A strategy for sensitive, large scale quantitative metabolomics

Introduction

Metabolomics, defined as an experiment that measures multiple metabolites simultaneously, has been an area of intense interest. Metabolomics provides a direct readout of molecular physiology and has provided insights into development and disease such as cancer^1–4. Nuclear magnetic resonance (NMR) and Gas chromatography-mass spectrometry (GC-MS) are among the most commonly used instruments^5–9. NMR, especially has been used for flux experiments since heavy isotope labeled compounds, such as ¹³C labeled metabolites, are NMR-active^10,11. However, this strategy requires relatively high sample purity and large sample quantity, which limits its applications in metabolomics. Meanwhile, data collected from NMR needs intensive analysis and compound assignment of complex NMR spectra is difficult. GC-MS has been widely used for polar metabolites and lipid studies, but it requires volatile compounds and therefore often derivatization of metabolites, which sometimes involves complex chemistry that can be time consuming and introduces experimental noise.

Liquid Chromatography (LC) coupled to triple quadrupole Mass Spectrometry uses the first quadrupole for selecting the intact parent ions, which are then fragmented in the second quadrupole, while the third quadrupole is used to select characteristic fragments or daughter ions. This method, which records the transition from parent ions to specific daughter ions, is termed multiple reaction monitoring (MRM). MRM is a very sensitive, specific and robust method for both small molecule and protein quantitation^{12–15, 21}. However, MRM does have its limitations. To achieve high specificity a MRM method needs to be built for each metabolite. This method consists of identifying a specific fragment and corresponding optimized collision energy, which requires pre-knowledge of the properties of the metabolites of interest, such as chemical structure information. Therefore, with some exceptions involving the neutral loss of common fragments, it is not possible to identify unknown metabolites with this method.

In the recent years, high-resolution mass spectrometry (HRMS) instruments have been released, such as the LTQ-orbitrap and Exactive series, the QuanTof, and TripleTOF 5600.^{16–18, 22} HRMS can provide a mass to charge ratio (m/z) of intact ions within an error of a few ppm. Therefore, an HRMS instrument operated by detecting all precursor ions (i.e. full scan mode) can obtain direct structural information from the exact mass and the resulting elemental composition of the analyte, and this information can be used to identify potential metabolites. Indeed, all information about a compound can be obtained with an exact mass, up to the level of structural isomers. Also, a full scan method does not require previous knowledge of metabolites and does not require method optimization. Moreover, since all ions with m/z falling into the scan range can be analyzed, HRMS has a nearly unlimited capacity in terms of the number of metabolites that can be quantified in a single run compared to the MRM method. HRMS is also comparable to a triple quadrupole MRM in quantitative capacity due to the short duty cycle resulting in a comparable number of data points that can be obtained in a full MS scan. Therefore, HRMS provides an alternative approach for quantitative metabolomics. Recently, an improved version of HRMS termed Q-Exactive mass spectrometry (QE-MS) can be operated under the switching between positive and negative modes with sufficiently fast cycle times in a single method, which expands the detection range¹⁹. Here we describe our metabolomics strategy using the QE-MS.

Protocol Text

Representative Results

The accuracy of metabolomics data highly depends on the LC-QE-MS instrument performance. To assess whether the instrument is operating in good condition, and whether the method applied is proper, several known metabolite LC peaks are extracted from the total ion chromatography (TIC), as shown in Figure 1. Polar metabolites, including amino acids, glycolysis intermediates, TCA intermediates, nucleotides, vitamins, ATP, NADP⁺ and so on have good retention on the column and good peak shapes in the amide column under current LC conditions. Meanwhile, a mass error test is done within 24 hrs after low mass calibration, as illustrated in Figure 2. 6 different concentrations of samples in triplicate are run twice after calibration, and the whole time range covers almost 24 hrs. The mass error is assessed by comparing the detected m/z to the theoretical m/z of targeted metabolites. Here the targeted metabolites have an m/z ranging from 74 (glycine) to 744 (NADP⁺). The Y axis here represents the accumulative percentage of metabolites within certain mass error range. The red curve shows the result in positive mode, while the blue colored curve shows the data collected in negative mode. Figure 2 clearly indicates that more than 90% of metabolites are within 5 ppm mass error in both positive and negative mode, which means the low mass range calibration method developed here is sufficient to maintain 5 ppm mass error for low mass range detection.

Another issue to be addressed is the sensitivity of the instrument with the current method and instrument setup. A serial dilution of triplicate samples from 10 cm petri dish was done 5 times with a dilution factor of 6, ending up with 6 different concentrations of samples. These samples represent the amount of metabolites extracted from 10⁷, 1.67×10⁶, 2.78×10⁵, 4.63×10^4, 7.72×10³, and 1.29×10³ of cells respectively. Since each concentration of sample is prepared in triplicate, a total of 18 samples are analyzed in LC-QE-MS. A targeted list is used to assess the number of metabolites detected at differing concentration of sample. The result in Figure 3 indicates that the optimal number of targeted metabolites detected is between 2.78×10⁵ and 1.67×10⁶ cells, while 1×10⁷ cells give a fewer number of detected metabolites, which is due to ion suppression effects. This result indicates that the optimal amount of cells to extract for this analysis is roughly that of a well of in a 6-well plate.

For untargeted metabolite analysis, a CV cutoff of 20% and an average intensity value of 10⁷ are used to filter the components table. These rigid CV and average intensity threshold values are used for this demonstration aim. CV cutoff values can be increased (for example, 30 %) while the average intensity values need to be decreased (for example, 10⁵) to include more peaks. After manually checking peaks, components with good shapes are selected and searched for in the human metabolome database. The results are shown in Table 2. Table 2A lists the results from data collected in positive mode, while Table 2B shows the results from negative mode. Some of the metabolites identified here overlap with the metabolites in the targeted list, such as glutathione, proline and so on, but meanwhile, additional metabolites absent from the targeted list are explored, such as methyglyoxal, which can be derived from glycolysis, and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, which is detected in positive with a retention time of 3.2 min, which is a reasonable retention for phospholipids on an amide column. A protocol on untargeted metabolite database searching has been previously reported²⁰.

Discussion

The most critical steps for successful metabolite profiling in cells using this protocol are: 1) controlling the growth medium and careful extraction of the cells; 2) adjusting the LC method based on MS method setup to ensure there are enough (usually at least 10) data points across a peak for quantitation; 3) doing a low mass calibration before running samples; 4) injecting no more than 5 μl to avoid retention time shifting and peak broadening caused by water; and 5) preparing and running samples for comparison in the same batch to minimize batch effects.

The standards (Table 1) chosen here for low mass range calibration are interchangeable. Any known compound with an m/z that falls into the mass scan range, is well behaved in a H-ESI source, and is soluble in water, methanol or acetonitrile are reasonable candidates for calibration standards. It is highly recommended to store all calibration solutions at 4 °C to stabilize caffeine and also to minimize the evaporation of methanol or acetonitrile in the calibration solution so that the calibration performance can be more reproducible. Compared to a regular mass range, m/z from 150 to 2000, low mass range calibration needs to be done more frequently, at least once every two days.

This workflow, from extraction solvent, reconstitution solvent, LC mobile phase, to low mass range calibration and MS scan range has been optimized to measure polar metabolites. This includes amino acids, acetyl amino acids, glycolysis pathway intermediates, nucleosides, TCA cycle intermediates, some one-carbon metabolism pathway intermediates and so on. However, modifications of this protocol for other classes of metabolites, such as Coenzyme A (CoA) species, folates, phospholipids are possible. For example, CoAs are more stable in acidic conditions, so the addition of an acid to 80% methanol/water will be helpful to improve CoA sensitivity. Also, CoAs and lipids tend to have much large molecular weights, thus the m/z scan range needs to be adjusted from 60–900 to the proper range which will cover those metabolites.

Even though some of the untargeted component database search results overlap with the targeted list, it is still of importance to build this targeted list based on the research priority. Since the metabolites in the targeted list are lower than the average intensity threshold, information on these metabolites will be removed during processing. The targeted list includes the retention time information, which gives us higher confidence for metabolite identification and quantitation. One further advantage with the QE-MS setup is that tandem mass spectrometry can allow for further identification of metabolites.

One issue associated with this workflow is that the H-ESI needle insert is sensitive to the salt content of the samples, as the sensitivity will be greatly compromised if there are high amounts of non-volatile salts. Therefore, minimizing salt content from samples, and routine cleaning of the column and the H-ESI needle insert will be helpful to ensure good quality data and to increase the column’s lifetime.

In summary, this protocol employs LC-QE-MS to successfully analyze polar metabolites from cultured cells, with minimal sample preparation steps and rapid data acquisition. Small modifications in sample preparation can be carried out to obtain data from other biological sources such as serum and tissue. For example, since pure methanol can be added to liquid serum added to make a final methanol concentration to 80% for polar metabolites extraction. For tissue samples, rigorous stirring and mixing is required to achieve better extraction efficiency. Usually 10 μl serum or 1 mg tissue is sufficient for metabolites analysis. The raw data can be analyzed both in targeted mode, if there are known metabolites in the samples, and in an un-targeted way followed by HRMS database searching. HRMS based metabolomics is still in its early stages. For future advances, the experimental techniques can be further optimized, additional metabolite HRMS information and MS/MS fragmentation patterns will be helpful and relevant algorithms, such as peak alignment, peak integration, isotope clustering and so on, can improve the efficiency and accuracy of data processing. Ultimately, however, many of the questions that our lab addresses in metabolism are limited by careful interpretation of data after processing. With these large-scale metabolomics techniques, we are often limited by our interpretation of the data and evaluation of hypotheses generated. Therefore, all metabolomics experiments need be formulated around specific questions.

Supplementary Material