Addressing the Bottlenecks in Metabolomics Making an Expedient
Addressing the Bottlenecks in Metabolomics: Making an Expedient Transition from Global Profiling to Targeted Quantitation Mark Sanders 1, Kevin Mc. Hale 1, Adam Takvam 2 , Michael Athanas 3, Mark Szewc 1, Jamie Humphries 2, Michael Pacold 4, David Sabatini 4 1 Thermo Fisher Scientific, Somerset, NJ, 2 Thermo Fisher Scientific, Austin, TX, 3 Thermo Fisher Scientific, San Jose, CA, 4 Whitehead Institute, Cambridge, MA Overview Results Purpose: Tools were developed to augment commercially available software packages (SIEVE and Trace. Finder) to provide automation and drive efficiency in the transition from global profiling to targeted quantitation. The challenge of moving from profiling to quantitation has been that traditionally two different instrument platforms were used, HRAM instrument and triple quads respectively. With the latest generation of HRAM instruments you can collect untargeted data with little to no compromise in quantitative performance 1. This provides new data processing possibilities, i. e. combining targeted and untargeted approaches on the same data set. In addition some studies , such as the one discussed here, can be further consolidated by using pos/neg switching. Whether the follow up quantitative analysis is performed on the same instrument or methods need to be developed on other instruments, the set up can be labor intensive. The software tools described here facilitate efficient transition from global profiling to targeted quantitation on a large number of endogenous metabolites. The full workflow is described in Figs. 4 -14. Methods: Breast cancer cell lines grown under different conditions were analyzed by high resolution/accurate mass (HRAM) LCMS on a Q Exactive. Using the SIEVE plugin for Trace. Finder components were accurately detected from full scan data, identified where possible and statistically evaluated. Components of interest were selected and quantitative methods were automatically generated for Trace. Finder. Results: The HRAM data from the Q Exactive facilitated the transition from profiling to quantitation with little to no compromise in quantitative performance, with the advantage of being able to retrospectively interrogate the data based on new information. Combining the capabilities of SIEVE with Trace. Finder enabled an automated transition from untargeted profiling/relative quantitation to targeted, full featured quantitation Introduction The goal of global metabolite profiling or untargeted metabolomics is to make a quantitative assessment of as many endogenous metabolites as possible. This approach is very effective for hypothesis generation and revealing the involvement of metabolic pathways that may not have been predicted. The downside of this approach is that massive amounts of data are collected and the unbiased data processing algorithms are relatively slow. This form of data processing is impractical for larger scale validation studies and in most cases, as you learn more about the biology under study, what starts as global profiling becomes a more targeted assay on the components of interest. Here we demonstrate some new tools to increase efficiency in this process Methods Sample Preparation. Breast cancer cell lines, MDA-MB-231 - low 3 -phosphoglycerate dehydrogenase (PHGDH) and MDA-MB-468 cells - high PHGDH were grown in DMEM (low amino acids) and RPMI (high amino acids) media. Aqueous extractions were performed on ~5 x 10^6 cells with ice-cold methanol. 500 µL of each extract was lyophilized and stored at -80 C. Prior to LCMS analysis samples were reconstituted in water/methanol 90: 10, 500 µL. Extractions were performed in triplicate. In this approach a subset of samples are subject to unbiased analysis. The samples could be from an initial discovery study or from pooled controls. The important factor is that these samples are representative of those in the full study. These results yield a list of both known and unknown components that can be evaluated statistically to determine which may be of interest. All information needed for a targeted assay is determined from the preliminary data analysis. For larger scale or follow up studies the levels of the components of interest can be measured in a targeted approach. The main benefits being processing speed and peak detection can be customized per component (Figs. 9, 10). FIGURE 4. Sample assignment and data acquisition is performed within Trace. Finder. Once acquisition is complete component detection is initiated using from the SIEVE icon on the Trace. Finder tool bar. FIGURE 10. Pos/neg switching data can be efficiently processed within the same method. SIEVE is the engine for unbiased component detection and the workflow is shown in Fig. 1. SIEVE removes non-sample ions and data is further reduced by extensive grouping of related ions (Fig. 2). This reduces statistical noise and improves the ability to detect differences between sample groups 2 (Fig. 3). While it is not absolutely necessary, the quantitative assessment is better performed when compound identifications are made. For this a novel database of MS n spectral trees, m/z. Cloud, was used to either identify or determine compound classes 3. Even if a compound is not fully identified knowing some structural information can aid in choosing a related surrogate for a suitable calibration curve (Fig. 13). FIGURE 1. SIEVE component detection workflow. . FIGURE 2. Component detection in SIEVE recognizes 17 related ions for Uric acid Mass Spectrometry MS: Source: Thermo Scientific Q Exactive mass spectrometer Heated Electrospray Ionization probe v. 2 (HESI 2) Cap. Temp: 275 Vap. Temp: 350 Sheath gas: 40 Aux gas: 15 S-Lens: 40 Polarity mode: Alternate positive, negative switching Mass Range: m/z 70 – 1000 Resolution: 70, 000 [at m/z 200] Calibration: External FIGURE 6. Using various filters and statistical tools, components of interest are identified. Data Analysis Data analysis was performed with Thermo Scientific Trace. Finder 3. 1 with the plug-in for Thermo Scientific SIEVE 2. 1. Neg Pos FIGURE 12. Data review by sample. Showing all compounds within a selected sample FIGURE 7. Components of interest are selected. FIGURE 3. SIEVE component detection removes noise allowing differences to be more easily identified Components Fed Large intra group variability Female Fasted FIGURE 8. Selected components are automatically imported into Trace. Finder as a compound database. Identified compounds are named, unknowns are labeled with a mass and retention time. Results (cont. ) Intelligent sequencing is another feature of Trace. Finder designed to maximize data quality. In a targeted quantitation mode. This allows the results from processed data to determine which sample/experiment the instrument will run next. It is particularly useful for large unattended studies with limited amounts of sample. A simple example of intelligent sequencing would be that if a blank sample showed significant analyte levels then the instrument could either reinject the blank sample until a blank was obtained, or the whole analysis could be paused, avoiding injecting precious sample into a compromised system. The custom report builder plug-in for Trace. Finder (Fig. 14) provides an easy, flexible way to generate reports customized for various workflows and/or studies. Working in an Excel-like environment, quantitative results, chromatograms/spectra, and charts/graphs can be automatically reported on a batch or single file basis. Conclusion FIGURE 11. Data review by compound. A selected compound viewed across all samples. m/z Features FIGURE 14. Rapidly build custom reports with the new report builder plug-in for Trace. Finder FIGURE 5. Component detection and identification is performed using the SIEVE plug-in for Trace. Finder. Liquid Chromatography LC: Thermo Scientific Ultimate 3000 RSLC system Column: Luna NH 2 100 A, 3µm, 150 x 2 mm Column temp. : 15°C AS tray temp. 10°C Inj. volume: 1 µL Wash solvent: 90% Me. OH Mobile phase: A = 10 m. M NH 4 OAc, p. H 9. 9 B = Acetonitrile Gradient: 90 -10% B in 20 min, hold at 10% B for 5 min. Flow rate: 250 µL/min FIGURE 9. Peak detection and integration parameters can be customized for each compound to accommodate the wide range of chromatographic performance that is expected in an untargeted method. FIGURE 13. The amount of any compound can be estimated based on a calibration curve from any other compound. Usually a compound in the same chemical class, or for endogenous compounds a calibration curve is often generated from a stable label standard. The software tools developed here were designed to make the transition from global profiling to targeted quantitation more efficient and automated. They include modules for data reduction, component detection and identification, automated information transfer, intelligent sample queuing and customized reporting. § The Q Exactive, HRAM platform can be used for both global profiling and targeted quantitation. § Global profiling has the benefit of an untargeted analysis so unexpected changes are not missed and data can be retrospectively interrogated as new information is gained. § Targeted quantitation has the benefit of processing speed for large studies and peak detection and integration parameters can be customized for each targeted metabolite, improving the quality of data. § The analytical strategies described here take advantage of the benefits of unbiased global profiling as well as the data quality aspects of targeted quantitation. The software tools provide an easy transition between the two. References 1. Accurate and sensitive all-ions quantitation using ultra-high resolution LCMS, Mark Sanders, Josef Ruzicka, Kevin Mc. Hale and Petia Shipkova, Thermo Fisher Scientific Apps Note 476. 2. Metabolomic profiling in drug discovery: understanding the factors that influence a metabolomics study and strategies to reduce biochemical and chemical noise, Mark Sanders, Serhiy Hnatyshyn, et. al. , Thermo Fisher Scientific Apps Note 562. No group separation Male 3. Untargeted metabolomics: From statistical objects to the efficient identification of "known unknowns“, Robert Mistrik and Juraj Lutisan, ASMS 2013, Th. OB am, 10: 10 am Excel is a trademark of Microsoft Corporation. All other trademarks are the property of Thermo Fisher Scientific and its subsidiaries This information is not intended to encourage use of these products in any manners that might infringe the intellectual property rights of others.
- Slides: 1