Determination of biomarkers for lipid peroxidation and oxidative stress: Development of analytical techniques and methods

(1)

1

DETERMINATION OF BIOMARKERS FOR LIPID PEROXIDATION AND OXIDATIVE STRESS

− Development of analytical techniques and methods −

Kristina Claeson Bohnstedt

Doctoral Thesis

Department of Analytical Chemistry Stockholm University

Stockholm, 2005

(2)

2

Akademisk avhandling som för framläggande av filosofie doktorsexamen vid Stockholms universitet offentligen försvaras i Magnélisalen, Kemiska övningslaboratoriet, Svante Arrhenius väg 12, torsdagen den 27 januari 2005.

ISBN 91-7265-988-2

Department of Analytical Chemistry Stockholm University

106 91 Stockholm

(3)

3

Abstract

Oxidative stress can be defined as a state of disturbance in the pro-oxidant/

antioxidant balance in favour of the former, leading to potential damage. Pro- cesses associated with oxidative stress involve reactive oxygen species and radicals and can result in elevated levels of oxidatively modified or toxic molecules that can cause cellular malfunction, and even cell death. Destruction of membrane lipids, lipid peroxidation, caused by reactive oxygen species and radicals has been coupled to many diseases and also normal ageing.

The measurement of low molecular weight biomarkers of lipid peroxidation present in complex matrices such as brain tissue, plasma, urine or cerebrospinal fluid is a delicate and difficult task and there is a need for improved analytical tools in this field of research.

The major foci of this thesis and the work underlying it are the development of analytical techniques and methods for determining biomarkers for oxidative stress and lipid peroxidation. Aspects of particular concern include the effects of sample treatments prior to analysis, evaluation of the developed methods with respect to possible artefacts, and the scope for results to be misinterpreted.

The specific research goals and issues addressed are detailed in five papers, which this thesis is based upon.

Paper I focuses on malondialdehyde, describing and evaluating two new sim- plified sample pre-treatment regimes for the determination of malondialdehyde in rat brain tissue by capillary electrophoresis with UV detection. The effects of sample storing and handling are also considered.

Paper II describes the synthesis, characterization and implementation of a new internal standard for the determination of malondialdehyde in biological samples using electrophoretic or chromatographic separation techniques. The useful- ness of the internal standard is demonstrated in analyses of rat brain tissue samples.

(6)

6

Paper III presents a method for the determination of 4-hydroxynon-2-enal in brain tissue from rats employing micellar electrokinetic chromatography separation and laser-induced fluorescence detection.

Paper IV is focused on the development of a new methodology for determin- ing the stereoisomeric F2-isoprostanes in human urine samples employing chromatographic separation on porous graphitic carbon and detection by electrospray ionization-tandem mass spectrometry. The results from this study conflict with the hypothesis that peripheral isoprostanes are elevated in patients with Alzheimer’s disease.

Paper V describes porous graphitic carbon chromatography-tandem mass spec- trometry for the determination of isoprostanes in human cerebrospinal fluid. A new simplified sample pre-treatment regime, involving a column switching technique, is presented that allows direct injection of a relatively large volume of CSF into the chromatographic system.

(7)

7

Preface

This thesis is divided into three parts. The first presents an introduction to oxidative stress, and then discusses types of biological samples and compounds that have been used to monitor such stress and to elucidate associated processes. The second is dedicated to the analytical tools used in the studies underlying the thesis, and summarises the results obtained. Finally, the third part discusses insights into oxidative stress provided by these and previous investi- gations and issues that remain to be clarified.

This thesis is based upon the following publications

I Kristina Claeson, Fredrik Åberg, Bo Karlberg.

Free malondialdehyde determination in rat brain tissue by capillary zone electrophoresis: evaluation of two protein removal procedures.

Journal of Chromatography B, 740 (2000) 87-92.

The author was responsible for the idea, the laboratory work, data evaluation and writing of this paper.

II Kristina Claeson, Gunnar Thorsén, Bo Karlberg.

Methyl malondialdehyde as an internal standard for the determination of malondialdehyde.

The author was responsible for the idea, the laboratory work (in collabo- ration with Gunnar Thorsén), data evaluation (except for the NMR and MS data) and writing of this paper.

III Kristina Claeson, Gunnar Thorsén, Bo Karlberg.

Micellar electrokinetic chromatography separation and laser induced fluorescence detection of the lipid peroxidation product 4-hydroxynonenal.

The author and Gunnar Thorsén contributed equally to the laboratory work, data evaluation and writing of this paper.

(8)

8

IV Kristina Claeson Bohnstedt, Bo Karlberg, Lars-Olof Wahlund, Maria Eriksdotter Jönhagen, Hans Basun, Staffan Schmidt.

Determination of isoprostanes in urine samples from Alzheimer patients using porous graphitic carbon liquid chromatography-tandem mass spectrometry.

The author and Staffan Schmidt were jointly responsible for the idea, and the author was responsible for the laboratory work, data evaluation and writing of this paper.

V Kristina Claeson Bohnstedt, Bo Karlberg, Hans Basun, Staffan Schmidt.

Porous graphitic carbon chromatography-tandem mass spectrometry for the detection of isoprostanes in human cerebrospinal fluid

Journal of Chromatography B, submitted.

The author and Staffan Schmidt were jointly responsible for the idea, and the author was responsible for the laboratory work, data evaluation and writing of this paper.

Paper not included in this thesis

Joanna Olsson, Kristina Claeson, Bo Karlberg, Ann-Caroline Nordström.

Determination of indole-3-acetic acid and indole-3-acetylaspartic acid in pea plant with capillary electrophoresis and fluorescence detection.

Journal of Chromatography A, 796 (1998) 231-239.

(9)

(10)

10

(11)

11

1. Aim

The aim of the studies this thesis is based upon was to develop improved methodology for the determination of selected biological markers of lipid peroxidation using modern analytical instruments and microseparation techniques. Such methods should ideally be fast and simple, keeping the number of steps in the analytical process at a minimum. Aspects of particular concern in both the experimental work and the thesis include the effects of sample treatments prior to analysis, evaluation of the developed methods with respect to possible artefacts, the scope for results to be misinterpreted, and other potential pitfalls when studying biological markers of lipid peroxidation in biological samples.

(12)

12

2. Introduction

Life processes, especially under stress conditions, generate a wide range of potentially damaging by-products as well as useful metabolites. Oxidants produced in this way, often free radicals, can cause extensive damage to DNA, proteins and lipids. Oxidative stress occurs when more oxidant by-products are produced than the cell can defend itself against. Oxidative stress can, therefore, occur when the production of free radicals increases, when quenching of free radicals or repair of damaged macromolecules decreases, or when these changes occur simultaneously. Sies similarly defined the process as “a disturbance in the pro-oxidant-antioxidant balance in favour of the former, leading to potential damage” [1]. Processes associated with oxidative stress can result in an elevated level of oxidatively modified or toxic molecules that can cause cellular malfunction and death. Oxidative stress and destruction caused by radicals have been coupled to many diseases and also normal ageing.

Oxygen-dependent deterioration of fats and oils, also called rancidity, causes problems when storing these substances and has been known since ancient times. In more recent times, it has been concluded that lipid peroxidation is one of the main processes induced by oxidative stress in vivo. Lipid peroxidation, i.e. the oxidative destruction of polyunsaturated fatty acids (PUFAs), is an autocatalytic, uncontrolled process leading to the formation of fatty acid hydroperoxides. These primary lipid peroxidation products are unstable and trans- form into more stable secondary products that can be used as biomarkers for the process. Animal cell membranes are prone to lipid peroxidation since they contain, among other constituents, the unsaturated fatty acids linoleic acid (18:2), linolenic acid (18:3), arachidonic acid (20:4) and docohexaenoic acid (22:6).

Detailed definitions of key terms like oxidative stress, oxidative damage and antioxidants can be found in a recent review by Halliwell and Whiteman [2].

The analytical challenges involved in determining minute concentrations of analytes in complex biologically relevant samples are well known. During lipid

(13)

13

peroxidation a multitude of degradation products are formed and, as will be discussed later, the samples are often very sensitive to harsh treatments. Mea- suring lipid peroxidation in biological samples is, therefore, a delicate and difficult task. Improved analytical tools, enabling reliable determinations of biomarkers associated with lipid peroxidation are urgently required in order to investigate and elucidate the fundamental mechanisms underlying this process.

(14)

14

3. What is a biomarker?

The word biomarker is widely employed and quoting Naylor may give insight into the extent of its usage; “[biomarker is]…an umbrella coalescence term which covers the usage and development of tools and technologies, monitoring drug discovery and development and understanding the prediction, causes, progression, regression, outcome, diagnosis and treatment of disease…“ [3].

A commonly used definition is the one proposed in 2001 by the Biomarkers Definitions Working Group from the National Institute of Health and the Food and Drug Administration (NIH/FDA) in the USA [4]: “A characteris- tic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention”.

Broadly, there are three types of biomarkers: (1) disease biomarkers – used to monitor and diagnose the progression of a disease; (2) drug efficacy/toxicity biomarkers – used to monitor the efficacy or toxicity of a treatment regime;

and (3) pharmacodynamic markers for monitoring pharmacological responses.

A biomarker needs to be validated for sensitivity, specificity and reproducibil- ity, and to be evaluated with respect to clinical endpoints, i.e. how patients feel, function and/or survive when changes in the level of the biomarker occur [5].

In 2004 Halliwell and Whiteman presented a list of criteria (shown in Table 1) that an ideal biomarker of oxidative damage should fulfil [2]. The same article also gives a comprehensive survey of the various methods and molecules used to measure and study oxidative damage to DNA, lipids and proteins.

(15)

15

Table 1 Criteria for an ideal biomarker of oxidative damage

A) Core criterion

The biomarker must be predictive for later development of disease.

B) Technical criteria

(i) The biomarker should detect a major part, or at least a fixed percentage of total ongoing oxidative damage in vivo.

(ii) The coefficient of variation between different assays of the same sample should be small in comparison with the difference between subjects.

(iii) Its levels should not vary widely in the same subjects under the same conditions at different times.

(iv) It must employ chemically robust measurement technology.

(v) It must not be confounded by diet.

(vi) It should ideally be stable during storage, i.e. neither lost nor formed artifactually in stored samples.

(16)

16

4. Biological samples

This thesis deals with a variety of biological matrices: brain tissue from rats and human plasma, cerebrospinal fluid (CSF) and urine. All types of samples are associated with specific analytical challenges depending on their origin and composition. A common feature of all the selected analytes is that they can be considered as small molecules, having molecular weights below 400 Da. The sample matrices studied in this work and some guide- lines on storing samples to prevent further decomposition of lipid sample constituents are described below.

4.1 Brain tissue

The total lipid content of the brain is approximately 10 %. The membranes in the brain are rich in polyunsaturated, highly peroxidable fatty acids. Oxygen consumption in the living brain is proportionally greater than in many other organs. In addition, brain tissue contains only moderate levels of both enzymatic and non-enzymatic quenchers for the reactive oxygen species. Due to the high lipid content in brain tissue, these samples must be stored at - 70 °C and an antioxidant such as butylated hydroxytoluene (BHT) should be added to the samples promptly.

4.2 Plasma

Blood sampling is one of the most convenient techniques for population screen- ing purposes since it is minimally invasive. Whole blood is composed of two fractions. The blood plasma accounts for 55 % of the volume, and formed elements, i.e. cells and cell fragments, comprise the remaining 45 %. When the formed elements are removed by centrifugation the plasma remains. Plasma consists of about 91.5 % water, 7 % proteins and 0.5-1 % lipids, the balance consisting of other solutes such as electrolytes, nutrients, gases, waste products, regulatory substances and vitamins. The storage of plasma samples has the same requirements as brain tissue samples.

(17)

17

4.3 Urine

Collection of urine samples is feasible for large clinical trials as it is noninvasive.

At a normal water intake, an adult person excretes 1-2 l of urine every day. The major urine components are water, urea, creatinine and sodium-, chloride- and potassium ions in the 50-250 mM range. Urine only contains minor amounts of proteins and negligible amounts of lipids. Urine is the primary medium for excretion of water-soluble waste products and other species that have been made water-soluble by metabolism. Compared to plasma and CSF, urine pro- vides an integrated index of analyte production over time. Since urine does not contain any significant levels of lipids there is no risk of artefactual generation of lipid peroxidation products by decomposition of sample constituents. It has even been shown that urinary isoprostane (see section 5.2) levels remain un- changed after a 5-day period at 37 °C [6]. The requirements for storing urine samples are thus less strict and samples can be kept at - 20 °C.

4.4 Cerebrospinal fluid

Cerebrospinal fluid, CSF, is the clear fluid that continuously circulates in the subarachnoid space (the space between the skull and the cortex), the ventricu- lar system of the brain and the spinal cord. It gives mechanical and chemical protection and is a medium for exchange of nutrients and waste products between the blood and the nervous tissue. The total amount of CSF is about 150 ml, and around 500 ml is produced every day, which indicates its very active circu- lation. CSF is in many ways similar to an ultrafiltrate of plasma and has a protein content below 430 mg/l. The CSF is a very useful matrix for studies of the central nervous system since the fluid reflects the metabolic state of the brain under both healthy and disease conditions. Usually, it is obtained by a procedure called lumbar puncture. CSF samples should be stored at - 70 °C.

(18)

18

5. Lipid peroxidation biomarkers

The targets of attack in lipid peroxidation are membrane lipids that surround cells and cell organelles. Fig. 1 shows a schematic diagram of the phospholipid bilayer of the outer membrane of a cell. When the lipids in the bilayer are changed oxidatively, not only are potentially harmful molecular species formed, but also the membrane fluidity is affected, which may lead to changes in or loss of cell functions or even cell death.

5.1 Aldehydes

The general lipid peroxidation process is illustrated in Fig. 2. The initiation of lipid peroxidation starts with a free radical attack. Hydrogen is abstracted from the target fatty acid (LH) to form a fatty acid radical (L·). The carbon radical is usually stabilized by a molecular rearrangement to form a conjugated diene.

Then, a peroxyl radical (LOO·) is produced by oxygen uptake. The peroxyl radical can abstract a hydrogen atom from another PUFA, thus forming a fatty acid hydroperoxide (LOOH). Hence, a new oxidation chain is initiated, in which new fatty acid radicals are generated. This is called the propagation stage of lipid peroxidation. The lipid hydroperoxides formed can then decompose to form, among other species, a great variety of aldehydes. Aldehydes are rela-

Cell Phospholipid bilayer

Phospholipid

O O

H₂C H2C O PO4

CH₃ Base

C H₃

O

Figure 1. The outer membrane of a cell, comprising a 60-100 Å thick phospholipid bilayer.

(19)

19

tively stable, compared with free radicals and hydroperoxides. This allows them to diffuse within or out of a cell, and to reach locations remote from the site where the initial free radical attack occurred. The stability of the aldehydes, and the fact that they are always produced when lipid hyderoperoxides break down in biological systems, make their identification and measurement valu- able. Aldehydes can thus work as biomarkers, providing an index of the extent of lipid peroxidation, and enable the role of aldehydes in specific pathological conditions to be examined.

Among the many different aldehydes that can be formed, the most intensively studied are malondialdehyde and 4-hydroxy-2-trans-nonenal [7]. Schauenstein and Esterbauer, in the 1960s, were among the first to recognize the importance of these aldehydes [8].

O2

Rearrangement to conjugated diene

Oxygen uptake Hydrogen abstraction

The peroxyl radical abstracts H from another PUFA causing an autocatalytic chain reaction.

Lipid hydroperoxide Cyclic peroxide Cyclic endoperoxide

Fragmentation to aldehydes e.g.

malondialdehyde and hydroxynonenal.

L

LOOH LOO

LH

C

O O

O O H

H

Part of fatty acid with three double bonds

Figure 2. Representation of the initiation and propagation reactions of lipid peroxidation.

(20)

20

5.1.1 Malondialdehyde – formation, properties and analysis

Malondialdehyde (MDA) is a volatile, 1,3-dicarbonyl compound of low molecular weight (72.07 g/mole). It is a weak acid, since the pKa value of the enolic OH group is 4.5. Hence, in neutral and alkaline conditions the predomi- nant form is the enolate ion (Fig. 3). The molecule absorbs light in the UV region in both acidic (λ_max = 245 nm, ε ≈ 13000) and neutral or basic solutions (λ_max = 267 nm, ε ≈ 31000). MDA is said to originate from the oxidative decomposition of fatty acids containing three or more double bonds, such as linolenic acid (18:3), arachidonic acid (20:4) and docosahexaenoic acid (22:6) [9].

Under physiological conditions MDA is moderately reactive and it can act both as an electrophile and as a nucleophile. It reacts with biomolecules containing primary amino groups, such as proteins, nucleic acids and amino phospholipids [9]. The correlation between MDA and various diseases is discussed in [10]

and its potential as a biomarker of oxidative damage to lipids in [2,10,11], amongst other texts.

Theories relating to MDA formation and methods for its quantification have been extensively discussed [7,9,12]. The techniques used for the determination of MDA can be divided into two classes: derivative and direct. The most common methods utilize the reaction (heated, in acidic conditions) between MDA and two molecules of thiobarbituric acid (TBA), which generates a red, fluo-

O O

H H

H

H H

O O

pKa 4.5

Figure 3. Malondialdehyde, MDA.

(21)

21

rescent product. This product is then measured, either directly by UV/Vis spec- troscopy or after liquid chromatography (LC) separation employing UV or fluorescence detection. The approach has several drawbacks, including a lack of selectivity and harsh derivatization conditions. As early as 1958 it was shown, in studies of fish oil, that 98 % of the MDA that reacts in the TBA-test was not originally present in the sample but was formed by decomposition of lipid per- oxides in the sample during the acid heating stage of the TBA assay [13,14].

The commonly recognised lack of selectivity of this assay, i.e. the fact that several compounds other than MDA yield products with similar absorption wavelengths on heating with TBA, has led to use of the term TBARS (thiobarbituric acid reactive species) [15]. This expression should be used when discussing results generated by this methodology. Apart from the TBA-based methods other derivative methods employ gas chromatography (GC) with electron capture (EC) [16-18] or mass spectrometric (MS) detection [19-23]. Gen- erally, the derivative methods have low limits of detection (LODs) but they have inherent problems, including the use of time-consuming sample treatments and the risk of producing MDA from unstable sample constituents during the derivatization step. As mentioned previously, MDA is known to bind to proteins [15,24]. For that reason, it is desirable to know whether a given method determines the free, bound or total MDA content. When a crude sample is exposed to harsh assay conditions, such as elevated temperatures or the addition of solutions with extreme pH values, changes in the relative amounts of free and bound forms of MDA can be expected. To avoid this, direct MDA analysis, performed under mild treatment conditions, is preferable, since it limits shifts in the free and bound equilibrium and minimizes the risk of MDA generation from sample constituents during analysis. Direct analytical methods include LC [25-28] and capillary electrophoresis (CE) [29], both in conjunc- tion with UV-absorbance detection of the native molecule.

Some researchers claim that malondialdehyde is less commonly monitored than it used to be. However, well over a thousand papers were published in 2004 in which “malondialdehyde” was listed as one of the key words.

(22)

22

There is quite a large range in the levels of MDA reported in the literature.

Depending on methodology and sample type, the reported values range from zero up to high micromolar concentrations. For example, reported plasma levels range from low nM to around 40 µM depending on methodology. MDA was the target molecule in studies I and II.

5.1.2 Hydroxynonenal – formation of, properties and analytical methods Fig. 4 illustrates the structure of 4-hydroxy-2-trans-nonenal (HNE), a mol- ecule with a formula weight of 156.2. It absorbs light in the UV region, with a λ_max at 221 nm (ε ≈ 14000). HNE is a major product of the peroxidative decomposition of ω6 PUFAs, such as linoleic acid (18:2) and arachidonic acid (20:4).

The HNE aldehyde group, the CC double bond and the hydroxy group are all able to take part in the chemical modification of biomolecules. HNE can react rapidly with thiol and amino groups at physiological pH levels [30]. This may be a significant factor in the claim that HNE is one of the most toxic substances produced during lipid peroxidation. It possesses cytotoxic, hepatotoxic, mu- tagenic and genotoxic properties [31]. It has also been hypothesized that HNE was one of the toxic agents in the “Spanish cooking oil syndrome” [32]. More recently, it has been suggested that HNE may not be merely a toxic product of lipid peroxidation, but may also function as a biological signal substance in both pathological and physiological conditions [33,34].

H O

OH Figure 4. 4–hydroxy-2-trans-nonenal, HNE.

(23)

23

Several authors have described methods for the determination of HNE. These techniques are mainly derivative, but at high levels (more than 2 µM), LC-UV can be used on the native molecule [7]. At lower levels, derivatization has so far been necessary to enable detection. Various hydrazine reagents have been used to facilitate its detection for LC-UV [7,35], LC with electrochemical detection [36] and GC-MS (after silylation) [37,38]. Several other methods have also been described [39,40]. Another relevant way to estimate HNE levels is to measure adducts to biomolecules using GC-MS or LC-MS [41]. The journal

“Molecular aspects of medicine” recently dedicated an entire issue to HNE and its role in lipid peroxidation [42].

As in MDA analysis, measures must be taken to minimize unintentional changes in sample composition caused by handling when analysing HNE. The levels of HNE in healthy tissues may be approximately 0.1 µM or lower, and the ratio between free and bound forms has not been comprehensively established. Dur- ing oxidative stress in vivo, increases to 1-20 µM, are possible [7]. HNE was investigated in studies described in Paper III.

5.2 Isoprostanes

5.2.1 Isoprostanes – formation, properties and analysis

In 1990 Morrow et al. demonstrated the formation of a group of prostaglandin F2-like lipid peroxidation products [6]. These substances constitute a family of lipids formed in vivo by the enzyme-independent and free radical catalyzed, peroxidation of arachidonic acid (AA) in membrane phospholipids. Since the molecules are isomeric to the prostaglandin F2α formed enzymatically by cyclooxygenase (COX), they have been called F2-isoprostanes. Isoprostanes were discovered serendipitously during studies of COX-derived prostaglandin D2. The researchers found, using GC-MS, that plasma samples from healthy volunteers that were processed and analyzed immediately contained peaks possessing characteristics of F-ring prostaglandins. After storage of the plasma samples for several months at

(24)

24

- 20 °C identical peaks were found, but this time at 100-fold higher levels [43], which had been formed from lipids in the plasma samples. The formation of the F2-isoprostanes from AA is outlined in Fig. 5. Following the discovery of the F2-isoprostanes, it has been shown that the isoprostane pathway can provide a route for the generation of other classes of isoprostanes from unsaturated fatty acids other than AA that have at least three double bounds. Examples are the “neuroprostanes” originating from docohexaenoic acid (22:6) [44]. Two structural features distinguish the F2-isoprostanes from their enzymatically derived relatives: the hydroxyl groups on their prostane ring have a cis orientation, and their side chains predominantly have a cis orientation in relation to the prostane ring.

COOH HO

HO

COOH

OH Class III

HO

HO OH

COOH

Class V

HO

COOH OH

Class VI HO

HO

COOH OH

Class IV

COX PG Arachidonic acid

FR + O₂ FR + O₂

Figure 5. Free radical (FR) attack on arachidonic acid generates four different classes of stereoisomers of prostaglandin F2 α called F2-isoprostanes (classes III-VI). Prostaglandins (PGs) are enzymatically formed from arachidonic acid by the cyclooxygenase (COX) pathway.

(25)

25

The F2-isoprostanes can be divided into four subgroups of regioisomers, called types III, IV, V and VI. Each regioisomer is comprised of eight racemic dias- tereoisomers (see Fig. 5). All the F2-isoprostanes have a molecular mass of 354.5 and the pKa of the acidic group is approximately 5. The nomenclature and abbreviations used for isoprostanes in general, and in the subgroup of F2- isoprostanes, are often confusing as the two major research groups working with isoprostanes (Morrow and FitzGerald, and their respective co-workers) use different terminology [45,46]. The two research teams also propose two slightly different explanations for the formation of F2-isoprostanes from AA [47,48].

Elevated levels of F2-isoprostanes have been reported in various physiological states associated with enhanced lipid peroxidation and oxidative stress, including a range of cardiovascular [49-53] and neurological diseases [54-56]. Some of the isoprostanes have proven to be biologically active, mediating vasocon- striction [57]. A large amount of work has been devoted to the field of isoprostane analysis since they were first described, and two main approaches have been adopted for their quantification in various biological samples. The first is an immunological approach involving radioimmunoassays (RIAs) and enzyme immunoassays (EIAs) that in many cases are inexpensive and easy to perform [58,59]. These methods are considered to give only a semi-quantitative estimate of isoprostane levels, since the risk of cross reactivity is significant. The second approach is based on chromatographic separation and detection by mass spectrometry (MS). One of the most frequently used methods of this kind involves gas chromatography-electron capture negative chemical ionization mass spectrometry (GC-ECNI-MS) [48,60]. This technique has low detection limits but is time consuming and labour intensive, requiring sample pre-treatment steps such as thin layer chromatography, solid phase extraction and derivatization. LC-MS has been used as an alternative to GC-ECNI-MS for the determination of F2-isoprostanes in urine and plasma samples. Li and co-workers have successfully separated F2-isoprostane isomers in each of the four classes using a C18 stationary phase, but their results have been very difficult to repeat [61]. Papers IV and V are assigned to F2-isoprostanes. The topic of isoprostanes has recently been reviewed in [57,62].

(26)

(27)

(28)

(29)

29

6. Preparation of biological samples for CE or LC

Table 2 presents a summary of the analytes, sample types, pre-treatments, separation and detection procedures used in the work underlying this thesis. To suit the demands of this type of sample matrices, the preparation procedures needed to be kept as simple as possible while generating samples that could be injected into CE systems (with UV or laser induced fluorescence detection instruments) or analysed by LC-MS.

Table 2

Paper Analyte Sample type Pre-treatment Separation Detection

I MDA Rat brain tissue Precipitation/ CE UV

Ultrafiltration

II MDA Rat brain tissue Ultrafiltration CE/GC/LC UV/EC/MS Unp.^a MDA Human plasma Ultrafiltration CE/ITP UV III HNE Rat brain tissue Derivatization MEKC LIF

IV IsoP Human urine Extraction PGC-LC ESI-MS

V IsoP Human CSF Ultrafiltration/ PGC-LC ESI-MS

Direct injection

aUnpublished results

As a biological matrix contains a myriad of compounds of various types and concentrations, the use of efficient separation steps such as CE or LC are very helpful for separating target compounds from a sample. Depending on the properties of the analyte and the sample, some kind of sample preparation is almost always needed before this final analysis step can be successfully performed.

(30)

30

The general reasons for subjecting biological samples to preparatory treatments (for the determination of low molecular weight molecules) are similar for CE and LC, and include;

{ Removal of unwanted compounds in the sample matrix, e.g. proteins and salts.

{ Enrichment of the analyte.

{ Facilitation of detection, i.e derivatization.

The sample components that cause the most problems in both CE and LC are often proteins, which readily adsorb onto the surface of the CE capillary walls, thereby impairing or completely destroying the separation. The flow through the capillary can even be irreversibly stopped. Another problem is when broad protein peaks interfere with the detection of the desired analyte peaks in the electropherogram. In LC, high protein contents in a sample can give rise to clogging of the column or irreversibly affect the stationary phase. Studies on deproteinization have been published for both CE [63] and LC applications [64]. Precipitation and ultrafiltration are two commonly used modes of protein removal and are discussed below.

6.1 Precipitation

Precipitation is a fast and simple means for protein removal and many samples can be processed at the same time. It can be accomplished by the addition of organic solvents, salts or acids to the sample followed by centrifugation of the precipitate. The clear supernatant is then either injected directly, or subjected to further preparation steps. Examples of precipitating agents used in literature include acetonitrile, methanol, trichloroacetic acid and ammonium sulphate.

The effect of the media used for precipitation on the chromatographic or electrophoretic system must be considered. One disadvantage with organic solvent precipitation is the dilution of the sample. However, the organic solvent lowers the sample’s conductivity, so the enhanced stacking of the charged sample components in CE may compensate for some of the dilution. For a discussion about

(31)

31

stacking, see section 7.3. In reversed phase LC, organic solvent in the injected sample, at a higher ratio than in the mobile phase, is detrimental to the separation. Concentration to dryness and reconstitution in an appropriate solvent can alleviate this problem. This, however, adds an additional step and could de- crease the total recovery as discussed later. The other two alternatives, precipitation by salt or acid, are not advantageous in either CE or LC. In CE, conductivity should be kept low in the injected sample zone to allow satisfactory stacking and low Joule heating. In LC, acid or salt can be detrimental to the column packing material or harm separation. Further, strong acid can cause degradation of sample components including the analytes. One additional alternative is to precipitate the proteins in a sample by heating. This treatment can be too harsh for thermolabile analytes, and may lead to artefactual generation of analytes from sample constituents, as is the case with lipid peroxidation markers. Precipitation by adding organic solvents (acetonitrile or methanol) was used in studies I and III.

6.2 Ultrafiltration

One convenient way to remove proteins from samples without dilution is ultrafiltration. Ultrafiltration involves filtration of the sample through a membrane of specific pore size, using a centrifuge. Particles and molecules larger than the threshold size (e.g. 30 kDa) are retained on the surface while smaller species are allowed to pass through. This significantly reduces the amount of protein in the sample [63,64]. Many samples can be run simultaneously and it is suitable for a wide range of sample volumes since the ultrafiltration devices come in different sizes. However, no sample enrichment is obtained with ultrafiltration. Furthermore, ultrafiltration allows only the free fraction of the analyte, not the protein-bound forms (if any), to pass through the filter. This facilitates studies of free versus protein-bound levels of the analyte. Consequently, it has been employed frequently for protein binding studies of drugs [65-67]. A vital aspect is to verify that the analytes of interest do not adsorb onto the membrane or the polymeric material in the devices. Further, it is important to maintain physiological settings regarding, for example, pH and temperature in free ver-

(32)

32

sus protein-bound form studies, since binding values can otherwise be affected.

Ultrafiltration was used in studies I, II and V.

If precipitation or ultrafiltration doesn’t generate sufficiently clean and concentrated samples, quite a wide range of alternative sample preparation methods can be applied. Such methods can also provide more sophisti- cated ways to remove proteins. Some examples of potentially useful procedures are described below.

6.3 Liquid–liquid extraction

Clean and concentrated samples can be obtained after applying liquid-liquid extraction (LLE). LLE is a separation process that exploits differences in the relative solubilities of the analytes in immiscible solvents. The compounds of interest can be extracted from, for example, an aqueous sample, into a volatile organic solvent such as pentane, hexane, diethyl ether or ethyl acetate. Impor- tant variables to drive the extraction in the desired direction are pH and the ionic strength of the aqueous solution. The organic solvents used for LLE can- not be injected into CE or, in most cases, reversed phase LC systems. Thus, after extraction, the organic phase is evaporated to dryness and the residue is then reconstituted into a suitable solvent. Loss of analyte can occur during the evaporation step, by adsorption to the equipment or incomplete reconstitution.

As LLE is not easily automated it can be a bottleneck in high-throughput analysis.

However, if a robot is used, many samples can be processed in parallel with reduced manual handling, thereby saving a considerable amount of time. In some cases, LLE can generate cleaner samples than solid phase extraction treatment (SPE). For example, Bonfiglio showed that ion suppression effects in electrospray ionization (see section 8.4) were reduced more by LLE than by ACN precipitation or SPE [68]. So, if the recovery in the evaporation/reconstitution step is high and a well-judged selection of solvents and pH is used, LLE can give very clean extracts with satisfactory selectivity for the studied analyte.

A robot-run LLE extraction technique was used in the work associated with Paper IV. Useful reviews on LLE of biological samples are presented in [69,70].

(33)

33

6.4 Solid phase extraction

Solid phase extraction (SPE) can be used to simultaneously enrich the analytes in samples and remove salts, proteins and a wide variety of other potentially interfering compounds. In SPE the sample is passed over a small tube filled with porous solid particles such as silica-C18. Alternatively, a membrane disc containing sorbent particles can be used. After preconditioning the sorbent, the sample is applied. The sorbent is then selectively washed to remove unwanted compounds without losing the analyte(s) of interest. Finally, the analytes are eluted using a minimum of solvent. The solvent used should be compatible with the next analytical step, otherwise it must be changed by evaporation/

reconstitution as in LLE. There are a great number of different, commercially available SPE phases and the process can be coupled on-line in order to save time and minimize the risk of sample loss. In recent years SPE has replaced many LLE extractions, but in some cases SPE eluates are not as clean as extracts obtained after a judiciously performed LLE . Examples of reviews on SPE of biological samples for CE and chromatographic applications are presented in [71,72]. SPE was a minor part of the analytical procedures described in Paper IV.

6.5 Column switching

The term column switching has been widely used in the literature and has been employed in applications ranging from on-line SPE to complex net- works of columns. In a review, Campins-Falco et al. define the term as encompassing “all techniques by which the direction of the flow of mobile phase is changed by valves, so the effluent of a primary column is passed to a secondary column for a defined period of time” [73]. Majors has described some practical considerations regarding the technique [74]. Col- umn switching can significantly facilitate analysis of biological samples by providing on-line sample clean-up, and enrichment or by improving resolution and selectivity through the use of different stationary phases.

To avoid band broadening the first column should have smaller, or at most, the same dimensions as the second column. Further, again to minimize band broadening, the retention capability of the first column should not exceed

(34)

34

that of the second column for the selected solvents, since efficient chromatography is promoted if the sample is re-concentrated at the top of the second column. An example of a column switching setup, described in paper V, is illustrated in Fig. 6.

An increasingly popular way to facilitate the direct injection of biological flu- ids in LC is to use restricted access materials (RAMs) in the pre-column. RAMs are porous packing materials that prevent macromolecules, like proteins, from penetrating the pores but allow free access of low molecular weight compounds.

The inert layer on the outside of the material does not interact with the sample matrix while the porous inner surface of the material is covered with a phase, for example C18, that retains analytes by reversed-phase interactions. Unretained components, such as proteins, are washed out to waste before the retained fraction is eluted onto the analytical column where the separation is performed.

Recent reviews on RAM have been presented by [75,76].

6.6 Derivatization

Derivatization is not a pure cleaning or concentration step in sample pre-treatment. Instead, it can facilitate meaningful determinations of compounds by increasing their volatility (prior to GC analysis), improving both the selectivity

Injector Pump I

Trap Waste

Pump II Analytical Column MS-MS

Figure 6. The column switching setup used in study V.

(35)

35

(especially in chiral separations) and chromatographic efficiency, and enhanc- ing detectability. The technique is often regarded as a necessary evil since it adds an additional step to the sample preparation procedure and thus intro- duces a further source of error in the analysis. The development of sensitive and universal detection methods, such as MS, has reduced its importance in many fields. Derivatization is used today, in most cases, prior to LC only if all other alternatives have failed. In the field of CE and capillary electrochromatography (CEC), where detection is often problematic due to the low loads that can be accommodated, and the limited cross-column path lengths for spectroscopic detection, derivatization can sometimes be beneficial.

Derivatization was used in study III to enable detection by laser-induced fluorescence after a micellar electrokinetic chromatography (MEKC) separation and in study II to increase the volatility and detectability of the target compounds in GC-ECD and GC-MS. Recent books on the topic of derivatization for chromatographic and electrophoresis based analyses include [77,78].

6.7 Other techniques

The sample preparation techniques mentioned above are some of the most commonly used. However, there are of course a large number of other ways to prepare a sample for analysis that have been described in the literature.

General reviews regarding sample pre-treatments of biological samples that are suitable before analysis by CE, LC or both appear continuously and helpful examples are [79-84]. Reviews concerning specific techniques such as the use of molecular imprinted polymers (MIP) [85,86], capillary ultrafiltration [87,88], solid phase micro extraction (SPME) [89], affinity techniques [90], microdialysis [91,92], and membrane-based methods [93] have also been published. The on-line sample concentration technique “stacking” for CE is discussed in section 7.3.

(36)

36

7. Capillary electrophoresis

Electrophoresis in micrometer bore capillaries was first presented in the early 1980s [94], since then capillary electrophoresis has rapidly become a well- known technique. Basic issues and applications of CE are covered in several good books that are recommended for further reading [95-98]. This section starts with a brief introduction to capillary electrophoresis, and is then dedicated to a discussion of the analyte characteristics and conditions used to solve problems associated with measuring lipid peroxidation in biological samples.

7.1 Introduction to CE

CE is performed in narrow tubes (20–100 µm i.d.) or on micro-fabricated chips.

The discussion here is focused on CE in fused silica capillaries, although most of the basic elements are applicable to electrophoresis in general. The basic instrumentation required in CE is inexpensive and simple, comprising the capillary mentioned above, two buffer vials, a high voltage supply, a pair of elec- trodes, a detector and a computer equipped with a data handling system. Fig. 7a shows a schematic illustration of a CE system. Separation in electrophoresis relies on the differences in the electrophoretic migration velocities of the solutes in an electric field. The silanol groups at the inner surface of the capillary have pKa values ranging from 2 to 5. When the capillary is filled with a background electrolyte (BGE) whose pH exceeds the pKa of these groups, a net negative charge will be created on the wall surface. The negative charge at- tracts layers of hydrated positively charged ions to the vicinity of the wall and when an electric field is applied the entire bulk liquid moves towards the cath- ode due to the viscosity of the solution. This phenomenon is called electroosmotic flow (EOF). Since the driving force for the EOF is evenly distributed along the capillary walls a flat velocity profile is obtained, compared to the parabolic flow arising from mechanical pumping. The flat profile eliminates a large contri- bution to bandbroadening, thereby increasing the theoretical platenumbers.

When the capillary is filled with a BGE, and small amount of sample solution

(37)

37

Figure 7. Schematic representation of a CE system. (a) EOF in the normal direction. (b) Reversed EOF.

(38)

38

is introduced into one end and an electric field is applied, sample molecules are transported by the joint action of electrophoretic migration and EOF. Some of the fundamental formulas applying to electrophoresis are shown below.

The velocity of an ion in an electric field is given by

v = µ_e E (1)

where v = ion velocity, µ_e= electrophoretic mobility and E = applied electric field.

The mobility µ_e of a given ion in a given medium is a constant which is charac- teristic of that ion. In an electric field a charged analyte is acted on by an elec- tric force F_Ethat is given by

F_E = q E (2)

and a frictional force F_F that is given (for a spherical hydrated ion) by

F_F = -6 π η r v (3)

where q = ion charge, η = solution viscosity, r = ion radius and v = ion velocity

During steady state electrophoresis these forces balance each other and are equal but opposite

q E = 6 π η r v (4)

Solving for velocity using the above formulas gives an equation that describes the electrophoretic mobility of an ion in physical parameters

µ_e = q / 6 π η r (5)

(39)

39

7.2 Injection

The two general methods that can be applied for injection in CE are hydrodynamic injection and electrokinetic injection. In the first case, a pressure difference is created across the inlet and outlet vials, forcing the sample solution into the capillary inlet end. The volume of the sample injected depends on the mag- nitude and duration of the pressure applied, the BGE viscosity, and the capillary dimensions. In the second case a voltage is applied which causes the sample ions to migrate into the capillary as a result of electroosmosis and electrophoretic mobility. The amount injected depends on the electrophoretic mobility of the solutes, the electroosmotic flow rate, the applied voltage, the capillary dimensions, the solute concentration and the duration of the injection.

Hydrodynamic injection is, conceptually, the simplest method. It is reproduc- ible, since the sample matrix has virtually no effect on the injected amount. In electrokinetic injection, variations in sample conductivity, which may arise due to matrix effects, influence the quantity loaded. This often affects repro- ducibility in a negative way. In all the work presented here involving CE, hydrodynamic injections were used.

7.3 In-capillary sample concentration

The term “sample stacking” was first used by Ornstein [99] to describe the stacking of proteins according to their mobilities in disc electrophoresis. An up-to-date and broad definition of stacking is that it covers “all on-line sample concentration techniques in electrophoresis” [100]. A number of techniques have been developed that concentrate samples on standard CE equipment. These techniques can be broadly divided into two classes: those based on electric field amplification and those based on isotachophoresis. When the simple term

“stacking” is used in the literature, it usually refers to field amplified stacking.

Refs [100-102] are recent review articles on stacking in CE. In studies I and II field amplified stacking was used and in studies outlined in the “Un- published results” chapter (see section 9.6), an isotachophoretic stacking state was induced.

(40)

40 7.3.1 Field amplified stacking

Mikkers et al. described on-capillary sample concentration, obtained through field strength differences between the sample zone and the BGE [103]. Chien and Burgi later continued the work [104-106]. The basis of stacking is to provide an electric field of high strength across the injection zone. When the conductivity of the injected sample zone is lower than that of the background electrolyte, the field strength increases. A lower conductivity is usually achieved by using dilute samples or by the addition of organic solvents that reduces the conductivity in the samples. Electrophoretic velocity is proportional to the electric field, and after injection the solute ions rapidly migrate through the dilute sample zone until they reach the concentration boundary between the sample zone and the run BGE. The solutes then encounter a lower electric field which makes them slow down, and a narrow stacked zone is formed. In the work presented here, efforts were made to obtain a good stacking effect by keeping the influencing variables, such as injected sample zone conductivity, BGE concentration and mobility at appropriate levels.

7.3.2 Isotachophoretic stacking

Everaerts et al. presented another major advance in sample concentration in CE, namely isotachophoretic stacking (ITP) [107-109]. ITP is a variation of electrophoresis that can also be used to increase sample concentration prior to ordinary CE. It is based on equation 1, which shows that the velocity of an ion in an electric field of strength E is dependent on the mobility (µ_e) of the ion.

The molecules to be separated are sandwiched between a leading electrolyte with a high-mobility ion and a terminating electrolyte with a low-mobility ion.

When the electric field is applied, the sample components begin to arrange themselves into zones, according to their mobility. The ions with highest mobility give rise to the highest conductivity, so the field strength is lowest across the zone that contains them. Conversely, the least mobile ions generate the highest field strength and their velocity is increased. A steady state velocity

(41)

41

thus develops. When an ion diffuses into a neighbouring zone, its velocity changes and it returns to its original zone. Since the current is constant there is a constant ratio between the concentration and the mobility in each zone. The concentration in each zone is determined by the concentration of the leading electrolyte. Zones that are less concentrated than the leading electrolyte become compressed, so that an appropriate concentration is achieved. The simplest way to create an ITP system is to dissolve the sample in a buffer that, together with the background electrolyte, will create ITP conditions. This is called transient ITP. If a sample contains high concentrations of a species that can serve as a leading electrolyte, ITP effects can occur. This state can either be induced by the addition of a highly mobile ion such as chloride, as suggested by Foret [96,110] or be self-generated by the sample constituents.

The latter is especially likely when complex samples are being processed.

The phenomenon is called “self-stacking” and is further discussed in Refs. [111,112]. An ITP state was probably induced during the analysis of the plasma samples as discussed in section 9.6. In this case, the analyte peak was equivalent to nearly 3×10⁶ theoretical plates, which is significantly more than can be explained by conventional CE stacking theory in such a high ionic strength sample.

7.4 Separation

CE for anionic species is most commonly run with the electrophoretic migration and EOF flowing in opposite directions, as in Fig. 7a. If the analytes of interest are small and/or highly charged a faster separation can be achieved by reversing the EOF, as shown in Fig. 7b. In this mode the directional vectors of electrophoretic migration and EOF combine, and thus increase the speed of the analysis. Reversal of EOF was demonstrated by Zare and co-workers for determination of low molecular weight carboxylic acids [113] and was also earlier presented by Terabe [114] and Tsuda [115]. Reversal of the EOF can be accomplished by adding quaternary amines such as the cationic surfactant cetyl trimethyl ammonium bromide (CTAB) to the BGE. CTAB monomers adhere