FiehnLab ASMS 2011 conference presentations
SessionPoster: MP18 - Metabolomics: Quantitative Analysis, poster number: 322, Monday, Poster Hall
Introduction
Targeted metabolomics quantifies a set of predefined metabolites in complex biological matrices.There are several key steps that need to be considered for the development of an accurate and comprehensive quantitation method. A number of platforms ( GC/MS, LC/MS and NMR) are used to address complex issues that arise. In this study, we are using GC triple quadrupole (QQQ) with chemical ionization source (CI) and use molecular ions to obtain MRM transitions of fifty common metabolites for quantitation of NIST human serum sample. Selectivity and sensitivity of this method is compared to electron impact ionization (EI) source with triple quadrupole and single quadruple results. Furthermore, we are exploring the intensity chemical property relationship by chemical properties and experimental slope information
Methods
15 µl aliquots of standard NIST human plasma were extracted by 1 ml of -20°C iPrOH/ACN/H2O. Aliquots were methoximated , trimethylsilylated and analyzed with Agilent GC/QQQ. GC conditions were set as injector temperature of 250oC, GC oven ramp from 60oC to 325oC by 20oC/min with 1.2 ml/min constant flow .The temperature set point for the transfer line, ion source, and quadrupoles were 280oC, 200oC and 150oC, respectively. A 30 m long, 0.25 mm i.d., 0.25 um film HP 5-MS column was used. QQQ mass spectrometer was operated with a helium quench gas flow of 1.5 ml/min and a nitrogen collision gas flow of 2.25 mL/min in the collision cell and for CI 35% methane flow.
Preliminary Data
GC/QQQ-CI: Fifty common serum metabolite standards are prepared in several mixes, analyzed in full scan mode and CI spectra and retention times are established for each metabolite. Based on the full scan analysis, MRM methods are set up for those metabolites. The precursor ions are selected from molecular ions or the [M-15]. Chromatograms of replicate standard human plasma (NIST) are processed for identification of detectable metabolites. Fifty compounds are unambiguously assigned by the Mass Hunter software using the Fiehn/Agilent Metabolomics Library. The MS acquisition method is divided into time segments to enhance the sensitivity and selectivity. For each compound, selectivity is validated by three different parameters, i.e., 6s - retention time and the two MRM transitions. Method detection limits and limit of quantitation for the common fifty metabolites were established.
GC/MS-EI: Chromatograms of replicate standard human plasma (NIST) are processed for identification of detectable metabolites. Subsequently, product and precursor ions were selected for each metabolite. The MS acquisition method is set to acquire 50 MRM transitions with narrow time segments. The validation is carried out by ion ratios as well as retention times. Method detection limits and limit of quantitation for the common fifty metabolites are established. Calibration curves of more than 100 common metabolites are calculated by Agilent Chemstation. These calibration curves are generated by scan mode and selective quant ions are used to the calculate slopes. Subsequently, all metabolites investigated for substructures that may contribute to sensitivity in GC/MS applications. Metabolite structures are downloaded from PubChem then trimethylsilylated in silico using ChemAxon standardize tool, topological, electronic and constitutional descriptors are calculated for each metabolite. Correlations of these descriptors are established in regards to experimental intensity results.
SessionPoster: TP09 - Metabolomics: Identification of Unknown Metabolites, poster number: 203, Tuesday, Poster Hall
The retention indices introduced by Kovats is independent of experimental parameters and give a reproducible value for a given compounds. Retention indices are calculated from the retention time in the presence of adjacently eluting standards. The retention behavior of any compound depends on the chemical structural features and boiling point. NIST database contains experimental retention indices for over 35,000 different molecules for building the RI prediction. In derivatization based metabolomics study, one metabolite can exist in different silylated structures thus prediction of exact silylated molecular formula is very crucial to obtain absolute molecular formula. There are very less reports on retention indices of silylated compounds and effect of number silylation towards retention indices behavior.
278 compounds having reported experimental RI values in our Fiehn library were downloaded from PubChem using the link for structure download. All the structures were subjected to one step silylation in standardiser. All the silylated structures with their PubChem CID were imported in the Instant J Chem and Si number were computed using chemical term field ‘atom number’ for all compounds. The experimental number of Si compared with the insilico Si number and compounds with different Si number than the experimental one are corrected through editing the structures of isomers. The curated structures were exported as SDF file from Instant J Chem. The curated structures were used to calculate the RI value using NIST RI algorithm.
Preliminary Data
The NIST RI algorithm largely ignores three class of compounds i.e. highly fluorinated compounds, cyclic siloxanes, and large or complex ring systems (steroids, adamantine analogs, and polynuclear aromatic compounds), thereby here we have also removed these compounds from our studies (to measure the error because of only silylation). We have taken difference of experimental RI values (converted in kovats units) and computational (from NIST RI algorithm) RI values (kovats) of 278 fully silylated compounds (having Si atom same for both experimental and computational measured value) and plotted them against the number of Silylation (Si) of corresponding compounds. The deviation of RI value from experimental value for compounds having zero silylation (Si) is found to be in range 33 to 174 kovats RI units with the absolute median error value of 80 which is in agreement with the absolute median error value reported earlier in original study. The error range (difference of experimental vs. computational) for compounds increases with the degree of silylation and negative deviation is obtained for the compounds silylated (≥3) (primarily oxygen donor site (hydroxyl and carboxyl)). The silyaltion on N rich compounds (terminal amino group) shows large positive deviation. Thus silylation (Si) has effect on RI values which cannot be explained by solely on the basis of group contribution method. The equation of calibration plot was used for calculation of correction factor corresponding to each Silylation number. Using this correction factor for Silylation, 30 quality control metabolites were studied and top 3 molecular formulas were constraint predicting correct molecular formula in 95% cases.
SessionPoster: WP16 - Metabolomics: General I, poster number: 303, Wednesday, Poster Hall
Introduction
Accurate mass information is critical for analyzing the wealth of metabolomics data. Besides mass accuracy, isotope ratio accuracy, MS and MS/MS spectral acquisition rate, and polarity switching, metabolomics data require powerful software for peak finding, mass spectral deconvolution, calculating elemental formulas, and comparisons of sample sets.
This research presents an unbiased exploration of the practical capabilities and limitations of the Agilent 6530 Q-TOF mass spectrometer and the Leco HRT mass spectrometer with their respective software packages in comparison to freeware data analysis software. In particular, the ability to obtain correct elemental formulas and MS/MS spectra as well as yielding acylcarnitines as biomarkers of physical training in unbiased metabolomics comparisons of mouse plasma was assessed.
Extraction/Chromatography: Plasma samples were collected from wild type C57BL/6 mice representing physically trained and untrained sample groups. Metabolites were extracted with a MeOH/H2O solvent system and separated on a Waters 1.7 μm BEH HILIC column with an acetonitrile/H2O gradient. LC eluents were analyzed with the Agilent Q-TOF with true MS/MS triggering and the Leco HRT which employs in-source CID with data deconvolution to align MS/MS spectra with parent ions.
Data Processing: Q-TOF raw data were analyzed by Agilent MassHunter and MassProfilerProfessional software. Processing of Leco HRT data was performed with Leco ChromaTOF HRT software. Results of both instrument data files were further compared with the open source mzMine tool and the Genedata Refiner MS Expressionist software.
Preliminary Data
Compound Identification: Analysis of eight biological replicates per group enabled detection of hundreds of accurate mass/retention time features. Both instruments yielded accurate masses at or below 1-2 mDa mass error in full scan MS mode. Using the LipidBLAST MS/MS library and accurate mass matching, over 50 complex lipids (triglycerides, phosphatidylethanolamines, phosphatidylcholines, and sphingomyelins), 12 acylcarnitines, and various other small molecules were identified such as free carnitine and choline. Identified structures covered three to four log units of signal intensity, eluting between 90% and 40% organic.
Software: None of the test software platforms (MassHunter, ChromaTOF HRT, Genedata Refiner MS and the open source mzMine tool) report isotope ratio abundances after peak finding in a fully automatic way, forcing laborious manual investigations. Depending on settings and parameters, a wide range of detected m/z-retention time features were retrieved, and examples of the balance between false positive/false negative metabolite detections are presented. Similarly, software was compared for automatic generation of correct elemental formulas for identified peaks, with the aim of retrieving the correct elemental formula as a top hit when constraining atom lists to C, H, N, O, P, S. Tables are presented for successful calculations of elemental formulas as a top hit, or within the top-5 hits, in comparison to the open source Fiehnlab 7-Golden-Rules tool. Lastly, mass accuracy and spectral purity of MS/MS spectra are compared between the Q-TOF MS/MS method and the HRT in-source CID method, presenting the best-hit identification accuracy using the Fiehnlab LipidBLAST software.
Exercise training effect: Trained and untrained mice differ in the ability to perform complete lipid oxidation. Long-chain acyl carnitines with 14-20 carbon fatty acid chains displayed a statistically significant (p < 0.05) decrease of >30% in the trained population when compared to the untrained control, suggesting changes relating to mitochondrial β-oxidation.
SessionPoster: WP20 - Informatics: Small Molecular Identification and Characterization, poster number: 365, Wednesday
The identification of small molecules using tandem mass spectrometry suffers from the non-existence of large MS/MS databases. Around 11.8 million chemicals are commercially available and approximately 100 million compounds are known. The largest MS/MS databases from NIST, MassBank and ReSpect only cover around 10,000 experimental CID tandem mass spectra obtained from small molecules. Computer generated (in-silico) mass spectral databases can be created to fill that gap. Unlike in proteomics where MS/MS information can be deduced from large genomic sequence databases, such an approach is not directly applicable for diverse small molecules. We used a bioinformatics algorithm to analyze molecule classes with consistent fragmentation patterns and generated in-silico tandem mass spectral libraries for such small molecule compound classes.
Methods
Experimental tandem mass spectra were collected from existing research publications. Compound structures were obtained from molecule databases or were generated using combinatorial algorithms. Structures were transferred using ChemAxon Instant-JChem into EXCEL templates. The generation of in-silico mass spectral databases requires the modeling of fragmentation patterns, peak abundances and peak annotations. Our algorithm uses deterministic models for fragmentation patterns and heuristic models for peak abundance modeling. Visual Basic for Applications was used to export the data to an external exchange format. For MS/MS library search we used the freely available NIST algorithm (NIST MS Search Software 2010) with accurate mass pre-filter and dot-product matching.
We exemplify our application with validation spectra from ceramide-1-phosphates, a series of sulfatides and the glycans of gangliosides including GM1, GM2 and GD1a. Many of those compounds and their related synthases are important regulators in signal transduction processes. They are ubiquitous in almost all vertebrates and bacterial cells. They recently have been overexpressed in various types of cancer such as breast, bladder and lung cancer and therefore represent an important molecular class in cancer research.
In order to validate the in-silico generated MS/MS databases a series of experiments were required. That included decoy database search, database search in itself and the search of independent experimental spectra. Experimental tandem mass spectra from triple quadrupole, Orbitrap, FT-ICR-MS and ion trap mass spectrometers covering ESI, MALDI, DESI and APCI ionization were investigated.
The use of MS/MS fragmentations, compared to simple accurate mass lookup, introduces an additional level of confidence because not only the accurate precursor mass is used, but additional molecule fragmentation data is included for comparison of spectra from different structural isomers. MS/MS database search therefore uses two selective filters. The first filter is the precursor filter that searches an accurate mass within a certain m/z window. The second filter is a classic mass spectral database matching algorithm that generates a search score.
High precursor mass accuracy usually yields fewer result candidates. However fragment rich MS/MS spectra from unit resolution mass spectrometers are also sufficient because the subsequent dot-product algorithm generates higher hit scores when fragment data matches. In case of very similar structures or stereoisomers an additional chromatographic separation with retention time matching is needed. Additional retention time data allows for compound annotations with the highest level of confidence.
SessionOral: ThOA pm - Biomarkers/Disease Signature, time: 3:10, Thursday, Wells Fargo
Introduction
Biomarkers are needed for many complex diseases, not simply for the presence/absence of disease, but for disease progression, correlation with severity and expected outcomes. In addition to the variability of disease phenotypes, patients have extremely variable response to drug treatment; while some of this variability can be attributed to cytochrome P450 activity, much is of unknown origin. We have investigated cardiovascular disease and the effect of cholesterol-lowering statin drugs, simvastatin and lovastatin. Here we integrate multiple mass spectrometry measurements on a variety of metabolomics and lipidiomics platforms with patient clinical data, clinical chemistry, transcriptomics, gut microbiome measures, and genetics, including GWAS. This meta-analysis approach has led to new understanding and prediction of patient response to statins.
Multiple mass spectrometry platforms were used to analyze plasma and cerebrospinal fluid. The metabolomics of polar compounds (amino acids, sugars, organic acids) utilized GCTOF mass spectrometry using BINBASE software and library. A lipidomics platform was used that combines lipid class separation with GC analysis of the derived fatty acids, and a third targeted platform measured cholesterols and bile acids. Mass spectrometry platforms were evaluated in a ring trial, by distributing NIST plasma to each laboratory.
A wide variety of data were integrated for this project, including clinical chemistry data such as cholesterol and C-reactive protein, imaging data, lipid profiles, patient outcomes. This facilitated finding new correlations via metaanalysis; new approaches were developed to analyze and integrate this data.
Preliminary Data
Specific biomarkers of cardiovascular disease, statin primary and secondary effects, and prediction of drug response have been found. For example, we found that statins induce highly specific changes in the amino acid profiles in plasma, indicating specific, coordinated changes in synthetic and degradative pathways. Specific amino acid transporters appear to also be in substrate competition with statin drugs. These changes in amino acids are also found in cerebrospinal fluid, suggesting that although cholesterol levels are not significantly altered in the brain with drug treatment, that there may be other, off-pathway effects in the brain that parallel systemic effects.
Another preliminary finding was that baseline cholesterol ester and phospholipid levels correlated with the response of LDL levels to simvastatin treatment. The magnitude of c-reactive protein (CRP) level response correlated with baseline plasmalogen levels. The lack of overlap between the metabolic predictors of LDL and CRP suggests that these effects occur by two separate pathways.
Compounds produced and/or transformed by the microbiome were found to be major predictors of the magnitude of LDL-C reduction in response to simvastatin. Simvastatin shares a liver and intestinal transporter molecule with gut microbiome-derived small molecule compounds, and plasma levels of simvastatin were found to correlate with levels of these microbiome molecules, suggesting a potential transport phenomenon. Polymorphisms in the statin transporter highlighted in a GWAS study correlate with microbiome metabolites that are also known substrates of this transporter. These results suggest that competition of drug for transport with endogenous metabolites may play a major role, not only in pharmacokinetics, but also in patient outcomes.
Preliminary results of the ring trial comparison using standardized plasma will be presented, in which the results from five metabolomics platforms are compared. Such inter-platform evaluations are essential to estabilish a rigorous basis for this type of meta-analysis.