Development and application of methodologies for non-targeted metabolomics in animal models of lung injury / a dissertation by Shama Naz; supervisor Coral Barbas Arribas and co-supervisor Antonia García Fernández.

Resumen

Non-targeted metabolomics approach offers the complete representation of a complex biological system through the combination of data-rich high throughput analytical techniques and MVDA. In comparison to other complex molecules such as proteins or nucleic acids, many metabolites changes faster, thus metabolites reflect the more adequate changes of any biochemical effects in an organism. Non-targeted metabolomics permits the analysis of all the small molecules (metabolites) that collectively constitutes the entire metabolome in a given biological fluids and tissues. Numerous analytical methods have been developed for non-targeted metabolomics based on NMR and MS coupled with different separation techniques such as LC, GC and CE. Working with these highly sensitive techniques requires development of robust and reliable method, which will be utilized for its intended use. The crucial points in any analytical method development are the validation of every step to get reliable and reproducible result and the challenges are completely new for non-targeted approach and different to target methods. Beside method validation, sample pre-treatment also requires careful consideration. Different sample types are being employed in metabolomics. However most commonly used sample types are bio-fluids such as urine and serum/plasma due to their minimally invasive nature. Although invasive, tissue analysis provides site specific information because in any altered conditions the changes are first seen at tissue site. ALI is an important cause of pulmonary and non-pulmonary morbidity in patients who survive hospitalization. Patients with severe breathing problem due to various factors/reasons admitted to ICU, where they are provided with mechanical ventilation to acquire proper respiration. The mechanical ventilation with high tidal volume could affect the lung through the process of continuous stretching and de-stretching, this condition is termed as VILI. Laterally due to other reasons, 50% severe septic patients admitted to ICU develop lung injury which is termed as SI-ALI. Upon injury the normal epithelial fluids transport disrupted and contribute to lung epithelial flooding. Interleukin-1ß and tumor necrosis factor-¿ have been evaluated to improve the understanding of ALI pathophysiology. However, the diagnostic or prognostic capability of these factors as a biomarker are very much non-specific. Thus there is a need for specific screening markers which can be used for easy diagnosis of ALI and also help in understanding the pathophysiology, irrespective of its heterogeneous characteristics. The study of metabolomics is of great clinical interest because of the urgent need to: (i) identify metabolic profile to extend the understanding of ALI pathophysiology and mechanism and (ii) identify new screening markers for disease diagnosis and helping the clinicians¿ in advance to follow a proper therapeutic treatment procedure, which is lacking in ICU. Considering the above mentioned facts the main goal of this dissertation is to develop and validate methodologies both in single and multiplatform aspects, suitable for serum and tissue non-targeted metabolomics approaches, providing reliable and effective data sets. Upon developing and validating methodologies for non-targeted metabolomics analysis, the following purpose of this project is to apply them on serum and lung samples from animal models of lung injury which will help in identifying markers associated to ALI as well its pathophysiology and molecular mechanisms. To attain these objectives the present work was divided in four chapters. Chapter-1: First of all an intensive review was done on the available validation strategies for MS-based non-targeted metabolomics approach. There are many methods existing for serum non-targeted analysis centered on LC-MS, GC-MS and NMR. Among the little published CE-MS serum metabolomics methods, none of them are adequately validated, involves high sample volume and time consuming sample preparation steps. In this section, a CE-MS method for non-targeted serum analysis was developed by using the high mass accuracy capabilities of TOF mass analyzer in combination with automated feature extraction and database search. A very easy and less time consuming serum sample pre-treatment, suitable for CE-MS analysis was also established. Traditional SPE extraction protocol which is usually time-consuming and also one step ultrafiltration protocol were tested. However, one step ultrafiltration with high protein cut-off, minimal sample volume, and lowest dilution, provided more accurate, precise with high recovery, better reproducibility and wider metabolite information compared to traditional SPE protocol. The initial sample volume is always a very crucial point specially while studying animal model. Different serum volumes were tried and to keep it as less with better reproducibility 100 µL was selected, which is much lower than previously published CE-MS based serum treatment protocols. Generally water is a very common solvent for CE-MS sample pre-treatment, however extraction with water provides identification of only free metabolites. While performing deproteinization, to remove proteins from serum, some metabolites could be retained by the protein. In order to improve the compound categories, an organic solvent, 5% acetonitrile was added and the signal instability generated from the addition of organic solvent was minimized by adding electrolyte (0.2 M formic acid). Several dilutions were tested, 1:1 was chosen for final sample treatment. Along with other instrumental parameter the injection time was also optimized to 35 seconds, to lessen the amount of sample for injection. A complete fingerprinting was performed in order to characterize rat serum using CE-MS, giving 35 identified compounds of different classes including organic acids, amino acid derivatives and carnitines. Many other metabolites were found but could not be identified due to the un-availability of authentic standards. The metabolite types prove the usefulness of CE-MS and its complementariness to other techniques when exploring altered metabolic pathways. The robustness of the developed methodology was tested by validating the method in terms of linearity, accuracy, precision, LOD and LOQ, choosing some identified compounds form the fingerprinting list. Different compound categories with various retention times covering the entire electropherogram were considered. All the metabolites used for validation fit the linear model (r > 0.99) without any biasness and with high accuracy and precision, demonstrating the reliability and reproducibility of the developed method. Eventually the validated method was then applied on serum samples of rat model of VILI, a mode of ALI that was not studied before using CE-MS. A set of distinctive metabolites were seen separating VILI from control groups. A decrease in arginine and an increase in asymmetric-di-methyl arginine and ornithine postulated that the condition ALI provoke lung airway remodeling, inflammation and oxidative stress. On the contrary, during mechanical ventilation lung also experiments a protective mechanism through cholinergic anti-inflammatory reflex. An increased level of choline was also found in the VILI animal model in this study, confirming that this protective mechanism is active in this pathology. Alongside this application, this developed method has been applied to several other application successfully and being routinely used in the laboratory for serum/plasma analysis. Chapter-2: LC-MS represents an important analytical platform for non-targeted metabolomics, advancing with high sensitivity and potentiality for biomarker identification. Thus the broad applicability of LC-MS to metabolites of all classes leads to accept it as a first choice of consideration. Previous studies have described the application of non-targeted metabolomics approach on lung injury induced by hyperoxic, gamma-irradiation and sepsis using NMR and LC-MS. However there are no publications available for non-targeted metabolomics application on VILI models. Thus an already developed method in our laboratory for serum analysis based on LC-QTOF-MS was applied on the VILI and control samples of rat to discover distinctive metabolite information than CE-MS. The repeatability of the methodology was tested by the clustering of QC samples, which were prepared by the aliquot of same sample volume of VILI and controls. This type of method validation is the most commonly used by researchers as an alternative validation approach. To find representative differentiating markers for ALI, obtained features were then filtered by choosing the data that had ¿present¿ calls in 100% of samples in any group. Applying univariate and MVDA, 44 significant masses were found, 15 of them were then confirmed by MS/MS analysis. The identified metabolites were belonging to the class of phospholipids, steroids, vitamin, amino acids, fatty amide and bile acids. The PLS-DA and OPLS-DA models showed the distinguishing separation of VILI from control group. The observed metabolites were then correlated with the biological pathways to find their relation with ALI. Sphingosine and sphingosine-1-phosphate are the metabolites of ceramide pathway which has anti-apoptotic activity, and it has been proposed that the balance of ceramide and sphingosine-1-phosphate determines the fate of the cell. Hexadecenal is also associated with cell apoptosis. The decreased sphingosine level could be involved in the cell apoptosis by the activation of ceramide synthase or by the increased degradation of hexadecenal. Decreased lyso-phospholipids and increased oleamide and phospholipids could be explained by the interruption in the lands cycle, which regulates the phospholipid metabolism. The changes in lysine, vitamin and bile acids proved the known facts about ALI which are altered collagen metabolisms and multi-organ failure. According to the results, the damage that occurs in the surfactant as a result of oxidative stress, promotes the generation of lysophosphatidylcholines that are degraded by lysophospholypase di-esterase on endocannabinoids that promote de novo synthesis of ceramides. This contributes to ALI by activating the metabolic pathway in which ceramide is converted into sphingosine-1-phosphate that inhibits apoptosis of neutrophils. One of its metabolites detected by LC-QTOF-MS is hexadecenal, which causes alterations in cytoskeletal reorganization and apoptosis. Lastly observations of this present study including the changes in certain metabolites and lipids shed light into the pathogenesis of ALI. Chapter-3: Before developing non-targeted metabolomics method for lung tissue analysis a critical review was performed to evaluate the available research based on animal/human tissue using MS detection approaches. Tissue provides more innovative information than bio-fluids; however the sample preparation is its main challenge. A LC-MS, GC-MS and CE-MS based multiplatform method has been developed for the non-targeted analysis of lung tissue, in combination with automated feature extraction and database search. An initial solvent for homogenization was optimized choosing from: water with 5% formic acid, water: methanol 50:50 and water: methanol 50:50 with 5% formic acid. Except CE-MS, LC-MS and GC-MS optimization results clearly reflected that water: methanol 50:50, was the best choice for these two platforms. In order to have an identical initial sample protocol (homogenization step) for lung tissue, as a compromise water: methanol 50:50, was also chosen for CE-MS as the homogenization solvent. Along with homogenization solvent, the sample volume was also optimized. Using this optimized method and with only 20 mg of tissue for the three techniques, a broader range of metabolites were putatively identified, covering polar to non-polar and different biochemical classes. This sample volume is very much applicable to biopsy sample in biomedical research. An extraction protocol was optimized targeting single phase extraction, using 80:20 methanol: MTBE, and injecting the same extract in all platforms. However the high quantity of organic solvent employed for non-polar metabolite extraction exerts less signal intensity in CE-MS. Hence to obtain better signal and reproducible analysis, the developed one step ultrafiltration protocol for CE-MS analysis was applied for lung tissue. Keeping other instrumental conditions similar, only the injection time was optimized to 50 seconds to increase the metabolite numbers. The 80:20 methanol: MTBE extract was split in two parts, 100 µL injected directly in LC-MS and 300 µL dried and derivatized for GC-MS analysis. A previously optimized instrumental condition for non-polar compounds was applied for LC-MS analysis; however the gradient was optimized to decrease the analysis time. Concomitant analyses with these multiplatform techniques a complete mouse lung fingerprinting, 1254 metabolites (1114, 69 and 85 metabolites from LC-MS, GC-MS and CE-MS respectively) of different biochemical classes were identified from only 20 mg of lung tissue. The importance of multiplatform approach was proved by the fact that only seven compounds were found in common. A particular technique is not sufficient to cover the entire metabolome which has also been proved by the fact that each technique analyses a unique category/ class of compounds. With the developed method, LC-MS provided a wide range of compounds from polar to non-polar, however the combination of CE-MS and GC-MS added more information about charged compounds and, free and volatile metabolites respectively. A complete validation was performed in all three platforms selecting ten different compounds from each technique from mouse lung profiling, considering the retention/ migration time, covering the entire chromatogram/ electropherogram and various functional groups, polarities and molecular masses. Compounds were chosen independent and common between and among analytical techniques. All the metabolites used for validation fit the linear model (r > 0.99) with high accuracy and precision. The optimized and validated method was then applied on a small set of lung samples of animal model of sepsis and control successfully. After chemometrics and statistical analysis 48 significant compounds were detected from three platforms, only propionyl-carnitine was common in LC-MS and CE-MS. Though sepsis was implemented through CLP but the significant compound from lung analysis proved the literature findings that sepsis ultimately results in multi-organ failure. The obtained result support the utility of multiplatform application in the rapid and simple screening for alterations in any type of biopsy tissue analysis. Chapter-4: In this section the developed non-targeted multiplatform method for lung tissue was applied on rat lung tissue, to find out the exact reason behind SI-ALI, whether it is the effect of mechanical ventilation or sepsis. To accomplish this study animal models of sepsis (induced by CLP), VILI (injury by mechanical ventilation and SI-ALI (injury due to sepsis as explained) along with their corresponding control groups were used. The developed method was successfully applied to the lung tissue of these models. The obtained features were filtered applying 80% filter by frequency and 30% CV on quality control for all platforms. The GC-MS metabolites were identified by comparing their mass fragmentation patterns with those available in the NIST mass spectral library and Fiehn Retention Time Library. In CE-MS, compounds were identified by matching the retention time from the in-house library and rest of them were kept as putative. However in LC-MS, metabolites were putatively identified not only using the accurate mass but also checking their isotopic pattern (higher score pattern) and retention time. Both univariate and MV statistical analyses are applied, 112, 35, 15 and 39 significant metabolites were obtained respectively from LC-MS (+), LC-MS (-), GC and CE-MS analysis, in total for all comparisons. The quality of the entire run was validated by the clustering of QCs in PCA models respective to the instrumental analysis. Both the PCA and PLS-DA models for the six groups distinctively showed that the separation is due to ventilation rather than sepsis. Although effect of sepsis, ventilation, and ventilation along with sepsis were checked separately, provided a well separation in all platforms. The identified significant metabolites were then correlated with the corresponding biological pathways. Metabolic changes indicated increased oxidative stress, changes in purine, energy, carnitine, amino acid, urea cycle, vitamins, collagen, ceramide-sphingomyelin and phospholipid metabolism. In summary, with the proposed research objectives rapid and simple methodologies have been developed and validated both in single and multiplatform aspects for non-targeted metabolomics. Validation parameters are adequate for bio-analysis and sample treatments are very simple permitting the detection of a wide range of compounds. A complete fingerprinting is obtained for rat serum and mouse lung. Finally the application of the validated methods was able to answer questions based on the need to solve a problem of great social and economic impact with high morbidity and mortality of ALI in ICU.