Anal. Chem. 2007, 79, 3843-3849
Sampling for Metabolome Analysis of Microorganisms Christoph J. Bolten,† Patrick Kiefer,‡,⊥ Fabien Letisse,‡,§ Jean-Charles Portais,‡,§ and Christoph Wittmann*,†
Biochemical Engineering, Saarland University, Saarbru¨cken, Germany, Laboratoire Biotechnologie-Bioproce´ de´ s, UMR-CNRS 5504, UMR INRA 792, Toulouse, France, and Universite´ Paul Sabatier, Toulouse, France
In the present work we investigated the most commonly applied methods used for sampling of microorganisms in the field of metabolomics in order to unravel potential sources of error previously ignored but of utmost importance for accurate metabolome analysis. To broaden the significance of our study, we investigated different Gramnegative and Gram-positive bacteria, i.e., Bacillus subtilis, Corynebacterium glutamicum, Escherichia coli, Gluconobacter oxydans, Pseudomonas putida, and Zymononas mobilis, and analyzed metabolites from different catabolic and anabolic intracellular pathways. Quenching of cells with cold methanol prior to cell separation and extraction led to drastic loss (>60%) of all metabolites tested due to unspecific leakage. Using fast filtration, Gram-negative bacteria also revealed a significant loss (>80%) when inappropriate washing solutions with low ionic strength were applied. Adapting the ionic strength of the washing solution to that of the cultivation medium could almost completely avoid this problem. Gram-positive strains did not show significant leakage independent of the washing solution. Fast filtration with sampling times of several seconds prior to extraction appears to be a suitable approach for metabolites with relatively high intracellular level and low turnover such as amino acids or TCA cycle intermediates. Comparison of metabolite levels in the culture supernatant and the cell interior revealed that the common assumption of whole broth quenching protocols attributing the metabolites found exclusively to the intracellular pools may not be valid in many cases. In such cases a differential approach correcting for medium-contained metabolites is required. The analysis of the metabolome, i.e., identification and quantification of intracellular metabolites, has recently emerged as an important tool for profiling of biological systems. Metabolomics, in combination with fluxomics, proteomics, or transcriptomics promises to support metabolic engineering strategies, toward optimized producer strains, as well as systems biology approaches. Recent years in metabolomics research exhibit a strong accelera* To whom correspondence should be addressed. Phone: +49-681-302-2205. Fax: +49-681-302-4572. E-mail:
[email protected]. † Saarland University. ‡ UMR-CNRS. § Universite´ Paul Sabatier. ⊥ Current address: ETH Zu ¨ rich, Zu ¨ rich, Switzerland. 10.1021/ac0623888 CCC: $37.00 Published on Web 04/06/2007
© 2007 American Chemical Society
tion of development. They have borne new analytical techniques for the various metabolites, as well as new data-mining and modeling tools to handle and interpret the large metabolome data sets. With the use of these novel tools metabolome analysis has been applied to different biological systems involving different microorganisms, plants, or mammalian cells. Sampling is especially critical in metabolome analysis due to high exchange rates and small pool sizes of the metabolites of interest.1 Due to this, quenching of the cells during sampling is usually applied. The most popular method for microbial cells is quenching with cold methanol, maintaining the sample temperature below -20 °C, followed by collection of the cells by a centrifugation step prior to extraction.2 This approach has been originally developed and applied for yeast, but has later also been used in studies of different bacteria.3-9 For animal and plant cells liquid nitrogen is commonly utilized for quenching.10 Also this technique is applied for microbial cells, whereby no separation of culture supernatant and cells is carried out.11 Other approaches for metabolite sampling are based on fast filtration1 or quenching of the cells through fast heating.12 Undoubtedly, metabolome analyses and also the conclusions drawn from the obtained data rise and fall with appropriate sampling. Despite this, potential problems connected to sampling in metabolome analysis have not been properly considered. In most studies sampling protocols are simply adapted from the literature without critically validating them for the given case and the investigated organism. This, however, can lead to enormous errors. As an example, the Gram-positive bacterium Corynebacterium glutamicum loses significant amounts of intracellular amino acids during cold methanol quenching due to unspecific cold (1) Wittmann, C.; Kro ¨mer, J. O.; Kiefer, P.; Binz, T.; Heinzle, E. Anal. Biochem. 2004, 327, 135-139. (2) de Koning, W.; van Dam, K. Anal. Biochem. 1992, 204, 118-123. (3) Buchholz, A.; Hurlebaus, J.; Wandrey, C.; Takors, R. Biomol. Eng. 2002, 19, 5-15. (4) Buchholz, A.; Takors, R.; Wandrey, C. Anal. Biochem. 2001, 295, 129137. (5) Li, M.; Ho, P. Y.; Yao, S.; Shimizu, K. J. Biotechnol. 2005. (6) Moritz, B.; Striegel, K.; De Graaf, A. A.; Sahm, H. Eur. J. Biochem. 2000, 267, 3442-3452. (7) Oldiges, M.; Kunze, M.; Degenring, D.; Sprenger, G. A.; Takors, R. Biotechnol. Prog. 2004, 20, 1623-1633. (8) Zhao, Y.; Lin, Y. H. Process Biochem. 2002, 37, 1455-1461. (9) Zhu, J.; Shimizu, K. Appl. Microbiol. Biotechnol. 2004, 64, 367-375. (10) Viant, M. R.; Pincetich, C. A.; de Ropp, J. S.; Tjeerdema, R. S. Metabolomics 2005, 1, 149-158. (11) Dominguez, H.; Rollin, C.; Guyonvarch, A.; Guerquin-Kern, J. L.; CocaignBousquet, M.; Lindley, N. D. Eur. J. Biochem. 1998, 254, 96-102. (12) Schaub, J.; Schiesling, C.; Reuss, M.; Dauner, M. Biotechnol. Prog. 2006, 22, 1434-1442.
Analytical Chemistry, Vol. 79, No. 10, May 15, 2007 3843
shock leading to severe underestimation of metabolite levels.1 Metabolite leakage was also previously shown for Gram-positive Lactococcus lactis when using cold methanol for quenching.13 Additionally, bacterial cells are known to be sensitive to osmotic pressure, i.e., require a compatible ionic strength of the medium to maintain their integrity,14 but many protocols involve treatment of the cells with solutions of extremely low ionic strength, such as deionized water.15 In this regard we have tested currently used sampling methods in metabolomics. This includes methods with cell separation such as cold methanol quenching or fast filtration. Additionally, the suitability of methods without cell separation such as liquid nitrogen quenching or fast heating of the whole broth was investigated by assessing metabolite levels in the medium that could potentially interfere. To broaden the significance of our study, we have investigated different bacteria receiving high interest as model organisms or as production strains in biotechnology, i.e., Bacillus subtilis, C. glutamicum, Escherichia coli, Gluconobacter oxydans, Pseudomonas putida, and Zymomonas mobilis. This includes both Gram-negative and Gram-positive bacteria to account for a possible influence of the cell wall structure. Hereby, we analyzed intracellular metabolites from different intracellular catabolic and anabolic pathways, i.e., glycolysis, the pentose phosphate pathway, the tricarboxylic acid cycle, and amino acid biosynthesis. MATERIALS AND METHODS Strains. The C. glutamicum ATCC 13032 was obtained from the American Type and Culture Collection (Manassas, VA). The E. coli K12 (DSMZ 2670), G. oxydans (DSMZ 3503), and Z. mobilis (DSMZ 424) were from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). The P. putida KT2440 was kindly donated by Ken Timmis, German Research Centre for Biotechnology (Braunschweig, Germany). The B. subtilis 168 was kindly provided by Jo¨rg Stu¨hlke, University of Go¨ttingen (Go¨ttingen, Germany). Chemicals. Yeast extract and peptone were obtained from Difco Laboratories (Detroit, MI). All other chemicals were from Sigma (Steinheim, Germany) and Merck (Darmstadt, Germany) and of analytical grade. Media and Cultivation. For preculture and main culture of B. subtilis, C. glutamicum, E. coli, and P. putida a minimal medium with glucose as a sole carbon source was used as described elsewhere.16 The medium for B. subtilis was additionally supplemented with tryptophan to a final concentration of 0.05 g L-1. The G. oxydans was always grown on a complex medium containing, per liter, 5 g of yeast extract, 3 g of peptone, and 25 g of mannitol. Precultivation of Z. mobilis was carried out on a complex medium containing, per liter, 10 g of peptone, 10 g of yeast extract, and 20 g of glucose. A minimal medium described in ref 17 was used for the main cultivation of Z. mobilis. In studies on whole culture quenching with liquid nitrogen the salt concentration in the medium used for B. subtilis, C. glutamicum, E. coli, and P. putida had to be lowered due to incompatibility of salts with the LC(13) Jensen, N. B.; Jokumsen, K. V.; Villadsen, J. Biotechnol. Bioeng. 1999, 63, 356-362. (14) Leder, I. G. J. Bacteriol. 1972, 111, 211-219. (15) Soga, T.; Ohashi, Y.; Ueno, Y.; Naraoka, H.; Tomita, M.; Nishioka, T. J. Proteome Res. 2003, 2, 488-494. (16) Becker, J.; Klopprogge, C.; Zelder, O.; Heinzle, E.; Wittmann, C. Appl. Environ. Microbiol. 2005, 71, 8587-8596. (17) Fuhrer, T.; Fischer, E.; Sauer, U. J. Bacteriol. 2005, 187, 1581-1590.
3844
Analytical Chemistry, Vol. 79, No. 10, May 15, 2007
MS/MS applied for analysis of metabolites from glycolysis, the pentose phosphate pathway, and the tricarboxylic acid cycle. In such studies, the concentration of potassium buffer and ammonium were reduced to 30% of the normal value. All cultivations were carried out in duplicate at 30 °C and 230 rpm on a rotary shaker (shaking diameter Ø 25 mm, Multitron II, Infors AG, Bottmingen-Basel, Switzerland). Aerobic cultures were grown in 500 mL baffled shake flasks with 50 mL of medium. Anaerobic cultures of Z. mobilis were grown in 100 mL shake flasks with 50 mL of medium, whereby the medium was gassed with filter-sterilized N2. Sampling of Cultivation Supernatant. Cell suspension (2 mL) was sucked into a 2 mL plastic syringe and directly squeezed through a sterile filter (polyvinylidene fluoride, 0.2 µm pore size, Roth, Karlsruhe, Germany). Sampling of Intracellular Metabolites. In the current work different protocols were applied for intracellular metabolite sampling. For sampling with cold methanol quenching 5 mL of cell suspension was sucked into a precooled plastic syringe filled with 10 mL of methanol (-58 °C, 60%, 10 mM HEPES).6 The temperature of the quenched sample was always below -20 °C as checked by temperature measurement. Subsequently, the mixture was immediately transferred into a falcon tube and centrifuged for 5 min at 10 000g and -19 °C (Biofuge Stratos, Kendro Laboratory Products, Langenselbold, Germany). For sampling via fast filtration 2 mL of cell suspension was harvested by vacuum filtration (cellulose nitrate, 0.2 µm pore size, 25 mm, Sartorius, Go¨ttingen, Germany) and washed four times each with 4 mL of NaCl (0.9%, room temperature) solution (the whole filtration procedure including the washing was finished in less than 30 s). In each experiment sampling was carried out fourfold in parallel. Intracellular Metabolite Extraction. Cells harvested via methanol quenching or via fast filtration as described above were incubated in 2 mL of R-aminobutyrate solution (200 µM) for 15 min at 100 °C for metabolite extraction. Subsequently, the extracts were cooled on ice, transferred into 2 mL tubes, and centrifuged (5 min, 16 000g, 4 °C, Biofuge fresco, Kendro Laboratory Products, Langenselbold, Germany) to remove cell debris. The obtained extract was directly used for metabolite quantification. Analytics. Cell concentration was assessed by measurement of optical density (OD) and cell dry weight (CDW) as described earlier (ref 16). The correlation of OD to CDW was CDW ) 0.289OD × g L-1 for B. subtilis and CDW ) 0.439OD × g L-1 for E. coli. The correlation factors for P. putida (CDW ) 0.565OD × g L-1) and C. glutamicum (CDW ) 0.353OD × g L-1) were taken from the literature.1,18 For G. oxydans and Z. mobilis the correlation factor was assumed to be that of E. coli. Amino acids were quantified by HPLC, using R-aminobutyrate as internal standard, as described previously.19 Glycolytic intermediates and organic acids of the tricarboxylic acid cycle were quantified by ion chromatography tandem mass spectrometry using a cell extract of E. coli, cultivated on 99% [13C6] glucose (Campro Scientific, Veenendaal, Netherlands), as internal standard.20 Liquid anion (18) Stephan, S.; Heinzle, E.; Wenzel, S. C.; Krug, D.; Mu ¨ ller, R.; Wittmann, C. Process Biochem. 2006, 41, 2146-2152. (19) Kro ¨mer, J. O.; Fritz, M.; Heinzle, E.; Wittmann, C. Anal. Biochem. 2005, 340, 171-173. (20) Wu, L.; Mashego, M. R.; van Dam, J. C.; Proell, A. M.; Vinke, J. L.; Ras, C.; van Winden, W. A.; van Gulik, W. M.; Heijnen, J. J. Anal. Biochem. 2005, 336, 164-171.
Figure 1. Quantification of glutamate loss from B. subtilis, C. glutamicum, E. coli, G. oxydans, P. putida, and Z. mobilis utilizing fast filtration (ref 1) and methanol quenching (ref 6). In the case of filtration (F), the obtained supernatant is the filtrate. In the case of methanol quenching it describes the obtained supernatant after cell separation by centrifugation. All data are corrected for glutamate in the medium so that the glutamate detected in the supernatant exclusively originates from the cell interior. The data are given in µmol (g cell dry weight)-1.
exchange chromatography was performed as described previously.21 Before analysis, all cell extracts were mixed with a defined volume of the internal standard solution. RESULTS In the present work we have tested the mainly applied sampling methods in metabolomics. This includes methods with cell separation such as cold methanol quenching or fast filtration. Additionally, the suitability of methods without cell separation such as liquid nitrogen quenching or fast heating of the whole broth was investigated by assessing metabolite levels in the medium that could potentially interfere. Overall, we have investigated different bacteria receiving high interest as model organisms or as production strains in biotechnology, i.e., B. subtilis, C. glutamicum, E. coli, G. oxydans, P. putida, and Z. mobilis. Metabolome Analysis Involving Cell Separation during Sampling. Amino Acids. Potential metabolite leakage during sampling and cell separation was assessed by quantifying metabolites in the normally discarded fractions of the methanol supernatant (from methanol quenching) and the filtrate (from fast filtration). In addition, amino acids were quantified in the culture broth to correct the data obtained for the methanol supernatant and the filtrate for molecules not stemming from the cell interior, but from the medium, and account exactly for the metabolite leakage. Glutamate is a significant candidate to evaluate the two methods of methanol quenching and fast filtration. This amino acid generally exhibits a high intracellular level and a high turnover time in the range of 1 h facilitating its quantification.1 For all bacterial strains tested the intracellular glutamate level measured was drastically lower, when employing the methanol quenching protocol as compared to fast filtration (Figure 1). Glutamate was generally detected in significant amounts in the methanol quenching supernatant. The corresponding data are corrected for the glutamate in the culture broth. It is obvious that a substantial fraction of the intracellular glutamate was released (21) Kiefer, P.; Nicolas, C.; Letisse, F.; Portais, J. C. Anal. Biochem. 2007, 360, 182-188.
from the cells during the methanol quenching. Concerning the filtration method all Gram-positive bacteria did show only rather low levels of glutamate in the filtrate, i.e., were not subjected to leakage during the filtration process. In contrast, the Gramnegative bacteria revealed significant leakage during filtration. The observations for glutamate were representative for other amino acids (Table 1). The intracellular amino acid levels determined for G. oxydans via methanol quenching were only 50% of those obtained by the filtration approach (Figure 2A). In all cases significant levels of amino acids were found in the quenching supernatant, clearly indicating leakage from the cells (data not shown). The fact that the extent of loss was rather similar for each amino acid indicates an unspecific process involved in metabolite leakage during methanol quenching (Figure 2A). For all investigated bacteria, irrespective of the cell wall structure, quenching resulted in significantly lower intracellular pool sizes in comparison to those of filtration (Table 1). Related to the total glutamate amount (the sum of extract and supernatant), both Gram-positive and Gram-negative revealed high average loss (>60%) using the quenching protocol. The values for B. subtilis (78%), E. coli (84%), C. glutamicum (64%), G. oxydans (75%), P. putida (98%), and Z. mobilis (92%) were all significant. The statement that the extent of loss was similar holds for most of the amino acids. In selected cases, amino acids synthesized from glycolytic intermediates such as alanine (in B. subtilis, E. coli, and Z. mobilis), glycine (in E. coli and P. putida), or valine (in P. putida) exhibited only a relatively small difference in the pool size determined by the two methods (Table 1). Overall, it can be concluded that methanol quenching with cell separation as commonly used is not applicable to accurately quantify in vivo levels of intracellular metabolites in Gram-negative and Grampositive bacteria. In contrast to the Gram-positive organisms, the Gram-negative species E. coli (72%), P. putida (82%), and Z. mobilis (80%) also showed severe leakage during fast filtration, a phenomenon that was further investigated by variation of the washing solution as described below. Intermediates from Glycolysis, the Pentose Phosphate Pathway, and the Tricarboxylic Acid Cycle. When the quenching protocol was applied, the levels of intracellular intermediates of the tricarboxylic acid cycle were significantly lower in all strains tested as compared to those of fast filtration (Table 2). This picture was thus very similar to that observed for the amino acids. In contrast, intermediates from glycolysis and pentose phosphate pathway generally did not show lower concentration when methanol quenching was applied (Table 2). Selected metabolites were concentrated even higher when methanol quenching was applied in comparison to fast filtration. Especially, intermediates of the upper glycolysis such as glucose 6-phosphate (G6P) and fructose 1,6-bisphosphate (FBP) showed this effect (Figure 2B). Impact of the Ionic Strength of the Washing Solution during Fast Filtration. A clear outcome of the studies applying fast filtration was the metabolite leakage observed for the Gram-negative bacteria (Figure 1). Since mechanical stress during filtration seems very unlikely, the washing solution used to remove residues of the medium from the filter remained as a potential source causing the leakage. To elucidate this in more detail additional experiments were performed with E. coli, one of the species showing this phenomenon. As it is known that bacterial cells require a compatible ionic strength of the medium to maintain their integrity, washing solutions with different ionic strengths, includAnalytical Chemistry, Vol. 79, No. 10, May 15, 2007
3845
