Quantitative Metabolomics of Saccharomyces Cerevisiae Using Liquid Chromatography Coupled with Tandem Mass Spectrometry

To improve a quantitative assessment of water-soluble metabolites within a yeast cell, we optimized the conditions of cell quenching for metabolite detection. Cell quenching for this purpose involves a rapid arrest of all enzymatic reactions within a cell^31,33,37,38. Such an arrest of cellular metabolic activity is an essential step of any method for the quantitation of water-soluble metabolites in vivo because it prevents the under-estimation of their intracellular concentrations^31,33,37,38. Cell quenching for metabolite assessment always impairs the integrity of the plasma membrane (PM) (and of the cell wall (CW), if present); this causes metabolite leakage from the cell^31,33,37,38. The method employs cell quenching conditions that minimize such impairment, thereby significantly decreasing the quenching-associated cell leakage of intracellular metabolites. Indeed, most current methods for cell quenching involve the use of a certain concentration of methanol (i.e., 40% (v/v), 60% (v/v), 80% (v/v), or 100% (v/v)) at specific temperature (i.e., -20 °C, -40 °C, or -60 °C), with or without a buffer^31,33,37,38. We compared the efficiency of the PM and CW impairment for one of the currently used cell quenching method (i.e., by cell treatment with 80% (v/v) methanol at -40 °C in the absence of a buffer³⁸) to that for the modified cell quenching method (i.e., by cell treatment with 60% (v/v) methanol at -20 °C in the presence of an isotonic buffered solution of ABC at pH = 8.0). PI is a fluorescent dye that is impermeable to intact cells; it can enter the cell only if the integrity of the PM (and of the CW, if present) is impaired³⁹. Moreover, the intensity of fluorescence emission by PI rises by 30-fold when it is bound to DNA or RNA³⁹. Thus, a PI staining assay can be used for assessing the efficiency of quenching-associated cell leakage of intracellular metabolites because these metabolites can leak into the extracellular space only if the PM and CW of a yeast cell are damaged³⁹. We found that the modified cell quenching method causes significantly lower damage to the PM and CW than the quenching method by cell treatment with non-buffered 80% (v/v) methanol at -40 °C (Figure 3). Indeed, almost all cells subjected to quenching using the method exhibited red fluorescence emission, which is characteristic of the yeast cells whose PM and CW are not damaged (Figure 3). In contrast, almost all cells subjected to quenching using the other method, displayed green fluorescence emission characteristic of the yeast cells whose PM and CW are significantly damaged (Figure 3). The most intense red fluorescence was converted to green fluorescence by a software to differentiate between the strong red fluorescence and mild red fluorescence emissions.

The modified cell quenching method caused significantly lower leakage of water-soluble metabolites from yeast cells than the – quenching method by cell treatment with non-buffered 80% (v/v) methanol at -40 °C. We subjected equal numbers of yeast cells to quenching using either the method (i.e., by cell treatment with 60% (v/v) methanol at -20 °C in the presence of an isotonic buffered solution of ABC at pH = 8.0; Figure 4) or the other method (i.e., by cell treatment with non-buffered 80% (v/v) methanol at -40 °C; Figure 5). The extent of quenching-associated cell leakage into the extracellular solution was assessed for specific water-soluble metabolites with the help of LC-MS/MS. The concentrations of different amino acid classes (i.e., large and small acidic, basic, neutral-nonpolar, and neutral polar amino acids) in the extracellular solution were measured before and after cell quenching. We found that the cell quenching method causes significantly lower leakage of all these amino acid classes into the extracellular solution (Figure 4) than the other method (Figure 5).

The LC-MS/MS method for a quantitative assessment of water-soluble metabolites within a yeast cell uses a single type of the column for chromatographic separation of all water-soluble metabolite classes. This column is the zwitterionic-phase column named in Table of Materials. We found that this column provides a much more efficient separation of different classes of water-soluble metabolites than the reverse-phase column named in Table of Materials (Supplemental Table 1). Indeed, the retention time (RT) shift values of water-soluble metabolite standards (i.e., NAD+, AMP, GMP, arginine, and glutamic acid) were significantly lower and the peak shapes were substantially sharper for the zwitterionic-phase column, as compared with the reverse-phase column (Supplemental Table 1). LC conditions used for chromatographic separation of all metabolites (i.e., water-soluble and water-insoluble) for the reverse-phase column are provided in Supplemental Table 2.

Another advantage of the LC-MS/MS method consists in the ability of chromatographic separation on the above zwitterionic-phase column to efficiently separate from each other different water-soluble metabolites with diverse structural, physical, and chemical properties. These water-soluble metabolites include the following metabolite classes: 1) acidic, basic, neutral polar, and non-neutral polar amino acids, including their different structural isomers (Figure 6); 2) stable and unstable nucleotides and their derivates that perform vital functions within a cell (Figure 7); and 3) various monosaccharides, including their different stereoisomeric forms (Figure 8).

Importantly, the LC-MS/MS method for a quantitative assessment of water-soluble metabolites within a yeast cell was versatile and robust. It allowed us to identify and quantify 374 water-soluble metabolites with diverse structural, physical, and chemical properties in S. cerevisiae cells that were cultured in the complete YP medium initially containing 2% glucose (Supplemental Table 3). 240 metabolites were detected in the positive ionization mode and 134 metabolites were detected in the negative ionization mode. The identities of all these water-soluble metabolites were confirmed by matching their data-dependent acquisition (DDA) MS2 fragments acquired both in the positive and negative ionization modes to the MS2 mzCloud spectral library. This online library includes the spectra of metabolite standards that differ in their MS1 and DDA MS2 parameters. To maximize the extent of matching the MS2 spectra of the sample with the online spectral library, we used different DDA MS2 parameters. These parameters are provided in Supplemental Table 4. Almost 6,000 features (putative metabolites) for the same sample were acquired with the help of the high-energy-induced-collision-dissociation (HCD) or collision-induced dissociation (CID) fragmentation method, using top 5 MS2 events, 35 collision energy values, and 10 ms activation times. After filtering the resulting files with > 95% MS2 matching and > 90% MS1 isotopic pattern matching criteria, only 162 metabolites under HCD fragmented condition and 142 metabolites under CID fragmented condition were identified. 81 out of 162 metabolites were unique to the HCD fragmentation method, whereas 42 out of 142 were unique to the CID fragmentation method (see a sheet named "T5_35E__10 ms_HCD vs CID" in Supplemental Table 4). Therefore, we concluded that the correct annotation of water-soluble metabolites with the help of the LC-MS/MS method requires the use of many different DDA MS2 parameters.

The LC-MS/MS method is also highly sensitive. It allows to identify and quantitate some water-soluble metabolites at concentrations as low as 0.05 pmol/µL (see data for phenylalanine in Table 4). This limit of quantitation varies within a wide range of concentrations for different metabolite classes (Table 4).

The MS system used here has a wide (at least two orders of magnitude) linear dynamic range for measuring the concentrations of various metabolites (Supplemental Table 5).

Figure 1: The total ion chromatogram (TIC) from liquid chromatography/mass spectrometry (LC-MS) data of water-soluble metabolites that were extracted from cells of the wild-type strain BY4742. The metabolites were separated by LC on the zwitterionic-phase chromatography column named in Table of Materials. The metabolites were detected by MS of parent ions (MS1) that were created using the positive electrospray ionization mode. Please click here to view a larger version of this figure.

Figure 2: A workflow used for the analysis of water-soluble metabolites with the help of the software named in Table of Materials. All the parameters were autocorrected by the software based on the MS raw data, except the followings: 1) "Detect Compounds" tab: set min, peak intensity 10,000; and 2) "Search ChemSpider" tab: 4 online databases were selected for the identification of metabolites, including the BioCyc, Human Metabolome Database, KEGG, and Yeast Metabolome Database. Please click here to view a larger version of this figure.

Figure 3: Differential interference contrast (DIC; top row) and fluorescence (bottom row) microscopic images of yeast cells quenched either with ammonium bicarbonate (ABC) buffer (pH = 8.0; control) or with a different quenching solution. After cell quenching, equal numbers of cells were incubated with the PI solution for 10 min in the dark and on ice. The most intense red fluorescence was converted to green fluorescence by software to differentiate between the strong red fluorescence and mild red fluorescence emissions. The efficiency of damage to the PM and CW was compared for the method of cell quenching (i.e., by cell treatment with 60% (v/v) methanol at -20 °C in the presence of an isotonic buffered solution of ABC at pH = 8.0) and one of the currently used methods (i.e., by cell treatment with 80% (v/v) methanol at -40 °C in the absence of a buffer). A scale bar is shown. Please click here to view a larger version of this figure.

Figure 4: The leakage percentage of different amino acid classes for the method of cell quenching. 5.0 x 10⁸ of yeast cells were subjected to quenching by the treatment with 60% (v/v) methanol at -20 °C in the presence of an isotonic buffered solution of ABC at pH = 8.0. The leakage percentage of different amino acid classes was assessed using LC-MS/MS to measure their concentrations in the extracellular solution before and after quenching. The mean values ± SD and individual data points are shown (n = 3). * p < 0.05. Please click here to view a larger version of this figure.

Figure 5: The leakage percentage of different amino acid classes for one of the currently used cell quenching methods. 5.0 x 10⁸ of yeast cells were subjected to quenching by the treatment with 80% (v/v) methanol at -40 °C in the absence of a buffer. The leakage percentage of different amino acid classes was assessed using LC-MS/MS to measure their concentrations in the extracellular solution before and after quenching. The mean values ± SD and individual data points are shown (n = 3). * p < 0.05. Please click here to view a larger version of this figure.

Figure 6: Efficient chromatographic separation of acidic, basic, neutral polar and non-neutral polar amino acid classes, including two structural isomers (i.e., leucine and isoleucine), on the zwitterionic-phase column named in Table of Materials. All these amino acids were detected by MS/MS in the positive ionization [ESI (+)] mode. Conditions for LC on the zwitterionic-phase chromatography column are described in Table 1. Conditions for MS/MS are described in Table 2 and Table 3. All amino acid standards are bought commercially (e.g., Sigma). The retention time shifts between 3 independent chromatography runs are less than ± 10 seconds. Please click here to view a larger version of this figure.

Figure 7: Efficient chromatographic separation of different classes of nucleotides, including energetically unstable nucleotides (nucleoside monophosphates, nucleoside diphosphates, and nucleoside triphosphates) and electron carrier molecules (NADH and NAD+), on the zwitterionic-phase columnnamed in Table of Materials. All these nucleotides were detected by MS/MS in the positive ionization [ESI (+)] mode. Conditions for LC on the zwitterionic-phase chromatography column are described in Table 1. Conditions for MS/MS are described in Table 2 and Table 3. All nucleotide standards are bought commercially (e.g., Sigma). The retention time shifts between 3 independent chromatography runs are less than ± 10 seconds. Please click here to view a larger version of this figure.

Figure 8: Efficient chromatographic separation of different classes of monosaccharides, including structural isomers and stereoisomeric forms of aldo- and ketohexoses (fructose, mannose, and galactose), and aldopentoses (ribose and arabinose), on the zwitterionic-phase columnnamed in Table of Materials. All these monosaccharides were detected by MS/MS in the positive ionization [ESI (+)] mode. Conditions for LC on the zwitterionic-phase chromatography column are described in Table 1. Conditions for MS/MS are described in Table 2 and Table 3. All monosaccharide standards are bought commercially (e.g., Sigma). The retention time shifts between 3 independent chromatography runs are less than ± 10 seconds. Please click here to view a larger version of this figure.

Table 1: LC condition used for the separation of water-soluble metabolites with the help of the zwitterionic-phase columnnamed in Table of Materials. These conditions were used in all experiments described here.

Table 2: The mass spectrometer settings used to analyze metabolites that were separated by LC. These conditions were used for the analysis of metabolites in all experiments described here. Abbreviations: FTMS = Fourier transform (FTMS); HCD = high-energy-induced-collision-dissociation; LTQ = linear trap quadrupole; AGC = automatic gain control, an ion population value for MS and MS/MS.

Table 3: The mass spectrometer settings used to detect secondary ions (MS2). Abbreviations: HCD = high-energy-induced-collision-dissociation; CID = collision induced dissociation; ms = milliseconds.

Table 4:The lowest concentrations of different water-soluble metabolite standards that the LC-MS/MS method can identify and quantitate. The MS1 peak area of each metabolite standard was used to estimate the lowest quantifiable concentration for this metabolite. Mean values of two independent experiments are shown. Three technical replicates were performed for each of the two independent experiments. NOTE: Threonine* can be detected but cannot be quantified due to its co-elution with homoserine, a chemical isomer of threonine. Glucose** cannot be identified because it creates multiple chromatography peaks. Metabolite standards*** can be identified and quantitated only in individual samples, but not in a mixture of metabolite standards or a biological mixture of metabolites.

Supplemental Table 1: Retention Time (RT) shift values of the same set of metabolites for the zwitterionic-phase and reverse-phase columns (see Table of Materials) after column equilibration. The table compares the retention reproducibility of the zwitterionic-phase and reverse-phase columns for a distinct set of metabolites the differ from each other in their hydrophilicity and hydrophobicity. These metabolites were identified with the help of LC-MS/MS. Note that the RT shift values of hydrophilic metabolites (i.e., NAD+, AMP, GMP, arginine, and glutamic acid) are significantly lower and the peak shapes are substantially sharper for the zwitterionic-phase column, as compared to the reverse-phase column. In contrast, the RT shift values of hydrophobic metabolites (i.e., stearic acid, lauric acid, and decanoic acid) are significantly lower and the peak shapes are substantially sharper for the reverse-phase column, as compared to the zwitterionic-phase column. Conditions for the chromatographic separation of metabolites with the help of the zwitterionic-phase and reverse-phase columns are detailed in Table 1 and Supplemental Table 2, respectively. The reported here RT shift values are based on the measurement of 20 different samples taken from different vials. The samples were analyzed after column equilibration. The metabolites in each sample were extracted from 5.0 × 10⁸ yeast cells. RT shift values are the means of 20 different samples (n = 20). The p values derived from the unpaired t test were used to compare the two columns with the equal variance between both sample types. Please click here to download this table.

Supplemental Table 2: LC conditions for the reverse-phase 8 column named in Table of Materials and used to separate different metabolites in this study. Column properties were as follows: 150 x 2.1 mm, 5 µm polymer. Please click here to download this table.

Supplemental Table 3: A list of all 374 water-soluble metabolites recovered and annotated using the LC-MS/MS method. All metabolites were recovered from the same LC gradient run, subjected to a data-dependent acquisition (DDA) fragmentation algorithm described in Supplemental Table 4, and annotated using the software named in Table of Materials. The MS1 peak shapes of 211 metabolites (whose status in the table is indicated as a blank) were appropriate to be used for quantification, and their respective MS2 spectra had full matches with the mzCloud spectral library. MS2 spectra of 38 metabolites (whose status in the table is indicated as a d) had full matches with the online spectral library. Still, their MS1 peak shapes were not appropriate to be used for quantification. MS2 spectra of 125 metabolites (whose status in the table is indicated as an n) did not have full matches with the online spectral library. These metabolites were annotated as follows: 1) using the "Predict composition node" (which annotates metabolites based on the exact match of MS1 values, number of matched and missed isotopes, and isotope % intensity matched with that of the theoretical reference standards); 2) using the "ChemSpider node" (which annotates metabolites based on the exact match of MS1 values and similarity match score of MS2 spectra with the online spectral library); and 3) using the retention time (RT) shift values of metabolites (these values depend on the physical structure and chemical properties of metabolites). The metabolites listed in the table were found in all 3 biological replicates performed, with the RT shift values of < 0.2 min. Please click here to download this table.

Supplemental Table 4: A data-dependent acquisition (DDA) fragmentation algorithm that was used here for the annotation of water-soluble metabolites with the help of the software named in Table of Materials. Please click here to download this table.

Supplemental Table 5: A typical linear dynamic range that we observed when we measured the concentrations of different amino acids with the help of the MS system named in Table of Materials. The MS system used here has a wide (at least two orders of magnitude) linear dynamic range for measuring various metabolites' concentrations. Please click here to download this table.