabolic activity and, at the same time remaining the cells’ integrity. However, it is reported that especially prokaryotic cells like Escherichia coli and Corynebacterium glutamicum tend to leak intracellular metabolites during cold methanol quenching. In this work leakage of adenylates is quantified for different quenching fluids. Further, a methanol/glycerol based quenching fluid is proposed, which reduces leakage drastically compared to the commonly applied methanol/water solution (16% ATP leakage compared to more than 70%).

Keywords

QuenchingCold shockCell leakageAdenylatesGlycerol

1 Introduction

Strategies for analyzing metabolites exist since the early days of biochemistry, but capturing metabolomic changes in the cellular milieu are only beginning to be appreciated (Vaidyanathan 2005). Developments in microbial metabolomics are currently reviewed by Mashego et al. (2007) as well as Oldiges et al. (2007). Takors et al. (2007) demonstrate the importance of metabolome data in the context of system biology as it is applied in modern biotechnology. Theoretical frameworks like metabolic control analysis (Kacser and Burns 1973; Fell 1992) and the biochemical systems theory (Savageau 1969) provide mathematical tools to analyse metabolome data and identify key regulatory points of biochemical pathways. Novel mathematical approaches consider the uncertainty and errors in data obtained from a biological sample (Wu et al. 2004; Wang and Hatzimanikatis 2006; Link and Weuster-Botz 2007).

A critical evaluation of metabolome data is necessary, because sampling protocols are not standardized and several challenging tasks have to be considered when capturing in vivo metabolic snapshots. Due to high turn over rates of many intracellular metabolite pools, rapid sampling techniques with simultaneous inactivation of metabolic activity were developed (Weuster-Botz 1997; Schaefer et al. 1999; Buziol et al. 2002; Visser et al. 2002; Schaub et al. 2006; Hiller et al. 2007). Subsequently, metabolites are extracted, concentrated and analysed using highly selective and sensitive methods like GC-MS or LC-MS (Luo et al. 2007). Systematic and statistical errors propagate during sampling, sample preparation and analysis. Whilst statistical errors are considered by methods like standard addition, the influence of systematic errors is hardly quantified.

A crucial step is an effective and instant inactivation of metabolic activity, which is either achieved by simultaneous quenching and extraction using high temperatures (Schaub et al. 2006). Alternatively, cells can be separated from the supernatant prior to metabolite extraction. The most popular approach is quenching the sample into 60% (v/v) methanol/H2O at −50°C (Jensen et al. 1999; Buchholz et al. 2001; Al Zaid Siddiquee et al. 2004; Wittmann et al. 2004; Hoque et al. 2005; Magnus et al. 2006). After a centrifugation step the supernatant is discarded and cells are prepared for metabolite extraction.

Several authors noticed leakage of intracellular metabolites into the quenching fluid during sampling (Wittmann et al. 2004; Faijes et al. 2007; Mashego et al. 2007; Bolten et al. 2007). Bolten et al. (2007) investigated leakage of amino acids and several intermediate metabolites for a variety of organisms. They concluded that a cold shock phenomenon during the phase of rapid cooling is responsible for a major loss of metabolites into the quenching fluid and methanol quenching is not a suitable method. Recently, Villas-Bôas and Bruheim (2007) proposed a glycerol-saline quenching fluid and showed, that most metabolites are detected at significantly higher levels when compared to methanol/H2O. Mashego et al. (2007) concluded that in case metabolite leakage occurs, leaked metabolites should be quantifiable. Compared to the amount of sample, the quenching fluid is in excess to assure mixing temperatures below −20°C, therefore already low concentrated intracellular metabolites leaking into the quenching fluid get further diluted. Depending on the biomass concentration in the sample and the percentage of lost metabolites, concentrations in the quenching fluid are expected to be below 1 μM and highly sensitive analytical methods are required to quantify leaked metabolites.

Nasution et al. (2006) proposed to use ATP as indicator metabolite for cell leakage. Bolten et al. (2007) and Faijes et al. (2007) proposed to check metabolome data by means of the adenylates energy charge.

In this work the focus is on leakage of ATP, ADP and AMP during quenching of E. coli cultures. Concentrations of adenylates are determined in vivo, in the quenching fluid and in the culture supernatant in order to quantify cell leakage. Effects of buffers used with cold methanol quenching are investigated and the commonly applied methanol/H2O solution is compared with a methanol/glycerol mixture.

2 Materials and methods

2.1 Micro-organism and medium

Escherichia coli K12 wild type strain (DSM 498, DSMZ, Braunschweig, Germany) was grown on a defined medium: 0.2 g l−1 NH4Cl, 2 g l−1 (NH4)2SO4, 3.25 g l−1 KH2PO4, 2.5 g l−1 K2HPO4, 1.5 g l−1 NaH2PO4 · H2O, 1 g l−1 MgSO4 · 7H2O, 10 mg l−1 CaCl2 · 2H2O, 0.5 mg l−1 ZnSO4 · 7H2O, 0.25 mg l−1 CuCl2 · 2H2O, 2.5 mg l−1 MnSO4 · H2O, 1.75 mg l−1 CoCl2 · 6H2O, 0.125 mg l−1 H3BO3, 2.5 mg l−1 AlCl3 · 6H2O, 0.5 mg l−1 Na2MoO4 · 2H2O, 18.3 mg l−1 FeSO4 · 7H2O. Magnesium and trace-element solution were added via a sterile filter.

2.2 Cultivation

Pre-cultures were grown overnight in 1,000 ml shake-flasks filled with 100 ml medium (37°C, 250 rpm). All shake-flasks were inoculated by addition of 500 μl culture from a glycerol stock.

Batch cultivation was performed in a 1.5 l stirred tank bioreactor (Infors) with 1 l sterile minimal medium containing 30 g l−1 glucose inoculated with 100 ml pre-culture.

A 42 l stirred tank bioreactor (Infors) with 22 l minimal medium was utilised for fed-batch cultivation with an open-loop control of biomass formation according to Jenzsch et al. (2006). An exponential glucose feeding was applied after inoculation with 500 ml pre-culture:
$$ F_{f} (t) = \frac{{\mu _{{set}} \cdot c^{0}_{X} \cdot V_{0} }} {{Y_{{XS}} \cdot c^{0}_{S} }} \cdot \exp (\mu _{{set}} \cdot t) $$
(1)
The initial biomass concentration \( c^{0}_{X} \) was about 0.1 g l−1, initial volume V0 was 22 l, biomass yield on glucose YXS was estimated to 0.45 g g−1, glucose concentration in the feed \( c^{0}_{S} \) was 100 g l−1 for the first 6.5 h and 300 g l−1 for the remaining time. The specific growth rate μset was set to 0.5 h−1 in the first 6.5 h and to 0.1 h−1 subsequently. Dissolved O2 was maintained above 40% by controlling aeration and stirrer speed. The pH was adjusted with 25% (v/v) NH4OH to a constant value of 7.0.

2.3 Sampling of cultivation supernatant.

Ten millitre of culture broth was centrifuged for 10 min at 4,500g and 4°C. The supernatant was squeezed through a sterile filter.

2.4 Biomass analysis

Optical density (OD) was determined at 660 nm. The correlation of OD to cell dry weight was CDW = 0.56 × OD660 g l−1. Cell volume, cell size and cell number was estimated by a cell size analyser via the Coulter principle in a Multisizer II (Beckman Coulter, Krefeld, Germany).

2.5 Quenching fluid

All quenching fluids used in this work had a methanol content of 60% (v/v). The remaining 40% was either de-ionised water or pure glycerol. Compared to the glycerol-saline solution proposed by Villas-Bôas and Bruheim (2007), the 60% (v/v) methanol/glycerol solution is less viscous and allows sample processing at lower temperatures (−50°C).

2.6 Sampling of intracellular metabolites

Rapid sampling was performed with the sampling device described by Hiller et al. (2007). Before sampling, sample containers were filled with quenching fluid, pressure was reduced to 200 mbar and the container was cooled down to −50°C. The quenched cell suspension was transferred into a test tube and centrifuged for 6 min at 6,000g and −19°C to separate cells. Five millilitre of the supernatant was sterile filtered and stored at −20°C until analysis. The cell pellet was resuspended in 60% (v/v) methanol/H2O (−20°C). OD and cell volume of the cell suspension was determined before metabolite extraction.

2.7 Extraction of intracellular metabolites

Two millilitre of cell suspension and 200 μl standard solution were added to 4 ml de-ionised water (30 mM TEA, pH = 7) in pre-heated test tubes and shaken for 5 min at 95°C on a thermomixer (Comfort, Eppendorf, Hamburg, Germany). The test tubes were cooled down on ice and then centrifuged for 10 min at 4,500g and 4°C. Five millilitre of the supernatant was transferred into a new test tube and cooled to −80°C. The frozen supernatant was lyophilised and stored at −20°C until analysis. For analysis the sample was re-suspended with 500 μl 50% (v/v) methanol/H2O.

2.8 Standard addition

To take into account the influence of the extraction procedure on the stability of metabolites, all extractions were performed with standard addition before extraction. To each extraction sample a standard mixture of analysed metabolites was added in different concentrations (0, 100, 200, 300, 400, 500, 600, 700 μM).

Samples of quenching fluid and culture supernatant were mixed 1:10 with standard solution, resulting in six samples with different concentrations of standards (0, 1, 2, 3, 4, 5 μM).

The standard mixture contained AMP, ADP and ATP with a purity of >99% purchased from Sigma. Concentrations of metabolites were calculated via linear regression of the calibration curve (MATLAB R2006b, The MathWorks Inc.). The size of the 95% confidence interval is calculated according to the method of standard addition in DIN 32633 (German Institute for Standardization). Since standard addition is an extrapolation method, confidence intervals are higher compared to an external calibration.

2.9 LC-MS analysis

Quantification of metabolites was performed by an LC-MS method adopted from Luo et al. (2007). A Synergi Hydro-RP (C18) 150 mm × 2.1 mm I.D., 4 μm 80 Å particles column (Phenomenex, Aschaffenburg, Germany) with eluent A (10 mM tributylamine aqueous solution adjusted pH to 4.95 with 15 mM acetic acid) and eluent B (methanol) was applied for chromatography. A degassed binary gradient at 0.2 ml min−1 was achieved with a P 1100 HPLC pump (Thermo Finnigan, Dreieich, Germany). The injection volume was 10 μl in case of cell extract and 20 μl in case of supernatant of quenching fluid and culture broth. Sample temperature was 4°C and column temperature was set to 35°C. HPLC flow was transferred directly to the mass spectrometer via the electro-spray ionisation (ESI) interface. ESI-MS analysis was performed using an LCQ Advantage Iontrap mass spectrometer (Thermo Finnigan, Dreieich, Germany). N2 was used as sheath gas and helium served as damping gas. Data acquisition and analysis were conducted using the Xcalibur software (Thermo Finnigan, Dreieich, Germany). The following ESI parameters were employed: temperature of heated capillary: 350°C; electrospray capillary voltage: 2.5 kV; sheath gas: 60 arbitrary units; auxiliary gas: 20 arbitrary units; detection of negative ions in selected ion monitoring mode.

2.10 Enzymatic analysis of ATP

Analysis of ATP in the quenching fluid supernatant was also performed using a bioluminescence assay catalysed by firefly luciferase (ATP determination kit, sensitive assay, Biaffin GmbH & Co KG, Germany).

3 Results and discussion

3.1 Screening of quenching fluids

Leakage of ATP in three kinds of quenching fluids was investigated:
  • 60% (v/v) methanol/H2O

  • 60% (v/v) methanol/H2O buffered with TEA or HEPES in two concentrations (10 mM and 70 mM)

  • and 60% (v/v) methanol/glycerol.

Samples from a E. coli batch culture were quenched into the different quenching fluids. After centrifugation and sterile filtration of the quenching fluid supernatant, ATP was determined enzymatically. In order to reduce effects of methanol on the enzymatic activity each sample was diluted 1:10 with de-ionised water.

It is assumed that each sample contains the same amount of culture broth and the level of intracellular ATP was constant during time of sampling. The first assumption was verified by checking the total sample volume that was 25 ml quenching fluid and 7.5 ml culture broth. The assumption of a constant intracellular ATP level is applicable, since all samples were taken in a short time interval during exponential growth.

Figure 1 shows the concentration of ATP in the quenching fluid supernatant related to the amount of biomass in the sample. Error bars indicate the standard deviation of three enzymatic tests of the same sample.
https://static-content.springer.com/image/art%3A10.1007%2Fs11306-008-0114-6/MediaObjects/11306_2008_114_Fig1_HTML.gif
Fig. 1

Concentration of ATP in the supernatant of different quenching fluids after sampling, centrifugation and sterile filtration. (A) 60% (v/v) methanol/H2O, (B) 60% (v/v) methanol/H2O 10 mM TEA, (C) 60% (v/v) methanol/H2O 10 mM HEPES, (D) 60% (v/v) methanol/H2O 70 mM TEA, (E) 60% (v/v) methanol/H2O 70 mM HEPES, (F) 60% (v/v) methanol/glycerol

ATP was present in the supernatant of all quenching fluids, whereas no ATP was detected in the culture supernatant (data not shown). Therefore, ATP in the quenching fluid is a result of leakage during sampling. Equal concentrations of ATP were present in the methanol solution (A) and the quenching fluids buffered with 10 mM of TEA (B) and HEPES (C). Lower concentrations of ATP are determined in the quenching fluids buffered with 70 mM TEA and HEPES (D and E). A considerable reduction of ATP leakage is observed in case of the methanol/glycerol buffer (F). As glycerol is known as protecting agent also used for cryostocks it might have a stabilizing effect during quenching.

3.2 Leakage of adenylates in a glycerol based quenching fluid

Cell leakage using a methanol/glycerol quenching fluid was quantified and compared to a methanol/H2O solution. The adenylates ATP, ADP and AMP were measured in the cell extract, the supernatant of the quenching fluid and of the culture broth during a fed-batch cultivation of E. coli using LC-MS. The theoretical growth resulting from formal Monod kinetics and the feeding strategy in Eq. 1 is shown together with measured values in Fig. 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs11306-008-0114-6/MediaObjects/11306_2008_114_Fig2_HTML.gif
Fig. 2

Theoretical (solid line) and measured (filled circles) biomass formation during fed-batch cultivation of E. coli (μ = 0.5 h−1 up to a process time of 6.5 h; μ = 0.1 h−1 after 6.5 h). Samples for metabolite analysis were taken two times (1 and 2)

Sampling for metabolome analysis was performed two times (indicated by the arrows in Fig. 2). Each time three types of samples are withdrawn: (i) 7.5 ml culture broth quenched into 25 ml of quenching fluid G: 60% (v/v) methanol/H2O with 30 mM TEA, (ii) 7.5 ml culture broth quenched into 25 ml of quenching fluid H: 60% (v/v) methanol/glycerol with 30 mM TEA, and (iii) 15 ml culture broth (without quenching) for analysis of extracellular metabolites and biomass. A concentration of 30 mM TEA (instead of 70 mM) was applied in order to prevent negative effects with the ESI interface of the mass spectrometer. The operation characteristics of the sampling device are detailed by Hiller et al. (2007) and a rapid inactivation of metabolism is assured. Samples at each time point are taken within less than 1 min. Due to the controlled conditions in the bioreactor a metabolic steady state can be assumed and the biological variation between samples G and H can be neglected. Samples with quenching fluid G and quenching fluid H were prepared for analysis in parallel. For this reason differences between sample G and H resulting from sample processing can be excluded.

OD660 and the size of cells was measured in triplicate for each sample. Figure 3 shows the specific cell volume, following from these measurements. Error bars result from error propagation of measured OD660 and size of cells.
https://static-content.springer.com/image/art%3A10.1007%2Fs11306-008-0114-6/MediaObjects/11306_2008_114_Fig3_HTML.gif
Fig. 3

Specific cell volume determined in the two direct samples (1, 2) and in the quenched samples using quenching fluid G and quenching fluid H (1 G, 2 G, 1 H and 2 H)

The specific cell volume is consistent for samples taken directly from the culture broth and the quenched samples using quenching fluids G (about 1.3 μl mgCDW−1). The size of the cells determined in the glycerol based quenching fluid H is higher in both cases. This might be a result of glycerol interacting with the cell wall. In the following, intracellular concentrations are computed with a specific cell volume of 1.3 μl mgCDW−1.

3.3 Intracellular concentrations of adenylates

In Fig. 4 intracellular concentrations of adenylates are shown for samples 1 and 2. Results obtained with quenching fluid G (1G and 2G) are opposed to those of quenching fluid H (1H and 2H). Error bars indicate the 95% confidence interval obtained from the linear regression analysis of standard addition to one sample.
https://static-content.springer.com/image/art%3A10.1007%2Fs11306-008-0114-6/MediaObjects/11306_2008_114_Fig4_HTML.gif
Fig. 4

Intracellular concentrations of adenylates. Samples were taken two times (1 and 2). Two different quenching fluids were used: 60% (v/v) methanol/H2O with 30 mM TEA (G) and 60% (v/v) methanol/glycerol with 30 mM TEA (H). Error bars indicate the 95% confidence interval obtained from the linear regression analysis of standard addition

In all cases intracellular concentrations were higher if quenching fluid H was applied. Differences are most apparent for ATP and still significant for ADP, whereas almost equal concentrations of AMP are detected in the four samples.

A metabolic steady state between the two sampling points can be assumed from a theoretical point of view (pseudo-steady-state hypothesis, Vallino and Stephanopoulos 1993). This assumption is clearly reflected in the constant concentrations of adenylates obtained with each method. Hence, the described sampling procedure proved to be robust against statistical errors. This is also reflected by the coefficient of correlation (R2) of the linear regression obtained from standard addition, which was higher than 0.95 in all cases. However, the tremendous impact of systematic errors resulting from cellular leakage becomes obvious when the concentrations of ATP are compared for both quenching fluids.

3.4 Concentrations of adenylates in the quenching fluid supernatant

The concentrations of adenylates found in the quenching fluid supernatant are shown in Fig. 5.
https://static-content.springer.com/image/art%3A10.1007%2Fs11306-008-0114-6/MediaObjects/11306_2008_114_Fig5_HTML.gif
Fig. 5

Concentrations of adenylates in the supernatant of two quenching fluids after sampling, centrifugation and sterile filtration. Sixty percent (v/v) methanol/H2O with 30 mM TEA (G) and 60% (v/v) methanol/glycerol with 30 mM TEA (H). Error bars indicate the 95% confidence interval obtained from the linear regression analysis of standard addition

In all cases concentrations in quenching fluid G are higher compared to quenching fluid H. In accordance with the results of intracellular concentrations, differences are most significant in case of ATP.

3.5 Concentrations of adenylates in the culture supernatant

Figure 6 shows the concentration of ATP in the supernatant of cultivation medium. ADP and AMP were not detected.
https://static-content.springer.com/image/art%3A10.1007%2Fs11306-008-0114-6/MediaObjects/11306_2008_114_Fig6_HTML.gif
Fig. 6

Concentration of ATP in the supernatant of the cultivation broth

In both samples about 2 μM ATP was detected in culture supernatant, coming very likely from cell lysis during cultivation.

3.6 Quantification of cell leakage

The measured concentrations of adenylates in cell extract, the quenching fluid and culture supernatant were used to quantify the percentage of leakage. Concentrations in the quenching fluid supernatant were first corrected for the amount in the culture supernatant and then related to the intracellular volume in the quenched sample (grey filled bars in Fig. 7). Together with intracellular concentrations determined from cell extract the real intracellular pools were estimated.
https://static-content.springer.com/image/art%3A10.1007%2Fs11306-008-0114-6/MediaObjects/11306_2008_114_Fig7_HTML.gif
Fig. 7

Concentrations of adenylates in cell extract (blank and black bars) and in the supernatant of quenching fluid G and H (grey bars) are shown related to the intracellular volume. The percentage of leaked adenylates is additionally given

Considering that the balances in Fig. 7 involve several measurements, which are all subject to errors, the total values are in good agreement. It is remarkable that the percentage of leaked adenylates is almost constant for each metabolite using quenching fluid H.

In general, leakage seems to be more specific than supposed so far (Bolten et al. 2007). In all cases, leakage is more apparent for ATP than for ADP and AMP. This might be an explanation why intracellular adenylates obtained by methanol quenching in other studies result in values for the energy charge (EC) which are not within physiological meaningful ranges (e.g., Buchholz et al. 2001). As the adenylates energy charge is computed as:
$$ {\text{EC}} = \frac{{({\text{ATP}} + {\text{ADP}}/2)}} {{({\text{ATP}} + {\text{ADP}} + {\text{AMP}})}}, $$
(2)
there would be no influence on the EC in case of unspecific leakage (e.g., 70% loss of ATP, ADP and AMP). Using the corrected values of adenylates found in this study the EC is between 0.72 (quenching fluid H) and 0.59 (quenching fluid G). For comparison, Schaub et al. (2006) determined an EC of 0.66 for E. coli growing also with μ = 0.1 h−1 by quenching the whole culture broth with a heat exchanger.

4 Concluding remarks

It was shown that ATP is an appropriate indicator for metabolite leakage, as it is not excreted from the cell interior under normal conditions and only present in low concentrations in the cultivation supernatant. Further, the highly sensitive luciferase enzyme assay is a convenient method to check samples for ATP leakage.

Severe leakage of ATP (72–79%) was observed using a buffered methanol/H2O quenching fluid, which is commonly applied in most sampling protocols reported so far. Screening of several quenching fluids revealed that the applied buffers like TEA and HEPES have only marginal influence on metabolite leakage. In contrast, a methanol/glycerol quenching fluid could significantly reduce leakage of ATP (15–16%). Further studies must investigate the effect of glycerol on other metabolites which were not assessed in this study. A problem is that the concentration of extracellular metabolites often exceeds the concentration of intracellular ones, so that the amount of leakage in the quenched sample is hard to quantify.

Especially in case of central metabolism most metabolites have high turn-over rates and cold methanol quenching is indispensable. Alternative methods like quenching of the whole culture broth have draw-backs, like correction for metabolites in cultivation medium and high salt concentrations interfering with LC-MS analysis. Fast filtration as proposed by Bolten et al. (2007) is only applicable if metabolites with high time constants are quantified. In accordance with the results of Villas-Bôas and Bruheim (2007) we conclude that glycerol is a promising component for quenching fluids, since it reduces cell leakage significantly.

Acknowledgement

This study is based on the work supported by the Deutsche Forschungsgemeinschaft DFG under Grant No. WE 2715/10-1.

Copyright information

© Springer Science+Business Media, LLC 2008