LC and LC-free mass spectrometry applications in non-targeted metabolomics for disease detection and early prediction
MetadataShow full item record
Metabolomics is the science of studying small molecule composition of biological systems. Non-targeted metabolomics, as the analytical technology for unbiased simultaneous measurement and analysis of the collection of low molecular weight metabolites within biological samples, has been widely adopted as a novel and powerful approach to study pathophysiological processes and discover potential biomarkers for disease diagnosis and preventive screening. By comparing and analyzing the global metabolome of different classes of samples with different phenotypes, non-targeted metabolomics can serve as a top-down strategy to discover disease related metabolic perturbations, and it has been applied in studies of various diseases. In this thesis work, mass spectrometry (MS) based non-targeted metabolomics was applied to discover potential biomarkers of two kinds of diseases: prostate cancer (PCa) and cystic fibrosis (CF) acute pulmonary exacerbations (APEs). Current clinical practices for prostate cancer (PCa) diagnosis focus on prostate-specific antigen (PSA) level. Although it exibits fair discriminating power for PCa detection, the PSA test for PCa screening remains controversial due to the risk of over-diagnosis and overtreatment. Another disease we have studied, CF lung disease, has intermittent episodes of acute worsening of symptoms termed acute pulmonary exacerbations (APEs), which is a major cause of morbidity for CF patients. To date, however, there is no consensus diagnostic criteria for CF APEs. Also, there is no preventive screening method for stable CF patients to signal an oncoming APE event, which hinders the initiation of early intervention before the establishment of substantial immune response. These drawbacks, together with a lack of in-depth information on the pathophysiology of these two diseases may prevent clinicians from making the best possible therapeutic interventions and treatment decisions to improve patient healthcare. Consequently, there has been a constant drive to discover novel biomarkers to improve PCa diagnosis and prediction of APE onset in CF patients via non-targeted metabolomics strategy. Mass spectrometry (MS) has been increasingly applied in metabolomics studies due to its high sensitivity. MS methods often include chromatography separation prior to ion detection, which helps to increase metabolite coverage and resolution, decrease spectral congestion and ion suppression (or enhancement) effects. As current metabolomics research focuses more on large scale studies with hundreds to thousands of samples, high-throughput metabolic profiling techniques with fast sample analysis speed become a pivotal necessity. Flow injection (FI) and direct infusion (DI) MS are alternative approaches involving direct introduction of biological samples into MS systems without prior chromatography separation, increasing sample analysis speed. The combination of FI or DI methods with ion mobility (IM) MS is generally appealing for its ability to simplify spectra, raise signal to noise ratio by eliminating chemical noise, produce cleaner MS/MS spectra and provide rapid separation of closely related compounds. Therefore this strategy has great potential in non-targeted metabolomics research demanding high sample throughput. In this thesis work, liquid chromatography (LC) MS method and LC-free FI-IM-MS and DI-IM-MS methods were employed for metabolic profiling of biological samples to find potential biomarkers for PCa and CF APEs and study the associated metabolic perturbations. An introduction to MS-based non-targeted metabolic profiling for human disease studies is provided in Chapter 1, with recent developments in disease biomarker discovery reviewed. Sample preparation, MS platforms utilized, metabolite identification, innovations in data analysis and pathway mapping were discussed. Part I of the dissertation consists of Chapters 2 and 3, which present LC-MS based non-targeted metabolomics studies of PCa and CF APE diseases. In Chapter 2, a metabolite-based in vitro diagnostic multivariate index assay (IVDMIA) was developed to predict PCa in serum samples with a panel of 40 metabolic features, yielding 92.1% sensitivity, 94.3% specificity, and 93.0% accuracy. The performance of the IVDMIA was demonstrated to be higher than the prevalent PSA test. The identification of amino acids, fatty acids, lysophospholipids, and bile acids provided insights into the metabolic alterations associated with the disease. In addition, several metabolites were mapped to the steroid hormone biosynthesis pathway, indicating its association with PCa. Chapter 3 discusses the feasibility of predicting APE in CF patients using EBC metabolites. In a pilot study, LC-MS was used to profile metabolites in exhaled breath condensate (EBC) samples in negative ion mode from 17 clinically stable CF patients, 9 CF patients with an APE severe enough to require hospitalization (termed APE), 5 CF patients during recovery from a severe APE (termed post-APE), and 4 CF patients who were clinically stable at the time of collection but in the subsequent 1 to 3 months developed a severe APE (termed pre-APE). Using multivariate analysis, a panel containing 2 metabolic discriminant features identified as 4-hydroxycyclohexylcarboxylic acid and pyroglutamic acid differentiated the APE from the stable CF samples with 84.6% accuracy. In addition, the pre-APE samples were distinguished from the stable CF samples with 90.5% accuracy using a panel of two discriminant features including lactic acid and pyroglutamic acid. In a larger EBC sample cohort (n=210) study, negative ion mode data and the combination of negative and positive ion mode data showed that classification was possible for age and gender-matched samples grouped into adult and pediatric patients. Negative ion mode data yielded acceptable sensitivities (83.3% and 76.2%), specificities (91.7% and 83.7%), and accuracies (88.9% and 81.3%) for discriminating APE from stable CF EBC samples, from pediatric and adult patients, respectively. For the pre-APE vs. stable CF comparison, good sensitivities (85.7% and 89.5%), specificities (88.4% and 84.1%), and accuracies (87.7% and 85.7%) were obtained for EBC samples from pediatric and adult patients, respectively. By combining positive with negative ion mode data, improved classification performance was achieved for most binary comparisons with accuracies enhanced between 3 and 9.6%. The discriminant metabolites identified in the pilot study were also selected in some of the discriminant metabolite panels. Some of the identified discriminant metabolites had microbial relevance, indicating a possible central role of bacterial metabolism in APE development. Part II of the dissertation includes Chapters 4 and 5, describing non-targeted metabolomics studies on PCa and CF APE disease using LC-free FI-IM-MS and DI-IM-MS. Chapter 4 presents the application of FI-IM-MS to the non-targeted metabolic profiling of serum extracts from 61 PCa patients and 42 controls from the same cohort in Chapter 2. Comprehensive data mining of the mobility-mass domain was used to discriminate compounds with various charges and filter matrix salt cluster ions. Specific criteria were set to ensure correct grouping of adducts, in-source fragments, and impurities in the dataset. Endogenous metabolites were identified with high confidence using tandem MS experiments and collision cross-section (CCS) matching with chemical standards or CCS databases. PCa patient samples were distinguished from control samples with good accuracies (88.3-89.3%), sensitivities (88.5-90.2%), and specificities (88.1%) using supervised multivariate classification methods. Results from this study show the potential of FI-IM-MS as a high throughput metabolic profiling tool for large scale metabolomics studies. In Chapter 5, transmission-mode direct analysis in real time (TM-DART) coupled to IM-MS was tested as a high-throughput alternative to conventional DI electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) methods, and the performances of the three ionization methods were compared. When using pooled EBC collected from a healthy control, ESI detected the most metabolites, and TM-DART the least. TM-DART-TWIM-TOF-MS was used to profile metabolites in the EBC samples from 5 healthy individuals and 4 CF patients, and a panel of 3 discriminant EBC metabolites was found to differentiate these two classes with excellent cross-validated accuracy. Appendix A presents a collaborative work that combined results from surface enhanced Raman spectroscopy (SERS), metabolomics and proteomics experiments, to study the molecular mechanisms of the cellular processes during the plasmonic photothermal therapy (PPTT) process. Our metabolomics results showed increased levels of phenylalanine and metabolites tentatively identified as its derivatives and phenylalanine-containing peptides, aiding in assignments of SERS bands with observed changes during PPTT. To better understand the mechanism of phenylalanine increase upon PPTT, we combined metabolomics and proteomics results using network analysis, which demonstrated that phenylalanine metabolism was perturbed. In addition, several apoptosis pathways were activated via key proteins (e.g. HADHA and ACAT1), which are consistent with the proposed role of altered phenylalanine metabolism in inducing apoptosis. At last, Chapter 6 summarizes the conclusions drawn from the thesis work, and also presents the outlook and possible future work.