Tu et al. 10.1073/pnas.0708365104. |
Fig. 6. Heat map of the LC-MS/MS metabolite data over two consecutive cycles.
Fig. 7. Heat map of the GCxGC-TOFMS metabolite data over two consecutive cycles.
Fig. 8. Cyclic changes in glycogen levels during the yeast metabolic cycle. Cells were harvested at the time points indicated and glycogen concentrations determined by using a method described in ref. 4. Figure kindly provided by B. Probst.
Fig. 9. Validation of Dzwf1, cys4 mutant strains by testing for sensitivity to oxidative stress. Serial 10-fold dilutions of the indicated strains where spotted on YEPD plates containing either 0 or 1 mM hydrogen peroxide and grown at 30° for 24 h.
Table 2. Normalized LC-MS/MS and GCxGC-TOFMS metabolite data values.
SI Appendix
LC-MS/MS analysis.
A library of common metabolites found in S. cerevisiae was constructed by using information provided on the Saccharomyces Genome Database (SGD) website (http://pathway.yeastgenome.org:8555/YEAST/class-instances?object=Pathways). For each metabolite, a 1 mM standard solution was infused into a Applied Biosystems 3200 QTRAP triple quadrupole-linear ion trap mass spectrometer for quantitative optimization detection of daughter ions upon collision-induced fragmentation of the parent ion [multiple reaction monitoring (MRM)]. The parent ion mass was scanned for first in positive mode (usually MW + 1). For positive mode, the infusion solvent contained either 50% methanol/0.1% formic acid or 50% methanol/5 mM NH4OAc, and for negative mode (usually MW - 1), 50% acetonitrile/10 mM NH4OAc/10 mM NH4OH, pH = 8.8. For each metabolite, the optimized parameters for quantitation of the two most abundant daughter ions (i.e., two MRMs per metabolite) were selected for inclusion in further method development. A list of all metabolites and parent ion/daughter ion masses detected are provided in SI ···.Metabolites were separated chromatographically on a C18-based column with polar embedded groups (Synergi Fusion, 150 ´2.0 mm 4 m, Phenomenex), using a Shimadzu Prominence LC20/SIL-20AC HPLC-autosampler coupled to the mass spectrometer. Flow rate was 0.5 ml/min using the following method: Buffer A: 99.9% H2O/0.1% formic acid, Buffer B: 99.9% methanol /0.1% formic acid. T = 0 min, 0% B; T = 4 min, 0% B; T = 11 min, 50% B; T = 13 min, 100% B; T = 15 min, 100% B, T = 16 min, 0% B; T = 20 min, stop. For those metabolites that were more optimally detected by using ammonium acetate, the same method was used except 0.1% formic acid was replaced with 5 mM NH4OAc. These conditions were chosen because the majority of the metabolites in the library could be quantitated reproducibly using these methods, as assessed by running a series of standards. For those metabolites that were only detectable in negative mode, we did not establish a reliable chromatography method to separate and quantitate many of them reproducibly. However, quite a few of them were detected successfully by the GCxGC-TOFMS analysis.
For running samples, extracts (typically 20 ml) were dried under vacuum and then resuspended in 100 ml 0.1% formic acid or 5 mM NH4OAc for injection (typically 30 ml injection volume). Mass spectrometer settings were: curtain gas = 45; collision gas = medium; IonSpray voltage = 5000; temperature = 600; ion source gas 1 = 70; ion source gas 2 = 55; interface heater = on. Many MRMs (up to 150, each targeting a parent/daughter ion pair) were conducted simultaneously (20 ms per MRM) throughout each chromatographic run. Often both MRMs for a given metabolite displayed high correlation (Fig. 1B), lending confidence to the developed methods. The retention time for each MRM peak was compared to an appropriate standard. The area under each peak was then quantitated by using Analyst software, reinspected for accuracy, and normalized against total ion count. The average CV for repeated injections was »10%. The periodicity of metabolites was determined by A. Kudlicki using a Fourier-type score, with P values based on the exponential distribution approximation (1).
GCxGC-TOFMS analysis.
Dried metabolite extracts from similar-prepared samples over two consecutive metabolic cycles were derivatized by using methoximation and trimethylsilylation. A mock sample was used as a background. Following derivatization, 1 ml of metabolite extracts were injected onto, and separated by, a comprehensive two dimensional gas chromatograph coupled to the time-of-flight mass spectrometer, referred to as a GCxGC-TOFMS instrument (LECO, St. Joseph, MI), following the method reported in ref. 2. Each sample was injected in triplicate to remove any variation due to the ordering of sample injection. The run time for each sample was 38 min in length. Each of the resulting 72 GCxGC-TOFMS chromatograms consisted of two gas chromatographic separation dimensions (column 1 at 38 min and column 2 at 1.5 seconds) as well as the mass spectral dimension (m/z range of 40-600), thus yielding three-dimensional data. Five thousand mass spectra per second were binned to unit resolution, and averaged along the column 2 axis to yield 100 mass spectra per second. The average CV for repeated injections was »8%.This three-dimensional data required chemometric "multivariate" data reduction tools to rapidly locate metabolites of interest, i.e., metabolites exhibiting periodic behavior. Metabolites with periodic changes in concentration were located by determining the signal-based depth of modulation over all time intervals at each point in the two dimensional gas chromatographic separation for each mass channel, m/z. The signal-based depth of modulation was defined as the largest signal for the 24 time points, divided by the smallest signal. The signal-based depth of modulation for each metabolite provided the information needed to locate metabolites that are either cycling with different periods, or with some other pattern, e.g., spiking at a given frequency. Initial identification of the metabolites of interest was achieved by using the Leco ChromaTOF software, Version 3.22 (LECO), followed by an identification confirmation based on the full mass spectral parallel factor analysis (PARAFAC) deconvolution. Retention time information was also used to confirm metabolite identification using an in-house library of previously analyzed derivatized metabolite standards. Accurate quantification of metabolites was determined by using PARAFAC in a graphical user interface (GUI) for high throughput calculations. To investigate the phase relationship of the metabolites with respect to dO2, relative concentrations for those metabolites found to cycle over the 24 sample points was normalized in intensity to range from zero to one to more readily compare one metabolite to another while preserving the depth of modulation and cycling phase information. A more comprehensive discussion of the methods are reported in ref. 3.
1. Kudlicki A, Rowicka M, Otwinowski Z (2007) Bioinformatics 23:1559-61.
2. Mohler RE, Dombek KM, Hoggard JC, Young ET, Synovec RE (2006) Anal Chem 78:2700-2709.
3. Mohler RE, Tu BP, Dombek KM, Hoggard JC, Young ET, Synovec RE (2007), in press.
4. Parrou JL, Francois J (1997) Anal Biochem 248:186-188.