Post on 17-Jan-2020
transcript
Multi-platform metabolomics assays to study the responsiveness of the human plasma and lung lavage metabolome
Masoumeh Karimpour
Doctoral thesis, Department of Chemistry,
Umeå University, 2016
This work is protected by the Swedish Copyright Legislation (Act 1960:729)
ISBN: 978-91-7601-506-3
Cover picture: online word cloud generator (Tagul)
Electronic version available at http://umu.diva-portal.org/
Printed at the KBC Service Centre, Umeå University
Umeå, Sweden, 2016
i
Table of Contents
ABSTRACT ........................................................................................................................................................ III
SAMMANFATTNING (SUMMARY IN SWEDISH) ................................................................................ V
LIST OF ABBREVIATIONS ......................................................................................................................... VII
LIST OF PUBLICATIONS ........................................................................................................................... VIII
BACKGROUND ................................................................................................................................................ 1
THE HISTORY OF METABOLOMICS ............................................................................................................................. 1 DEFINITION OF KEY CONCEPTS .................................................................................................................................. 2 AIM OF THE WORK ..................................................................................................................................................... 3
METHODS ......................................................................................................................................................... 4
MULTI-PLATFORM METABOLOMICS .......................................................................................................................... 4 Non-targeted metabolomics ............................................................................................................................. 4 Targeted metabolomics ..................................................................................................................................... 5 Advantages and disadvantages of different metabolomics assays ........................................................... 5
ANALYTICAL WORKFLOW ........................................................................................................................................... 6 Biofluids, sampling and exposure setup ......................................................................................................... 8 Chromatography techniques coupled to mass spectrometry .................................................................... 11
Gas chromatography-time-of-flight mass spectrometry (GC-TOF-MS) ................................................................. 11 Liquid chromatography-quadrupole time-of-flight mass spectrometry (LC-QTOF-MS) ..................................... 12 Liquid chromatography-tandem mass spectrometry (LC -MS/MS) ....................................................................... 13
Nuclear magnetic resonance (NMR) spectroscopy .................................................................................... 13 DATA ANALYSIS ........................................................................................................................................................ 15
Univariate analysis (UVA) ............................................................................................................................. 15 Multivariate analysis (MVA) ......................................................................................................................... 16
Principal component analysis (PCA) ........................................................................................................................... 16 Orthogonal projections to latent structures-discriminant analysis (OPLS-DA) .................................................... 18
Statistical validation of multivariate models .............................................................................................. 18 Receiver operating characteristic (ROC) curve .......................................................................................... 20
RESULTS AND DISCUSSION .................................................................................................................... 22
A METABOLOMICS PILOT STUDY OF THE POSTPRANDIAL PHASE (PAPER I) ............................................................. 22 Objective .......................................................................................................................................................................... 22 Main findings and discussion of results in Paper I .................................................................................................... 23
AIR POLLUTION EXPOSURE STUDIES (PAPER II, III AND IV) .................................................................................. 24 Paper II .............................................................................................................................................................. 26
Objective .......................................................................................................................................................................... 27 Main findings and discussion of results in Paper II .................................................................................................. 27
Paper III ............................................................................................................................................................ 28 Objective .......................................................................................................................................................................... 29 Main findings and discussion of results in Paper III ................................................................................................. 30
Paper IV ............................................................................................................................................................. 30 Objective .......................................................................................................................................................................... 30 Main findings and discussion of results in Paper IV ................................................................................................. 30
CONCLUSION AND FUTURE PERSPECTIVES .................................................................................... 32
ii
FUTURE PERSPECTIVES ............................................................................................................................................ 33
ACKNOWLEDGMENT .................................................................................................................................. 34
REFERENCES ................................................................................................................................................. 35
iii
Abstract
Metabolomics as a field has been used to track changes and perturbations in the human body by
investigating metabolite profiles indicating the change of metabolite levels over time and in
response to different challenges. In this thesis work, the main focus was on applying
multiplatform-metabolomics to study the human metabolome following exposure to
perturbations, such as diet (in the form of a challenge meal) and exhaust emissions (air pollution
exposure in a controlled setting). The cutting-edge analytical platforms used for this purpose were
nuclear magnetic resonance (NMR), as well as gas chromatography (GC) and liquid
chromatography (LC) coupled to mass spectrometry (MS). Each platform offered unique
characterization features, allowing detection and identification of a specific range of metabolites.
The use of multiplatform-metabolomics was found to enhance the metabolome coverage and to
provide complementary findings that enabled a better understanding of the biochemical processes
reflected by the metabolite profiles. Using non-targeted analysis, a wide range of unknown
metabolites in plasma were identified during the postprandial stage after a well-defined challenge
meal (in Paper I). In addition, a considerable number of metabolites were detected and identified
in lung lavage fluid after biodiesel exhaust exposure compared to filtered air exposure (in Paper
II). In parallel, using targeted analysis, both lung lavage and plasma fatty acid metabolites were
detected and quantified in response to filtered air and biodiesel exhaust exposure (in Paper III
and IV).
Data processing of raw data followed by data analysis, using both univariate and multivariate
methods, enabled changes occurring in metabolites levels to be screened and investigated. For the
initial pilot postprandial study, the aim was to investigate the plasma metabolome response after
a well-defined meal during the postprandial stage for two types of diet. It was found that
independent of the background diet type, levels of metabolites returned to their baseline levels
after three hours. This finding was taken into consideration for the biodiesel exhaust exposures
studies, designed to limit the impact of dietary effects. Both targeted and non-targeted approaches
resulted in important findings. For instance, different metabolite profiles were detected in
bronchial wash (BW) compared to bronchoalveolar lavage (BAL) fluid with mainly NMR and LC-
MS. Furthermore, biodiesel exhaust exposure resulted in different metabolite profiles as observed
by GC-MS, especially in BAL. In addition, fatty acid metabolites in BW, BAL, and plasma were
shown to be responsive to biodiesel exhaust exposure, as measured by a targeted LC-MS/MS
iv
protocol. In summary, the new analytical methods developed to investigate the responsiveness of
the human plasma and lung lavage metabolome proved to be useful in an analytical perspective,
and provided important biological findings. However, further studies are needed to validate these
results.
v
Sammanfattning (Summary in Swedish)
Metabolomik har använts för att spåra förändringar och störningar i kroppens funktioner genom
undersökning av metabolit-profiler. I detta avhandlingasarbete har huvudfokus varit på
tillämpning av flera olika analytiska plattformar för metabolomikstudier av det mänskliga
metabolomet efter exponering för olika kost och avgasutsläpp från biodieselbränsle. De
sofistikerade analytiska plattformarna som användes för detta ändamål var kärnmagnetisk
resonans (NMR), samt gaskromatografi (GC) och vätskekromatografi (LC) kopplat till
masspektrometri (MS). Varje plattform erbjöd unika karakteriseringsmöjligheter med detektion
och identifiering av specifika grupper av metaboliter. Användningen av multipattform-
metabolomik förbättrade täckningen av metabolomet och genererade kompletterande resultat
som möjliggjorde en bättre förståelse av de biokemiska processer som reflekteras av
metabolitprofilerna. Med hjälp av breda analyser har ett stort antal okända metaboliter i plasma
identifierats under den postprandial fasen efter en väldefinerad måltid (i Paper I). Dessutom har
ett stort antal metaboliter påvisats och identifierats i lungsköljvätska efter exponering av
biodieselavgaser jämfört med kontollexponering med filtrerad luft (i Paper II). Parallellt med
dessa breda analyser har också riktade analyser genomförts av både lungsköljvätska och plasma.
Därigenom har bioaktiva lipider detekterats och kvantifieras efter avgasexponering och
resultaten har jämförts med filtrerad luft som kontrollexponering (Paper III och IV).
Processning av rådata följt av dataanalys, med både univariata och multivariata metoder
möjliggjorde screening och fördjupad undersökning av förändringen i metabolitnivåer. I den
första pilotstudien av postprandiala nivåer var syftet att undersöka responsen i
plasmametabolomet efter en väldefinierad måltid under den postprandiala fasen vid två olika
typer av kost. Resultaten visade att oberoende av kosten, så återvände metabolitnivåerna till sina
baslinjenivåer tre timmar efter måltiden. Detta togs i beaktande vid exponeringsstudierna för
biodieselavgaser, som designades så att dietens inverkan minimerades. Både breda och riktade
analyser resulterade i viktiga resultat. Exempelvis så detekterades olika metabolitprofiler i
bronkiell sköljvätska (BW) jämfört med bronkoalveolär sköljvätska (BAL), speciellt med NMR och
LC-MS. Dessutom resulterade avgasexponering i förändrade metabolitprofiler, observerade med
GC-MS, särskilt i BAL. Dessutom uppvisade fettsyrametaboliter i BW, BAL och plasma
förändrade halter efter avgasexponering, uppmätt genom en riktad LC-MS/MS-analys.
Sammanfattningsvis så visade sig de nya metoderna som utvecklats för att undersöka
vi
förändringar i metabolithalterna i plasma och lungsköljvätska fungera väl ur ett analytiskt
perspektiv och resulterade i viktiga biologiska fynd. Fördjupade studier behövs dock för att
validera resultaten.
vii
List of Abbreviations
AUC Area Under the Curve
CV-ANOVA Cross Validation- ANalysis Of VAriance
EDTA Ethylene Diamine Tetraacetic Acid
GC-MS Gas Chromatography-Mass Spectrometry
LC-MS Liquid Chromatography-Mass Spectrometry
MVA MultiVariate Analysis
NMR Nucleic Magnetic Resonance
OPLS Orthogonal Projections to Latent Structures
OPLS-DA Orthogonal Projections to Latent Structures - Discriminant Analysis
PCA Principal Component Analysis
RME Rapeseed Methyl Ester
ROC Receiver Operating Characteristic
RSD Relative Standard Deviations
SMC Swedish Metabolomics Centre
SUS Shared and Unique Structure
TOF Time Of Flight
UVA UniVariate Analysis
UPSC Umeå Plant Science Centre
viii
List of publications This thesis is based on the following papers, which will be referred to by Roman numerals.
I. Karimpour M, Surowiec I, Wu J, Gouveia-Figueira S, Pinto R, Trygg J, Zivkovic A.L, L.
Nording M.L. Postprandial metabolomics: a pilot mass spectrometry and NMR study of
the human plasma metabolome in response to a challenge meal. Analytica Chimica Acta.
2016 Feb; 908: 121-131.
II. Surowiec I*, Karimpour M*, Gouveia-Figueira S, Wu J, Unosson J, Bosson J.A,
Blomberg A, Pourazar J, Sandström T, Behndig A.F, Trygg J, Nording M.L. Multi-platform
metabolomics assays for human lung lavage fluids in an air pollution exposure study. Anal
Bioanal Chem. 2016 Apr 25. [Epub ahead of print]
III. Gouveia-Figueira S, Karimpour M, Bosson J.A, Blomberg A, Unosson J, Pourazar J,
Sandström T, Behndig A.F, Nording M.L. Mass spectrometry profiling of oxylipins,
endocannabinoids and N-acylethanolamines in human lung lavage fluids reveal
responsiveness of prostaglandin E2 and associated lipid metabolites to biodiesel exhaust
exposure. (Submitted manuscript)
IV. Gouveia-Figueira S, Karimpour M, Bosson J.A, Pourazar J, Blomberg A, Unosson J,
Sandström T, Behndig A.F, Nording M.L. Effect of controlled exposure to biodiesel exhaust
on human plasma bioactive lipid profiles. (Manuscript)
*these authors contributed equally to this work
The published papers have been reprinted with the kind permission from the original
publishers.
1
“The body and its parts are in a continuous state of dissolution and nourishment, so they are
inevitably undergoing permanent change”
Ibn al-Nafis (1213–1288)
Background
The word metabolism originates from the Greek word “μεταβολισμός” (metavolismós), which
means “change”. Input from the environment and lifestyle can affect human metabolic pathways.
Hence, the study of metabolism and its constituents, the metabolites, is a powerful way to elucidate
factors behind the influence of the environment on humans [1]. Furthermore, integrative analysis
of metabolites plays an important role in describing the biochemical and biological mechanisms
in any complex regulatory system [1-3]. Thus, metabolomics is an emerging field of “omics”
science aimed at comprehensive detection and identification of the enormous metabolite content
in biological samples by using advanced analytical techniques and data analysis methods. Thereby,
metabolomics is an important tool for the identification of disease biomarkers, as well as drug
discovery, in addition to the elucidation of environmental influences.
The history of metabolomics
The first study to investigate a number of metabolites simultaneously was published in 1966 by
Dalgliesh et al., who applied gas chromatography (GC) with flame ionization detection (FID) for
the separation of a wide range of metabolites [4]. Mamer and Crawhall [5] and Horning and
Horning [6] performed the first mass spectrometry (MS) based metabolomics experiments. These
studies represented the beginning of the metabolomics development process, but the name
“metabolomics” had not yet been conceived. In 1971, Horning and Horning used GC-MS for
metabolite profiling and identification of metabolites in human samples [6]. In the same year,
Pauling et al. published a quantitative and qualitative analysis of human breath, which resulted in
a rich dataset of metabolites that were integrated by means of a computer program [7]. Although,
the latter study was not labeled as a metabolomics study at the time, this paper is generally
considered to be one of the first to give birth to the concept of metabolomics.
2
Metabonomics was introduced as a scientific field in 1999 by Nicholson et al. [3], and shortly after,
the term “metabolomics” was coined by Fiehn and colleagues to refer to investigations of the
metabolome [8]. Today the term “metabonomics” is widely used to describe multiple metabolic
changes caused by a biological perturbation. Metabolomics has a similar definition. However, it
places a greater emphasis on comprehensive metabolic profiling of various species [3, 8-10].
Definition of key concepts
The metabolome comprises a large number of small molecular weight compounds, such as lipids,
amino acids, nucleotides, organic acids, etc. Metabolome components span a diverse range of
compounds with different properties. For example, within lipids alone, there are high abundance
compounds, such as fatty acids, triglycerides or phospholipids, but other compounds with lower
abundance, such as eicosanoids derived from arachidonic acid, also have significant regulatory
effects [11, 12]. Owing to the large diversity of physicochemical properties and abundance levels of
metabolites, metabolomics requires a broad range of instrumentation and special protocols and
techniques for sample preparation, separation and detection.
Analytical techniques have been developed for metabolomics analysis of endogenous and
exogenous metabolites in many types of biological samples, such as human plasma, saliva, serum,
urine and lung lavage fluids, enabling identification of key metabolites and mechanisms behind
different pathologies [4, 13-20]. For instance, lung lavage analysis was successfully performed to
investigate lung inflammation in air pollution studies and pulmonary diseases, such as cystic
fibrosis, asthma, and respiratory distress syndrome [21-26]. Both non-targeted and targeted
analytical protocols have been used extensively under names such as untargeted metabolomics,
global metabolomics, metabolic fingerprinting, lipidomics, targeted metabolomics, MRM
metabolomics, etc., as reviewed by Wishart [1], Griffiths [10] and Lu [27].
3
Aim of the work
In this thesis work, the general goal was to study the human metabolome in plasma and lung
lavage fluid following different interventions, such as a challenge meal or exposure to air
pollutants (biodiesel exhaust). We hypothesized that applying multi-platform metabolomics
would enable detection and identification of a wide swath of metabolites and that targeted/non-
targeted analysis of the human metabolome during the postprandial stage and/or post exposure
would deepen the knowledge on systemic and pulmonary responses in these situations.
The specific aims were as follows:
I. To investigate the human plasma metabolome during the postprandial stage.
II. To evaluate human lung lavage metabolite profiles after filtered air and biodiesel exhaust
exposure.
III. To investigate bioactive lipid (oxylipins, endocannabinoids and N-acylethanolamines)
responses in human lung lavage fluid after exposure to filtered air and biodiesel exhaust.
IV. To explore bioactive lipid profiles in human plasma at different time points pre- and post-
exposure to filtered air and biodiesel exhaust.
4
Methods
This chapter presents and explains the methods used throughout this thesis work. The focus is on
describing all of the metabolomics steps used in the described studies. In addition, examples will
be given from Papers I-IV.
Multi-platform metabolomics
At present, there is no single-instrument platform that can fully analyze all metabolites in a given
sample. Both MS and nuclear magnetic resonance (NMR) are advanced techniques suitable for
metabolomics analysis. However, they have different analytical advantages and disadvantages. To
date, different protocols and methods have been introduced in metabolomics and it has been
shown that depending on the specific aims and hypotheses, different platforms can be employed.
Each platform has specific characterization features, which enables analysis of a particular range
of metabolites. Therefore, applying several platforms has the potential for obtaining more
comprehensive metabolite coverage [9, 28-31].
Non-targeted metabolomics
Non-targeted metabolomics, or global metabolome analysis, is a powerful approach that aims to
discover and identify a wide range of both unknown and known metabolites in a biological sample
[32]. Therefore, non-targeted metabolomics can allow discovery of novel metabolite species and
analysis of pathways and metabolism in biological and complex systems. The most used platforms
in non-targeted metabolomics are NMR, GC-MS and liquid chromatography-MS (LC-MS).
However, wide-range metabolomics analysis of samples is often challenging owing to the huge
number of compounds present from different chemical classes and at different concentrations.
Consequently, platforms such as quadrupole time-of-flight-MS (QTOF-MS) could be useful for
detecting metabolites of different chemical classes [20, 33-36]. In non-targeted metabolomics,
determination is achieved by using a semi-quantitative or quantitative concentration measure for
each metabolite. Identification of a metabolite, in addition to de novo structure elucidation, entails
the integrated analysis of exact mass, MS/MS fragmentation patterns and searches in libraries and
databases to find the right structure [10, 14].
5
Targeted metabolomics
Targeted metabolomics aims to measure a set of specific and predefined metabolites. Knowledge
of metabolites and their biochemical pathways plays a pivotal role in the discovery of novel
biomarkers [37, 38]. Based on the aims and hypotheses, special instruments can be selected for
analysis and specific analytical methods have been developed. LC-MS-based targeted
metabolomics analysis is performed using a triple quadrupole (QqQ)-MS, which enables multiple
reaction monitoring (MRM) of specific product ions and their unique one or two fragment ions to
achieve a high sensitivity and selectivity of detection [37, 39, 40]. To date, MRM targeted analyses
have been developed and optimized for a set of desired metabolites to enable analysis of low
abundant metabolites and provide a high sensitivity and wide dynamic range [12, 13, 41]. The
acquired data from targeted metabolomics is quantitative and can be used for biomarker discovery
and validation, and for constructing pathway maps [10]. MRM targeted analysis has some
limitations, e.g., it only measures known metabolites [37]. However, MRM-targeted analysis has
been used for the quantitative analysis of analytes such as oxylipins and endocanabinoids in
human plasma and lung lavage fluid [13, 21, 41, 42].
Advantages and disadvantages of different metabolomics assays
NMR has an important role in metabolomics owing to its easy and rapid sample preparation, non-
destructiveness, no need for chromatographic separation and high degree of reproducibility [18,
22]. NMR based metabolomics has been extensively used in drug toxicity studies [13],
environmental assessment [14], pharmacological drug discovery [15, 16] and pulmonary and
nutrition research [24, 29, 43]. However, it has lower sensitivity and resolution compared to MS-
based techniques.
MS-based metabolomics platforms combined with GC or LC separation methods have been
frequently used in metabolomics [10, 15, 20, 44-47] because of their higher sensitivity to a wide
range of compounds compared with other detection techniques, wide metabolome coverage and
possibility for metabolite identification. In metabolite profiling, LC-MS has several advantages
over GC-MS, e.g., no need for sample derivatization and capabilities for analyzing more polar and
higher molecular weight compounds. In contrast, compound deconvolution and compound
identification are more challenging in LC-MS analysis. GC-MS electron impact libraries are
standardized and comparable between different instruments, and retention indices are to some
6
extent comparable between different laboratories. In contrast, in LC, retention times depend on
the column used and MS/MS fragmentation spectra can differ between different instruments.
Hence, LC-MS libraries are best acquired on the same instrument as used for the metabolomics
analysis.
Analytical workflow
A typical metabolomics workflow is shown in Figure 1. It starts with sampling (A) and sample
preparation (B), continues with metabolite separation and detection (C) and data processing (D),
and ends with data analysis (E).
7
*gas chromatography-time of flight-mass spectrometry
**liquid chromatography-tandem mass spectrometry ***nuclear magnetic resonance
Figure 1. General overview of the metabolomics workflow.
(E) Data analysis
Univariate analysis Multivariate analysis Validation
(D) Data processing
Quantification Identificaion Normalization
(c) Metabolite separation and detection
Targeted metabolomics
(LC-MS/MS**)
Non-taregetd metabolomics
(GC-TOF-MS, NMR***, LC-TOF-MS)
(B) Sample preparation
Methanol/water extraction Solid phase extraction Derivitization ( for GC-TOF-MS*)
(A) Sampling
8
As mentioned above, the most widely used techniques in metabolomics are based on XC-MS,
where X is a chromatographic separation method, or NMR [9, 45, 48]. The studies in this thesis
are also based on these types of assays, as discussed in more detail below.
Biofluids, sampling and exposure setup
Almost 60 percent of the human body is made up of water, circulating between different body
compartments and units. The circulation plays a vital role in human life in delivering nutrients to
all cells in the body. Most human metabolomics measurements to date have been carried out on
biofluids because they contain physiological information due to their circulation. Plasma, serum,
saliva, urine, cerebrospinal fluid (CSF) and respiratory tract lining fluid (RTLF) are some
examples of human biofluids. Blood is a special biofluid owing to its central circulation, which
means it reflects processes taking place in organs [48].
Plasma is an extracellular matrix of blood cells that consists of dissolved proteins, clotting factors,
hormones, electrolytes, carbon dioxide, etc. Plasma is often collected by venipuncture into
standard vials containing anti-coagulant, e.g., ethylene diamine tetra acetate (EDTA), heparin or
citrate. Thus, extra resonances may be observed in NMR spectra due to the formation of complexes
between the anti-coagulants, such as EDTA, and ions, such as Ca2+ and Mg2+, in plasma [49].
Plasma was used in Paper I to investigate the postprandial response after a defined meal by
means of multiplatform metabolomics, and in Paper IV to explore the response of bioactive lipids
following exposure to biodiesel exhaust (rapeseed methyl ester (RME)-biodiesel) compared to
filtered air.
RME-biodiesel exhaust was generated from an idling diesel engine. More than 90% of the exhaust
gas was shunted away, and the remainder was diluted with filtered air at a temperature of 20°C
(relative humidity 50%) before being fed into a whole-body exposure chamber (3.0 x 3.0 x 2.4 m).
The chamber was monitored continuously for pollutants. Figure 2 shows the exposure chamber
setup used in the studies in Paper II-IV. A detailed description of the exposure facility and
bronchoscopy sampling procedure used is given in Paper II-IV.
In Paper II and III, airway responses following exposure to RME-biodiesel exhaust and filtered
air were assessed. Subjects were exposed to RME-biodiesel exhaust and filtered air in a
randomized fashion, at least three weeks apart. During exposure, the subjects alternated between
15 minutes of exercise on a bicycle ergometer (ventilation of 20 L/m2 of body surface) and 15
9
minutes of rest. Bronchoscopy was performed six hours after each exposure. Bronchial wash (BW)
(2 × 20 mL) and bronchoalveolar lavage (BAL) (3×60 mL) with sterile sodium chloride were
extracted in either the middle lobe or lingual lobes. The aspirates recovered were collected on ice
and filtered and centrifuged. The resulting supernatants and cell pellets were separated and the
supernatant was used for metabolomics analysis. Bronchoscopy is a safe and widely used routine
procedure for the diagnosis of many pulmonary diseases, as well as for research to evaluate airway
responses following exposure to, e.g., air pollutants [50, 51]. The characterization of the
metabolome within this compartment provides a novel opportunity for identifying biomarkers and
pathways in order to, e.g., investigate adverse health effects of air pollution in the human
respiratory tract.
Figure 2. Exposure chamber facility (A) and the engine (B) at SMP Svensk Maskinprovning AB, Umeå,
Sweden. Photos by Karimpour, June, 2013.
A)
B)
10
The spatial resolution of sampling for different lung compartments during metabolite profiling
can be improved by using both BW and BAL samples: the contents of BW fluid reflect the
metabolite profile of the central airways, whereas BAL fluid reflects the metabolite profile of the
lungs’ more peripheral regions. This dual approach has previously been applied to investigate
compartment-specific fatty acid metabolite (oxylipin) profiles [51]. BAL and BW samples were
used in Paper II and III to investigate the effects of biodiesel exposure on the lung lavage
metabolite profile and compare them to results obtained after filtered air exposure. Figure 3
shows the procedure used for collecting the BAL and BW fluids.
Figure 3. Procedure for collecting bronchoalveolar lavage (BAL) and bronchial wash (BW) fluids using a
flexible video bronchoscope. Photo by Ester Roos-Engstrand, 2006.
11
Chromatography techniques coupled to mass spectrometry
Chromatography techniques coupled to MS are powerful approaches whereby compounds are first
separated based on their chemical properties by chromatography and then eluted compounds can
be identified based on their mass spectra and quantified based on peak intensities. Depending on
the instrument setup, a wide and diverse range of compounds can be reliably detected and
analyzed.
Gas chromatography-time-of-flight mass spectrometry (GC-TOF-MS)
GC-TOF-MS is a common technique for generating metabolomics data. For GC-TOF-MS analysis,
pre-treatment of samples (i.e., extraction and derivatization) is required to isolate metabolites and,
if needed, transfer them into volatile derivatives [20]. Furthermore, protein removal from plasma
samples is needed prior to analysis. Vaporization of the samples is carried out before injection to
the GC column. A carrier gas (helium) is used as a mobile phase to transport the sample
components through the column containing the stationary phase. The molecules interact with the
stationary phase during their passage and are separated based on their strength of interaction with
the column, which is dependent on the chemico-physical properties of the molecule and type of
stationary phase. The time for the molecules to pass through the column will be different for each
compound, resulting in separation of the compounds. GC ionization is usually carried out using a
hard ionization technique, known as electron impact (EI), in which analytes are ionized and
fragmented through a bombarding electron stream, generating radical cation species
characteristic of the ionized compound. In TOF-MS, generated ions are accelerated in an electric
field and vacuum, reaching the detector at different times dependent on the m/z ratio of the ion.
[52]. To identify specific molecules, e.g., metabolites in plasma, mass spectra as well as
information about the compound’s retention index are needed [53].
GC-TOF-MS analysis utilizes three dimensional data, including the intensity, spectral, and
chromatographic dimensions. A multivariate deconvolution method is applied to translate the
three dimensions into a two dimensional data table, i.e., integrated area under the resolved
chromatographic profile. In the work presented in this thesis, sample files from GC-TOF-MS
analysis were exported to MATLAB 8.1 (R20013a) (Mathworks, Natick, MA, USA) in NetCDF
format. Matlab-based scripts were successfully customized to accomplish alignment of
12
chromatograms, peak detection and identification based on maximum covariance between the
samples’ total ion current chromatograms, retention indices and full MS spectra from the in-house
mass spectra library established by the Umeå Plant Science Centre (UPSC) and Swedish
Metabolomics Centre (SMC) in Umeå, Sweden. The advantage of this method is that it allows for
the possibility of deconvoluting many samples in a short time, as well as subsequent identification
of important metabolites. Data were normalized in Excel by dividing analyte peak areas by the
peak area of the internal standard [30, 54].
Liquid chromatography-quadrupole time-of-flight mass spectrometry (LC-QTOF-MS)
In LC, analytes are injected into a column and then separated due to different interactions with
the stationary and mobile phases [55]. Eluted analytes are ionized by an electrospray ionization
(ESI) technique in an ionization source. It should be noted that ESI is one of many ionization
techniques that have been used in LC-MS studies [56]. ESI is a soft ionization technique, which
means that, contrary to EI ionization used in GC, analytes are not fragmented in the ion source.
Identification of metabolites is achieved by accurate mass measurement of their ionized molecular
ions and their subsequent fragmentation to obtain fragment ion patterns characteristic for each
molecule.
LC-QTOF-MS has been used for characterization, identification and quantification of small
molecules in complex biological samples and offers several advantages, such as higher sensitivity
for many compounds and improved resolution compared with other detection methods [38]. LC
is often coupled to QTOF-MS to deliver a wide dynamic range for non-targeted analysis, and ESI
is the most commonly used ionization method in LC-MS analysis [57]. In addition, tandem mass
spectrometry (MS/MS) offers improved selectivity of detection and possibility for identification of
unknown metabolites. Prior to LC-MS analysis, pre-treatment of samples (i.e., protein
precipitation in plasma samples) is needed. In specific applications, the sample complexity can be
reduced by customizing the extraction or matrix removal steps to enhance metabolite detection,
increasing the signal-to-noise ratio in the LC-MS data, as well as decreasing ion-suppression
effects.
13
Liquid chromatography-tandem mass spectrometry (LC -MS/MS)
Sample preparation for targeted and quantitative LC-MS/MS usually begins with solid phase
extraction (SPE), followed by evaporation of eluates and reconstitution with a solvent [41].
For quantitative analysis of small molecules and metabolites by LC-MS/MS, triple quadrupole
(QqQ) mass spectrometry (MS/MS) in MRM mode is usually used owing to its good sensitivity,
reproducibility and broad dynamic range [27, 45]. In QqQ, quadrupoles are arranged in series.
The first quadrupole selects the parent ion of interest to be further fragmented in a second
quadrupole that works as a collision cell. Selected product ions are then analyzed in the third
quadrupole. For analysis of multiple metabolites, this process is cyclically repeated for each
compound. Therefore, specificity and signal-to-noise ratios are increased.
Nuclear magnetic resonance (NMR) spectroscopy
NMR spectroscopy is one of the versatile technologies that can be used for biological samples.
NMR has been significantly advanced in metabolomics through developments in sample
preparation, spectral processing and multivariate analysis. In this thesis work, the focus was on
the application of 1H NMR spectroscopy to profile metabolite levels in human samples, i.e., plasma
and lung lavage [49].
Because both low and high molecular weight components may be present in some biofluids, e.g.,
plasma, 1H NMR spectra may have a wide range of signal line widths. Larger molecules, e.g.,
proteins and lipoproteins, contribute to broad bands, which are often overlaid by sharp peaks due
to smaller molecules. Therefore, microcentrifuge filters were used to remove proteins and
insoluble impurities in the plasma samples. A deuterated sodium salt of 3-trimethylsilylpropionic
acid (TSP) was used as a reference compound in the studies of Paper I and II. Water in all
biofluids has to be removed by appropriate standard NMR solvent suppression methods in order
to eliminate its large interference in the signal. Various techniques can be applied for monitoring
as well as quantification of different groups of metabolites in a given sample. For processing 1H
NMR spectra, spectral editing techniques, such as phase and baseline correction, as well as
interrogation of spectral databases followed by a data reduction step are crucial. NMR data
processing might involve normalization by the sum of the integrals of each spectrum. Each spectral
14
integral is helpful as a metabolic descriptor for determining similarities or differences in the
acquired data. Biomarkers can be identified in plasma and lung lavage fluid based on their
chemical shifts, signal multiplicities and effect of adding authentic material.
15
Data analysis
Metabolomics studies generate complex multivariate datasets. Therefore, in addition to univariate
analysis, chemometric and bioinformatic tools are required for data analysis, visualization and
interpretation [58]. Hence, the statistical analysis may consist of univariate analysis [59] and
several multivariate data analysis (MVA) methods, such as principal components analysis (PCA)
[60], partial least squares-discriminant analysis (PLS-DA) and orthogonal projections to latent
structures (OPLS) [61]. Chemometrics, is an interdisciplinary science incorporating tools from
chemistry, statistics and mathematics fields, that has been acknowledged and established in the
field of metabolomics [32, 62].
Univariate analysis (UVA)
A variety of statistical tests are used to show whether metabolites are at significantly different
levels between different studied groups. Student’s t-test is a parametric test for normally
distributed data. However, for non-normally distributed datasets, as typical for clinical
metabolomics studies, the Wilcoxon’s test is used [57]. In t-tests, each metabolite is investigated
separately to determine whether the two groups have significantly different mean values. The null
hypothesis for the test is:
H0: μgroup1 = μgroup2
If the test p-value is smaller than a cutoff value, usually 0.05, the null hypothesis is rejected. In
contrast, if the p-value is larger than the cutoff value, the null hypothesis is approved, meaning
that the mean values of the two groups are not significantly different and the investigated
metabolite is not able to differentiate them. For an individual metabolite t-test, a 0.05 cutoff value
is normally used. However, for evaluating a set of metabolites, a smaller cutoff value is
recommended because of the increased risk of false positive findings when multiple tests are
performed. Multiple comparison correction procedures can be used to control for the t-test errors.
Simple multiple comparison procedures, such as Bonferroni correction, do not work well for
datasets with a very large number of metabolites; the set cutoff value is too close to zero as the
number of tests increases and becomes a too conservative approach to be of practical relevance
(with higher risk of reporting false negative results) [63]. A simple 0.05 cutoff can be used if
16
metabolites with small p-values are further evaluated by building a discrimination model and t-
tests alone do not give conclusive results [64].
Multivariate analysis (MVA)
Multivariate analysis (MVA) methods are crucial in metabolomics studies because all metabolites
are considered at the same time, allowing trends to be detected between both samples and
metabolites, as well as within samples and metabolites. MVA is performed to investigate
metabolite patterns and their ability for classifying the dataset [57].
Principal component analysis (PCA)
Principal component analysis (PCA) [60] is an unsupervised projection method commonly used
in chemometrics to reduce multidimensional data complexity in order to visualize and interpret
relations between samples and between studied variables (e.g. levels of metabolites). This method
summarizes the variation within a dataset by a smaller number of variables, called principal
components (PCs). PCs are linearly weighted combinations of the original variables calculated in
such a way that each PC consecutively models the maximum variation in the data, and at the same
time is orthogonal to the other PCs. This summary into PCs results in a bi-linear decomposition,
represented by two matrices, known as the scores (T) and loadings (P); Equation (1). Scores, linear
combinations of the original variables (X), may be considered as new variables, whereas loadings
describe how the old variables are linearly combined to form the new variables. Each point in a
score plot and loading plot represents a single sample and a single variable (metabolite),
respectively [9, 65]. Figure 4 shows examples of PCA score and loading plots.
(1) 𝑿 = 𝑻𝑷′ + 𝑬
Equation (1) where P' is transposed P, and E represents the residuals, i.e., unexplained variation
in the data.
17
In this thesis work, PCA was used to screen the data and detect clustered samples for further data
analysis.
Figure 4. Principal component analysis (top) score plot and (bottom) loading plot for GC-TOF-MS
(Paper I) (R2X = 0.80, Q2 (cum) = 0.62) t[1] is the first component and explains the largest variation,
t[2] is independent of t[1] and explains the second largest variation, which is orthogonal to t[1]. p[1] and
p[2] display the loadings of the first and second components, respectively.
18
Orthogonal projections to latent structures-discriminant analysis (OPLS-DA)
Information about samples in the dataset (e.g. class or other parameter of interest) can be used as
a response variable (Y) in a supervised method, such as PLS, OPLS [66], PLS-DA, and OPLS-DA
[67]. In a two-class situation, Y is a binary dummy vector (1/0) that contains information about
the pre-defined sample class ID. In OPLS-DA, the Y is used to divide systematic variation in X into
two parts, related and unrelated (orthogonal) to Y, which simplifies data interpretation. The
mathematical relationship is described in Equation (2).
(2) 𝑿 = 𝑻𝒑𝑷𝒑′ + 𝑻𝒐𝑷𝒐′ + 𝑬
Equation (2) where Tp is the predictive score matrix for X, Pp is the predictive loading matrix for
X, To is the corresponding Y-orthogonal score matrix, Po is the corresponding Y-orthogonal
loading matrix, Pp' is transposed Pp, Po' is transposed Po, and E is the predictive matrix for X.
OPLS-DA is utilized to map the variable patterns in the X matrix that discriminate the pre-defined
classes. To build a multivariate model for discriminating the sample classes, correlation between
the metabolite data matrix (X) and (Y) is undertaken. Examples of OPLS-DA score and loading
plots from Paper I are illustrated in Figure 5.
Statistical validation of multivariate models
To build a reliable multivariate model, statistical validation step is needed. Two types of
validations are commonly performed: external and internal. In external validation, new data are
collected/measured and used, and for internal validation the data are either divided into two sets:
a training set for modeling and a test/validation set or a permutation technique is used. The
purpose of validation is to assess the predictive ability of the model. Preferably, both types of
methods should be used in parallel to confirm the reliability of the model.
19
Figure 5. Orthogonal projections to latent structures-discriminant analysis (OPLS-DA) score plot (top) and
loading plot (bottom) obtained from GC-TOF-MS spectra showing differences between usual and modified
diets (R2X = 0.37, Q2 (cum) = 0.72, CV-ANOVA p-value = 0.00004); t[1] shows the direction of class
separation; to[1] expresses within-class variability; p[1] displays the loadings of the predictive component;
po[1] displays the loadings of the orthogonal component.
20
Cross-validation techniques include leave-n-out and K-fold methods [68]. In K-fold cross-
validation, the dataset is partitioned into k sized subsets and then iteratively k−1 subsets are
combined as a training set, with the remaining subset functioning as a test set. In leave-n-out, the
data are divided into N choose-n subsets and each subset serves as a validation set in each iteration.
If the sample size is n, in leave-one-out, n−1 samples are used as a training set to fit a classification
model and the remaining sample is used as test data. Therefore, every sample functions as a test
set just once. The model built on n-1 samples has the same accuracy as a model built on all (n)
samples. The estimated error rate is based on the misclassified test data. Leave-one-out techniques
are computationally desirable because they involve fitting the classification model n times [59].
This process is repeated until all samples have been left out and predicted once. The prediction
error sum of squares (PRESS) is defined as the sum of the square of the differences between the
observed and predicted value for each sample; Equation (3).
(3) 𝑷𝑹𝑬𝑺𝑺 = ∑ (𝒚𝒊 − 𝒚�̂�)𝟐𝒏
𝒊=𝟏
Equation (3) where 𝑦 is the true value for each sample, 𝑦̂ is the predicted value, 𝑖 is the number
of samples and n is the total number of samples.
In this thesis, CV-score plots are shown based on CV values indicating the predicted values.
Receiver operating characteristic (ROC) curve
Receiver operating characteristic (ROC) curve analysis is a validation tool that does not assume a
normal distribution of the studied variable and measures a variable’s predictive accuracy by
showing the relation between the true positive rate (sensitivity) and true negative rate (specificity)
[59, 69, 70]. Sensitivity is defined as the fraction of positive observations correctly classified by
the model into the positive class. Specificity is defined as the fraction of negative observations
correctly assigned by the model to the negative class. The area under the curve (AUC) of a plot of
sensitivity vs. specificity indicates the diagnostic value of each analyzed feature. The AUC can be
used as a criterion for assessing the success of the classification model. The closer the area under
the ROC curve (AUC) to 1 (maximum value), the more successful the classification model is; a
21
value of 0.5 indicates no diagnostic value. In addition, the ability of the metabolite to differentiate
between two groups of samples can be investigated from the shape of the ROC curve. The most
desirable curve has a sharp increase in true positive rate and slight increase in false positive rate.
22
Results and discussion
In this chapter, the main findings from the papers included in the thesis are summarized and
discussed in chronological order (Papers I - IV). The studies were designed to investigate the
responsiveness of the metabolome following external challenges in the form of meal intake and air
pollution exposure. As described above, the human body may respond to any stimuli or change in
the environment and metabolomics platforms are capable of investigating the net result of the
response by detecting alteration of metabolic profiles.
A metabolomics pilot study of the postprandial phase (Paper I)
Nutrition plays a vital role in human life and is obviously closely linked to metabolism. Thus,
nutritional metabolomics has been extensively investigated for its application to therapeutics,
medical foods and dietary supplements and as a confounding factor in studies of other external
stimuli. Studies have occasionally focused on metabolite profiling and fingerprinting to track
changes during the postprandial phase, directly after a meal intake [17, 71]. Postprandial
metabolomics has been shown to be useful, e.g., for metabolite quantification and identification
in plasma in response to a challenge meal [30], and assessment of individual metabolic
responsiveness to a lipid challenge [72], altered metabolic pathways in diabetes [16] and
postprandial insulin demand in postmenopausal women [71]. Potentially, postprandial
metabolomics can provide a comprehensive health assessment and allow optimization to different
dietary challenges, thus helping in the prevention of diseases and disorders in humans.
Objective
In this study, we aimed to detect a wide range of analytes in the postprandial human plasma
metabolome after usual and modified diet regimens, and investigate the overlapping and unique
extracted features from each of the investigated non-targeted metabolomics assay (LC-TOF-MS,
GC-TOF-MS and NMR), as well as the postprandial response (Figure 6).
23
Figure 6. Graphical abstract illustrating the non-targeted metabolomics study steps in Paper I.
Main findings and discussion of results in Paper I
Among the detected compounds, 54 GC-TOF-MS detected metabolites and 36 NMR detected
metabolites were identified, with 19 common to both assays. As explained in the methods section,
identification of LC-TOF-MS features is challenging and we were not able to identify the unknown
compounds detected by LC-TOF-MS. Therefore, we performed data analysis using the
unannotated LC-TOF-MS data.
Metabolome analysis during the postprandial stage by NMR and GC-TOF-MS indicated that
independent of background diet, samples were clustered into two groups; samples at 0.5 and 1 h
were separated from the baseline and samples at 3 h. However, this trend was not obvious in the
corresponding LC-TOF-MS data. To further investigate the trend, we constructed OPLS-DA
models (by defining samples at 0.5 and 1 h as one class, and the baseline and samples at 3 h as a
second class) and explored the metabolite profiles. To validate the model, we applied a cross-
validation test, which indicated that the metabolite profiles were significantly different between
24
fasting (0 + 3 h), and response (0.5 h + 1 h) to both background diets and for both the NMR and
GC-TOF-MS assays. The 19 common metabolites showed more significant changes in the NMR
OPLS-DA model compared to the GC-TOF-MS, which may at least be partly explained by the
higher within group variation in the GC-TOF-MS data. Increased levels of amino acids, organic
acids and sugars at the response stage, as well as decreased levels of fatty acids, acetoacetate and
3-hydroxybutyric acid, were observed in the OPLS-DA models independent of background diet.
Higher levels of amino acids and sugars could be related to absorption and metabolism processes
following meal (banana) intake. The meal contained no lipids, which may have caused the
decreased levels of fatty acids. In addition, this decrease could be linked to insulin secretion in
adipose tissue, which can inhibit lipolysis. In the LC-TOF-MS analysis, only one significant OPLS-
DA model was obtained according to CV-ANOVA (for the modified diet in negative ionization
mode; 21 compounds were upregulated and seven compounds downregulated).
Metabolites showing common or unique behavior after the usual or modified diet during the
postprandial stage were explored by plotting the correlation loading vectors (p(corr)) from the
OPLS-DA models on shared and unique structure (SUS)-plots. In the GC-TOF-MS SUS plot,
cysteine upregulation during the postprandial stage was uniquely observed after the modified
background diet, which could be related to the high protein foods consumed, such as eggs, meat
and dairy products.
OPLS-DA models classifying the background diets were constructed for all three analytical
techniques. The GC-TOF-MS diet-dependent OPLS-DA plots showed higher linoleic acid, oleic
acid, glycine and 3-hydroxybutyric acid levels following the usual diet, whereas the modified diet
samples displayed higher cholesterol, tyrosine, phenylalanine and taurine levels. The univariate
analysis results were in agreement with those obtained from multivariate analysis, with similar
trends observed for the majority of the studied compounds.
Air pollution exposure studies (Paper II, III and IV)
Air pollution originating from fossil fuel combustion contributes considerably to the worldwide
increasing prevalence of respiratory and cardiovascular diseases [73-78]. Owing to the chemical
and toxicological properties of particulate matter (PM) generated from combustion, it can cause
substantial damage to the lungs if it penetrates deep into the respiratory tract during inhalation
25
[79]. Because of limited oil reserves and concerns about the adverse health effects and
environmental damage of fossil fuel combustion, there is an increasing trend toward replacing
fossil fuels with biofuels derived from renewable sources [80]. RME is considered renewable and
carbon dioxide neutral. Therefore, it is regarded as ecologically less damaging than petrodiesel.
However, there is currently a lack of studies investigating how biodiesel replacement can affect
human health [81].
The filtered air and biodiesel exhaust exposure study were performed at the Department of Public
Health and Clinical Medicine, Division of Respiratory Medicine and Allergy, Norrlands University
Hospital. The exposure chamber at SMP Svensk Maskinprovning AB, Umeå, Sweden was
connected to a Volvo engine (Volvo TD40 GJE, 4.0 L, 4 cylinders), which was designed to generate
biodiesel exhaust emissions from RME100 fuel (Figure 2). Many studies have been conducted
using this set-up to investigate the adverse health effects of exhaust exposure on cardiorespiratory
status and inflammatory processes [82-84]. In this thesis, we complement these studies applying
different analytical platforms (based on targeted and non-targeted metabolomics assays) to
analyze healthy human BW, BAL and plasma samples (Paper II-IV) gathered after bio-diesel
exhaust and filtered air exposure. The experimental pipeline applied in Paper II-IV is presented
in Figure 7.
26
Figure 7. Graphical abstract illustrating the air pollution exposure studies (Paper II, III and IV).
Paper II
Non-targeted metabolomics of BAL fluid has been successfully used to study human lung injuries
and pulmonary diseases, such as cystic fibrosis, asthma, and respiratory distress syndrome, as well
as for animal lung inflammation assessment [21, 24, 25, 85-87]. In such studies, key metabolic
markers were detected and identified to develop an understanding of the studied pathologies. It
was reported that LC-MS was able to detect and quantify the maximum number of 23 unique
metabolites in BAL samples [21]. Nevertheless, no study has yet applied GC-MS, LC-MS and NMR
together for analysis of aliquots of the same sample, which might be the reason for the limited
number of metabolites identified in previous studies. Metabolite profiling of lung lavage fluids
using multi-platforms, i.e., GC-TOF-MS, LC-TOF-MS and NMR, can provide more extensive
coverage of the lung metabolome. However, such studies are challenging because of the extremely
low metabolite levels and high salt content in this type of sample [26, 50].
27
Objective
In Paper II, we hypothesized that multi-platform metabolite profiling (GC-TOF-MS, LC-TOF-MS
and NMR) would enable wider detection and identification of metabolites in lung lavage fluids
and demonstrate biodiesel exhaust exposure effects on the human lung lavage metabolome. For
this purpose, we developed an extraction protocol to detect and identify a wide range of
metabolites in BW and BAL fluids collected from subjects exposed to biodiesel exhaust and filtered
air.
Main findings and discussion of results in Paper II
The developed and optimized extraction protocol allowed us to identify 53 metabolites in the GC-
TOF-MS assay and 14 in the LC-TOF-MS assay, with a signal-to-noise ratio above three (relative
to blanks). In parallel, Chenomx software was used for the NMR assay to identify 23 metabolite
omitting interferences related to the anesthetic process in the bronchoscopy. Interestingly, the
three assays yielded a wide range of metabolites, e.g., fatty acids, sugars, amino acids and small
organic acids, with no single molecule in common between all three platforms. Seven metabolites
were in common between the NMR and GC-TOF-MS assays (glucose, glutamate, glycine, lactic
acid, pyruvic acid, taurine and valine), and one between the LC-TOF-MS and NMR assays
(creatinine).
Pathway analysis was carried out using the MetaboAnalyst 3.0 software to investigate the pathway
coverage for the identified metabolites. A few pathways were significant according to the
calculated p-values, with most of the pathways connected to amino acid and fatty acid metabolism.
OPLS-DA modeling for all three platforms indicated that the metabolite profiles differed between
the BW and BAL samples after exposure to filtered air. By focusing only on samples after filtered
air exposure, variations between the lung compartments related to biodiesel exhaust exposure
were omitted.
In addition, univariate analysis revealed that the metabolite concentrations in the BW and BAL
samples were at different levels after biodiesel exhaust exposure. For instance, in the GC-TOF-MS
assay, pentadecanoic acid levels were significantly higher in the BW samples, whereas
ethanolamine, inosine and nonanoic acid were significantly higher in the BAL samples. Increased
levels of compounds such as ethanolamine, phosphate, glycerol-3-phosphate and unsaturated
fatty acids and decreased levels of some fatty acid glycerol esters after biodiesel exposure in the
28
BAL samples indicated that biodiesel exposure could have an effect on the lipid degradation of the
cell membrane. Further, the LC-TOF-MS assay analysis indicated that following biodiesel exhaust
exposure, chenodeoxycholic acid glycine conjugate levels were decreased in BW fluid, whereas
niacinamide was increased in BAL fluid. Increased levels of niacinamide, which is a component of
the coenzyme NAD, may be connected to changes in the redox environment in airway cells. In the
NMR assay analysis, only changes in lactic acid concentrations were significant and it had lower
levels after biodiesel exhaust exposure in BW samples. This trend was also observed in the GC-
TOF-MS analysis, but the results were not significant. Under a shortage of oxygen, lactic acid may
be produced in aerobic organisms to meet the need for its consumption. Some other processes,
e.g., glycolysis and/or gluconeogenesis, may also affect levels of lactic acid.
Furthermore, to monitor changes in the metabolite profiles induced by biodiesel exhaust exposure,
we used OPLS-DA modeling. The results indicated that the only assay capable of differentiating
between the filtered air exposure and biodiesel exhaust exposure was GC-TOF-MS. The OPLS-DA
models for the BW and BAL samples were built separately and both models were significant based
on cross-validation; the strongest model (CV-ANOVA p-value = 0.0009) was obtained for BAL
samples. Independent of the sample type (BW or BAL), most subjects, with only a few exceptions,
showed the same direction of the metabolic response after biodiesel exhaust exposure. To further
confirm the OPLS-DA models, we plotted ROC curves. The corresponding AUC (0.75) verified that
the OPLS-DA models were able to accurately classify the groups.
Paper III
In parallel to the non-targeted study of BW and BAL samples (as discussed in Paper II), for the
study in Paper III, we used a targeted LC-MS/MS approach to detect a wide range of oxylipins,
endocannabinoids, N-acylethanolamines and related compounds in lung lavage fluid after
exposure to filtered air and biodiesel exhaust exposure.
Bioactive lipid mediators play a significant role in pulmonary inflammation, i.e., in initiation,
propagation and resolution of inflammation [88, 89]. Oxidation of polyunsaturated fatty acids
(PUFA) leads to the biosynthesis of a group of compounds called oxylipins. Eicosanoids, such as
prostaglandin E2 (PGE2), PGD2, PGF2α and 15-hydroxyeicosatetraenoic acid (15-HETE), are
oxylipins produced by the oxidation of ω6 arachidonic acid (20:4n6). Oxidation of linoleic acid
29
(LA) leads to production of 9-hydroxy-octadecadienoic acid (9-HODE), 13-HODE, 9,10-
dihydroxy-12Z-octadecenoic acid (9,10-DiHOME) and 12,13-DiHOME, etc. Three possible
pathways for this oxidation are shown in Figure 8 based on cyclooxygenase (COX), lipoxygenase
(LOX) and cytochrome (CYP) P450 as oxylipin sources [89, 90]. Endocannabinoids are lipid
mediators that comprise ligands to the cannabinoid (CB) receptors, e.g., anandamide (AEA) and
2-arachidonoyl glycerol (2-AG) [91, 92]. N-acylethanolamines and glycerol fatty acid derivatives
are examples of endocannabinoid-related lipids which influence the activity of the CB1 and CB2
receptors [93].
Figure 8. Scheme showing oxylipins and their metabolic pathways.
Objective
In the study described in Paper III, we hypothesized that the lipid mediator profile (consisting
of eicosanoids, endocannabinoids and related lipids) in the human lung changes in response to
biodiesel exhaust exposure. To test this hypothesis, we performed a targeted analysis using LC-
MS/MS to profile the levels of bioactive lipid mediators in both BW and BAL fluids.
30
Main findings and discussion of results in Paper III
Both univariate and multivariate analyses indicated that there was a significant difference between
the BW and BAL samples, reflecting different locations in the lung (proximal or bronchial region
vs. distal or alveolar region). Investigating the exhaust exposure effect, it was found that nine
metabolites in the human lung lavage samples were altered following biodiesel exhaust exposure,
six in BAL samples and three in BW samples. Of these, three compounds (PGE2, 12,13-DiHOME
and 13-HODE) in the BAL samples and no compounds in the BW samples passed the Bonferroni
correction (p-value ≤ 0.0019), which however is a conservative approach (discussed above).
Finally, the exposure-dependent profile of the detected compounds was investigated by OPLS-DA
modeling of the data for the biodiesel exhaust samples vs. filtered air exposure samples. The
peripheral airways were most responsive to exhaust exposure, showing increased concentrations
of arachidonic acid (PGE2) and linoleic acid (12, 13-DiHOME and 13-HODE) derived oxylipins.
Paper IV
Objective
In the work presented in Paper IV, we extended the lipid mediator profiling by LC-MS/MS to
plasma samples from the same individuals at multiple time points (pre-exposure and 2, 6 and 24
h post-exposure) to follow temporal trends in plasma lipid mediators. We hypothesized that the
local effect on bioactive lipid profiles would be found at the systemic level, in the blood stream.
Main findings and discussion of results in Paper IV
Overall, 35/38 oxylipins (in addition to 13 endocanabinoids) were quantified, of which 20 were
present in all plasma samples. We used the quantified data for further analysis. Independent of
the exposure, compounds derived from the LOX pathway showed the highest concentrations. Five
metabolites from the arachidonic acid LOX pathway (5-HETE, 11-HETE, 12-HETE, 15-HETE and
5-oxo-ETE) as well as one from the linoleic acid LOX pathway (9-HODE) and one from the CYP
31
pathway (9,10-DiHOME) showed decreased levels at 24 h post-exposure, except 5-oxo-ETE,
which was decreased at 6 h. Additionally, 5-HETE was also decreased at 6 h post-exposure.
Furthermore, the arachidonic acid COX metabolite PGF2α was elevated at 6 h and the N-
acylethanolamine DEA was decreased at 2 h. The majority of the responsive fatty acid metabolites
were monohydroxy fatty acids, linked to cardiovascular outcomes [94]. Thus, we concluded that it
is possible to detect alterations in circulating bioactive lipids in response to biodiesel exhaust
exposure using LC-MS/MS by focusing on compounds with known or suspected effects on
cardiovascular health.
32
Conclusion and future perspectives
Using three different non-targeted analytical methods (GC-TOF-MS, NMR and LC-TOF-MS), a
wide range of metabolites were detected and identified in the studies of Paper I (in plasma) and
Paper II (in lung lavage fluid). Paper I presents a pilot study where all three assays showed a
stable postprandial response over time that was largely independent of the background diet. Both
multivariate and univariate analyses indicated that 0.5 h after food intake, levels of amino acids
were increased, whereas levels of fatty acids were decreased, but both types of metabolites
returned to baseline levels 3 h after the meal intake. Comparing the platforms used in the Paper
I work, GC-TOF-MS seemed to be the most appropriate owing to the wide range of detected
metabolites, well established and robust deconvolution approaches and straightforward
compound identification. For robust detection of more polar compounds, NMR may be the best
choice. Multivariate analysis yielded a variety of information about the studied system, including
information about trends in samples and variables and connections between them. Univariate
analysis was used as a complementary statistical method to the multivariate analysis. The study
discussed in Paper II showed for the first time that by applying complementary analytical
techniques, a large number of metabolites (82) could be detected and identified in BW and BAL
fluid samples after air and biodiesel exhaust exposure, providing important and novel information
on the metabolome of the respiratory tract. We observed that there were significant differences in
the levels of specific metabolites between BW and BAL samples using non-targeted metabolomics
assays, in line with previous targeted metabolomics protocols. Furthermore, metabolite profiles
of BW and BAL samples, particularly the BAL samples, changed after exposure to biodiesel
exhaust compared to those following exposure to filtered air. Notably, this responsiveness was
best monitored by GC-TOF-MS, while NMR and LC-TOF-MS provided better data for assessment
of differences between the BW and BAL metabolite profiles.
Since metabolite levels in BAL and BW fluids reflect the molecular status of the lung epithelium,
the alterations observed after exposure to air pollution derived from biodiesel exhaust may be
related to changes in the human lung condition. The techniques used for this study may serve as
a starting point for future metabolomics studies on the respiratory tract system and further
evaluation of air pollution effects on human health.
The work presented in Paper III represents the first investigation of bioactive lipid responses of
human lung lavage fluid following exposure to RME biodiesel exhaust. By applying a MS based
33
approach, we were able to detect and quantify multiple oxylipins, endocannabinoids and
endocannabinoid-related lipids (36), and found compartment-dependent responsiveness of
individual species after biodiesel exhaust exposure.
In Paper IV, among the 35 oxylipins and 13 endocannabinoids and related lipids detected in
plasma by LC-MS/MS, nine metabolites were affected, mainly at 24 h post-exposure. Thus, we
concluded that it is possible to detect alterations in circulating bioactive lipids in response to
biodiesel exhaust exposure using LC-MS/MS by monitoring compounds with known or suspected
effects on cardiovascular health. However, we were not able to detect responsiveness among the
same metabolites as those detected in BW and BAL from the same study.
Following the presented papers in this thesis, we performed non-targeted metabolomics profiling
of the plasma samples, in addition to the bioactive lipid profiling, to investigate the bloodstream
response to biodiesel exhaust exposure vs. filtered air exposure of the more abundant metabolites.
This is currently work in progress.
Future perspectives
It would be very interesting to monitor metabolite changes in plasma and lung lavage fluids after
diesel exhaust or other exposures to investigate whether the local effects are more evident than
systemic effects, and extend the knowledge to air pollutants in general. To that end, targeted and
non-targeted metabolomics investigations of plasma samples after exposure to biodiesel vs. diesel
exhaust exposure is planned, with the aim of exploring in more detail the local vs. systemic
response after biodiesel exhaust and diesel exhaust exposure. A follow-up validation study would
offer an excellent opportunity for further improvement and development of the applied
methodologies, as well as confirmation of the biological findings. Extension of metabolomics
strategies and analysis methods to detect and identify more metabolites and at lower levels will
help to elucidate the mechanisms behind the effects of interventions and environmental influences
on the human body. Specifically, exposure-related effects on cardiorespiratory or other similar
diseases could facilitate the identification of disease specific biomarkers, drug development and
illness treatment.
34
Acknowledgment
First of all, I would like to thank my supervisor, Malin L. Nording, for giving me the opportunity
to study as a PhD student at Umeå University and supporting me to experience new things. Special
thanks go to my co-supervisor, Izabella Suroweic, who helped and strongly supported me
throughout these years. I would also like to acknowledge my co-supervisors, Johan Trygg and
Annelie Behndig, for their helpful guidance and suggestions during my projects. Thanks too to Rui
for helping me start the data analysis and Kate for kind support.
Thanks to Sandra for her fantastic help and companionship during my projects. I am also grateful
to Junfang for her valuable comments and help. Huge thanks go to my co-authors at Norrland
University Hospital Umeå and the University of California, Davis, for their contribution to
publications. Acknowledgment also goes to Hans, Jonas, Krister and other colleagues at the
Swedish Metabolomics Centre (SMC) for their great help with the laboratory and data processing
works. I would also like to thank Patrik Andersson, Jerker Fick, and Richard Lindberg for giving
me the opportunity to be involved in lab work prior to my PhD.
I sincerely thank Yaser, Mar, Shery, Qiuju, Sandra, Joao, Junfang, Jana, Mirva, Matyas, Lan,
Marcus, Jin and other friends at MKL for their companionship and happy times we had together
throughout these years, making me feel awesome. Thanks to our Iranian friends here in Umeå for
all the great times we shared.
Finally, sincere thanks go to my family. Thanks to my parents for their great support and love
throughout my whole life. Thanks to my brother, Reza, and sister, Marisa, for the wonderful
moments and their unlimited kindness. Thanks too to my brother-in-law, Ashkan, for his sincerity.
A very special thanks to my dearest Mehdi for all the happy and sad moments we have experienced
together and all the support and amiability that made me feel lucky throughout our shared lives.
Lastly, thanks to Mehdi’s family for their tenderness.
35
References
1. Wishart, D.S., Emerging applications of metabolomics in drug discovery and precision medicine. Nat Rev Drug Discov, 2016. advance online publication.
2. Tweeddale, H., L. Notley-McRobb, and T. Ferenci, Effect of Slow Growth on Metabolism of Escherichia coli, as Revealed by Global Metabolite Pool (“Metabolome”) Analysis. Journal of Bacteriology, 1998. 180(19): p. 5109-5116.
3. Nicholson, J.K., J.C. Lindon, and E. Holmes, 'Metabonomics': understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica, 1999. 29(11): p. 1181-1189.
4. Dalgliesh, C.E., et al., A gas–liquid-chromatographic procedure for separating a wide range of metabolites occuring in urine or tissue extracts. Biochemical Journal, 1966. 101(3): p. 792-810.
5. Mamer, O. and J. Crawhall, The identification of urinary acids by coupled gas chromatography-mass spectrometry. Clinica Chimica Acta, 1971. 32(2): p. 171-184.
6. Horning, E.C. and M.G. Horning, Metabolic Profiles: Gas-Phase Methods for Analysis of Metabolites. Clinical Chemistry, 1971. 17(8): p. 802-809.
7. Pauling, L., et al., Quantitative Analysis of Urine Vapor and Breath by Gas-Liquid Partition Chromatography. Proceedings of the National Academy of Sciences of the United States of America, 1971. 68(10): p. 2374-2376.
8. Fiehn, O., et al., Metabolite profiling for plant functional genomics. Nat Biotech, 2000. 18(11): p. 1157-1161.
9. Lindon, J.C., J.K. Nicholson, and E. Holmes, The handbook of metabonomics and metabolomics. 2011: Elsevier.
10. Griffiths, W.J., et al., Targeted Metabolomics for Biomarker Discovery. Angewandte Chemie International Edition, 2010. 49(32): p. 5426-5445.
11. Harrigan, G.G. and R. Goodacre, Metabolic profiling: its role in biomarker discovery and gene function analysis. 2012: Springer Science & Business Media.
12. Yang, J., et al., Quantitative Profiling Method for Oxylipin Metabolome by Liquid Chromatography Electrospray Ionization Tandem Mass Spectrometry. Analytical Chemistry, 2009. 81(19): p. 8085-8093.
13. Gouveia-Figueira, S. and M.L. Nording, Development and Validation of a Sensitive UPLC-ESI-MS/MS Method for the Simultaneous Quantification of 15 Endocannabinoids and Related Compounds in Milk and Other Biofluids. Analytical Chemistry, 2014. 86(2): p. 1186-1195.
36
14. Wikoff, W.R., et al., Metabolomics analysis reveals large effects of gut microflora on mammalian blood metabolites. Proceedings of the National Academy of Sciences, 2009. 106(10): p. 3698-3703.
15. Underwood, B.R., et al., Huntington disease patients and transgenic mice have similar pro-catabolic serum metabolite profiles. Brain, 2006. 129(4): p. 877-886.
16. Dutta, T., et al., Concordance of Changes in Metabolic Pathways Based on Plasma Metabolomics and Skeletal Muscle Transcriptomics in Type 1 Diabetes. Diabetes, 2012. 61(5): p. 1004-1016.
17. Pellis, L., et al., Plasma metabolomics and proteomics profiling after a postprandial challenge reveal subtle diet effects on human metabolic status. Metabolomics, 2012. 8(2): p. 347-359.
18. Beckonert, O., et al., Metabolic profiling, metabolomic and metabonomic procedures for NMR spectroscopy of urine, plasma, serum and tissue extracts. Nat Protoc, 2007. 2(11): p. 2692-703.
19. Gupta, A., N. Bansal, and B. Houston, Metabolomics of urinary tract infection: a new uroscope in town. Expert Rev Mol Diagn, 2012. 12(4): p.361-369.
20. A, J., et al., Extraction and GC/MS Analysis of the Human Blood Plasma Metabolome. Analytical Chemistry, 2005. 77(24): p. 8086-8094.
21. Evans, C.R., et al., Untargeted LC–MS Metabolomics of Bronchoalveolar Lavage Fluid Differentiates Acute Respiratory Distress Syndrome from Health. Journal of Proteome Research, 2014. 13(2): p. 640-649.
22. Cribbs, S.K. and G.S. Martin, Biomarkers in acute lung injury: are we making progress? Crit Care Med, 2008. 36(8): p. 2457-2459
23. Rai, R.K., et al., Metabolic profiling in human lung injuries by high-resolution nuclear magnetic resonance spectroscopy of bronchoalveolar lavage fluid (BALF), in Metabolomics. 2013, 9(3): p. 667-676.24. Singh, C., et al., Mini-bronchoalveolar lavage fluid can be used for biomarker identification in patients with lung injury by employing 1H NMR spectroscopy. Critical Care, 2013. 17(2): p. 1-3.
25. Wolak, J.E., C.R. Esther, and T.M. O'Connell, Metabolomic analysis of bronchoalveolar lavage fluid from cystic fibrosis patients. Biomarkers, 2009. 14(1): p. 55-60.
26. Peng, J., et al., Metabolomic profiling of bronchoalveolar lavage fluids by isotope labeling liquid chromatography mass spectrometry: a promising approach to studying experimental asthma. Metabolomics, 2014. 10(6): p. 1305-1317.
27. Lu, W., B.D. Bennett, and J.D. Rabinowitz, Analytical strategies for LC-MS-based targeted metabolomics. Journal of chromatography. B, Analytical technologies in the biomedical and life sciences, 2008. 871(2): p. 236-242.
28. Zhang, A., et al., Modern analytical techniques in metabolomics analysis. Analyst, 2012. 137(2): p. 293-300.
37
29. Tredwell, G.D., et al., Metabolomic Characterization of Nipple Aspirate Fluid by 1H NMR Spectroscopy and GC-MS. Journal of Proteome Research, 2014. 13(2): p. 883-889.
30. Karimpour, M., et al., Postprandial metabolomics: A pilot mass spectrometry and NMR study of the human plasma metabolome in response to a challenge meal. Analytica Chimica Acta, 2016. 908: p. 121-131.
31. Dunn, W.B., N.J. Bailey, and H.E. Johnson, Measuring the metabolome: current analytical technologies. Analyst, 2005. 130(5): p. 606-25.
32. Shulaev, V., Metabolomics technology and bioinformatics. Briefings in Bioinformatics, 2006. 7(2): p. 128-139.
33. Chorell, E., et al., Impact of probiotic feeding during weaning on the serum lipid profile and plasma metabolome in infants. British Journal of Nutrition, 2013. 110(01): p. 116-126.
34. Koek, M.M., et al., Quantitative metabolomics based on gas chromatography mass spectrometry: status and perspectives. Metabolomics, 2011. 7(3): p. 307-328.
35. Wiklund, S., et al., Visualization of GC/TOF-MS-Based Metabolomics Data for Identification of Biochemically Interesting Compounds Using OPLS Class Models. Analytical Chemistry, 2008. 80(1): p. 115-122.
36. Liesenfeld, D.B., et al., Review of Mass Spectrometry–Based Metabolomics in Cancer Research. Cancer Epidemiology Biomarkers & Prevention, 2013. 22(12): p. 2182-2201.
37. Roberts, L.D., et al., Targeted Metabolomics. Current Protocols in Molecular Biology, 2012. CHAPTER: p. Unit30.2-Unit30.2.
38. Vinayavekhin, N. and A. Saghatelian, Untargeted Metabolomics, in Current Protocols in Molecular Biology. 2001, Chapter 30: p. Unit 30.1.1-24.
39. Beynon, J.H., et al., Design and Performance of a Mass-analyzed Ion Kinetic Energy (MIKE) Spectrometer. Analytical Chemistry, 1973. 45(12): p. 1023A-1031A.
40. Yost, R.A. and C.G. Enke, Selected ion fragmentation with a tandem quadrupole mass spectrometer. Journal of the American Chemical Society, 1978. 100(7): p. 2274-2275.
41. Gouveia-Figueira, S., et al., Profiling the Oxylipin and Endocannabinoid Metabolome by UPLC-ESI-MS/MS in Human Plasma to Monitor Postprandial Inflammation. PLoS ONE, 2015. 10(7): p. e0132042.
42. Fiehn, O., Metabolomics — the link between genotypes and phenotypes, in Functional Genomics, C. Town, Editor. 2002, Springer Netherlands: Dordrecht. p. 155-171.
43. Wishart, D.S., Metabolomics: applications to food science and nutrition research. Trends in Food Science & Technology, 2008. 19(9): p. 482-493.
44. Smilowitz, J.T., et al., Nutritional Lipidomics: Molecular Metabolism, Analytics, and Diagnostics. Molecular nutrition & food research, 2013. 57(8): p. 1319-1335.
45. Dettmer, K., P.A. Aronov, and B.D. Hammock, MASS SPECTROMETRY-BASED METABOLOMICS. Mass spectrometry reviews, 2007. 26(1): p. 51-78.
38
46. Lei, Z., D.V. Huhman, and L.W. Sumner, Mass Spectrometry Strategies in Metabolomics. Journal of Biological Chemistry, 2011. 286(29): p. 25435-25442.
47. Fiehn, O., Extending the breadth of metabolite profiling by gas chromatography coupled to mass spectrometry. Trends in analytical chemistry : TRAC, 2008. 27(3): p. 261-269.
48. Lutz, N.W., Jonathan V. Sweedler, and e. Ron A. Wevers, Methodologies for Metabolomics. 2013: Cambridge University Press.
49. Smolinska, A., et al., NMR and pattern recognition methods in metabolomics: From data acquisition to biomarker discovery: A review. Analytica Chimica Acta, 2012. 750: p. 82-97.
50. Walters, E.H., C. Ward, and X. Li, Bronchoalveolar lavage in asthma research. Respirology, 1996. 1(4): p. 233-45.
51. Larsson, N., et al., Lipid mediator profiles differ between lung compartments in asthmatic and healthy humans. European Respiratory Journal, 2014. 43(2): p. 453-463.
52. El-Aneed, A., A. Cohen, and J. Banoub, Mass spectrometry, review of the basics: electrospray, MALDI, and commonly used mass analyzers. Applied Spectroscopy Reviews, 2009. 44(3): p. 210-230.
53. Schauer, N., et al., GC-MS libraries for the rapid identification of metabolites in complex biological samples. FEBS Lett, 2005. 579(6): p. 1332-7.
54. Jonsson, P., et al., High-Throughput Data Analysis for Detecting and Identifying Differences between Samples in GC/MS-Based Metabolomic Analyses. Analytical Chemistry, 2005. 77(17): p. 5635-5642.
55. Zhou, B., et al., LC-MS-based metabolomics. Mol Biosyst, 2012. 8(2): p. 470-81.
56. de Hoffmann, E. and V. Stroobant, Mass Spectrometry: Principles and Applications. 2007: Wiley.
57. Raftery, D., Mass Spectrometry in Metabolomics: Methods and Protocols. 2014: Springer New York.
58. Madsen, R., T. Lundstedt, and J. Trygg, Chemometrics in metabolomics—a review in human disease diagnosis. Analytica Chimica Acta, 2010. 659(1): p. 23-33.
59. Xi, B., et al., Statistical analysis and modeling of mass spectrometry-based metabolomics data. Methods Mol Biol, 2014. 1198: p. 333-53.
60. Wold, S., K. Esbensen, and P. Geladi, Principal component analysis. Chemometrics and intelligent laboratory systems, 1987. 2(1-3): p. 37-52.
61. Bylesjö, M., et al., OPLS discriminant analysis: combining the strengths of PLS‐DA and SIMCA classification. Journal of Chemometrics, 2006. 20(8‐10): p. 341-351.
62. Trygg, J., E. Holmes, and T. Lundstedt, Chemometrics in Metabonomics. Journal of Proteome Research, 2007. 6(2): p. 469-479.
39
63. Bretz, F., T. Hothorn, and P. Westfall, Multiple Comparisons Using R. 2010: Taylor & Francis.
64. Bender, R. and S. Lange, Adjusting for multiple testing--when and how? J Clin Epidemiol, 2001. 54(4): p. 343-9.
65. Weimer, B.C. and C. Slupsky, Metabolomics in food and nutrition. 2013: Elsevier.
66. Trygg, J. and S. Wold, Orthogonal projections to latent structures (O-PLS). Journal of Chemometrics, 2002. 16(3): p. 119-128.
67. Bylesjö, M., et al., OPLS discriminant analysis: combining the strengths of PLS-DA and SIMCA classification. Journal of Chemometrics, 2006. 20(8-10): p. 341-351.
68. Worley, B. and R. Powers, Multivariate Analysis in Metabolomics. Curr Metabolomics, 2013. 1(1): p. 92-107.
69. Egan, J.P., Signal Detection Theory and ROC-analysis. 1975: Academic Press.
70. Brown, C.D. and H.T. Davis, Receiver operating characteristics curves and related decision measures: A tutorial. Chemometrics and Intelligent Laboratory Systems, 2006. 80(1): p. 24-38.
71. Moazzami, A.A., et al., Metabolomics Reveals Differences in Postprandial Responses to Breads and Fasting Metabolic Characteristics Associated with Postprandial Insulin Demand in Postmenopausal Women. The Journal of Nutrition, 2014. 144(6): p. 807-814.
72. Zivkovic, A., et al., Assessing individual metabolic responsiveness to a lipid challenge using a targeted metabolomic approach. Metabolomics, 2009. 5(2): p. 209-218.
73. Hoek, G., et al., Long-term air pollution exposure and cardio- respiratory mortality: a review. Environmental Health, 2013. 12(1): p. 1-16.
74. Atkinson, R.W., et al., Short-term exposure to traffic-related air pollution and daily mortality in London, UK. J Expos Sci Environ Epidemiol, 2016. 26(2): p. 125-132.
75. Beelen, R., et al., Long-term effects of traffic-related air pollution on mortality in a Dutch cohort (NLCS-AIR study). Environ Health Perspect, 2008. 116(2): p. 196–202.
76. Zheng, X.-y., et al., Association between Air Pollutants and Asthma Emergency Room Visits and Hospital Admissions in Time Series Studies: A Systematic Review and Meta-Analysis. PLoS ONE, 2015. 10(9): p. e0138146.
77. Thibodeau, L.A., et al., Air pollution and human health: a review and reanalysis. Environmental Health Perspectives, 1980. 34: p. 165-183.
78. Mazzarella, G., et al., Effects of diesel exhaust particles on human lung epithelial cells: An in vitro study. Respiratory Medicine. 101(6): p. 1155-1162.
79. Löndahl, J., et al., Measurement Techniques for Respiratory Tract Deposition of Airborne Nanoparticles: A Critical Review. Journal of Aerosol Medicine and Pulmonary Drug Delivery, 2014. 27(4): p. 229-254.
40
80. Robbins, M., Policy: Fuelling politics. Nature, 2011. 474(7352): p. S22-S24.
81. Bünger, J., et al., Potential hazards associated with combustion of bio-derived versus petroleum-derived diesel fuel. Critical Reviews in Toxicology, 2012. 42(9): p. 732-750.
82. Langrish, J.P., et al., Altered Nitric Oxide Bioavailability Contributes to Diesel Exhaust Inhalation-Induced Cardiovascular Dysfunction in Man. Journal of the American Heart Association, 2013. 2 2:e004309.
83. Behndig, A.F., et al., Effects of controlled diesel exhaust exposure on apoptosis and proliferation markers in bronchial epithelium - an in vivo bronchoscopy study on asthmatics, rhinitics and healthy subjects. BMC Pulm Med, 2015. 15: p. 99.
84. Jamshid, P., et al., Airway Inflammatory Response in Healthy Subjects Following Chamber Exposure to 100% RME Biodiesel, in C104. INDOOR AND OUTDOOR POLLUTION: EPIDEMIOLOGY AND MECHANISMS. 2015, American Thoracic Society. p. A5252-A5252.
85. Connett, G.J., Bronchoalveolar lavage. Paediatric Respiratory Reviews. 1(1): p. 52-56.
86. Bino, R.J., et al., Potential of metabolomics as a functional genomics tool. Trends in Plant Science. 9(9): p. 418-425.
87. Reynolds, Y.H., Use of Bronchoalveolar Lavage in Humans—Past Necessity and Future Imperative. Lung. 178(5): p. 271-293.
88. Serhan, C.N., Novel Lipid Mediators and Resolution Mechanisms in Acute Inflammation: To Resolve or Not? The American Journal of Pathology, 2010. 177(4): p. 1576-1591.
89. Harizi, H., J.-B. Corcuff, and N. Gualde, Arachidonic-acid-derived eicosanoids: roles in biology and immunopathology. Trends in Molecular Medicine. 14(10): p. 461-469.
90. Dennis, E.A. and P.C. Norris, Eicosanoid storm in infection and inflammation. Nat Rev Immunol, 2015. 15(8): p. 511-523.
91. Devane, W.A., et al., Isolation and structure of a brain constituent that binds to the cannabinoid receptor. Science, 1992. 258(5090): p. 1946-1949.
92. Sugiura, T., et al., 2-Arachidonoylgylcerol: A Possible Endogenous Cannabinoid Receptor Ligand in Brain. Biochemical and Biophysical Research Communications, 1995. 215(1): p. 89-97.
93. Ueda, N., K. Tsuboi, and T. Uyama, Metabolism of endocannabinoids and related N-acylethanolamines: Canonical and alternative pathways. FEBS Journal, 2013. 280(9): p. 1874-1894.
94. Maayah, Z.H. and A.O.S. El-Kadi, The role of mid-chain hydroxyeicosatetraenoic acids in the pathogenesis of hypertension and cardiac hypertrophy. Archives of Toxicology, 2016. 90(1): p. 119-136.