Advances in drug-protein adduct analysis using LC MS based proteomics Linda Switzar

Transcription

1 Advances in drug-protein adduct analysis using LC MS based proteomics Linda Switzar Advances in drug-protein adduct analysis using LC MS based proteomics Linda Switzar

2 ADVANCES IN DRUG PROTEIN ADDUCT ANALYSIS USING LC MS BASED PROTEOMICS Linda Switzar

3 The research described in this thesis was performed within the framework of project D3-201 Towards novel translational safety biomarkers for adverse drug toxicity of the Dutch Top Institute Pharma. Printed by Wöhrmann Print Service, Zuthpen, The Netherlands. Financial support for printing of this thesis was kindly provided by Matrix Science Ltd.. Copyright 2013 Linda Switzar, Amsterdam. All rights reserved. No part of this thesis may be reproduced in any form or by any means without permission from the author.

4 VRIJE UNIVERSITEIT Advances in drug protein adduct analysis Using LC MS based proteomics ACADEMIS PROEFSRIFT ter verkrijging van de graad Doctor aan de Vrije Universiteit Amsterdam, op gezag van de rector magnificus prof.dr. F.A. van der Duyn Schouten, in het openbaar te verdedigen ten overstaan van de promotiecommissie van de Faculteit der Exacte Wetenschappen op vrijdag 20 september 2013 om uur in de aula van de universiteit, De Boelelaan 1105 door Linda Switzar geboren te Leidschendam

5 promotoren: prof.dr. H. Irth prof.dr. W.M.A. Niessen copromotor: dr. H. Lingeman

6 Voor mijn moeder

7 leescommissie: prof.dr. I.D. Wilson prof.dr. R. Bischoff prof.dr. G.W. Somsen dr. R.H.H. Pieters dr. Y.E.M. van der Burgt The research described in this thesis was carried out at the Division of BioMolecular Analysis, Department of Chemistry and Pharmaceutical Sciences, Faculty of Sciences, VU University Amsterdam, De Boelelaan 1083, 1081 HV, Amsterdam, The Netherlands.

8 TABLE OF CONTENTS Chapter 1 9 General introduction Chapter 2 25 Protein digestion: An overview of the available techniques and recent developments Chapter 3 47 Protein digestion optimization for characterization of drug protein adducts using response surface modeling Chapter 4 67 A high-throughput sample preparation method for cellular proteomics using 96-well filter plates Chapter 5 81 Identification and quantification of drug albumin adducts in serum samples from a drug exposure study in mice Chapter Summary, conclusions and perspectives Appendices Nederlandse samenvatting 121 List of abbreviations 127 List of publications 129 Curriculum Vitae 131 Dankwoord 133

9

10 Chapter 1 General introduction

11 1 10

12 1.1 ADVERSE DRUG REACTIONS AND DRUG PROTEIN ADDUCTS The number of new drugs reaching the market has been in decline for decades despite ever increasing investments into pharmaceutical R&D. [1] Drug toxicity and more stringent regulatory requirements are seen as a major cause of the high attrition rates in drug discovery and development. [2] Even approved drugs may eventually be withdrawn from the market due to adverse events that only become apparent when the drug is used by a larger population. [3-4] 1 Traditionally, adverse drug reactions (ADRs) were classified into type A and B reactions, which have been described in several ways, such as predictable and non-predictable, respectively. [5] Type A reactions are related to the pharmacological effect of the drug and are usually dose-related and predictable. These reactions occur quite often, especially with individuals that are at the extremes of the dose-response curves, but they are seldom life-threatening. [5] Conversely, type B reactions are bizarre and abnormal effects that are not related to the pharmacological effect of the drug. These often severe effects occur only in a few individuals (idiosyncratic) and are, therefore, difficult to predict. [5] Due to the diverse range of ADRs, these two classes are insufficient to describe the diverse range of ADRs. As a result, other types of adverse drug reactions were distinguished over the years, leading to a total of six categories labeled A-F, see Table 1.1. [6] Type C reactions are a combination of dose- and time-related (chronic) effects that are uncommon and related to the cumulative dose. [6] Type D reactions, originally proposed by Royer in 1997 [7], are very uncommon delayed effects that become apparent some time after use of the drug, such as adenocarcinoma in daughters of women who have taken stilbestrol during pregnancy. Type E reactions can also be considered as a delayed response to drug treatment, but are seen as a separate category because the reaction occurs directly after drug withdrawal. [6] The most recent addition to the classification of ADRs is the type F reaction that is described as therapeutic failure which may be caused by a number of reasons, such as inadequate dosage that is too low to obtain an effect or drug-drug interactions. [6, 8] Even with this expanded classification, not every ADR can be classified into one of the proposed categories. However, the classification is under constant revision and may be adapted further as more becomes known about the mechanisms underlying adverse effects. Several theories describing the mechanisms behind ADRs have been proposed in literature and have been compiled in reviews. [9-10] ADRs have long since been associated with drug metabolism. [11,12] During this biological process, the drugs are metabolized mainly into more polar metabolites that can be easily excreted via the kidneys and urine (Figure 1.1). However, many drugs also undergo bioactivation, which leads to reactive metabolites that may bind to other molecules present in the liver. The reactive metabolites can be detoxified via covalent binding to glutathione (GSH). This conjugation reaction results in increased polarity and the drug-gsh conjugates are subsequently excreted via the urine (and feces, via the bile). When large amounts of reactive metabolites are produced, it may lead to depletion of GSH. Once the GSH stores are depleted, the surplus of reactive metabolites may covalently bind to proteins. Reversible binding 11

13 1 Table 1.1 Classification of adverse drug reactions, adapted from Edwards et al.. [6] Type of reaction Mnemonic Features A: Dose-related Augmented - Common - Related to the pharmacological action of the drug - Predictable - Low mortality Examples Toxicity: Hepatotoxicity due to acetaminophen overdose Side effects: Anticholinergic effects of tricyclic antidepressants Immunological reactions: Penicillin hypersensitivity Idiosyncratic reactions: Pseudoallergy (ampicillin rash) Hypothalamic-pituitary-adrenal axis suppression by corticosteroids Carcinogenesis Teratogenesis (vaginal adenocarcinoma with diethylstilbestrol) Opiate withdrawal syndrome Myocardial ischaemia (ß-blocker withdrawal) Inadequate dosage of an oral contraceptive, particularly when used with specific enzyme inducers B: Non-dose-related Bizarre - Uncommon - Not related to the pharmacological action of the drug - Unpredictable - High mortality C: Dose- and time-related Chronic - Uncommon - Related to the cumulative dose D: Time-related Delayed - Uncommon - Usually dose-related - Occurs or becomes apparent some time after use of the drug E: Withdrawal End of use - Uncommon - Occurs soon after withdrawal of drug F: Unexpected failure of therapy Failure - Common - Dose-related - Often caused by drug-drug reactions 12

14 Bioactivation Phase I/II enzymes Metabolism 1 Reactive metabollites Detoxification (Glutathione) Stable metabollites Covalent binding to proteins Drug-protein adducts Excretion via urine ADRs Figure 1.1 Drug metabolism and bioactivation pathways. of drugs to proteins is considered a critical factor in drug efficacy [13], whereas covalent binding of reactive drug metabolites to proteins and the hereby formed drug protein adducts have been associated with drug toxicity. [3-4] This hypothesis is known as the hapten theory, where it is assumed that low-molecular weight compounds (< 1000 Da), such as drugs and their metabolites are not to able to elicit ADRs because, presumably, they are too small to be recognized by the immune system. [14] On the other hand, it has been observed that chemically inert drugs, not susceptible to bioactivation and the formation of reactive metabolites, still may elicit an immune response. This observation has led to the formation of the pharmacological interaction with immune receptors or p-i concept that was introduced more recently by Pichler. [15] Pichler proposed that the drugs themselves may bind directly and reversibly to immune receptors and, subsequently, to T-cell receptors, thereby stimulating a T-cell response. The p-i model has proven to be useful in explaining the observed drug hypersensitivity reactions after treatment with chemically inert drugs, such as lidocaine, but also some drugs that are known to be bioactivated, for example sulfamethoxazole and carbamazepine. [16] 13

15 1 Other theories, such as the danger hypothesis and non-immune-mediated mechanisms have also been proposed, but the hapten theory remains the major working hypothesis. [10] Consequently, drug metabolism and bioactivation as well as covalent binding of reactive drug metabolites to proteins have received a great deal of scientific attention, which has been summarized in an extensive review by Zhou. [13] Perhaps the best known example and most likely the basis of the hapten theory is the covalent protein binding of N-acetyl-p-benzoquinoneimine (NAPQI), the reactive metabolite of acetaminophen (APAP). [11] In the 1970s, APAP was one of the first drugs of which its metabolism and covalent protein binding was studied. [12] A decade later, free protein thiols were determined to be the major binding sites of NAPQI in in vitro experiments using bovine serum albumin (BSA). [17] Similar to human serum albumin (HSA), BSA contains a free cysteine residue on position 34 that is not involved in a disulphide bridge and, thus, is available for binding NAPQI. [17] Due to its high abundance in vivo, serum albumin was considered as a promising target of NAPQI. Identification of 27 potential target proteins of NAPQI was achieved several years later in mouse studies [18-21] but the list of identified proteins did not include serum albumin. Recently, the NAPQI albumin adduct was identified in human serum samples from patients that were exposed to high levels of APAP [22]. Apart from cysteine thiols, also other amino acids in proteins are susceptible to nucleophilic attack by reactive metabolites, including lysine amines, histidine imidazoles and protein N-terminal amines. [23] Many adducts of reactive metabolites of drugs and other xenobiotics, and their protein targets have been identified over the years and are listed in the reactive metabolite target protein database, a web-accessible resource. [24] In addition to elucidation of the mechanisms behind ADRs, covalent binding of reactive drug metabolites to proteins also is a critical factor for the evaluation of lead compounds in pharmaceutical R&D. [4, 25] The assessment of covalent protein binding and identification of drug protein adducts is, therefore, of utmost importance in the pharmaceutical industry. 1.2 ALLENGES AND METHODOLOGIES FOR THE DETECTION AND IDENTIFICATION OF DRUG PROTEIN ADDUCTS The analysis of drug protein adducts is faced with considerable challenges. The major complication for the detection and identification of drug protein adducts is their extremely low abundance. Drugs often do not exclusively bioactivate to a single reactive metabolite, but may produce several reactive metabolites as well as be metabolized into one or more stable metabolites. [26] Additionally, the reactive metabolites that were formed are, for the most part, detoxified by GSH, which leaves only a small portion, if anything, for covalent binding to proteins. Even in the case of a drug overdose where large amounts of reactive metabolites are produced and GSH is depleted [27], only a minute amount of the target protein is modified in vivo. This is illustrated by the detection of low pmol/ml serum levels of NAPQI HSA in a patient exposed to high levels of APAP and suffering from severe liver toxicity [22]. 14

16 Since only a very small percentage of the target protein is modified, an excess of 99% of this protein will still be present in its non-modified form, which represents another difficulty in the detection of drug protein adducts. The separation of the modified from non-modified protein is extremely challenging because the small reactive metabolite that has covalently bound to the protein often has a negligible effect on the molecular characteristics of the macromolecule, i.e., mass, charge and size. Conventional separation techniques, such as liquid chromatography (LC), do not provide the required selectivity for this purpose. Capillary electrophoresis (CE) can efficiently separate protein isoforms caused by small modifications. [28-29] Recent improvements of the CE mass spectrometry (MS) interface allows for accurate detection and identification of the separated isoforms and protein modifications. [30] CE MS has also been applied to the analysis of synthetically prepared drug protein conjugates, present at relatively high levels of 5% of the total protein content [31]. Although CE MS possesses the potential for the detection and identification of clinically relevant levels of drug protein adducts of 0.1% or less, this has never been actually demonstrated. 1 Another method of separation is based on thiol affinity resins to achieve the selective enrichment of HSA-cysteine34 adducts. [32] Using this method, non-modified HSA was efficiently removed from a freshly isolated HSA sample through binding to the resin via the available cysteine34. In the remaining non-bound fraction, constituting 2.9% of the total HSA, the cysteine34 residues were modified by reactive systemic electrophiles, i.e., by cysteinylation and glycosylation. Although this methodology is promising for selective extraction of drug-hsa adducts, it remains to be seen if it is applicable to 15-fold or even lower adduct levels, as is the case with drug protein adducts. Separation of the modified from non-modified protein in itself is not straight-forward, but biological tissues and fluids also contain many other proteins that are present in a very wide dynamic range and may mask the presence of the drug protein adduct. Consequently, analytical methodologies for the detection of drug protein adducts must possess a high degree of selectivity and sensitivity. Analytical methods for the detection and identification of drug protein adducts can roughly be divided into global and targeted approaches, see Figure 1.2. Global approaches are usually based on the use of radiolabeled drugs in combination with two-dimensional gel electrophoresis (2D-GE) and bottom-up proteomics, and have been successfully applied to the identification of protein targets of reactive drug metabolites [20, 33-34]. In this way, 42 cytosolic and 24 microsomal protein targets of thiobenzamide (TB) were identified in rat liver. [33] The identified proteins serve a broad range of biological functions, but, in the context of cytotoxicity, several groups could be distinguished, i.e., enzymes of intermediary metabolism and heat shock and stress response proteins. Covalent binding of reactive drug metabolites to these types of proteins may play a role in enzyme inhibition and misfolding of proteins, which may have grave consequences for cell viability or may lead to apoptosis. Similarly, 15 rat liver proteins were identified as likely targets of tienilic acid, 12 of which were previously unidentified targets. [34] Again, the majority of the identified proteins are enzymes that participate in metabolic and catabolic cellular processes and their modification or inhibition may lead to cellular stress that may be used, in combination with the immune system, to explain the mechanism involved in tienilic acid ADRs. 15

17 1 Global analysis Radiolabeled drug In vitro/in vivo metabolism of radiolableled drug in hepatocytes, cells and animal models Targeted analysis Drugs with or without label Samples from in vitro/in vivo experiments, clinical studies, patient samples 2D-gel electrophoresis and radioactivity assay Selective protein extraction from complex matrix Radioactive protein spots are excised, reduced, alkylated and digested in-gel Purified protein is denatured, reduced, alkylated and digested in-solution Analysis of proteolytic peptides with LC ESI-MS or (LC )MALDI-MS Analysis of proteolytic peptides with LC ESI-MS or (LC )MALDI-MS Identification of (adducted) peptides and proteins using protein database search based on known or predicted drug metabolism Identification of (adducted) peptides and protein using protein database search based on known or predicted drug metabolism and adduct formation Identification of protein targets of reactive drug metabolites Identification of site of adduct formation and reactive drug metabolite Figure 1.2 Main approaches for the analysis of drug protein adducts. Currently, more than 540 proteins of various sources are listed in the reactive metabolite target protein database, 358 of which are non-redundant. [24] A considerable overlap in protein targets of different drugs exists, for example protein disulfide-isomerase A3 and glyceraldehyde-3-phosphate dehydrogenase are known targets of the reactive metabolites of APAP, TB, mycophenolic acid (MA) and several other xenobiotics. These three drugs combined have 147 protein targets of which two are shared by all three drugs, a further five targets are shared only by APAP and TB, five other targets are shared only by TB and MA, and three are shared only by APAP and MA, see Figure 1.3. Commonalities in protein targets may imply that the protein adducts of different drug metabolites may affect similar pathways, leading to the same ADRs. More commonalities between target proteins and drugs may be discovered in the future when the methods become more sensitive and repeatable, perhaps by standardization. Hepatic protein targets of reactive drug metabolites have been studied most extensively, since covalent binding is primarily evaluated 16

18 5 APAP TB 50 5 MA 42 Figure 1.3 Overview and commonalities between the protein targets of the reactive metabolites of acetaminophen (APAP), mycophenolic acid (MA) and thiobenzamide (TB). in liver and mostly in animal models. The list of identified proteins, therefore, contains mainly hepatic targets, but may easily be expanded with other protein targets by application of global approaches to other tissues and biofluids. Expansion of the list of protein targets may contribute to the elucidation of the role of drug protein adducts in biological pathways, such as inflammation, and to a better understanding of the mechanism behind ADRs. Unfortunately, radiolabeled drugs can only be used in in vitro experiments or in vivo animal models and this type of approach cannot be applied to patient samples or samples from clinical trials in humans. Additionally, global approaches often do not shed any light about the identity of the drug metabolite or the site of adduct formation due to lack of detection of the proteolytic peptide containing the adduct site. [26] Such detailed characterization of drug protein adducts and unequivocal confirmation that adduct formation has taken place requires a higher degree of selectivity and sensitivity that may be provided by a targeted approach. The in-depth information obtained by this type of approach is crucial for understanding how reactive drug metabolites modify proteins, how the modification of specific amino acid residues may alter the protein function and affects downstream signaling pathways. [35] A targeted adduct-proteomics method is focused on a single protein target that is selectively purified from a complex matrix, e.g., tissue or serum. Selection of potentially interesting protein targets is performed based on information obtained through global approaches using selective labeling, such as radiolabels. Selective purification of the target protein, e.g., by affinity chromatography, is often a prerequisite for qualitative or quantitative proteomic approaches and has proven to be useful for studying low-abundant proteins, as well as protein interactions, e.g. protein-protein complexes, and modifications, such as post-translational modifications (PTMs) and glutathione conjugation. [36-37] With all extraction methods, the selected protein is purified to a certain extent and although there will always be some low-abundant proteins in the 17

19 1 background, the sample complexity and interferences from other proteins will be significantly reduced. The purified protein fraction, i.e., containing both the adducted and non-modified species, is subsequently treated further using conditions specifically chosen for this target in order to maximize the information content obtained from the analysis. A priori knowledge of the number of cysteine residues and the protein sequence enables educated selection and optimization of the treatment conditions, such as reduction, alkylation and digestion of the protein. Trypsin is considered as the golden standard in enzymatic protein digestion due to its unparalleled specificity and efficiency [38], but a lack or over-abundance of tryptic cleavage sites may call for the use of another enzyme or possibly a multiple enzyme digestion [39-40]. Assessment of the proteolytic enzyme to be used in combination with knowledge of the protein sequence and prediction of possible adduct formation sites may lead to improved detection and identification of drug protein adducts. In contrast to the great number of identified target proteins of reactive drug metabolites, only a few drug protein adducts have actually been characterized in vivo. A targeted approach employing albumin affinity chromatography was developed for the identification of NAPQI albumin adducts in human serum [22]. Different enzymatic digestion approaches were developed, using either pronase for NAPQI HSA digestion or trypsin for NAPQI MSA digestion. In another publication, HSA adducts of several sulfur mustards were identified in whole blood after precipitation of the albumin fraction in a multi-step procedure followed by pronase digestion. [41] In all three above mentioned cases, the cysteine34 residue of serum albumin was identified as the site of adduct formation and the respective reactive metabolite or drug could be identified based on the observed shift in the peptide mass. Finally, adducts of flucloxacillin and its 5-hydroxymethyl metabolite primarily to lysine190 and lysine212 residues of serum albumin were successfully identified in plasma samples of patients treated with this synthetic penicillin using albumin affinity chromatography and peptide mass mapping. [42] In addition, more adduct modification sites on different proteins have been characterized in detail using other xenobiotics, such as bromobenzene (BB) [26], acrylonitrile [43] and acrylamide [44]. For example, bromobenzoquinone adducts to cysteine-111 of GST-A1 and -A2 were identified in liver tissue samples obtained from a BB metabolism study performed in rats. [26] For this purpose, the researchers used 14 C-labeled BB and developed an analytical strategy employing GST affinity chromatography and further separation of the GST subclasses by reversed-phase HPLC, tryptic digestion and performed the identification by a combination of 1D-GE, intact protein MS and peptide mass mapping. These potentially relevant target proteins of other xenobiotics may aid in the identification of targets of reactive drug metabolites. Whereas identification of protein targets in global approaches is more efficiently achieved by protein database searching, in targeted approaches manual searches may be performed based on a priori predicted adducts. The structure and mass of many reactive metabolites were identified in numerous drug bioactivation studies, predominantly performed in vitro using (mutant) cytochrome P450 enzymes. [45-46] Subsequent characterization of the covalent binding of reactive metabolites to clinically relevant proteins, such as GST [47], greatly aids the detection and identification of drug protein adducts in vivo. Information regarding potential adduct formation sites on proteins may also be obtained from selective labeling of protein mixtures, for instance using radiolabels, as discussed earlier. The enhanced selectivity of this type of strategy and the information obtained from its application may contribute to the confirmation 18

20 of drug protein adduct formation in vivo. Therefore, the two types of approaches for drug protein adduct identification described here are not mutually exclusive, in fact, their power lies in their combined or successive use and the increased level of information gained from their combined results. The identified potential protein targets of (labeled) reactive (drug) metabolites in global approaches are the starting point of targeted approaches where actual drug protein adduct confirmation may be achieved. The combination of these two methods is very powerful and may be the only route to understanding the mechanisms behind ADRs and how they are linked to drug protein adducts. 1 As stated earlier, some ADRs are dose-dependent, usually meaning that an overdose of a drug will cause the adverse reaction, while others are assumed not to be related to the dose. The latter may imply that an individual may respond adversely to a drug due to genetic predisposition and/or that even a small dose may lead to severe ADRs. Therefore, the quantitative relationship between ADRs and drug protein adduct formation requires scientific attention. Quantification of protein biomarkers in vivo as readout of biological response in clinical proteomics has been the subject of research for many years. [48] Indeed, up- or down-regulation of proteins presented in a patient may be an indication for a disease state [49], e.g. HER2/NEU for staging of breast cancer [50]. The measurement of certain biomarkers, such as serum albumin in plasma as a marker for (mal)nutrition and other disease states, is straightforward and routinely performed in the clinical setting. [51] However, the readout of other protein biomarkers may require more complex proteomics-based quantification approaches involving MS techniques, such as selected reaction monitoring (SRM), and/or isotopic labeling strategies, such as isotope coded affinity tags (ICAT) and isobaric tags for relative and absolute quantification (itraq), that are not yet standardized or unsuited for clinical use. [49, 52] Additionally, the inter- (during the day) and intra-individual biological variation, the wide dynamic range of proteins in biological samples, and the (absence of) linearity of the relationship between the disease and biomarker are hurdles that still have to be overcome in the analysis and quantification of biomarkers. [52-53] In recent years, changes in the post-translational modification of proteins have been implied as a possible biomarker for disease states, such as cancer [54-55] and Alzheimer s disease [56-57] Determination of the level of PTMs of proteins may be useful for early detection of a disease and following its progress, but is extremely challenging due to absence of standard reference peptides, inefficiency of isotopic labeling, and large variation. [58] Many of the challenges associated with the quantification of proteins and protein PTMs also apply to drug protein adducts. To our best knowledge, only one method for drug protein adduct quantification in vivo has previously been reported in literature, specifically for the NAPQI HSA adduct [22]. Damsten et al. utilize a synthetic reference peptide approach for absolute quantification, where a NAPQI HSA pronase digest sample spiked with the synthetic NAPQI CPF peptide is compared to a non-spiked sample. [22] A NAPQI CPF serum concentration 35 pmol/ml was detected in a sample from a patient that had taken an APAP overdose of 40 g, while for two other patients having taken 10 or 12 g of APAP this level was extrapolated to be 3-4 pmol/ml serum. From these results, it appears that the NAPQI CPF serum concentration is related to the dose, but this relationship may not be linear. Absolute quantification has advantages over other approaches, however, the one-point calibration described in this paper is generally not very accurate, but could easily be improved by extending it to a standard addition 19

21 1 method. Application of the above described methodology resulted in the detection of extremely low concentrations of the NAPQI HSA adduct in serum, but could not establish a clear relationship to the dose. Taking into account that the subjects in these studies received relatively high drug (over)doses means that the detection and quantification methods for drug protein adducts in vivo require further improvement of the detection limit and sensitivity. Especially in cases where severe ADRs occur at therapeutic doses of a drug, the drug protein adducts may be present at even lower levels that may be undetectable with the current methodologies. To summarize, more research is still needed to unravel the precise role of drug protein adducts in the mechanisms behind ADRs. Their analysis, especially in the low-abundant protein fractions of biological tissues and fluids, is challenging, but essential for identification of the affected biological pathways involved in adverse responses. The development of analytical methodologies specifically for drug protein adduct detection, identification and quantification is indispensible for gaining a better understanding of ADRs. 1.3 SCOPE OF THE THESIS The work in this thesis was performed within the framework of project D3-201 Towards novel translational safety biomarkers for adverse drug toxicity of the Dutch Top Institute Pharma. The project consortium, consisting of five academic and seven industrial partners, studied the relationship between drug bioactivation and immune system-dependent adverse drug reactions. The research groups performed translational research in their respective expertise areas aimed at correlation of pre-clinical in vitro and in vivo models to ADR patient data. This research includes: 1) Mechanistic studies of drug-induced liver injury through investigation of intracellular stress signaling using high content imaging assays, 2) Characterization of reactive drug metabolites through low-molecular weight trapping agents and adduct formation to selected proteins using in vitro bioactivation systems, 3) Precision-cut liver slices as an ex vivo model for translation of human data to in vitro human and mouse in vitro and in vivo models, 4) Validation of mechanism-based mouse in vivo models of drug-induced hepatotoxicity and systemic immunosensitization, and 5) Identification of predictive biomarkers of ADRs in plasma and urine of ADR patients through protein profiling. This thesis is focused on the development and optimization of proteomics-based methods for the detection and identification of drug protein adducts in vivo to be applied to samples from the other research groups. Due to the challenges involved in the detection of low-abundant drug protein adducts in complex biological samples, sample preparation plays a crucial role. Protein digestion is a fundamental part of bottom-up proteomics, which is the method of choice for the analysis of proteins and their modifications. An overview of digestion strategies and recent novel developments in this area is given in the review in Chapter 2. 20

22 The analysis of low-abundant proteins and their modifications in general and, more relevant to the current context, drug protein adducts can benefit considerably from efficient sample preparation protocols. Chapter 3 describes the optimization of sample pretreatment and digestion conditions specifically for enhanced detection of cysteine34 adducts of HSA. In this study, an experimental design approach was applied to the optimization of tryptic and thermolytic digestion of a model HSA adduct, which led to optimum conditions that differed from those found in literature, but performed significantly better. 1 Sample preparation for bottom-up proteomics consists of multiple treatment steps, prior to digestion, including protein purification, denaturation, reduction, alkylation and multiple sample clean-up steps. Especially for large sample sets, these workflows are often very laborious and time consuming, in addition to the high consumption of chemicals and reagents. Chapter 4 concerns the development of a high-throughput sample preparation method for bottom-up proteomics employing 96-well filter plates that reduces sample transfer to a minimum and allows for the simultaneous preparation of 96 samples. Chapter 5 describes the application of a targeted approach for identification of albumin adducts, incorporating the optimized sample preparation and digestion conditions, to mouse serum samples resulting from an APAP exposure experiment to study the kinetics of ADRs caused by this drug. An adduct quantification methodology was developed for effective evaluation of the relationship between the ADRs and the formation of NAPQI MSA adducts. Chapter 6 presents a summary of this thesis as well as a discussion of the conclusions and directions for further research in the context of analytical challenges. REFERENCES [1] Scannell, J.W., Blanckley, A., Boldon, H., Warrington, B., Diagnosing the decline in pharmaceutical R&D efficiency. Nat. Rev. Drug Discov. 2012, 11, (3), [2] Dickson, M., Gagnon, J.P., Key factors in the rising cost of new drug discovery and development. Nat. Rev. Drug Discov. 2004, 3, (5), [3] Baillie, T.A., Future of toxicology-metabolic activation and drug design: challenges and opportunities in chemical toxicology. Chem. Res. Toxicol. 2006, 19, (7), [4] Evans, D.C., Watt, A.P., Nicoll-Griffith, D.A., Baillie, T.A., Drug protein adducts: an industry perspective on minimizing the potential for drug bioactivation in drug discovery and development. Chem. Res. Toxicol. 2004, 17, (1), [5] Rawlins, M.D., Clinical pharmacology. Adverse reactions to drugs. Br. Med. J. (Clin. Res. Ed.). 1981, 282, (6268), [6] Edwards, I.R., Aronson, J.K., Adverse drug reactions: definitions, diagnosis, and management. Lancet. 2000, 356, (9237), [7] Royer, R.J., Mechanism of action of adverse drug reactions: an overview. Pharmacoepidemiol. Drug Saf. 1997, 6 Suppl 3, S [8] Hartigan-Go, K.J., Wong, J.Q., Inclusion of therapeutic failures as adverse drug reactions, Elsevier, Amsterdam, [9] Uetrecht, J., Idiosyncratic drug reactions: Current understanding. Annu. Rev. Pharmacol. 2007, 47, [10] Uetrecht, J., Idiosyncratic drug reactions: Past, present, and future. Chem. Res. Toxicol. 2008, 21, (1),

23 1 [11] Jollow, D.J., Mitchell, J.R., Potter, W.Z., Davis, D.C., Gillette, J.R., et al., Acetaminophen-induced hepatic necrosis. II. Role of covalent binding in vivo. J. Pharmacol. Exp. Ther. 1973, 187, (1), [12] Mitchell, J.R., Jollow, D.J., Gillette, J.R., Brodie, B.B., Drug metabolism as a cause of drug toxicity. Drug Metab. Dispos. 1973, 1, (1), [13] Zhou, S., Chan, E., Duan, W., Huang, M., Chen, Y.Z., Drug bioactivation, covalent binding to target proteins and toxicity relevance. Drug. Metab. Rev. 2005, 37, (1), [14] Kevin Park, B., Coleman, J.W., Kitteringham, N.R., Drug disposition and drug hypersensitivity. Biochem. Pharmacol. 1987, 36, (5), [15] Pichler, W.J., Pharmacological interaction of drugs with antigen-specific immune receptors: the p-i concept. Curr. Opin. Allergy Clin. Immunol. 2002, 2, (4), [16] Pichler, W.J., Beeler, A., Keller, M., Lerch, M., Posadas, S., et al., Pharmacological interaction of drugs with immune receptors: the p-i concept. Allergol. Int. 2006, 55, (1), [17] Streeter, A.J., Dahlin, D.C., Nelson, S.D., Baillie, T.A., The covalent binding of acetaminophen to protein. Evidence for cysteine residues as major sites of arylation in vitro. Chem. Biol. Interact. 1984, 48, (3), [18] Landin, J.S., Cohen, S.D., Khairallah, E.A., Identification of a 54-kDa mitochondrial acetaminophen-binding protein as aldehyde dehydrogenase. Toxicol. Appl. Pharmacol. 1996, 141, (1), [19] Pumford, N.R., Halmes, N.C., Martin, B.M., Cook, R.J., Wagner, C., et al., Covalent binding of acetaminophen to N-10-formyltetrahydrofolate dehydrogenase in mice. J. Pharmacol. Exp. Ther. 1997, 280, (1), [20] Qiu, Y., Benet, L.Z., Burlingame, A.L., Identification of the hepatic protein targets of reactive metabolites of acetaminophen in vivo in mice using two-dimensional gel electrophoresis and mass spectrometry. J. Biol. Chem. 1998, 273, (28), [21] Zhou, L., McKenzie, B.A., Eccleston, E.D., Jr., Srivastava, S.P., Chen, N., et al., The covalent binding of [14C] acetaminophen to mouse hepatic microsomal proteins: the specific binding to calreticulin and the two forms of the thiol:protein disulfide oxidoreductases. Chem. Res. Toxicol. 1996, 9, (7), [22] Damsten, M.C., Commandeur, J.N., Fidder, A., Hulst, A.G., Touw, D., et al., Liquid chromatography/tandem mass spectrometry detection of covalent binding of acetaminophen to human serum albumin. Drug. Metab. Dispos. 2007, 35, (8), [23] Labenski, M.T., Fischer, A.A., Lo, H.H., Monks, T.J., Lau, S.S., Protein electrophile-binding motifs: Lysine-rich proteins are preferential targets of quinones. Drug Metab. Dispos., 2009, 37, (6), [24] Hanzlik, R.P., Koen, Y.M., Theertham, B., Dong, Y., Fang, J., The reactive metabolite target protein database (TPDB)--a web-accessible resource. BMC Bioinformatics. 2007, 8, 95. [25] Park, B.K., Laverty, H., Srivastava, A., Antoine, D.J., Naisbitt, D., et al., Drug bioactivation and protein adduct formation in the pathogenesis of drug-induced toxicity. Chem. Biol. Interact. 2011, 192, (1-2), [26] Koen, Y.M., Yue, W., Galeva, N.A., Williams, T.D., Hanzlik, R.P., Site-specific arylation of rat glutathione s-transferase A1 and A2 by bromobenzene metabolites in vivo. Chem. Res. Toxicol. 2006, 19, (11), [27] Bessems, J.G., Vermeulen, N.P., Paracetamol (acetaminophen)-induced toxicity: molecular and biochemical mechanisms, analogues and protective approaches. Crit. Rev. Toxicol. 2001, 31, (1), [28] Ryvolova, M., Adam, V., Eckschlager, T., Stiborova, M., Kizek, R., Study of DNA-ellipticine interaction by capillary electrophoresis with laser-induced fluorescence detection. Electrophoresis. 2012, 33, (11), [29] Taichrib, A., Pioch, M., Neususs, C., Toward a screening method for the analysis of small intact proteins by CE- ESI-TOF MS. Electrophoresis. 2012, 33, (9-10), [30] Haselberg, R., de Jong, G.J., Somsen, G.W., Capillary electrophoresis-mass spectrometry for the analysis of intact proteins Electrophoresis. 2011, 32, (1), [31] Haselberg, R., Harmsen, S., Dolman, M.E., de Jong, G.J., Kok, R.J., et al., Characterization of drug-lysozyme conjugates by sheathless capillary electrophoresis-time-of-flight mass spectrometry. Anal. Chim. Acta. 2011, 698, (1-2), [32] Funk, W.E., Li, H., Iavarone, A.T., Williams, E.R., Riby, J., et al., Enrichment of cysteinyl adducts of human serum albumin. Anal. Biochem. 2010, 400, (1), [33] Ikehata, K., Duzhak, T.G., Galeva, N.A., Ji, T., Koen, Y.M., et al., Protein targets of reactive metabolites of thiobenzamide in rat liver in vivo. Chem. Res. Toxicol. 2008, 21, (7), [34] Methogo, R.M., Dansette, P.M., Klarskov, K., Identification of liver protein targets modified by tienilic acid metabolites using a two-dimensional Western blot-mass spectrometry approach. Int. J. Mass Spectrom. 2007, 268, (2-3),

24 [35] Park, K.B., Dalton-Brown, E., Hirst, C., Williams, D.P., Selection of new chemical entities with decreased potential for adverse drug reactions. Eur. J. Pharmacol. 2006, 549, (1-3), 1-8. [36] Azarkan, M., Huet, J., Baeyens-Volant, D., Looze, Y., Vandenbussche, G., Affinity chromatography: a useful tool in proteomics studies. J. Chromatogr. B. 2007, 849, (1-2), [37] Lee, W.C., Lee, K.H., Applications of affinity chromatography in proteomics. Anal. Biochem. 2004, 324, (1), [38] Burkhart, J.M., Schumbrutzki, C., Wortelkamp, S., Sickmann, A., Zahedi, R.P., Systematic and quantitative comparison of digest efficiency and specificity reveals the impact of trypsin quality on MS-based proteomics. J. Proteomics. 2012, 75, (4), [39] Swaney, D.L., Wenger, C.D., Coon, J.J., Value of Using Multiple Proteases for Large-Scale Mass Spectrometry- Based Proteomics. J. Proteome Res. 2010, 9, (3), [40] Wisniewski, J.R., Mann, M., Consecutive proteolytic digestion in an enzyme reactor increases depth of proteomic and phosphoproteomic analysis. Anal. Chem. 2012, 84, (6), [41] Noort, D., Hulst, A.G., Jansen, R., Covalent binding of nitrogen mustards to the cysteine-34 residue in human serum albumin. Arch. Toxicol. 2002, 76, (2), [42] Jenkins, R.E., Meng, X., Elliott, V.L., Kitteringham, N.R., Pirmohamed, M., et al., Characterisation of flucloxacillin and 5-hydroxymethyl flucloxacillin haptenated HSA in vitro and in vivo. Proteomics Clin. Appl. 2009, 3, (6), [43] Nerland, D.E., Cai, J., Pierce, W.M., Jr., Benz, F.W., Covalent binding of acrylonitrile to specific rat liver glutathione S-transferases in vivo. Chem. Res. Toxicol. 2001, 14, (7), [44] Barber, D.S., Stevens, S., LoPachin, R.M., Proteomic analysis of rat striatal synaptosomes during acrylamide intoxication at a low dose rate. Toxicol. Sci. 2007, 100, (1), [45] Damsten, M.C., van Vugt-Lussenburg, B.M., Zeldenthuis, T., de Vlieger, J.S., Commandeur, J.N., et al., Application of drug metabolising mutants of cytochrome P450 BM3 (CYP102A1) as biocatalysts for the generation of reactive metabolites. Chem. Biol. Interact. 2008, 171, (1), [46] Rea, V., Dragovic, S., Boerma, J.S., de Kanter, F.J., Vermeulen, N.P., et al., Role of residue 87 in the activity and regioselectivity of clozapine metabolism by drug-metabolizing CYP102A1 M11H: application for structural characterization of clozapine GSH conjugates. Drug. Metab. Dispos. 2011, 39, (12), [47] Boerma, J.S., Vermeulen, N.P., Commandeur, J.N., Application of CYP102A1M11H as a tool for the generation of protein adducts of reactive drug metabolites. Chem. Res. Toxicol. 2011, 24, (8), [48] Qian, W.J., Jacobs, J.M., Liu, T., Camp, D.G., 2nd, Smith, R.D., Advances and challenges in liquid chromatographymass spectrometry-based proteomics profiling for clinical applications. Mol. Cell. Proteomics. 2006, 5, (10), [49] Sparbier, K., Wenzel, T., Dihazi, H., Blaschke, S., Muller, G.A., et al., Immuno-MALDI-TOF MS: new perspectives for clinical applications of mass spectrometry. Proteomics. 2009, 9, (6), [50] Ludwig, J.A., Weinstein, J.N., Biomarkers in cancer staging, prognosis and treatment selection. Nat. Rev. Cancer. 2005, 5, (11), [51] Margarson, M.P., Soni, N., Serum albumin: touchstone or totem? Anaesthesia. 1998, 53, (8), [52] Simpson, K.L., Whetton, A.D., Dive, C., Quantitative mass spectrometry-based techniques for clinical use: biomarker identification and quantification. J. Chromatogr. B. 2009, 877, (13), [53] Thomas, C.E., Sexton, W., Benson, K., Sutphen, R., Koomen, J., Urine Collection and Processing for Protein Biomarker Discovery and Quantification. Cancer Epidem. Biomar. 2010, 19, (4), [54] He, Y., Korboukh, I., Jin, J., Huang, J., Targeting protein lysine methylation and demethylation in cancers. Acta. Biochim. Biophys. Sin. (Shanghai). 2012, 44, (1), [55] Wang, Z.Q., Protein S-nitrosylation and cancer. Cancer Lett. 2012, 320, (2), [56] Francis, Y.I., Fa, M., Ashraf, H., Zhang, H., Staniszewski, A., et al., Dysregulation of histone acetylation in the APP/ PS1 mouse model of Alzheimer s disease. J. Alzheimers Dis. 2009, 18, (1), [57] Zhang, K.L., Schrag, M., Crofton, A., Trivedi, R., Vinters, H., et al., Targeted proteomics for quantification of histone acetylation in Alzheimer s disease. Proteomics. 2012, 12, (8), [58] Farley, A.R., Link, A.J., Identification and quantification of protein posttranslational modifications. Methods Enzymol. 2009, 463,

25

26 Chapter 2 Protein digestion: An overview of the available techniques and recent developments Linda Switzar, Martin Giera and Wilfried M.A. Niessen Journal of Proteome Research, 2013, 12 (3),

27 ABSTRACT 2 Several proteomics approaches are available that are defined by the level (protein or peptide) at which analysis takes place. The most widely applied method still is bottom-up proteomics where the protein is digested into peptides that can be efficiently analyzed with a wide range of LC MS or MALDI-TOF MS instruments. Sample preparation for bottom-up proteomics experiments requires several treatment steps in order to get from the protein to the peptide level and can be very laborious. The most crucial step in such approaches is the protein digestion, which is often the bottleneck in terms of time consumption. Therefore, a significant gain in throughput may be obtained by speeding up the digestion process. Current techniques allow for reduction of the digestion time from overnight (~15 h) to minutes or even seconds. This advancement also makes integration into online systems feasible, thereby reducing the number of tedious sample handling steps and the risk of sample loss. In this review, an overview is given of the currently available digestion strategies and recent developments in the acceleration of the digestion process. Additionally, tailored approaches for classes of proteins that pose specific challenges are discussed. Protein (mixture) Digestion Peptide separation a 1 b 2 c 2 H O H O H 2 N C C NH C C OH R 1 R 2 x 1 y 2 z 2 Intensity VAHRFKDLGE m/z Time Database search MS/MS LC MS Intensity MS analysis 26

28 2.1 INTRODUCTION Protein digestion, either enzymatically or nonenzymatically, is an important and (almost) indispensable tool in protein identification, characterization, and quantification by proteomics strategies. [1] Proteomics plays a vital role in major research areas, including disease biomarker discovery and systems biology, and, as such, significantly contributes to the understanding of biological processes that are essential for life. [2] Selection of the proteomics approach should be based on the type of question to be answered. Global proteomics, such as finding a biomarker in a highly complex sample, concerns the identification of an often low abundant, unknown protein that is present in a complex sample with a wide dynamic range of proteins. This requires a very different approach than the targeted analysis of a protein, for example, characterization of post-translational modification (PTM) states, where a detailed investigation and complete sequence mapping of a known protein is required for localization of the possibly low-abundant PTMs. Additionally, (classes of) proteins with special characteristics may pose specific requirements in their analysis. 2 Proteomics approaches can be discriminated by the level at which analysis takes place (see Figure 2.1). Advances in mass spectrometry (MS) instrumentation now allow for the direct analysis of proteins. In such a so-called top-down experiment, purified proteins are detected intact and following fragmentation using collisional-activated dissociation (CAD), electron-capture dissociation (ECD), or electron transfer dissociation (ETD), providing information on intact protein mass and amino acid sequence. [3-4] Top-down analysis of intact proteins reduces sample preparation to a minimum and preserves information that is sometimes lost in other proteomics strategies, such as the connectivity of multiple PTMs, but is relatively insensitive [5]. Due to the large size of the analytes, the requirements for MS instruments in terms of resolution and mass accuracy are only provided by high-end mass spectrometers such as the Fourier transform ion-cyclotron resonance MS (FTICR MS) and Orbitrap MS. [6] However, intact protein analyses have also been reported using more accessible instruments, such as the tandem quadrupole and quadrupole-time-of-flight (QTOF) MS. [7] In practice, the mass range of proteins that can be analyzed using top-down proteomics is limited to ~50 kda, thus approximately 500 amino acids. [6] Otherwise, only the C- and N-termini are sequenced. [8-9] Despite the clear advantages of top-down proteomics, further development of MS instrumentation is necessary before it will become a mainstream technique. The vast majority of proteomics experiments rely on digestion of the protein into peptides prior to MS analysis, which is the main focus of this review. The analysis of peptides has several advantages over proteins, including more efficient separation by liquid chromatography (LC), a lower molecular mass and fewer charge states, leading to improvements in sensitivity. [10] Depending on the size of the produced peptides, the approach is referred to as either bottom-up proteomics or middle-down proteomics. In a bottom-up strategy, the protein is digested to peptides within the range of ~ Da. These peptides are subsequently analyzed with liquid chromatography electrospray ionization MS (LC ESI-MS) or matrix-assisted laser desorption ionization time-of-flight MS (MALDI-TOF MS). Protein identification is performed based on peptide mass fingerprinting or peptide sequence analysis. [3] Sequencing of the peptides can effectively be achieved by collision-induced dissociation (CID) in more widely available ESI-ion trap and ESI-QTOF mass spectrometers or using MALDI-TOF/TOF instruments. [11] 27

29 2 Intensity m/z m/z MS/MS Protein sequence MS Intact protein mass Intensity m/z Time Intensity Top-down proteomics Middle-down proteomics Bottom-up proteomics Protein mixture (Mass 50 kda) Protein mixture (No mass limit) Protein mixture (No mass limit) Digestion Digestion Separation of proteins MS analysis of intact protein ( 50 kda) Intensity Separation of peptides MS analysis of peptides (~ Da) Intensity Separation of peptides MS analysis of peptides (~ Da) Figure 2.1 Overview of the proteomics approaches. MS/MS Peptide sequences LC MS Intact peptide masses Intensity m/z Time MS/MS Peptide sequences LC MS Intact peptide masses 28

30 Digestion of a complex protein sample, such as a whole proteome, with a bottom-up approach produces a vast amount of peptides, more than even the most efficient instrument can analyze. The peptide sample complexity can be reduced without compromising the information content by producing fewer, but larger peptides. This, so-called middle-down proteomics approach combines the best of top-down and bottom-up by taking advantage of the improvements in MS instrumentation and the availablity of electronbased fragmentation methods [12], while retaining the level of sensitivity associated with the analysis of peptides [5]. The mid-range peptides (~ Da) show improved separation by LC, [13] and after ESI carry a higher number of charges, which enhances fragmentation by CID [13], ETD [14] or ECD [15] in Orbitrap MS, quadrupole-linear ion trap (QTrap) MS and quadrupole-fticr MS instruments. In comparison with smaller peptides, more confident peptide identifications are obtained, leading to improved protein sequence coverage and identification of PTMs. [16] 2 The predictable nature of peptide fragmentation allows matching of experimental MS/MS spectra with predicted spectra from in silico digestion of known protein sequences to establish protein identity. [17] In that respect, advanced bioinformatics tools play a pivotal role in any proteomics strategies. However, these tools rely on the assumption that the digestion process (including reduction and alkylation of disulfide bridges) is optimal. If this is not the case, for example, when peptides are connected via an intact disulfide bridge, unpredicted peptides and/or complex fragmentation spectra are generated that will not be identified in an automated database search. Digestion of proteins may result in a loss of information, such as, the presence and connectivity of PTMs [7] or the ability to distinguish closely related proteins, because of failure in detection of certain parts of the protein sequence due to inadequate size or unfavorable ionization properties of the generated peptides. Finally, the quality of the obtained protein identifications and modifications can be monitored via false discovery rates, protein and peptide score thresholds, but this requires a critical review of the obtained results. Protein digestion is an essential step in both bottom-up and middle-down proteomics strategies and has a large influence on the quality of protein identification. [18] Over the years, protein digestion has been improved through the development of novel techniques in order to increase throughput and reproducibility. This paper provides an overview of the available protein digestion techniques and reviews recent developments in protein digestion for proteomics. 2.2 PROTEIN DIGESTION The classical approaches for protein digestion are enzymatic digestion, involving proteolytic enzymes, and nonenzymatic digestion, utilizing chemicals, and are mostly performed in-solution or in-gel. A wide range of proteolytic enzymes with varying cleavage specificities and efficiencies is available for enzymatic digestion to which new enzymes are added regularly. Chemical digestion mainly utilizes acids and small chemical reagents, but recently also instrumental techniques have been introduced as a method for nonenzymatic digestion. An overview of the available approaches is given in this chapter. 29

31 2.2.1 Enzymatic digestion 2 The most widely applied method for protein digestion involves the use of enzymes. Many proteases are available for this purpose, each having their own characteristics in terms of specificity, efficiency and optimum digestion conditions. Trypsin is most widely applied in bottom-up proteomics and can be considered as the gold standard in proteomics. Over the years, this enzyme has been modified to a highly efficient and autolysis-resistant protease. [19] It is available in large quantities at low cost and has a very high degree of specificity, cleaving the peptide bonds C-terminal to the basic residues Lys and Arg, except when followed by Pro. [20] Lys and Arg are relatively abundant amino acids in the human proteome (see Figure 2.2) and are usually well distributed throughout a protein. [18] This leads to tryptic peptides with an average length of ~14 amino acids that carry at least two positive charges, which is ideally suited for CID-MS analysis. [20] The advantageous properties of tryptic peptides lead to high quality MS/MS fragmentation spectra and confident peptide identification in protein database searches. This in turn increases the accuracy of inference of protein identity. Standardized protocols have been described for in-solution and in-gel protein digestion by trypsin (and other proteases). [21] A typical protocol involves denaturation of the protein using chaotropic agents like urea or guanidine, reduction of disulfide bridges using dithiothreitol (DTT), and subsequent alkylation of the cysteines by iodoacetic acid or iodoacetamide. After reagent removal and buffer exchange, the trypsin digestion is typically performed at neutral ph in an ammonium bicarbonate buffer at 37 C. Depending on the way the digestion is performed, it may take up to 18 h (overnight digestion). The Frequency (%) L S E A G P V K R T Q D I N F Y H C M W Amino acid Figure 2.2 Amino acid composition of the human proteome. Amino acid abundances were obtained from the proteome analysis database. [47] 30

32 experimental conditions for trypsin digestion can be optimized for a specific application, for instance using a design of experiments approach. [22] The digestion is stopped by the addition of (formic) acid. Such digestion protocols are often very laborious and require many sample handling steps. High-throughput 96-well formats allow for automation of sample treatment steps via robotics. [23] Sample preparation for MALDI-MS analysis can also be easily automated by depositing (separated) proteins directly onto the MALDI target plate for on-plate digestion. [24] In-solution digestion may also be integrated into an online digestion LC system to increase throughput and reduce sample handling. [25] In addition, digestion may be achieved via immobilized enzyme reactors (IMERs) [26] or other formats, which are discussed below. Despite the many advantages of trypsin, it may be necessary to use other proteases in specific cases, such as a lack or an over-abundance of Lys and Arg in the protein sequence or ph incompatibility. A wide range of alternative proteases are available with different cleavage specificities, see Table 2.1. For instance, aspartic proteases like pepsin are active under acidic conditions and are thus often utilized in hydrogen/deuterium exchange (HDX) experiments that favor a low ph. [27-28] Pepsin proteases are less specific, but they allow reproducible digestion when five-to-six replicate digestions are performed. [28] However, a compromise is made due to the wider specificity or even nonspecific nature of the protease. Less specific proteases may lead to a larger number of peptides, even with overlapping amino acid sequences, which may result in more complex MS data. In addition, many smaller peptides may be formed, which are more difficult to annotate and are thus of little use for protein identification. 2 In contrast, the endoproteinases Arg-C, Asp-N, Glu-C and Lys-C provide high cleavage efficiency and specificity, and are often used as an alternative to trypsin. Like trypsin, Arg-C and Lys-C have the advantage to retain at least two basic amines in the peptide, N-terminal and Lys or Arg side chain, leading to doubly protonated peptides. Lys-N is the most recent addition to this group. [35] It has lower specificity and more observed missed cleavages than Lys-C, but it can be used under severe denaturing conditions, such as, elevated temperatures, in the presence of 8 M urea or 80% acetonitrile. [36] Since the endoproteinases have a high selectivity for a single residue, they are often employed in middle-down proteomics. Global proteome and phosphoproteome analysis of whole cell lysates using Lys-N, strong cation exchange enrichment and ETD with supplemental collisional activation may serve as an example of this. [37] As an alternative, limited proteolysis using common enzymes, for example, rapid digestion with trypsin, chymotrypsin or pepsin for only a few minutes, can be used to produce large protein fragments. [38] The recently introduced outer membrane protease T (OmpT), which cleaves specifically between two consecutive basic amino acid residues (Lys/Arg-Lys/Arg) and produces larger peptides (on average >6.3 kda) than other enzymes, is also ideally suited for middle-down proteomics. [39] In addition, highly specific enzymes for a single (class of) target protein, such as the immunoglobulin-degrading enzyme of Streptococcus pyogenes (IdeS). [40] This bacterial cysteine protease specifically cleaves immunoglobulin G (IgG) under its hinge domain and cleaves the heavy chain into two fragments [41], whereas the streptococcal cysteine proteinase streptococcal exotoxin B (SpeB, from the same bacterium) cleaves the heavy chains of all human immunoglobulins [42]. IdeS digestion of IgG results in three protein fragments of ~25 kda (the light chain, and the VH he1 and 2 nd3 domains of the heavy chain) that could easily be separated and characterized by LC ESI-QTOF MS. 31

33 2 Table 2.1 Commonly used proteases and chemicals for protein digestion. Enzyme Organism Specificity ph range Chemical Specificity ph range acidic acidic acidic 2.0 d 9-10 e 9.0 f M D c D D d C d N G CNBr HAc FA HCl NTCB Hydroxylamine a a a a 8.0 b 8.0 a a 1.3 > g a a R D E a K Kb K, R F,W,Y F,L,W,Y F,L A,F,I,L,M,V R,K,D,H,G,Y a A,E,F,I,L,T,V,W,Y Clostridium histolyticum Pseudomonas fragi Staphylococcus aureus Lysobacter enzymogenes Lysobacter enzymogenes Bos taurus Bos taurus Sus scrofa Arg-C Asp-N Glu-C Lys-C Lys-N Trypsin Chymotrypsin Pepsin Bacillus thermoproteolyticus Carica papaya Streptomyces griseus Thermolysin Papain Pronase All data obtained from the Expasy bioinformatics resource portal [29] ( except a Roche website ( b Raijmakers et al. [30], c Swatkoski et al. [31], d Smith [32], e Tang et al. [33], f Crimmins et al. [34] and g Sigma-Aldrich website ( 32

34 2.2.2 Nonenzymatic digestion Chemical cleavage is an alternative to enzymatic digestion. It can be achieved by treatment with dilute solutions of formic acid (FA) [43], hydrochloric acid (HCl) [44], or acetic acid (HAc) [31], or with other chemicals such as cyanogen bromide (CNBr) [45], 2-nitro-5-thiocyanobenzoate (NTCB) [33], and hydroxylamine [46]. These reagents have a high specificity for a single cleavage site (see Table 2.1) with similar abundances to Lys and Arg (in the case of Asp) or much lower (Met or Cys), see Figure 2.2. [47] Although the abundance is not necessarily correlated to the distribution of these residues in the sequence, chemical digestion may lead to high-mass peptides that are suitable for middle-down proteomics. [39] Electrochemical oxidation is one of the latest additions to the array of nonenzymatic cleavage strategies, resulting in specific cleavage at Tyr and Trp. [48-49] Both these residues are low abundant in the human proteome, see Figure 2.2, at less than half of the occurrence of Lys and Arg in the entire Swiss-Prot database. In silico digestion of the proteins in the SwissProt database showed that the average size of electrochemically generated peptides would be 2.4 kda [48], thus ideally suited for analysis with a wide range of LC MS instruments. Clear advantages of electrochemical cleavage of proteins are the speed of the reaction (minutes) and the possibility for online coupling to MS. [48] Electrochemistry, eventually online with MS, may also be used in the reduction of disulfide bridges, and thus may aid in achieving higher protein coverage. [50] Multiple digestion strategies Multiple enzyme digestion is a strategy to increase protein and proteome coverage through the utilization of a combination of proteases. The combined, parallel or successive use of multiple enzymes has been suggested as the only way to realize 100% sequence coverage. [51] Especially in whole proteome sequencing, the use of the multiple enzymes in parallel has significantly improved the number of identified peptides and proteins. [52] In this way, different portions of the proteome can be made visible, as demonstrated for complex samples like cerebrospinal fluid [53] and plasma [54]. Several combinations of enzymes have been applied, although mostly trypsin is used in combination with an endoproteinase, that is, Lys-C [54], Lys-N [55] and Glu-C [53], or less specific enzymes [56]. Similar to multiple enzyme digestion, combinations of chemical and enzymatic treatments have also been reported and were found especially useful for the analysis of membrane proteins (see below). [57-59] 2.3 ACCELERATED DIGESTION TENIQUES Bottom-up proteomics sample preparation protocols are often lengthy procedures, also requiring a number of pretreatment steps prior to protein digestion (see above). Although the sample preparation process can be automated, the digestion step remains the bottleneck in terms of time consumption. The advantages and drawbacks of various approaches for acceleration of bottom-up proteomics workflows have been 33

35 2 Table 2.2 Overview of techniques for accelerated digestion. Accelerated technique Digestion time Online High temperature Microwave Minutes (~15) Minutes ( 15) Not done, but possible Possible Ultrasound Specific applications Compatibility Wide application area Membrane proteins (increased solubility), glycoproteins (decreased sterical hindrance) Wide application area Often applied to chemical digestion, not all proteases are thermostable Compatible with proteases and chemical cleavage reagents Compatible with proteases and chemical cleavage reagents Mostly done with enzymes HDX experiments (speed of online digestion-ms) Wide application area Membrane proteins (increased solubility) Wide application area Wide application area Wide application area Only advantageous for enzymes due to increased interaction with protein Chemical digestion is often done in the presence of solvents, but some enzymes also tolerate relatively high percentages of organic solvent Compatible with each protease that retains activity when immobilized Compatible with each protease that retains activity when immobilized Compatible with each protease that retains activity when immobilized High Pressure Infrared Solvent IMER Magnetic particle immobilized enzyme On-chip immobilized enzyme Minutes ( 5) Seconds (<60) Minutes (~5) Hours ( 5) Minutes ( 20) Seconds (~30) Seconds (5) Not feasible Yes Not done Not done. Possible, but requires stop-flow strategy due to long digestion time Yes Yes Yes 34

36 reviewed several years ago. [60] In this chapter, approaches for acceleration of protein digestion in order to improve the throughput and recent applications of these techniques are briefly discussed. An overview of the accelerated techniques is given in Table Assisted digestion Whereas optimum digestion conditions, such as, ph and temperature, have been established for many proteases and applications [22, 61], some proteases are known to perform well under a wider range of conditions. This has been catalogued in the BRENDA enzyme database ( org/). It is well known that reaction rates improve at elevated temperature and, in some cases, a simple increase in digestion temperature may accelerate the protein digestion process. This may partially be caused by thermal denaturation of the protein, which increases the accessibility of cleavage sites. [62] Especially, the use of a thermostable protease allows for a significant increase in digestion temperature and a concurrent reduction in digestion time. For example, thermolysin digestion at 65 C for 15 min or less resulted in unequivocal protein identification. [63] Reductive methylation of trypsin enhances the rigidity of its secondary structure and thereby increases its thermostability. This modification shifts the temperature for optimal catalytic activity to C, thereby allowing for faster digestion at elevated temperature. [64] 2 Obviously, not all proteases can withstand prolonged exposure to elevated temperatures. [65] As an alternative, microwave irradiation, which is known to efficiently accelerate organic reactions, [66] has already proven its usefulness in accelerating protein proteolysis to several minutes, without significant protease degradation. [67] In comparison to a conventional incubation at 37 C for 18 h, microwave-assisted tryptic digestion of a glycoprotein mixture was performed in only 15 min at 45 C thereby achieving an even higher number of identified unique peptides and sequence coverage.[68] Alternatively, acid hydrolysis of proteins is generally performed at temperatures above 100 C for several hours and is thus ideal to be microwave accelerated. [69] Recently, microwave-assisted acid digestion was combined with electrochemistry-ms to achieve a higher reproducibility and to introduce a third cleavage site, thus yielding smaller peptides with an average length of 10 amino acids, similar to tryptic digestion. [70] Power ultrasound is known to have a positive effect on the reaction rate of various processes in the analytical laboratory. [71] High-intensity focused ultrasound can also assist in acceleration of digestion. [72] The digestion can be completed within minutes, even for complex samples, such as endothelial cell extracts [72], soy bean protein extracts [73], and bacterial protein extracts. [74] An evaluation of the effect of temperature on ultrasound-assisted digestion using an ultrasonic bath showed that ultrasound appears to be especially effective at temperatures below 55 C. [75] However, it is know that ultrasonic baths are less efficient in the generation of sonic energy than sonoreactors or ultrasonic probes. [76] A comparison of the latter two methods showed that the sonoreactor has some advantages over the ultrasonic probe in terms of sample throughput and reproducibility, but both methods produce similar sequence coverages and numbers of peptide identifications. [76] 35

37 2 Although the exact mechanism of microwave-assisted and ultrasound-assisted digestion is still unknown, elevated pressure and temperature may play a role. Therefore, it may not be surprising that not only higher temperature [64], but also high pressure has a beneficial effect on digestion efficiency. [65] Using a simple syringe to apply a pressure of 6 bar reduced the duration of a tryptic digestion to 30 min while obtaining similar or improved results as compared to atmospheric overnight digestion. [77] The beneficial effect of pressure may be due to increased protein denaturation since higher charge states were obtained or the intact MS analysis of myoglobin after application of psi (~690 bar) whereas a pressure of psi (~2400 bar) was demonstrated to reduce the tryptic digestion time of bovine serum albumin (BSA) to 60 s. [65] Pressure-assisted pepsin digestion of BSA or whole cell lysates at psi (~1725 bar) for 60 s improved the number of identified peptides and sequence coverage. [78-79] Online pressure-assisted digestion was also advantageous for HDX experiments in preventing back exchange of the reversible label. [80] In this study, proteins that are difficult to digest under normal conditions, such as amyloid β-peptide 1-42 and an HIV-1 capsid mutant protein, were successfully digested with pepsin in an online pressure-assisted digestion system employing a single UPLC gradient pump for digestion and delivering the sample to the MS for HDX studies. Infrared energy has recently been suggested as a simple and reproducible means for accelerated digestion and was shown to improve protein identification for trypsin or chymotrypsin digestion compared to conventional digestion. [81-82] The beneficial effect of infrared energy on protein digestion may arise from increased interaction between enzyme and protein and higher exposure of the cleavage sites due to bond vibrations that fall within the wavenumber range of the infrared red. [82] A comparison between conventional and ultrasound- and infrared-assisted tryptic in-gel digestion showed that the accelerated techniques reduced the total sample preparation time from 19 to 3 h, while improving sequence coverage, number of identified peptides, and peptide scores. [83] In the study, infrared-assisted digestion was favored over ultrasound-assisted digestion because of shorter digestion time (5 vs 10 min) and slightly better Mascot results. Solvent-assisted digestion has been described as well, but mainly in relation to membrane proteins (see below). High organic solvent concentrations, such as 80% of methanol or acetonitrile, decrease enzyme activity and may lead to irreproducible digestion [84], but lower concentrations may have a beneficial effect on protein digestion due to improved unfolding of proteins while still retaining sufficient enzyme activity [65]. In the quantification of proteins in complex samples, solvent-assisted digestion was also shown to reduce the digestion time to 30 min and improve the throughput. [85] The digestion time may be further reduced by combining a solvent-assisted and a pressure-assisted digestion protocol, as was shown for the 60 s digestion of BSA in the presence of 20% methanol. [65] Immobilized enzyme digestion Improved throughput in protein digestion may also be achieved by immobilization of the protease onto a solid support. [86] Immobilization leads to increased enzyme stability through reduced autolysis and improved digestion efficiency due to the increased enzyme-to-protein ratio. [26] In addition, immobilized 36

38 enzyme systems can be regenerated and reused for several times without loss of catalytic activity. [26] In principle, any enzyme can be immobilized on beads made of different materials, including (activated) agarose [87], polystyrenedivinylbenzene [88], silica [89], and glass [90]. The enzyme-immobilized beads may be used in an offline sample preparation method or packed into a column to be used in flow systems. [90] Most commonly, enzymes are directly immobilized in columns, also termed immobilized enzyme reactors (IMERs). This has been demonstrated for several enzymes, including trypsin [91-93], pepsin [94-95], chymotrypsin [96-97], and alkaline phosphatase [97]. IMER digestion of a single protein can be achieved in less than 20 min with any of the aforementioned enzymes and is very versatile. IMERs can be used offline and even in tandem [97], but are also easily integrated into online systems [98-99], for instance preceded by a protein separation step [ ] or following the immunoaffinity extraction of a target protein from a complex mixture. [92, 94] 2 For offline use, magnetic nanoparticles and polymer fibers for immobilized enzyme digestion have several advantages over the conventional solid supports, including a large surface area and a uniform and well-controlled size distribution, which improves enzyme loading and (apparent) activity. [ ] After digestion, the magnetic beads are easily removed by means of a magnet. Their intrinsic properties, such as magnetism and conductivity, can be exploited in designing sample digestion protocols. Trypsin-coated magnetic nanoparticles in combination with elevated temperature have been employed in rapid digestion protocols requiring only 30 s for digestion of single proteins with high sequence coverage. [104] As magnetic particles are excellent microwave absorbers, microwave-assisted digestion using trypsin-coated magnetic particles allowed for rapid digestion of a fractionated rat liver extract within 15 s. [ ] Pressure-assisted digestion involving trypsin-coated magnetic particles resulted in equal or better efficiency and reproducibility than conventional in-solution digestion. [107] For online protein digestion in a capillary electrophoresis (CE) system, trypsin-coated magnetic beads were captured by two magnets. [108] Separated proteins to be digested were kept at the distal tip of a first CE capillary. Periodically, the resulting peptides are separated in a second CE capillary prior to MS detection. This two-dimensional system required 30 min for the digestion and separation of 30 protein fractions. Polymer nanofibers are an alternative to beads. Trypsin-coated nanofibers were found to maintain a high activity level even after one year of repeated use and recycling. [109] Using nanofibers, a yeast proteome was digested in 6 h leading to the identification of 850 peptides corresponding to 400 proteins. As nanofibers have improved stability at high temperature and in the presence of solvents and various ph values, digestion under such rigorous conditions may reduce the required digestion time. [110] Trypsin coated magnetic particles [111] and fibers [112] may also be used for on-chip digestion followed by MALDI-MS analysis. Alternatively, the microchannel of a microchip modified with zeolite nanoparticles can be used to immobilize trypsin. [ ] This again provides accelerated digestion, for example, within 5 s for several standard proteins. [113] The microchip, online coupled to FTICR MS, was applied to digest and separate a protein extract from a mouse macrophage. This led to the identification of 191 distinct proteins. [114] Even more complex protein analysis strategies may be performed on microfluidic devices, including integrated tryptic digestion, SPE enrichment, CE separation, and ESI-MS. [115] 37

39 2.4 DIGESTION STRATEGIES FOR SPECIFIC APPLICATIONS 2 Certain classes of proteins possess characteristics that prevent their analysis through conventional digestion approaches and require extra attention. Membrane proteins are notoriously difficult to solubilize and digest, which complicates their analysis with bottom-up proteomics. [116] On the other hand, the presence of post-translational modifications (PTMs), such as phosphorylation and glycosylation, may adversely influence the efficiency of the protein digestion as well as the ionization and fragmentation efficiency of post-translationally-modified peptides. [117] Specifically designed protein digestion strategies for these protein classes are described below Membrane proteins Membrane proteins play an important role in cell signaling, cell-cell interactions, and transport across the membrane. Due to their hydrophobicity and limited solubility in aqueous solutions, they pose significant challenges to protein identification strategies. [116,118] Top-down proteomics seems to be ideally suited for the analysis of membrane proteins and their PTMs because the aqueous/organic solvent mixtures containing high acid concentrations (up to 90%) required for dissolution are highly compatible with ESI-MS. [ ] However, as stated earlier, the size of the proteins may prevent full sequence coverage which may hinder complete mapping of PTM sites. Bottom-up proteomics of membrane proteins is also not straightforward. As membrane proteins contain fewer Lys and Arg residues, tryptic digestion cannot effectively be employed in their bottom-up analysis. Improved sequence coverage of the membrane proteome can be achieved using other proteases like chymotrypsin and staphylococcal peptidase I [122] or pepsin [123]. A combination of chemical and enzymatic treatments was also found to achieve better sensitivity and membrane proteome coverage. Acidic conditions during in-solution chemical digestion improve the solubilization of membrane proteins and the resulting, more soluble, protein fragments can subsequently be digested with trypsin [57-58] or chymotrypsin [57]. In an in-gel digestion protocol for membrane proteins, trypsin digestion was followed by CNBr cleavage for improved coverage of the membrane proteome. [59] Assisted digestion techniques (see above) can obviously be very beneficial in the digestion of membrane proteins. Microwave irradiation has been found to increase their solubility [124] and accelerate their digestion. [ ] The solubilization of membrane proteins can be improved using detergents, such as sodium dodecyl sulfate (SDS). However, detergents compromise the protease efficiency [127] and need to be removed prior to digestion and MS analysis, for instance using the filter-aided sample preparation (FASP) protocol [128]. Membrane proteins would more readily solubilize in organic-aqueous buffers and, as stated above, low concentrations of organic solvent do not affect enzyme activity and may even have a beneficial effect on digestion efficiency. 38

40 A methanol-assisted solubilization and digestion protocol was applied to the analysis of the membrane fraction of E. coli cells. [127] The solvent-assisted protocol, in which the membrane proteins were dissolved and digested in an ammonium bicarbonate buffer containing 60% methanol, was compared to a protocol using 1% SDS for solubilization and 0.1% SDS during digestion. The methanol-assisted method allowed for more protein identifications (358 vs 299), higher reproducibility, and more efficient identification of integral membrane proteins (159 vs 120). Moreover, the methanol-assisted digestion was completed in 5 h, compared to overnight digestion in SDS Post-translational modifications Microwave-assisted acid hydrolysis has demonstrated its usefulness for the simultaneous determination of the protein sequence as well as the identification of phosphorylation or glycosylation sites of several standard proteins. [129] This digestion technique was found to be especially beneficial for glycoproteins because the glycan moieties may sterically hinder enzymatic digestion. [68] Other assisted digestion techniques have not been applied to the study of protein PTMs. It is possible that the harsh conditions associated with these techniques, such as elevated temperature or pressure, may cause degradation and/or loss of the labile PTMs, although this did not seem to be the case for microwave-assisted digestion. Alternatively, specific enzymes (or chemicals) can be used to remove the PTMs prior to digestion of the protein backbone, for example, alkaline phosphatases and peptide N-glycosidases (such as PNGase F [Peptide-N4-(N-acetyl-β-glucosaminyl) asparagine amidase], PNGase A or endoglycosidases H or D). [130] A tandem microreactor setup, in which the first reactor contained immobilized chymotrypsin for protein digestion and the second reactor contained alkaline phosphatase for dephosphorylation, was successfully applied to the characterization of phosphoproteins. [97] 2 Advanced strategies for glycoprotein characterization have been described, including separate strategies to sequence the peptide backbone and to characterize the glycan structure. [131] For instance, CAD-MS/MS allows for sequencing of the monosaccharide residue, but (depending on the chosen sample pretreatment strategy) may not be able to pinpoint the site of the glycan attachment due to the loss of the sugar which is inherent to this type of fragmentation. [132] However, ETD fragmentation favors backbone cleavage of glycosylated peptides and, thus, does more readily allow for the identification of glycosylation sites. [132] As such, the application of middle-down proteomics, using the OmpT enzyme to generate larger peptides, was beneficial for the identification of post-translationally modified peptides. [39] Incomplete digestion with trypsin or Lys-C enhanced the determination of multiple phosphorylation sites of twenty recombinant nucleotide-binding proteins in Escherichia coli, including kinases and cystathionine beta-synthase domain containing proteins. [133] Middle-down analysis of post-translationally modified proteins gives insight into the connectivity and combinatorial effect of multiple PTMs in the same polypeptide chain, which can be explored more fully with top-down proteomics. [7] 39

41 2.5 CONCLUSIONS AND PERSPECTIVES 2 Efficient and reproducible protein digestion is of utmost importance in bottom-up and middle-down proteomics. The selection of the most suitable enzyme or chemical reagent depends on the characteristics of the target proteins, such as amino acid composition and hydrophobicity, and compatibility with the digestion technique that is being used, in terms of ph or solvent compatibility. In addition, a wide variety of methods exist for acceleration of the digestion process to as less than seconds. Some of these techniques may be beneficial for certain applications, such as the digestion of membrane proteins. However, great care and consideration has to be taken to sufficiently optimize these accelerated digestion protocols, especially for protein or biomarker quantification, as they may lead to less reproducible results. In-solution or in-gel protein digestion using trypsin in a bottom-up approach is still considered the gold standard in proteomics, although in particular applications, other enzymes or chemical digestion protocols may be useful. Middle-down analysis of proteins serves as a distinct alternative to bottom-up proteomics and has gained popularity in recent years due to the availability of specialized enzymes and continuous developments in MS instrumentation. The latter also enables the analysis of intact proteins in top-down proteomics, which is still somewhat limited in its application, but can provide information that is sometimes lost in digestion strategies. Selection of the suitable proteomics approach is guided by the types of research questions that need to be answered, the properties of the proteins to be analyzed, and the availability of expertise and instrumentation. In fact, due to their complementary nature, a combination of these approaches may be the only way to understand the biological processes that are essential to life. REFERENCES [1] Angel, T.E., Aryal, U.K., Hengel, S.M., Baker, E.S., Kelly, R.T., et al., Mass spectrometry-based proteomics: existing capabilities and future directions. Chem. Soc. Rev. 2012, 41, (10), [2] Marko-Varga, G., Fehniger, T.E., Proteomics and disease--the challenges for technology and discovery. J. Proteome Res. 2004, 3, (2), [3] Calligaris, D., Villard, C., Lafitte, D., Advances in top-down proteomics for disease biomarker discovery. J. Proteomics. 2011, 74, (7), [4] Chait, B.T., Chemistry. Mass spectrometry: bottom-up or top-down? Science. 2006, 314, (5796), [5] Wu, S.-L., Hühmer, A.F.R., Hao, Z., Karger, B.L., Online LC MS Approach Combining Collision-Induced Dissociation (CID), Electron-Transfer Dissociation (ETD), and CID of an Isolated Charge-Reduced Species for the Trace-Level Characterization of Proteins with Post-Translational Modifications. J. Proteome Res. 2007, 6, (11), [6] McLafferty, F.W., Breuker, K., Jin, M., Han, X., Infusini, G., et al., Top-down MS, a powerful complement to the high capabilities of proteolysis proteomics. FEBS J. 2007, 274, (24), [7] Lanucara, F., Eyers, C.E., Top-down mass spectrometry for the analysis of combinatorial post-translational modifications. Mass Spectrom. Rev. 2013, 32, (1), [8] Han, X., Jin, M., Breuker, K., McLafferty, F.W., Extending top-down mass spectrometry to proteins with masses greater than 200 kilodaltons. Science. 2006, 314, (5796),

42 [9] Chi, A., Bai, D.L., Geer, L.Y., Shabanowitz, J., Hunt, D.F., Analysis of intact proteins on a chromatographic time scale by electron transfer dissociation tandem mass spectrometry. Int. J. Mass Spectrom. 2007, 259, (1-3), [10] Compton, P.D., Zamdborg, L., Thomas, P.M., Kelleher, N.L., On the scalability and requirements of whole protein mass spectrometry. Anal. Chem. 2011, 83, (17), [11] Aebersold, R., Mann, M., Mass spectrometry-based proteomics. Nature. 2003, 422, (6928), [12] Boyne, M.T., Garcia, B.A., Li, M., Zamdborg, L., Wenger, C.D., et al., Tandem Mass Spectrometry with Ultrahigh Mass Accuracy Clarifies Peptide Identification by Database Retrieval. J.Proteome Res. 2008, 8, (1), [13] Cannon, J., Lohnes, K., Wynne, C., Wang, Y., Edwards, N., et al., High-Throughput Middle-Down Analysis Using an Orbitrap. J. Proteome Res. 2010, 9, (8), [14] Hauser, N.J., Han, H., McLuckey, S.A., Basile, F., Electron transfer dissociation of peptides generated by microwave D-cleavage digestion of proteins. J. Proteome Res. 2008, 7, (5), [15] Kalli, A., Hakansson, K., Electron capture dissociation of highly charged proteolytic peptides from Lys N, Lys C and Glu C digestion. Mol. Biosyst. 2010, 6, (9), [16] Wu, S.-L., Kim, J., Hancock, W.S., Karger, B., Extended Range Proteomic Analysis (ERPA): A New and Sensitive LC MS Platform for High Sequence Coverage of Complex Proteins with Extensive Post-translational ModificationsComprehensive Analysis of Beta-Casein and Epidermal Growth Factor Receptor (EGFR). J. Proteome Res. 2005, 4, (4), [17] Deutsch, E.W., Lam, H., Aebersold, R., Data analysis and bioinformatics tools for tandem mass spectrometry in proteomics. Physiol. Genomics. 2008, 33, (1), [18] Brownridge, P., Beynon, R.J., The importance of the digest: proteolysis and absolute quantification in proteomics. Methods. 2011, 54, (4), [19] Freije, J.R., Mulder, P.P.M.F.A., Werkman, W., Rieux, L., Niederlander, H.A.G., et al., Chemically Modified, Immobilized Trypsin Reactor with Improved Digestion Efficiency. J. Proteome Res. 2005, 4, (5), [20] Burkhart, J.M., Schumbrutzki, C., Wortelkamp, S., Sickmann, A., Zahedi, R.P., Systematic and quantitative comparison of digest efficiency and specificity reveals the impact of trypsin quality on MS-based proteomics. J. Proteomics. 2012, 75, (4), [21] Kinter, M., Sherman, N.E., Protein sequencing and identification using tandem mass spectrometry, John Wiley, New York, [22] Switzar, L., Giera, M., Lingeman, H., Irth, H., Niessen, W.M., Protein digestion optimization for characterization of drug protein adducts using response surface modeling. J. Chromatogr. A. 2011, 1218, (13), [23] Richardson, J., Shah, B., Xiao, G., Bondarenko, P.V., Zhang, Z., Automated in-solution protein digestion using a commonly available high-performance liquid chromatography autosampler. Anal. Biochem. 2011, 411, (2), [24] Zheng, S., Yoo, C., Delmotte, N., Miller, F.R., Huber, C.G., et al., Monolithic column HPLC separation of intact proteins analyzed by LC-MALDI using on-plate digestion: An approach to integrate protein separation and identification. Anal. Chem. 2006, 78, (14), [25] Bruyneel, B., Hoos, J.S., Smoluch, M.T., Lingeman, H., Niessen, W.M., et al., Trace analysis of proteins using postseparation solution-phase digestion and electrospray mass spectrometric detection of marker peptides. Anal. Chem. 2007, 79, (4), [26] Massolini, G., Calleri, E., Immobilized trypsin systems coupled online to separation methods: recent developments and analytical applications. J. Sep. Sci. 2005, 28, (1), [27] Rey, M., Man, P., Brandolin, G., Forest, E., Pelosi, L., Recombinant immobilized rhizopuspepsin as a new tool for protein digestion in hydrogen/deuterium exchange mass spectrometry. Rapid Commun. Mass Spectrom. 2009, 23, (21), [28] Ahn, J., Cao, M.J., Yu, Y.Q., Engen, J.R., Accessing the Reproducibilityand Specificity of Pepsin and other Aspartic Proteases. Biochim. Biophys. Acta [29] Artimo, P., Jonnalagedda, M., Arnold, K., Baratin, D., Csardi, G., et al., ExPASy: SIB bioinformatics resource portal. Nucleic Acids Res. 2012, 40, (Web Server issue), W [30] Raijmakers, R., Neerincx, P., Mohammed, S., Heck, A.J., Cleavage specificities of the brother and sister proteases Lys-C and Lys-N. Chem. Commun. (Camb). 2010, 46, (46), [31] Swatkoski, S., Gutierrez, P., Ginter, J., Petrov, A., Dinman, J.D., et al., Integration of residue-specific acid cleavage into proteomic workflows. J. Proteome Res. 2007, 6, (11), [32] Smith, B.J., Chemical cleavage of polypeptides. Methods Mol. Biol. 2003, 211, [33] Tang, H.Y., Speicher, D.W., Identification of alternative products and optimization of 2-nitro-5- thiocyanatobenzoic acid cyanylation and cleavage at cysteine residues. Anal. Biochem. 2004, 334, (1), [34] Crimmins, D.L., Mische, S.M., Denslow, N.D., Chemical cleavage of proteins in solution. Curr. Protoc. Protein Sci. 2005, Chapter 11, Unit [35] Taouatas, N., Drugan, M.M., Heck, A.J., Mohammed, S., Straightforward ladder sequencing of peptides using a Lys-N metalloendopeptidase. Nat. Methods. 2008, 5, (5),

43 2 [36] Taouatas, N., Heck, A.J., Mohammed, S., Evaluation of metalloendopeptidase Lys-N protease performance under different sample handling conditions. J. Proteome Res. 2010, 9, (8), [37] Taouatas, N., Altelaar, A.F., Drugan, M.M., Helbig, A.O., Mohammed, S., et al., Strong cation exchange-based fractionation of Lys-N-generated peptides facilitates the targeted analysis of post-translational modifications. Mol. Cell. Proteomics. 2009, 8, (1), [38] Tan, Y.J., Wang, W.H., Zheng, Y., Dong, J., Stefano, G., et al., Limited Proteolysis via Millisecond Digestions in Protease-Modified Membranes. Anal. Chem. 2012, 84, (19), [39] Wu, C., Tran, J.C., Zamdborg, L., Durbin, K.R., Li, M., et al., A protease for middle-down proteomics. Nat. Methods. 2012, 9, (8), [40] von Pawel-Rammingen, U., Johansson, B.P., Bjorck, L., IdeS, a novel streptococcal cysteine proteinase with unique specificity for immunoglobulin G. EMBO J. 2002, 21, (7), [41] Chevreux, G., Tilly, N., Bihoreau, N., Fast analysis of recombinant monoclonal antibodies using IdeS proteolytic digestion and electrospray mass spectrometry. Anal. Biochem. 2011, 415, (2), [42] Hytonen, J., Haataja, S., Gerlach, D., Podbielski, A., Finne, J., The SpeB virulence factor of Streptococcus pyogenes, a multifunctional secreted and cell surface molecule with strepadhesin, laminin-binding and cysteine protease activity. Mol. Microbiol. 2001, 39, (2), [43] Li, A., Sowder, R.C., Henderson, L.E., Moore, S.P., Garfinkel, D.J., et al., Chemical cleavage at aspartyl residues for protein identification. Anal. Chem. 2001, 73, (22), [44] Lo, L.H., Huang, T.L., Shiea, J., Acid hydrolysis followed by matrix-assisted laser desorption/ionization mass spectrometry for the rapid diagnosis of serum protein biomarkers in patients with major depression. Rapid Commun. Mass Spectrom. 2009, 23, (5), [45] Patwa, T.H., Wang, Y., Simeone, D.M., Lubman, D.M., Enhanced detection of autoantibodies on protein microarrays using a modified protein digestion technique. J. Proteome Res. 2008, 7, (6), [46] Park, H.B., Pyo, S.H., Hong, S.S., Kim, J.H., Optimization of the hydroxylamine cleavage of an expressed fusion protein to produce a recombinant antimicrobial peptide. Biotechnol. Lett. 2001, 23, (8), [47] Apweiler, R., Biswas, M., Fleischmann, W., Kanapin, A., Karavidopoulou, Y., et al., Proteome Analysis Database: online application of InterPro and CluSTr for the functional classification of proteins in whole genomes. Nucleic Acids Res. 2001, 29, (1), [48] Permentier, H.P., Bruins, A.P., Electrochemical oxidation and cleavage of proteins with online mass spectrometric detection: development of an instrumental alternative to enzymatic protein digestion. J.Am. Soc. Mass Spectrom. 2004, 15, (12), [49] Permentier, H.P., Jurva, U., Barroso, B., Bruins, A.P., Electrochemical oxidation and cleavage of peptides analyzed with online mass spectrometric detection. Rapid Commun. Mass Spectrom. 2003, 17, (14), [50] Zhang, Y., Dewald, H.D., Chen, H., Online mass spectrometric analysis of proteins/peptides following electrolytic cleavage of disulfide bonds. J. Proteome Res. 2011, 10, (3), [51] Meyer, B., Papasotiriou, D.G., Karas, M., 100% protein sequence coverage: a modern form of surrealism in proteomics. Amino Acids [52] Swaney, D.L., Wenger, C.D., Coon, J.J., Value of using multiple proteases for large-scale mass spectrometrybased proteomics. J. Proteome Res. 2010, 9, (3), [53] Biringer, R.G., Amato, H., Harrington, M.G., Fonteh, A.N., Riggins, J.N., et al., Enhanced sequence coverage of proteins in human cerebrospinal fluid using multiple enzymatic digestion and linear ion trap LC MS/MS. Brief. Funct.Genomic. Proteomic. 2006, 5, (2), [54] Choudhary, G., Wu, S.L., Shieh, P., Hancock, W.S., Multiple enzymatic digestion for enhanced sequence coverage of proteins in complex proteomic mixtures using capillary LC with ion trap MS/MS. J. Proteome Res. 2003, 2, (1), [55] Gauci, S., Helbig, A.O., Slijper, M., Krijgsveld, J., Heck, A.J., et al., Lys-N and trypsin cover complementary parts of the phosphoproteome in a refined SCX-based approach. Anal. Chem. 2009, 81, (11), [56] Chen, R., Jiang, X., Sun, D., Han, G., Wang, F., et al., Glycoproteomics analysis of human liver tissue by combination of multiple enzyme digestion and hydrazide chemistry. J. Proteome Res. 2009, 8, (2), [57] Kwon, J., Oh, J., Park, C., Cho, K., Kim, S.I., et al., Systematic cyanobacterial membrane proteome analysis by combining acid hydrolysis and digestive enzymes with nano-liquid chromatography-fourier transform mass spectrometry. J. Chromatogr. A. 2010, 1217, (3), [58] Lee, J.E., Kwon, J., Baek, M.C., A combination method of chemical with enzyme reactions for identification of membrane proteins. Biochim. Biophys. Acta. 2011, 1814, (3), [59] van Montfort, B.A., Doeven, M.K., Canas, B., Veenhoff, L.M., Poolman, B., et al., Combined in-gel tryptic digestion and CNBr cleavage for the generation of peptide maps of an integral membrane protein with MALDI-TOF mass spectrometry. Biochim. Biophys. Acta. 2002, 1555, (1-3), [60] Capelo, J.L., Carreira, R., Diniz, M., Fernandes, L., Galesio, M., et al., Overview on modern approaches to speed up protein identification workflows relying on enzymatic cleavage and mass spectrometry-based techniques. Anal. Chim. Acta. 2009, 650, (2),

44 [61] Norrgran, J., Williams, T.L., Woolfitt, A.R., Solano, M.I., Pirkle, J.L., et al., Optimization of digestion parameters for protein quantification. Anal. Biochem. 2009, 393, (1), [62] Park, Z.Y., Russell, D.H., Thermal denaturation: a useful technique in peptide mass mapping. Anal. Chem. 2000, 72, (11), [63] Bark, S.J., Muster, N., Yates, J.R., 3rd, Siuzdak, G., High-temperature protein mass mapping using a thermophilic protease. J. Am. Chem. Soc. 2001, 123, (8), [64] Havlis, J., Thomas, H., Sebela, M., Shevchenko, A., Fast-response proteomics by accelerated in-gel digestion of proteins. Anal.Chem. 2003, 75, (6), [65] Lopez-Ferrer, D., Petritis, K., Hixson, K.K., Heibeck, T.H., Moore, R.J., et al., Application of pressurized solvents for ultrafast trypsin hydrolysis in proteomics: proteomics on the fly. J. Proteome Res. 2008, 7, (8), [66] Larhed, M., Moberg, C., Hallberg, A., Microwave-accelerated homogeneous catalysis in organic chemistry. Acc. Chem. Res. 2002, 35, (9), [67] Pramanik, B.N., Mirza, U.A., Ing, Y.H., Liu, Y.H., Bartner, P.L., et al., Microwave-enhanced enzyme reaction for protein mapping by mass spectrometry: a new approach to protein digestion in minutes. Protein Sci. 2002, 11, (11), [68] Segu, Z.M., Hammad, L.A., Mechref, Y., Rapid and efficient glycoprotein identification through microwaveassisted enzymatic digestion. Rapid Commun. Mass Spectrom. 2010, 24, (23), [69] Hua, L., Low, T.Y., Sze, S.K., Microwave-assisted specific chemical digestion for rapid protein identification. Proteomics. 2006, 6, (2), [70] Basile, F., Hauser, N., Rapid online nonenzymatic protein digestion combining microwave heating acid hydrolysis and electrochemical oxidation. Anal. Chem. 2010, 83, (1), [71] Chemat, S., Lagha, A., Ait Amar, H., Chemat, F., Ultrasound assisted microwave digestion. Ultrason. Sonochem. 2004, 11, (1), 5-8. [72] Lopez-Ferrer, D., Capelo, J.L., Vazquez, J., Ultra fast trypsin digestion of proteins by high intensity focused ultrasound. J. Proteome Res. 2005, 4, (5), [73] Dominguez-Vega, E., Garcia, M.C., Crego, A.L., Marina, M.L., First approach based on direct ultrasonic assisted enzymatic digestion and capillary-high performance liquid chromatography for the peptide mapping of soybean proteins. J. Chromatogr. A. 2010, 1217, (42), [74] Carreira, R.J., Cordeiro, F.M., Moro, A.J., Rivas, M.G., Rial-Otero, R., et al., New findings for in-gel digestion accelerated by high-intensity focused ultrasound for protein identification by matrix-assisted laser desorption ionization time-of-flight mass spectrometry. J. Chromatogr. A. 2007, 1153, (1-2), [75] Shin, S., Yang, H.J., Kim, J., Effects of temperature on ultrasound-assisted tryptic protein digestion. Anal. Biochem [76] Rial-Otero, R., Carreira, R.J., Cordeiro, F.M., Moro, A.J., Santos, H.M., et al., Ultrasonic assisted protein enzymatic digestion for fast protein identification by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Sonoreactor versus ultrasonic probe. J. Chromatogr. A. 2007, 1166, (1-2), [77] Yang, H.J., Hong, J., Lee, S., Shin, S., Kim, J., Pressure-assisted tryptic digestion using a syringe. Rapid Commun. Mass Spectrom. 2010, 24, (7), [78] Lopez-Ferrer, D., Petritis, K., Robinson, E.W., Hixson, K.K., Tian, Z., et al., Pressurized pepsin digestion in proteomics: an automatable alternative to trypsin for integrated top-down bottom-up proteomics. Mol. Cell. Proteomics. 2011, 10, (2), M [79] Lopez-Ferrer, D., Petritis, K., Lourette, N.M., Clowers, B., Hixson, K.K., et al., Online digestion system for protein characterization and proteome analysis. Anal. Chem. 2008, 80, (23), [80] Jones, L.M., Zhang, H., Vidavsky, I., Gross, M.L., Online, high-pressure digestion system for protein characterization by hydrogen/deuterium exchange and mass spectrometry. Anal. Chem. 2010, 82, (4), [81] Wang, S., Zhang, L., Yang, P., Chen, G., Infrared-assisted tryptic proteolysis for peptide mapping. Proteomics. 2008, 8, (13), [82] Wang, S., Liu, T., Zhang, L., Chen, G., Efficient chymotryptic proteolysis enhanced by infrared radiation for peptide mapping. J. Proteome Res. 2008, 7, (11), [83] Dycka, F., Bobal, P., Mazanec, K., Bobalova, J., Rapid and efficient protein enzymatic digestion: an experimental comparison. Electrophoresis. 2012, 33, (2), [84] Wall, M.J., Crowell, A.M., Simms, G.A., Carey, G.H., Liu, F., et al., Implications of partial tryptic digestion in organicaqueous solvent systems for bottom-up proteome analysis. Anal. Chim. Acta. 2011, 703, (2), [85] Li, F., Schmerberg, C.M., Ji, Q.C., Accelerated tryptic digestion of proteins in plasma for absolute quantitation using a protein internal standard by liquid chromatography/tandem mass spectrometry. Rapid Commun. Mass Spectrom. 2009, 23, (5), [86] Girelli, A.M., Mattei, E., Application of immobilized enzyme reactor in online high performance liquid : a review. J. Chromatogr. B. 2005, 819, (1),

45 2 [87] Dodds, E.D., Seipert, R.R., Clowers, B.H., German, J.B., Lebrilla, C.B., Analytical performance of immobilized pronase for glycopeptide footprinting and implications for surpassing reductionist glycoproteomics. J. Proteome Res. 2009, 8, (2), [88] Lopez-Ferrer, D., Hixson, K.K., Smallwood, H., Squier, T.C., Petritis, K., et al., Evaluation of a high-intensity focused ultrasound-immobilized trypsin digestion and 18O-labeling method for quantitative proteomics. Anal. Chem. 2009, 81, (15), [89] Casadonte, F., Pasqua, L., Savino, R., Terracciano, R., Smart trypsin adsorption into N-(2-aminoethyl)-3- aminopropyl-modified mesoporous silica for ultra fast protein digestion. Chemistry. 2010, 16, (30), Lim, L.W., Tomatsu, M., Takeuchi, T., Development of an online immobilized-enzyme reversed-phase HPLC method for protein digestion and peptide separation. Anal. Bioanal. Chem. 2006, 386, (3), [90] Calleri, E., Temporini, C., Perani, E., De Palma, A., Lubda, D., et al., Trypsin-based monolithic bioreactor coupled online with LC/MS/MS system for protein digestion and variant identification in standard solutions and serum samples. J. Proteome Res. 2005, 4, (2), [91] Hoos, J.S., Sudergat, H., Hoelck, J.P., Stahl, M., de Vlieger, J.S., et al., Selective quantitative bioanalysis of proteins in biological fluids by online immunoaffinity chromatography-protein digestion-liquid chromatography-mass spectrometry. J. Chromatogr. B. 2006, 830, (2), [92] Nicoli, R., Gaud, N., Stella, C., Rudaz, S., Veuthey, J.L., Trypsin immobilization on three monolithic disks for online protein digestion. J. Pharm. Biomed. Anal. 2008, 48, (2), [93] Cingoz, A., Hugon-Chapuis, F., Pichon, V., Total online analysis of a target protein from plasma by immunoextraction, digestion and liquid chromatography-mass spectrometry. J. Chromatogr. B. 2010, 878, (2), [94] Geiser, L., Eeltink, S., Svec, F., Frechet, J.M., In-line system containing porous polymer monoliths for protein digestion with immobilized pepsin, peptide preconcentration and nano-liquid chromatography separation coupled to electrospray ionization mass spectroscopy. J. Chromatogr. A. 2008, 1188, (2), [95] Temporini, C., Calleri, E., Campese, D., Cabrera, K., Felix, G., et al., Chymotrypsin immobilization on epoxy monolithic silica columns: development and characterization of a bioreactor for protein digestion. J. Sep. Sci. 2007, 30, (17), [96] Yamaguchi, H., Miyazaki, M., Kawazumi, H., Maeda, H., Multidigestion in continuous flow tandem proteaseimmobilized microreactors for proteomic analysis. Anal. Biochem. 2010, 407, (1), [97] Spross, J., Sinz, A., A capillary monolithic trypsin reactor for efficient protein digestion in online and offline coupling to ESI and MALDI mass spectrometry. Anal. Chem. 2010, 82, (4), [98] Sun, L., Ma, J., Qiao, X., Liang, Y., Zhu, G., et al., Integrated device for online sample buffer exchange, protein enrichment, and digestion. Anal. Chem. 2010, 82, (6), [99] Slysz, G.W., Schriemer, D.C., Blending protein separation and peptide analysis through real-time proteolytic digestion. Anal.Chem. 2005, 77, (6), [100] Yuan, H., Zhang, L., Hou, C., Zhu, G., Tao, D., et al., Integrated platform for proteome analysis with combination of protein and peptide separation via online digestion. Anal. Chem. 2009, 81, (21), [101] Gao, M., Deng, C., Zhang, X., Magnetic nanoparticles-based digestion and enrichment methods in proteomics analysis. Expert Rev. Proteomics. 2011, 8, (3), [102] Kim, J., Kim, B.C., Lopez-Ferrer, D., Petritis, K., Smith, R.D., Nanobiocatalysis for protein digestion in proteomic analysis. Proteomics. 2010, 10, (4), [103] Jeng, J., Lin, M.F., Cheng, F.Y., Yeh, C.S., Shiea, J., Using high-concentration trypsin-immobilized magnetic nanoparticles for rapid in situ protein digestion at elevated temperature. Rapid Commun. Mass Spectrom. 2007, 21, (18), [104] Lin, S., Lin, Z., Yao, G., Deng, C., Yang, P., et al., Development of microwave-assisted protein digestion based on trypsin-immobilized magnetic microspheres for highly efficient proteolysis followed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry analysis. Rapid Commun. Mass Spectrom. 2007, 21, (23), [105] Lin, S., Yao, G., Qi, D., Li, Y., Deng, C., et al., Fast and efficient proteolysis by microwave-assisted protein digestion using trypsin-immobilized magnetic silica microspheres. Anal. Chem. 2008, 80, (10), [106] Lee, B., Lopez-Ferrer, D., Kim, B.C., Na, H.B., Park, Y.I., et al., Rapid and efficient protein digestion using trypsincoated magnetic nanoparticles under pressure cycles. Proteomics. 2011, 11, (2), [107] Li, Y., Wojcik, R., Dovichi, N.J., A replaceable microreactor for online protein digestion in a two-dimensional capillary electrophoresis system with tandem mass spectrometry detection. J. Chromatogr. A [108] Kim, B.C., Lopez-Ferrer, D., Lee, S.M., Ahn, H.K., Nair, S., et al., Highly stable trypsin-aggregate coatings on polymer nanofibers for repeated protein digestion. Proteomics. 2009, 9, (7),

46 [109] Ahn, H.K., Kim, B.C., Jun, S.H., Chang, M.S., Lopez-Ferrer, D., et al., Robust trypsin coating on electrospun polymer nanofibers in rigorous conditions and its uses for protein digestion. Biotechnol. Bioeng. 2010, 107, (6), [110] Slovakova, M., Minc, N., Bilkova, Z., Smadja, C., Faigle, W., et al., Use of self assembled magnetic beads for onchip protein digestion. Lab. Chip. 2005, 5, (9), [111] Fan, H., Chen, G., Fiber-packed channel bioreactor for microfluidic protein digestion. Proteomics. 2007, 7, (19), [112] Huang, Y., Shan, W., Liu, B., Liu, Y., Zhang, Y., et al., Zeolite nanoparticle modified microchip reactor for efficient protein digestion. Lab Chip. 2006, 6, (4), [113] Ji, J., Zhang, Y., Zhou, X., Kong, J., Tang, Y., et al., Enhanced protein digestion through the confinement of nanozeolite-assembled microchip reactors. Anal. Chem. 2008, 80, (7), [114] Wang, C., Jemere, A.B., Harrison, D.J., Multifunctional protein processing chip with integrated digestion, solidphase extraction, separation and electrospray. Electrophoresis. 2010, 31, (22), [115] Helbig, A.O., Heck, A.J., Slijper, M., Exploring the membrane proteome--challenges and analytical strategies. J. Proteomics. 2010, 73, (5), [116] Temporini, C., Calleri, E., Massolini, G., Caccialanza, G., Integrated analytical strategies for the study of phosphorylation and glycosylation in proteins. Mass Spectrom. Rev. 2008, 27, (3), [117] Lu, B., McClatchy, D.B., Kim, J.Y., Yates, J.R., 3rd, Strategies for shotgun identification of integral membrane proteins by tandem mass spectrometry. Proteomics. 2008, 8, (19), [118] Ryan, C.M., Souda, P., Bassilian, S., Ujwal, R., Zhang, J., et al., Post-translational modifications of integral membrane proteins resolved by top-down Fourier transform mass spectrometry with collisionally activated dissociation. Mol. Cell. Proteomics. 2010, 9, (5), [119] Souda, P., Ryan, C.M., Cramer, W.A., Whitelegge, J., Profiling of integral membrane proteins and their post translational modifications using high-resolution mass spectrometry. Methods. 2011, 55, (4), [120] Whitelegge, J., Halgand, F., Souda, P., Zabrouskov, V., Top-down mass spectrometry of integral membrane proteins. Expert. Rev. Proteomics. 2006, 3, (6), [121] Fischer, F., Poetsch, A., Protein cleavage strategies for an improved analysis of the membrane proteome. Proteome Sci. 2006, 4, 2. [122] Rietschel, B., Bornemann, S., Arrey, T.N., Baeumlisberger, D., Karas, M., et al., Membrane protein analysis using an improved peptic in-solution digestion protocol. Proteomics. 2009, 9, (24), [123] Ye, X., Li, L., Microwave-assisted protein solubilization for mass spectrometry-based shotgun proteome analysis. Anal. Chem. 2012, 84, (14), [124] Zhong, H., Marcus, S.L., Li, L., Microwave-assisted acid hydrolysis of proteins combined with liquid chromatography MALDI MS/MS for protein identification. J. Am. Soc. Mass Spectrom. 2005, 16, (4), [125] Vaezzadeh, A.R., Deshusses, J.M., Waridel, P., Francois, P., Zimmermann-Ivol, C.G., et al., Accelerated digestion for high-throughput proteomics analysis of whole bacterial proteomes. J. Microbiol. Methods. 2010, 80, (1), [126] Zhang, N., Chen, R., Young, N., Wishart, D., Winter, P., et al., Comparison of SDS- and methanol-assisted protein solubilization and digestion methods for Escherichia coli membrane proteome analysis by 2-D LC MS/MS. Proteomics. 2007, 7, (4), [127] Wisniewski, J.R., Zielinska, D.F., Mann, M., Comparison of ultrafiltration units for proteomic and N-glycoproteomic analysis by the filter-aided sample preparation method. Anal. Biochem. 2011, 410, (2), [128] Lill, J.R., Ingle, E.S., Liu, P.S., Pham, V., Sandoval, W.N., Microwave-assisted proteomics. Mass Spectrom. Rev. 2007, 26, (5), [129] Lazar, I.M., Lazar, A.C., Cortes, D.F., Kabulski, J.L., Recent advances in the MS analysis of glycoproteins: Theoretical considerations. Electrophoresis. 2011, 32, (1), [130] Pan, S., Chen, R., Aebersold, R., Brentnall, T.A., Mass spectrometry based glycoproteomics--from a proteomics perspective. Mol. Cell. Proteomics. 2011, 10, (1), R [131] Hitchen, P.G., Twigger, K., Valiente, E., Langdon, R.H., Wren, B.W., et al., Glycoproteomics: a powerful tool for characterizing the diverse glycoforms of bacterial pilins and flagellins. Biochem. Soc. Trans. 2010, 38, (5), [132] She, Y.M., Xu, X., Yakunin, A.F., Dhe-Paganon, S., Donald, L.J., et al., Mass spectrometry following mild enzymatic digestion reveals phosphorylation of recombinant proteins in Escherichia coli through mechanisms involving direct nucleotide binding. J. Proteome Res. 2010, 9, (6),

47

48 Chapter 3 Protein digestion optimization for characterization of drug protein adducts using response surface modeling Linda Switzar, Martin Giera, Henk Lingeman, Hubertus Irth and Wilfried M.A. Niessen Journal of Chromatography A, 2011, 1218 (13),

49 ABSTRACT 3 The formation of drug protein adducts in vivo may have important clinical and toxicological implications. Consequently, there is a great interest in the detection of these adducts and the elucidation of their role in the processes leading to adverse and idiosyncratic drug reactions. Enzymatic digestion is a crucial step in bottom-up proteomics strategies for the analysis of drug protein adducts. The chosen proteolytic enzyme and digestion conditions have a large influence on the protein coverage of the modified protein and identification of its modification site. In this work, the enzymatic digestion conditions (ph, temperature and time) of trypsin and thermolysin were optimized specifically for the characterization of Human Serum Albumin (HSA) adducts. Using a Design of Experiments (DOE), it was found that of the three optimized parameters mainly ph and temperature showed strong effects on both responses. The optimized digestion conditions were different from those obtained from the suppliers or literature. Their application to HSA adducts resulted in improved protein coverage and signal intensity regarding the peptide containing the modification site, thereby highlighting the importance of a detailed optimization of digestion conditions. 48

50 3.1 INTRODUCTION Drug protein adducts are suggested to play a role as mediators of Adverse Drug Reactions (ADRs) and Idiosyncratic Drug Reactions (IDRs). [1] Therefore, their detection and identification is crucial within the framework of drug safety. [2] In the last decade, several Liquid Chromatography Mass Spectrometry (LC MS) based strategies have been developed for the determination of drug protein adducts [3], i.e., the screening of reactive drug intermediates trapped by small molecules such as glutathione (GSH) [4 6] and proteomics based methods analyzing the adduct formed by a drug and its protein target [7 11]. The latter strategies are mostly based on enzymatic digestion of the modified protein followed by LC MS(/MS) analysis of the resulting proteolytic peptides. These approaches allow for the detection of clinically relevant drug protein adducts and their simultaneous identification thereby giving insight into the mechanisms underlying ADRs. Two major factors influencing the success of such methods are protein coverage, linked to successful identification of the modified protein, and the detection of the specific peptides that contain the modification site. The latter defines the actual sensitivity of the method and, naturally, achieving high protein coverage increases the chance of detecting the modified peptides. The most delicate step in this respect is the digestion of drug protein adducts. It is not only critical to choose the appropriate enzyme, but also to apply the right digestion conditions, such as buffer ph, digestion temperature and time. Enzyme suppliers usually provide optimal conditions for the delivered enzyme. In addition, a wide range of digestion conditions obtained with different substrates are available from literature and enzyme databases such as BRENDA ( For example, the optimal digestion conditions of bovine trypsin (EC ) according to several suppliers are 2-18 h digestion time (depending on the amount of protein) at a temperature of 37 C in 50 mm ammonium hydrogencarbonate or 100 mm Tris-HCl, ph 8.5. However, other optima can be found in the literature, such as overnight digestion at 37 C in 50 mm ammonium hydrogencarbonate buffer ph 7.8 [12] and 45 min digestion at 37 C in 10 mm ammonium hydrogencarbonate buffer ph 8.5 [13], while BRENDA displays an optimal ph range of and an optimal temperature range of C. 3 The wide variety in published digestion optima complicates the selection of the correct digestion conditions based on literature data. Furthermore, digestion conditions often are optimized for specific protein targets, such as monoclonal antibodies [14], polyclonal bovine immunoglobulin G [15] and membrane proteins [16,17], or specific applications, such as online bioreactors [18] and are mostly focused on improving the peptide yield and protein identification rate. Taken together, this underlines the need for a detailed and systematic optimization of enzymatic digestion conditions for drug protein adducts. Optimization of chemical processes is traditionally carried out using a One-Variable-At-a-Time (OVAT) approach. Commonly, a limited number of OVAT experiments are carried out in which the levels of one variable are changed while the others are kept constant [19]. A major disadvantage of OVAT approaches is the disregard of interactions between variables. Therefore, this methodology often does not lead to the true optimum and may even lead to different end results depending on the starting point [20]. In order 49

51 to avoid the local optima, more experiments need to be performed, which makes this approach more costly in terms of analysis time and consumption of chemicals [19]. In contrast, DOE techniques, such as the Response Surface Methodology (RSM), change combinations of variables simultaneously which does allow for incorporation of the interaction effects [21]. Another advantage of this technique is the concurrent optimization of multiple responses in order to find the optimal compromise between them. Additionally, RSM only requires a small subset of experiments from all possible variable combinations to cover the design space, which significantly reduces the number of necessary experiments. These advantages allow for a more efficient and more accurate determination of the optimum conditions. 3 In this study, a RSM approach was applied to the optimization of the three above mentioned conditions (buffer ph, digestion temperature and time) for digestion of HSA adducts with trypsin and thermolysin. These enzymes were selected because of their varying specificities and efficiencies [22]. HSA is the most abundant serum protein and often a target for reactive intermediates of drugs because of the free thiol on cysteine34 (Cys34) [23]. A wide range of drugs, or their metabolites, including the N-acetyl-p-benzoquinoneimine (NAPQI) intermediate from acetaminophen [7] and several intermediates of diclofenac [3], are known to covalently bind to this site in vivo, thereby causing severe ADRs [1]. For the RSM optimization experiments, a model adduct was prepared by modification of HSA with monochlorobimane (MCB), which was selected for the simplicity of the adduct formation [24]. The two responses used to evaluate the optimization were the protein coverage of HSA and the peak area of the modified Cys34 peptide. For comparison, the digestion optima obtained from the RSM and selected literature conditions were applied to the digestion of NAPQI HSA adducts. 3.2 MATERIALS AND METHODS Reagents and materials Human serum albumin (HSA), monochlorobimane (MCB), guanidine HCl (GHCl), ethanol, dl-dithiothreitol (DTT), iodoacetic acid (IAA), thermolysin from Bacillus thermoproteolyticus rokko (EC ), Tris-HCl, silver nitrate, sodium hydroxide, acetaminophen and the HPLC peptide standard mixture were purchased from Sigma Aldrich (Schnelldorf, Germany). Ammonium hydrogencarbonate, hydrochloric acid (HCl) 37% and diethyl ether were obtained from Riedel-de Haën (Seelze, Germany). Methanol, formic acid (FA), acetonitrile and chloroform came from Biosolve (Valkenswaard, The Netherlands). Trypsin from bovine pancreas (EC ) was supplied by Roche (Almere, The Netherlands), acetone by Interchema (Oosterzee, The Netherlands), Bradford reagent by Biorad (Veenendaal, The Netherlands) and the synthetic peptide H-Pro-Pro-Pro-Pro-OH (Pro4) by Bachem (Weil am Rhein, Germany). Illustra NAP-25 gel-filtration columns with a bed volume of 2.5 ml, prepacked with G-25 DNA grade Sephadex, were obtained from GE Healthcare (Diegem, Belgium). Water was purified by a Millipore (Amsterdam, The Netherlands) Milli-Q unit. 50

52 3.2.2 Design of Experiments A RSM was applied for the optimization of digestion conditions of both enzymes with respect to the digestion of HSA adducts. A face-centered Central Composite Design (CCD) with uniform precision was created using JMP from SAS Institute Inc. (Cary, NC, USA). The CCD design was used to maximize two responses (protein coverage and peak area of the adducted Cys34 peptide) by optimization of three factors (buffer ph, digestion temperature and digestion time). The factor ranges were selected based on protease supplier s instructions and the BRENDA enzyme database. The applied factor ranges were ph 6-10, C and 1-12 h for trypsin and ph 5-9, C and h for thermolysin. The complete DOE consisted of 40 randomized experiments per enzyme, including 6 center points and 1 replicate Sample preparation Preparation of MCB HSA adduct samples For the RSM experiments, the MCB HSA adduct was formed by adding a 50-fold molar excess of a 0.1 M solution of MCB in methanol to 5.5 ml of a M HSA solution in 50 mm ammonium hydrogencarbonate buffer ph 7.4. The reaction mixture was kept at 40 C for 4 h after which the excess MCB was removed with a NAP-25 gel filtration column using 2 M GHCl at ph 8.5 as the eluting buffer. The 35 cysteine residues of the denatured HSA were reduced by the addition of a 50-fold molar excess of 1 M DTT and alkylated using a 75-fold molar excess of 1 M IAA. The reduced and alkylated MCB HSA sample was split into three aliquots before being desalted using NAP-25 columns. As eluting buffers, three 50 mm ammonium hydrogencarbonate solutions were used with ph values corresponding to the three levels of the RSM design. A 100 µl aliquot of the desalted MCB HSA was then digested with either trypsin or thermolysin (0.01 mg/ml in 0.1 mm HCl) using protein:enzyme ratios of 100:1 and 50:1, respectively. 3 The enzymatic digestion was stopped with the addition of 10 µl of 10% FA. From a 12.3 mm internal standard solution of Pro4 in water, 20 µl was added to the digested MCB HSA samples to achieve a final concentration of 1.23 mm. The final volume of the samples was adjusted to 200 µl with water. The RSM experiments for trypsin and thermolysin were performed on different days and with different batches of the MCB HSA adduct. A series of confirmation experiments were performed in triplicate to test whether the determined optimum digestion conditions lead to the predicted responses. These experiments were performed using the same batch of MCB HSA for both enzymes, in order to guarantee comparability of the results. Preparation of NAPQI HSA adduct samples The optimum digestion conditions obtained from the RSM of both enzymes were compared to literature conditions using the NAPQI HSA adduct, which was prepared according to Hoos et al. [7]. This HSA adduct sample subsequently received the same treatment as described above for the MCB HSA adducts applying either the optimum RSM digestion conditions or conditions obtained from literature. The selected literature values for trypsin digestion were taken from Aldini et al. [12] and consisted of overnight (13 h) digestion 51

53 at 37 C in 50 mm ammonium hydrogencarbonate buffer ph 7.8 and a protein:enzyme ratio of 20:1. For thermolysin, the reference digestion conditions were obtained from Bark et al. [25] and consisted of 15 min digestion at 65 C in 100 mm ammonium hydrogencarbonate buffer ph 7.5 and a protein:enzyme ratio of 50: LC MS/MS analysis of digested HSA adduct samples 3 The digested HSA samples were analyzed with a 1200 series Rapid Resolution LC system coupled to a 6520 QTOF mass spectrometer (Agilent, Amstelveen, The Netherlands), that was controlled by the Agilent Masshunter Workstation Acquisition software (version B.02.00). The proteolytic peptides were separated on an Agilent XDB-C18 column (4.6 mm 50 mm, 1.8 µm particles) that was protected by a guard column (4 mm 2 mm) from Phenomenex (Utrecht, The Netherlands). Mobile phase A consisted of 5% acetonitrile and 0.1% FA in water, mobile phase B consisted of 95% acetonitrile, 5% water and 0.1% FA. The flow rate was set at 0.6 ml/min and the thermostated column compartment was maintained at 40 C. Gradient elution was performed as follows: 0% B for the first 2 min, linearly increased to 40% B in 23 min, then set to 100% B and held constant for 4 min, followed by a re-equilibration at 0% B for 9 min. Using an internal switching valve, the LC flow from 2 to 25 min was directed to the mass spectrometer, which was operated in 2 GHz, extended dynamic range mode. The electrospray ionization source was operated in positive mode (ESI+), the capillary voltage was set to 3500 V and nitrogen ( %) was used as the drying (350 C) and nebulizer gas at s flow rate of 12 L/min and a pressure of 60 psig, respectively. Profile data were acquired in data-dependent mode where the most intense ion (m/z ) was selected for fragmentation and subsequently excluded from fragmentation for 0.2 min. MS/MS spectra were recorded from m/z 50 to 3000 at a rate of 1.02 spectra/s using a fixed collision energy voltage of 20 V and nitrogen was used as the collision gas. A blank sample and the HPLC peptide standard mixture were analyzed alternatively after every four sample runs to check the stability of the LC MS system throughout the 45 h sequences Data analysis Peak extraction (using a 20 ppm half-width m/z window) and integration was performed with Agilent Masshunter Qualitative Analysis software (version B.02.00). The peak areas were normalized to the internal standard. Mascot Distiller (version 2.3.2) and Mascot server (version 2.2) (Matrix Science Ltd., London, UK, were used for peak picking of the data, identification of the peptide sequences and calculation of the protein coverage. General peak picking settings, such as a minimum precursor mass of 500 Da, maximum precursor mass of 16,000 Da, a maximum intermediate scan count of 2, a signal-to-noise (S/N) of 10, were the same for all enzymes. The only enzyme-dependent setting was the default precursor charge range, which was set to 1-5 charges for tryptic digests and to 1-4 charges for thermolytic digests. The obtained peak lists were searched against the SwissProt database with carboxymethyl (cysteine) and chlorobimane (cysteine) (monoisotopic delta mass of Da) or NAPQI (cysteine) (monoisotopic delta mass of Da) selected as variable modifications. The peptide and MS/MS tolerance was set to 0.2 Da and the number of allowed missed cleavages was set to 3 for tryptic 52

54 digests and to 4 for thermolytic digests. The cleavage definitions of thermolysin (N-terminal to A, F, I, L, and V) were determined experimentally (data not shown) and added to the enzyme database manually. Mascot peptide summary reports were formatted using an ion score cut-off value of 20 to remove random peptide matches and requiring a protein hit to include at least one top-ranking peptide match. Statistical evaluation of the RSM data was done with ANOVA functions embedded in JMP The optimum digestion conditions for maximization of both responses were determined using the Maximize Desirability function in the prediction profiler. 3.3 RESULTS AND DISCUSSION Method improvements for RSM experiments A DOE was used for efficient optimization of a multi-parameter analytical method. A RSM significantly reduces the number of experiments, which is advantageous in terms of instrumental analysis time and sample handling. Depending on the number of variables to be optimized, a predefined number of experiments must be performed for each enzyme and this number increases rapidly with each added variable. For a replicated RSM design with three variables, the number of experiments amounts to 40 and with four variables to 62. Critical and time-consuming steps in such a procedure are the preparation of the samples, the time needed to analyze them, and the time needed for data interpretation and processing. Robotic sample preparation would simplify the first step, whereas the sample analysis can be performed in unattended overnight experiments. With this in mind, it is imperative that the general analytical procedure is sufficiently well developed and optimized with respect to the parameters that are not evaluated in the DOE. Otherwise, no reliable data may be obtained within the series of experiments. Therefore, prior to starting with the RSM experiments, a number of parameters were optimized and considered, such as the chromatography, the use of an internal standard, sample handling steps prior to the actual digestion (denaturation, reduction, alkylation and removal of reagents involved in these steps). 3 LC MS separation and detection Protein digests are complex samples containing a multitude of proteolytic peptides. Successful identification of the protein under investigation relies on the acquisition of high-quality MS/MS spectra, which in turn depends on the resolution of the preceding LC separation step. Separation of protein digest samples with conventional LC systems requires long gradient runs, often exceeding 90 min [14,26], in order to achieve a satisfactory chromatographic resolution. However, it is preferable to complete the analysis of all samples from a DOE within a reasonable time in order to avoid effects of extraneous factors. The use of Ultra High Performance Liquid Chromatography (UHPLC) or Rapid Resolution Liquid Chromatography (RRLC) systems with analytical columns containing sub-2 µm particles offer substantial benefits over conventional LC separations in terms of speed and resolution. Using this technology, a fast and high-resolution separation method with a total run time of 38 min was developed for the separation of HSA digests, which 53

55 is significantly shorter than achieved with lower resolution instruments. The increased separation speed resulted in narrow peaks with peak widths at half height 0.2 min. The MS settings were, therefore, adjusted to ensure the collection of sufficient data points to accurately detect each peak to enable appropriate peak area determination for the Cys34 adducted peptide, and to obtain sufficient MS/MS data for Mascot database searches. This was achieved by selecting only the highest-intensity ion from the full scan spectrum as the precursor ion for fragmentation. Subsequent exclusion of this ion for 0.2 min allowed for selection of lower-intensity precursor ions. 3 Selection of the internal standard Since one of the measured responses of the DOE was the peak area of the MCB Cys34 peptide, it is essential to normalize the obtained peak areas. For correct mapping of the behavior of the studied analyte, the most appropriate internal standard for application in proteomics experiments is a synthetic peptide with an amino acid sequence that cannot be attributed to the protein that was digested. Considering the nature of this study, a synthetic peptide had to be selected taking into account the cleavage sites of the proteolytic enzymes for which the digestion conditions were optimized. Even though the internal standard was added after stopping the enzymatic reaction, residual enzyme activity could cleave the internal standard and thereby negatively influence peak area normalization. The best option for an internal standard was found to be a peptide consisting solely of proline residues, as this is not a cleavage site for the investigated enzymes. Sample preparation With the fast LC MS/MS method in place, reduction of sample handling steps was investigated next. From a series of preliminary experiments (data not shown), it was found that denaturation, reduction and alkylation of the protein were crucial for obtaining a high digestion efficiency. Deletion of any of these steps from the sample preparation protocol resulted in a significant decrease in the protein coverage. Moreover, failure to remove the applied denaturation, reduction and alkylating agents prior to enzymatic digestion negatively influenced the enzyme activity. Removal of the excess modifying agent after adduct formation is also necessary to prevent the formation of other cysteine adducts during the reduction of the protein S S bonds. Gel-filtration columns were used for these sample clean-up steps because of their compatibility with a wide range of common buffers, including GHCl. Therefore, denaturation and removal of the excess modifying agent could be combined in a single step by using 2.0 M GHCl as the eluting buffer. A 6.0 M GHCl buffer (ph 6) is commonly used for denaturation of proteins, but the high salt concentration was found to block the gel-filtration columns and reduced their lifetime. Using a lower concentration of the GHCl buffer did not noticeably affect the denaturation of the adducted protein. Increasing the ph value of the GHCl buffer to ph 8.5 facilitated the subsequent reduction (DTT is active at ph > 7) and alkylation reactions (IAA is selective for cysteines at ph 8.5). The alkylation process was further improved by adding a larger excess of alkylating agent over reducing agent. Previously, a 50-fold molar excess of both reagents was added, in sequence, to the denatured protein, but the presence of reducing agents during the alkylation step quenches the alkylating agent. It was observed that under these conditions all S-S bridges were reduced, but less than 35% of the detected cysteines were alkylated. By increasing the molar excess of alkylating agent to 75-fold (1.5 the excess of reducing agent), the number of detected alkylated cysteine residues increased to >90%. The denaturation, reduction and alkylation were followed by a desalting and simultaneous solvent exchange step to the appropriate digestion buffer using the gel-filtration columns. 54

56 3.3.2 RSM The above described analytical procedure was developed to facilitate the optimization of the enzymatic digestion conditions of trypsin and thermolysin. The varying specificities and efficiencies of these enzymes result in characteristic peptide profiles and corresponding protein coverages. While trypsin generally displays high protein coverage, it has only two cleavage sites and, therefore, produces larger peptide fragments. When using less than optimum digestion conditions, the increased occurrence of missed cleavages may prevent detection of these large peptides, resulting in decreased coverage. On the other hand, thermolysin has a broader specificity, having five main cleavage sites, and produces peptides with smaller chain lengths that are easily detectable. However, when the chain lengths become too small (less than 5 amino acids) elucidation of the amino acid sequence becomes more difficult, which, in turn, also leads to decreased protein coverage. Therefore, the digestion conditions of the above mentioned enzymes were optimized to maximize the coverage. The three factors to be optimized, buffer ph, digestion temperature and time, were chosen based on their influence on the digestion efficiency. Another advantage of applying a DOE is the possibility to optimize multiple responses simultaneously. Generally, protein coverage is the only response that is optimized in protein digestion experiments, but in this particular case detection of a specific part of the protein is essential, namely the site of adducting formation. In this respect, sensitivity for detection of the modification site is paramount since drug protein adducts are often low-abundant. Therefore, the peak areas of the various modified Cys34 peptides generated by the two enzymes were chosen as the second response for which the digestion process was optimized. After optimization of the digestion conditions using the MCB HSA adduct, the obtained optimum digestion protocols of the two enzymes were compared by application to a clinically relevant drug protein adduct to determine the best candidate for identification of HSA adducts. 3 Trypsin RSM results Trypsin is a widely applied proteolytic enzyme because of its high efficiency and specificity, cleaving predominantly at the C-terminal side of lysine and arginine residues. Under optimal conditions, tryptic digestions are characterized by a low number of missed cleavages and high protein coverage, which was also observed in the RSM experiments. In general, trypsin performed well under a wide range of conditions producing a protein coverage ranging from 81 to 93%, except when a combination of high ph and high temperature was used. These extreme conditions significantly reduced the digestion efficiency, regardless of the length of digestion time (1-12 h), which decreased the protein coverage to 48%. This is further exemplified by the deviating chromatographic peptide profile as compared to profiles obtained using less extreme digestion conditions (Figure 3.1). Due to the reduced digestion efficiency, the MCB Cys34 peptide was not detected in these incomplete tryptic digestion samples. In the remaining 36 experiments, the MCB Cys34 peptide was identified predominantly without missed cleavages, while the other peptides identified by Mascot exhibited 0, 1 or 2 missed cleavages. The amino acid sequences of the identified tryptic MCB Cys34 peptides can be found in Table 3.1. The sequence coverage of the MCB Cys34 peptide without missed cleavages was 71% allowing for accurate identification of the modification site. 55

57 3 x Counts vs. Acquisition Time (min) Figure 3.1 LC MS chromatograms showing the peptide profiles that are obtained through tryptic digestion of MCB HSA under different conditions. Solid line: digestion at ph 10, 50 C for 12 h. Dotted line: digestion at ph 8, 37 C for 6 h. Table 3.1 Sequences of the detected MCB Cys34 peptides obtained through tryptic or thermolytic digestion of MCB HSA. Enzyme MCB Cys34 peptide # Missed cleavages m/z [M+nH] n+ Sequence cov. (%) Trypsin ALVLIAFAQYLQQCaPFEDHVK ALVLIAFAQYLQQCaPFEDHVKLVNEVTEFAK Thermolysin LQQCaP 0 nd LQQCaPFEDH LQQCaPFEDHVK LQQCaPFEDHVKL nd Not detected a Indicates the MCB Cys34 modification site Amino acids in italics represent missed cleavages 56

58 Evaluation of the RSM results showed that ph and temperature influenced both responses. Figure 3.2a shows a surface plot of the factors ph and temperature for the response protein coverage. It reveals that the digestion ph should be decreased accordingly when a higher digestion temperature is applied and vice versa, in order to achieve similar protein coverage. The highest predicted protein coverage (96%) is obtained when performing the digestion at ph 6 and a temperature of 47 C. The factor time by itself did not seem to have an influence on either of the responses. However, the surface plot of the effect of time and temperature on the peak area of the MCB Cys34 peptide shows that a combination of a low digestion temperature and long digestion time leads to the highest peak area of this peptide. Protein coverage, on the other hand, would benefit from a higher digestion temperature, but digestion at elevated temperatures significantly decreases the peak area of the MCB Cys34 peptide, which would mean a loss in sensitivity of the method. Therefore, a compromise has to be made to obtain the optimal digestion conditions for both responses. Using the Maximize Desirability function of the prediction profiler, the most desirable factor settings to maximize both responses simultaneously were obtained through an iterative process. The prediction profiler in Figure 3.3 shows that for this specific optimization experiment the optimum tryptic digestion conditions are digestion at ph 8.4 and 24 C for 11.6 h. The predicted maximum response values for these conditions are protein coverage of 91% and a peak area of the ALVLIAFAQYLQQC(MCB)PFEDHVK peptide of The optimum ph and time values are comparable with the manufacturer s instructions, but the found optimum temperature is lower than expected. The discrepancy between the RSM optima and those indicated by the manufacturers could be explained by the fact that the conditions for optimal enzyme activity are often determined using relatively small substrates. Judging by the wide range of trypsin digestion optima applied to different substrates published in the literature, the enzyme activity optima cannot be directly extrapolated to every substrate or analytical system. In addition, protein digestion is generally evaluated by protein coverage and the use of a second response in this study may also lead to different optimum conditions. 3 Although the protein coverage remained reasonably stable and the MCB Cys34 peptide could be identified under most digestion conditions, the peak area of this peptide changed significantly with varying conditions. Therefore, increasing the temperature to 37 C would lead to a significant loss in peak area of the MCB Cys34 peptide. Additionally, the prediction profiler also showed that a shorter digestion time would lead to a slight increase in protein coverage. However, this would negatively affect the peak area of the MCB Cys34 and, thus, is not advantageous for the whole model. Furthermore, a significant decrease in MS signal was observed during the long LC MS sequence of the 40 trypsin RSM samples. A possible explanation for this observation could be the analysis of incomplete tryptic digests obtained under less than optimal conditions. In general, trypsin produces larger peptide fragments and with an incomplete digestion these even larger fragments may pollute the ion source. Statistical evaluation of the RSM by Analysis of Variance (ANOVA) showed that a well-fitted model (R2 > 0.8 [27]) was obtained for both responses. The R 2 values for protein coverage and peak area of the MCB Cys34 peptide were and 0.894, respectively, indicating that the model could explain 85.9% and 89.4% of the variation in the respective responses. Additionally, F- tests revealed that the regression for both 57

59 a b 3 Figure 3.2 Surface plots showing the effects of buffer ph and digestion temperature on protein coverage (a) and temperature and time on the peak area of the MCB Cys34 peptide (b). Protein coverage % ± Peak area Cys341m.cl ± Desirability Temperature ( C) Time (hrs) Buffer ph Desirability Figure 3.3 Prediction profiler plot showing the optimum tryptic digestion conditions and the predicted responses for the complete model under these conditions. The dashed lines indicate the 95% confidence intervals, the dotted vertical lines indicate the optimum level for each of the factors and the dotted horizontal lines show the predicted value of both responses ± their standard deviations. 58

60 responses (F-values of and 28.04) was statistically significant at a confidence level of >99.9% and that an insignificant proportion of the pure error is explained by variation due to lack of fit (F-values of and 5.03 with P-values of and ). The centerpoint experiments showed limited variation for protein coverage with a %RSD of 2.4% and slightly more for the peak area of the MCB Cys34 peptide with a %RSD of 8.1%. The predictive quality of the model was exemplified by the confirmation experiments performed at the optimum tryptic digestion conditions. An average (n = 3) protein coverage of 86 ± 1% and a peptide peak area of ± 15% were obtained, which is in good agreement with the predicted responses. Thermolysin RSM results Thermolysin is a thermostable enzyme that is able to withstand elevated temperatures. The optimum digestion temperature, according to the supplier, is 70 C at a ph value of 8.0, but an optimal digestion time is not supplied and BRENDA does not contain information about the digestion time either. However, thermolysin is suggested to exhibit an accelerated rate of reaction at elevated temperatures [25]. Therefore, a wider temperature range of C and shorter digestion times of h were chosen for this design. Table 3.2 shows the complete design and the obtained response values. From this table, it can already be seen that the results can be clustered into three groups consisting of low, intermediate and high response values. Especially for the low values, the response can be correlated to the application of a specific combination of conditions, in this case high ph and high digestion temperature. 3 As anticipated, the peptide profile resulting from thermolytic digestion of MCB HSA is different from that obtained with tryptic digestion. The LC MS chromatograms shown in Figure 3.4 contains a higher number x Counts vs. Acquisition Time (min) Figure 3.4 LC MS chromatograms showing the peptide profiles that are obtained through thermolytic digestion of MCB HSA under different conditions. Solid line: digestion at ph 7, 55 C for 4.25 h. Dotted line: digestion at ph 9, 30 C for 0.5 h. 59

61 3 Table 3.2 RSM design and experimental data of thermolysin digestion optimization. Exp. # ph C Hours Protein coverage (%) Peak area MCB Cys34 ( 10 6 ) 1 7a a a a a a a a a a a a a Centerpoint 60

62 of peptides due to the lower specificity of this enzyme and a larger number of missed cleavages. Due to the higher number of missed cleavages, several MCB Cys34 peptides with different chain lengths were detected (see Table 3.1). Although, for this enzyme, a proline residue in this position should not interfere with the cleavage, the shortest possible peptide with a sequence of LQQCP was not detected. This could be due to the fact that the MCB adduct is located on the cysteine next to the proline or because of the digestion efficiency of this enzyme.three other MCB Cys34 peptides with an increasing number of missed cleavages, LQQCPFEDHV, LQQCPFEDHVK and LQQCPFEDHVKL, were detected. Of these three, the MCB Cys34 peptide with two missed cleavages (LQQCPFEDHVK) had the highest intensity and sequence coverage of 100% allowing for unambiguous identification of the modification site. Therefore, this peptide was used as the second response of the RSM, besides the protein coverage. Similar to trypsin, the application of extreme conditions in experiments 6, 12, 37 and 40 led to a low protein coverage of <15% and failure to detect the MCB Cys34 peptide. Although the protein coverage in the remaining experiments, ranging from 44 to 74%, was lower than that obtained with trypsin, still high-confidence identification of the target protein could be achieved. Again, the main parameters temperature and ph proved to have an effect on the response values, as shown in Figure 3.5a for the peak area of the MCB Cys34 peptide. This surface plot also shows that there is a true optimum temperature and buffer ph for achieving maximum peak area, which is at ph 7.4 and 38 C. Somewhat surprisingly, as shown in Figure 3.5b, there was little effect of time on protein coverage for this specific system. Moreover, lower, rather than elevated temperatures resulted in the highest obtained protein coverage. Although the factor time did not have a large effect on either of the two responses (Figure 3.6), the highest peak area of the MCB Cys34 peptide was achieved after 8 h of digestion. 3 Slightly different optimum conditions were obtained for each of the responses separately, but the optimum conditions for the whole model were digestion at ph 7.5 and 34 C for 8 h. The predicted protein coverage and peak area of the MCB Cys34 peptide at these conditions were 69% and , respectively. As mentioned above, the optimum digestion temperature is much lower than expected based on supplier s instructions and literature, and a longer digestion time is preferred. The optimum ph also does not concur with the provided instructions, but was an exact match with the literature ph value. Statistical analysis was also performed on the thermolysin data revealing that the RSM model fitted well, represented by R 2 values of and for protein coverage and peak area of the MCB Cys34 peptide, respectively. Similar to the trypsin RSM model, F-tests at a >99.9% confidence level indicated that the regression for both responses was statistically significant (F-values of and 41.48) and that an insignificant proportion of the pure error is explained by variation due to lack of fit (F-values of and ). The centerpoint experiments showed limited variation for both responses with %RSD values of 5.4% for protein coverage and 6.3% for peak area of the MCB Cys34 peptide. The confirmation experiments resulted in protein coverage of 62 ± 2% and a peak area of ± 3%, which is in good agreement with the RSM optima as well. 61

63 a b 3 Figure 3.5 Surface plots showing the effects of buffer ph and digestion temperature on the peak area of the MCB Cys34 peptide (a) and temperature and time on the protein coverage (b). Protein coverage % ± Peak area MCB- Cys34 2 m.cl ± Desirability Temperature ( C) Time (hrs) Buffer ph Desirability 1 Figure 3.6 Prediction profiler plot showing the optimum thermolytic digestion conditions and the predicted responses for the complete model under these conditions. The dashed lines indicate the 95% confidence intervals, the dotted vertical lines indicate the optimum level for each of the factors and the dotted horizontal lines show the predicted value of both responses ± their standard deviations. 62

64 3.3.3 Comparison experiments using NAPQI HSA Since the RSM optimization was performed using a model HSA adduct, the optimized conditions for tryptic and thermolytic digestion were applied to another, clinically relevant HSA adduct. Additionally, a comparison was made with conditions obtained from literature to assess whether the optimized protocols lead to better results. These values, as well as the RSM optima and the responses are shown in Table 3.3. Tryptic digestion of NAPQI HSA using the RSM optimum conditions leads to similar protein coverage and peak area of the modified Cys34 peptide, as compared to those obtained with MCB HSA. The same can be concluded for the protein coverage obtained with thermolytic digestion of either MCB HSA or NAPQI HSA. However, the peak area of the NAPQI Cys34 peptide is 4-fold lower than the peak area of the MCB Cys34 peptide. This may most likely be attributed to different ionization efficiencies of the adducted peptides. These results show that the optimized digestion protocol can be successfully applied to different HSA Cys34 adducts, as long as the modification itself does not interfere with cleavage of the protein due to sterical hindrance, for instance. For both enzymes, the optimized digestion protocols lead to higher protein coverage and peak area of the NAPQI Cys34 peptide than obtained with values chosen from literature. Trypsin is a very versatile enzyme and will lead to sufficiently high protein coverage even at less than optimal conditions, which was also evident from the RSM results. However, when investigating a certain part of the protein or modification site, it is worthwhile to optimize the digestion conditions and test different enzymes to improve the limit of detection for the peptide containing this site. With the optimized conditions from the RSM, the peak area of the tryptic NAPQI Cys34 peptide was 50% higher than that obtained with the conditions taken from literature. In the case of thermolysin, the applied digestion conditions had a large effect on both responses. Using the RSM optima for thermolytic digestion, the protein coverage improved by 1.5-fold and the NAPQI Cys34 peak area by 7-fold, which may become crucial when analyzing lower concentrations of the adducted protein. 3 Table 3.3 Digestion conditions and results from the comparison experiment using NAPQI HSA. Digestion conditions ph Temp. (C ) Time (h) Protein cov. (% ± %RSD) Peak area NAPQI Cys34 ( 10 5 ± %RSD) Trypsin RSM optima ± ± 2 Literature ± ± 5 Thermolysin RSM optima ± ± 1 Literature ± ± 30 63

65 3.4 CONCLUSION 3 The results presented in this study highlight the importance of detailed optimization of protein digestion conditions for each analytical system. In particular, when specific target peptides have to be detected, such as in proteomics-based drug protein adduct studies where detection and identification of the modification site is paramount, much stands to be gained. By applying a DOE, interactions between the different factors were taken into account and a suitable compromise could be made to obtain optimum conditions for both responses, thereby rendering OVAT approaches inadequate. The optimum digestion conditions found for trypsin and thermolysin showed discrepancies with the optima given by the supplier and the values found in the literature. Especially for thermolysin, these differences were significant, rendering a thorough optimization of the digestion conditions for this enzyme mandatory, as shown here in the field of drug protein adduct research. In both cases, the optimized digestion protocols lead to improved digestion of the studied HSA adducts. ACKNOWLEDGEMENTS We thank Filipe Santos and Johan van Heerden from the Molecular Cell Physiology department at the Faculty of Earth and Life Sciences of the VU University Amsterdam for their help with the statistical evaluation of the data. This work was performed within the framework of the Dutch Top Institute Pharma project Towards novel translational safety biomarkers for adverse drug toxicity [Project no. D ]. REFERENCES [1] Zhou, S., Chan, E., Duan, W., Huang, M., Chen, Y.Z., Drug bioactivation, covalent binding to target proteins and toxicity relevance. Drug Metab. Rev. 2005, 37, (1), [2] Evans, D.C., Watt, A.P., Nicoll-Griffith, D.A., Baillie, T.A., Drug protein adducts: an industry perspective on minimizing the potential for drug bioactivation in drug discovery and development. Chem. Res. Toxicol. 2004, 17, (1), [3] Zhou, S., Separation and detection methods for covalent drug protein adducts. J. Chromatogr. B. 2003, 797, (1-2), [4] Zheng, J., Ma, L., Xin, B., Olah, T., Humphreys, W.G., et al., Screening and identification of GSH-trapped reactive metabolites using hybrid triple quadruple linear ion trap mass spectrometry. Chem. Res. Toxicol. 2007, 20, (5),

66 [5] Ma, S., Zhu, M., Recent advances in applications of liquid chromatography-tandem mass spectrometry to the analysis of reactive drug metabolites. Chem. Biol. Interact. 2009, 179, (1), [6] Ma, S., Subramanian, R., Detecting and characterizing reactive metabolites by liquid chromatography/tandem mass spectrometry. J. Mass Spectrom. 2006, 41, (9), [7] Hoos, J.S., Damsten, M.C., de Vlieger, J.S.B., Commandeur, J.N.M., Vermeulen, N.P.E., et al., Automated detection of covalent adducts to human serum albumin by immunoaffinity chromatography, online solution phase digestion and liquid chromatography - mass spectrometry. J. Chromatogr. B. 2007, 859, (2), [8] Damsten, M.C., Commandeur, J.N.M., Fidder, A., Hulst, A.G., Touw, D., et al., Liquid chromatography/tandem mass spectrometry detection of covalent binding of acetaminophen to human serum albumin. Drug Metab. Dispos. 2007, 35, (8), [9] Carol-Visser, J., van der Schans, M., Fidder, A., Huist, A.G., van Baar, B.L.M., et al., Development of an automated online pepsin digestion-liquid chromatography-tandem mass spectrometry configuration for the rapid analysis of protein adducts of chemical warfare agents. J. Chromatogr. B. 2008, 870, (1), [10] Badghisi, H., Liebler, D.C., Sequence mapping of epoxide adducts in human hemoglobin with LC-tandem MS and the SALSA algorithm. Chem. Res. Toxicol. 2002, 15, (6), [11] Yang, X.X., Hu, Z.P., Chan, S.Y., Zhou, S.F., Monitoring drug protein interaction. Clin. Chim. Acta. 2006, 365, (1-2), [12] Aldini, G., Vistoli, G., Regazzoni, L., Gamberoni, L., Facino, R.M., et al., Albumin is the main nucleophilic target of human plasma: a protective role against pro-atherogenic electrophilic reactive carbonyl species? Chem. Res. Toxicol. 2008, 21, (4), [13] Allardyce, C.S., Dyson, P.J., Coffey, J., Johnson, N., Determination of drug binding sites to proteins by electrospray ionisation mass spectrometry: the interaction of cisplatin with transferrin. Rapid Commun. Mass Spectrom. 2002, 16, (10), [14] Dick, L.W., Jr., Mahon, D., Qiu, D., Cheng, K.C., Peptide mapping of therapeutic monoclonal antibodies: improvements for increased speed and fewer artifacts. J. Chromatogr. B. 2009, 877, (3), [15] Cresswell, C., Newcombe, A.R., Davies, S., Macpherson, I., Nelson, P., et al., Optimal conditions for the papain digestion of polyclonal ovine IgG for the production of bio-therapeutic Fab fragments. Biotechnol. Appl. Biochem. 2005, 42, (Pt 2), [16] Rietschel, B., Bornemann, S., Arrey, T.N., Baeumlisberger, D., Karas, M., et al., Membrane protein analysis using an improved peptic in-solution digestion protocol. Proteomics. 2009, 9, (24), [17] Fischer, F., Poetsch, A., Protein cleavage strategies for an improved analysis of the membrane proteome. Proteome Sci. 2006, 4, 2. [18] Calleri, E., Temporini, C., Perani, E., Stella, C., Rudaz, S., et al., Development of a bioreactor based on trypsin immobilized on monolithic support for the online digestion and identification of proteins. J. Chromatogr. A. 2004, 1045, (1-2), [19] Bezerra, M.A., Santelli, R.E., Oliveira, E.P., Villar, L.S., Escaleira, L.A., Response surface methodology (RSM) as a tool for optimization in analytical chemistry. Talanta. 2008, 76, (5), [20] Tye, H., Application of statistical design of experiments methods in drug discovery. Drug Discov. Today. 2004, 9, (11), [21] Gooding, O.W., Process optimization using combinatorial design principles: parallel synthesis and design of experiment methods. Curr. Opin. Chem. Biol. 2004, 8, (3), [22] Chen, R., Jiang, X., Sun, D., Han, G., Wang, F., et al., Glycoproteomics analysis of human liver tissue by combination of multiple enzyme digestion and hydrazide chemistry. J. Proteome Res. 2009, 8, (2), [23] Funk, W.E., Li, H., Iavarone, A.T., Williams, E.R., Riby, J., et al., Enrichment of cysteinyl adducts of human serum albumin. Anal. Biochem. 2010, 400, (1), [24] Fernandez-Checa, J.C., Kaplowitz, N., The use of monochlorobimane to determine hepatic GSH levels and synthesis. Anal. Biochem. 1990, 190, (2), [25] Bark, S.J., Muster, N., Yates, J.R., 3rd, Siuzdak, G., High-temperature protein mass mapping using a thermophilic protease. J. Am. Chem. Soc. 2001, 123, (8), [26] Qiu, Y., Benet, L.Z., Burlingame, A.L., Identification of the hepatic protein targets of reactive metabolites of acetaminophen in vivo in mice using two-dimensional gel electrophoresis and mass spectrometry. J. Biol. Chem. 1998, 273, (28), [27] Lundstedt, T., Seifert, E., Abramo, L., Thelin, B., Nystrom, A., et al., Experimental design and optimization. Chemom. Intell. Lab. Syst. 1998, 42, (1-2),

67

68 Chapter 4 A high-throughput sample preparation method for cellular proteomics using 96-well filter plates Linda Switzar*, Jordy A. van Angeren*, Martijn Pinkse, Jeroen Kool and Wilfried M.A. Niessen Accepted for publication in Proteomics *Authors contributed equally

69 ABSTRACT A high-throughput sample preparation protocol based on the use of 96-well molecular weight cut-off (MWCO) filter plates was developed for bottom-up proteomics. All sample preparation steps, including cell lysis, buffer exchange, protein denaturation, reduction/alkylation and proteolytic digestion are performed in a 96-well plate format, making it an ideal platform to process large number of samples and is also directly compatible with functional assays for cellular proteomics. In addition, the usage of a single plate for all sample preparation steps following cell lysis reduces potential samples losses and allows for convenient automation. The MWCO filter also enables sample concentration, thereby increasing the overall sensitivity, and implementation of washing steps involving organic solvents, for example, to remove cell membranes constituents. The optimized protocol was found to be superior in sample throughput and sensitivity in terms of the number of identified cellular proteins when compared to a protocol employing gel-filtration columns. 4 68

70 4.1 INTRODUCTION Proteomics studies often require the analysis of a large number of samples, such as the large sample sets needed for disease biomarker discovery [1] or due to fractionation of complex proteomes [2]. However, sample preparation for bottom-up proteomics involves many steps and, as such, is very labor-intensive and time-consuming. [3] Procedures based on treatment of individual samples, for instance using solid-phase extraction (SPE) [4] or gel-filtration columns [5], result in significant time and buffer consumption. These disadvantages can be overcome by the use of the filter-aided sample preparation (FASP) protocol employing spin filters. [6] The molecular weight cut-off (MWCO) membrane filters used in such a protocol can be used for the treatment of complex samples, such as cell lysates, while reducing the sample preparation time and sample losses. However, the number of samples that can be treated simultaneously is still limited due to the manual steps involved. High-throughput preparation of large sample sets is more efficiently achieved in the 96-well format. For in-gel digestion, protocols based on a standard 96-well plate [7] or filter plate [8] have been described. In these formats, buffer-exchange and desalting can easily be achieved via evaporation or vacuum filtration, respectively, because the proteins are contained within the gel pieces. More complex digestion protocols that include protein denaturation, reduction, alkylation and buffer-exchange/desalting steps, can also be performed in 96-well format. One protocol for in-gel digestion that employs 96-well capillary plates allows for most sample preparation steps to be performed in a single plate, but requires a second plate for clean-up of the digest. [9] For in-solution digestion, a protocol was developed employing a 96-well plate for protein denaturation, reduction and alkylation, followed by buffer-exchange via gel-filtration in a second 96-well plate and, finally, elution and digestion in a third 96-well plate. [10] These protocols require multiple sample transfer steps and may result in sample loss. To our best knowledge, only one 96-well protocol, based on strong cation exchange [11], allows for all sample preparation steps to take place in a single plate. However, all aforementioned protocols were developed for pre-fractionated protein samples, but not for complex samples, such as whole cell lysates that may also require the use of organic solvents. 4 The aim of this study was to develop a high-throughput sample preparation method in 96-well format for cellular proteomics that limits the number of sample transfer steps. For this purpose, a 96-well filter plate with a 10 kda MWCO membrane was selected for its ease of use, organic solvent compatibility, low dead volume and universal application through the size-exclusion mechanism. MWCO membranes are compatible to all of the above mentioned sample pre-treatment steps and the 96-well format would also allow for automation using robotics. The final procedure (Figure 1) includes cell lysis in a 96-well plate followed by transfer of the cell lysates to the filter plate in which all further sample treatment steps, including removal of cell debris, buffer-exchange, denaturation, reduction, alkylation, and digestion, are performed prior to collection of the peptides into a 96-well plate and subsequent LC MS analysis. The methodology was optimized using human embryonic kidney (HEK293T) cell lysates and compared to an existing column-based gel-filtration protocol [5]. 69

71 Cell culture and lysis 96-well plate Sample transfer 1:1 Transfer to 96- well filter plate Dilution in Eppendorf tubes Desalting and buffer-exchange Centrifugation and resuspension in filter plate Gel-filtration 4 Denaturation, reduction and alylation Desalting, buffer-exchange and concentration Incubation reactions in filter plate Centrifugation and resuspension in filter plate Incubation reactions in Eppendorf tubes Gel-filtration Freeze-drying and resuspension in Eppendorf tubes Incubation in filter plate Incubation in Eppendorf tubes. Digestion Stop digestion by centrifugation and collection of peptides in 96-well plate Stop digestion by addition of acid Figure 4.1 Overview of the filter plate and gel-filtration protocols. Each arrow represents a sample transfer step. An additional methanol wash or elution step may be included in the filter plate protocol following sample transfer or peptide collection, respectively. 70

72 4.2 MATERIALS AND METHODS Chemicals Acetone, dichloromethane, methanol, [Met 5 ]enkephalin, HPLC standard peptide mixture, dithiothreitol (DTT), guanidine HCl (GHCl), human serum albumin (HSA), iodoacetic acid (IAA), sodium deoxylate, sodium dodecyl sulfate, dibasic sodium phosphate, monobasic potassium phosphate, potassium chloride and sodium chloride were obtained from Sigma-Aldrich (Schnelldorf, Germany). Chloroform, formic acid (FA) and LC MS grade acetonitrile (ACN) were purchased from Biosolve (Valkenswaard, The Netherlands). Ammonium bicarbonate (ABC) was supplied by Riedel-de Haën (Seelze, Germany). Illustra NAP-5 gel filtration columns prepacked with Sephadex G-25 DNA grade resin were purchased from GE Healthcare (Diegem, Belgium). The Pierce bicinchoninic acid (BCA) Protein Assay Kit was obtained from Pierce (Rockford, IL, USA). Nonidet P-40 and trypsin from bovine pancreas (EC ) were obtained from Roche (Almere, The Netherlands). Multiscreen 96-well filter plates containing an Ultracel-10 membrane with a 10 kd MWCO were purchased from Millipore (Amsterdam, The Netherlands). Water was obtained from an in-house Millipore MilliQ system Cell culture and lysis HEK293T cells were cultured in dulbecco s modified eagle medium with high glucose concentration supplemented with 10% fetal bovine serum, 1% penicillin and 1% streptomycin at 37 C in 5% CO2. Plating of 2 million cells per Petri dish was achieved. Prior to cell lysis, the culture medium was carefully removed from the cells in the Petri dishes through vacuum pipetting at room temperature. The cells were washed three times at room temperature with 3 ml chilled phosphate buffered saline (PBS, 8.1 mm dibasic sodium phosphate, 1.5 mm monobasic potassium phosphate, 2.7 mm potassium chloride and 137 mm sodium chloride) for the removal of traces of serum and medium. Subsequently, 300 µl lysis buffer (1% Nonidet P-40, 0.5% sodium deoxylate and 0.1% sodium dodecyl sulfate in PBS, ph 7.4) was added to the cells followed by incubation on ice for 10 min. Afterwards the resulting suspension was swirled in the Petri dishes. Using a cell scraper, the bottom of the culture dishes was scraped to assure suspension of the complete content in the dish. The cell solution was transferred to a 2 ml Eppendorf vial and centrifuged at 15,000 rpm for 10 min at 4 ºC using an Eppendorf centrifuge 5415 R from Eppendorf (GmbH, Engelsdorf, Germany) for removal of cell debris. The supernatant was collected and the protein content (3.5 mg/ml) was determined using a BCA assay. Prior to sample preparation, the cell lysate samples were 10-fold diluted with lysis buffer. 4 For cell lysis in 96-well plate, the cells from a Petri dish were divided over the wells of a 96-well plate, which is the normal procedure for functional assays. Plating of 70,000 cells/well of the 96-well plate was achieved. The cells in each well were washed three times with 50 µl chilled PBS. Subsequently, 200 μl of lysis buffer was added to each well and the 96-well plates were placed in a shake plate at 100 rpm for 30 min at 4 ºC. The plates were subsequently centrifuged for 1 h at 4 ºC to spin off cell debris using a Multifuge 3 S-R centrifuge (Heraeus Instruments, Kendro, Newtown, CT, USA). The average protein content in the wells was determined to be 99 µg/ml (n = 3) using the BCA assay. 71

73 4.2.3 Sample preparation Initial filter plate protocol with additional methanol washing and elution steps The filters were first equilibrated with 100 μl of lysis buffer followed by centrifugation at 4,000 rpm for 1 h at 4 ºC. Subsequently, 100 μl of a 10-fold diluted cell lysate sample obtained from a Petri dish culture (35 µg of protein) was spiked with HSA (0.1, 0.01 or 0 mg/ml) and added to the wells of the filter plate followed by 1 h of centrifugation. The resulting protein pellets were then washed using 100 µl MilliQ. For one set of samples, the protein pellets in the wells were resuspended in 100 μl of denaturation buffer (2 M GHCl in 50 mm ABC, ph 8.5) and incubated in the dark at 4 ºC for 1 h. For the other set of samples, the protein pellet was first washed with 50% methanol prior to resuspension in denaturation buffer. 4 Afterwards, the proteins were reduced by the addition of 50 μl 0.1 M DTT to the wells and incubation of the filter plates at 50 ºC for 30 min. The reduced cysteines were then alkylated by the addition 75 μl of 0.1 M IAA to the wells, followed by incubation in the dark for 30 min on a shake plate at 4 ºC. The excess reducing and alkylation agents were removed through centrifugal filtration for 2 h at 4 ºC. The proteins were washed using 100 μl digestion buffer (50 mm ABC, ph 8.5) and then resuspended in 100 µl digestion buffer containing 30 μl of 0.1 mg/ml trypsin and the plates were incubated at 37 ºC overnight. The following day, the digestion was quenched by the addition of 10 µl of a 10% formic acid (FA) solution followed by the recovery of the peptides in a 96-well plate through filtration of the filter plate for 1.5 h at 4 ºC. A second elution step with 100 µl of 1% FA in 50% methanol was applied to a subset of samples. Finally, the internal standard, [Met5]enkephalin, was added to all elution fractions at a final concentration of 1.7 µm. Final high-throughput 96-well filter plate protocol The initial filter plate protocol, without washing and elution steps, was slightly modified for the final comparison experiment. Following cell lysis in 96-well format, the cell lysates (n = 3, 200 µl, 20 µg of protein) were transferred from the wells of the 96-well plate to the corresponding wells of the equilibrated filter plate. Following 1 h of centrifugation to remove the lysis buffer, the proteins in the wells were resuspended in 100 μl of denaturation buffer and incubated in the dark at 4ºC for 30 min. Afterwards, reduction, alkylation, buffer-exchange, digestion and peptide collection was performed as above. Column-based gel-filtration protocol The cell lysate content (200 µl, 20 µg) of a well from the 96-well culture plate was diluted to 750 µl (n = 3) and applied to the gel-filtration columns that were equilibrated with 10 ml of denaturation buffer. The proteins were subsequently eluted into an Eppendorf tube using 750 µl of denaturation buffer. The denatured proteins were reduced by the addition of 10 µl of 1 M DTT and alkylated by the addition of 20 µl of 1 M IAA using the same incubation conditions as above. The samples were subsequently applied to gel-filtration columns equilibrated with 10 ml of MilliQ water and eluted with 750 µl of MilliQ into Eppendorf vials. These samples were freeze-dried for 3 h and resuspended and digested as above. Following overnight incubation at 37 ºC, 10 µl of a 10% FA solution and 10 µl of a 10 µm internal standard solution was added to the samples. 72

74 4.2.4 LC MS analysis In initial experiments, the digested cell lysate samples were analyzed with a Series 1200 Rapid Resolution LC system coupled to a 6520 QTOF mass spectrometer equipped with an electrospray ionization source operated in positive ion mode (Agilent, Amstelveen, The Netherlands), that was controlled by the Agilent Masshunter Workstation Acquisition software (version B.02.00). Separation of the peptides was achieved using an Agilent XDB-C18 column (50 x 4.6 mm I.D., 1.8 μm particle size) protected by a SecurityGuard C18 guard column (4 x 2 mm I.D.) from Phenomenex (Utrecht, The Netherlands) operated at a constant flow rate of 0.6 ml/min and a temperature of 40 ºC. The mobile phases consisted of (v/v) 5% ACN and 0.1% FA in MilliQ for mobile phase A and 95% ACN and 0.1% FA in MilliQ for mobile phase B. Gradient elution was performed using a method with a run time of 42 min that held the %B constant at 0% for the first 2 min, then linearly increased to 40% B in 23 min, followed by a wash step at 100% B for 7 min and a re-equilibration step at 0% B for 10 min. Using an internal switching valve, the LC flow was only directed to the MS from 2 to 25 min, which was operated in 2 GHz extended dynamic range mode. The capillary voltage was set to 3500 V and nitrogen ( %) was used as the drying (350 C) and nebulizer gas at a flow rate of 12 L/min and a pressure of 60 psig, respectively. Profile data were acquired in data-dependent mode where the most intense ion in the range of m/z was selected for fragmentation and subsequently excluded for the next 0.2 min. MS/MS spectra were recorded from m/z 50 to 3000 at a rate of 1.02 spectra/s using a fixed collision energy voltage of 20 V; nitrogen was used as the collision gas. The digested cell lysate samples from the comparison experiment were analyzed in duplicate on a reversedphase nanolc coupled to a LTQ-Orbitrap Velos MS system (Thermo Fisher Scientific, Bremen, Germany). A vented column nanolc setup [12] was configured on a Agilent Series 1200 HPLC system using an in-house packed capillary trapping column (20 x 0.1 mm I.D., 5 μm particle size) and analytical column (300 x 0.05 mm I.D., 5 μm particle size) filled with Reprosil Pur 120 C18-AQ (Dr. Maisch, Ammerbuch-Entringen, Germany). The mobile phases consisted of 0.6% acetic acid in MilliQ for mobile phase A and 80% ACN and 0.6% acetic acid in MilliQ for mobile phase B. One μl of digested cell lysate was loaded on the trapping column at constant flow of 5 μl/min for 5 min of 100% mobile phase A. Gradient elution was performed at a flow of nl/min using a method with a run time of 85 minutes that started from 0% B to 40% B in 65 min, then to 95% B in 5 min which was held constant for 5 min, followed by a re-equilibration step at 0% B for 10 min The column effluent was directly electrosprayed in the ion source of the linear ion trap operating in the positive ion mode. The mass spectrometer was programmed to operate in data-dependent mode, automatically switching between MS and MS/MS. Survey full spectrum MS spectra were acquired from m/z 400 to 1500 in the Orbitrap analyzer at a resolution of 30,000 at m/z 400 after accumulation of ions to a target value of 1 x The ten most intense multiply charged ions above a threshold of 1000 were isolated, fragmented and analyzed in the linear ion trap after accumulation to a target value of 1 x 104. The isolation width was set to 2.5 amu, the normalized collision energy at 35% and dynamic exclusion was set to 90 s. 4 73

75 4.2.5 Data analysis The LC MS data files were imported into Mascot distiller (version ; Matrix Science Ltd., London, United Kingdom) for generation of peak lists. General peak picking settings for trypsin digestion and the Agilent QTOF MS system included a minimum precursor mass of 500 Da, a precursor m/z tolerance of 1.2 Da and a maximum intermediate scan count of 2. For MS processing, the maximum charge per peak was set to 4 and for MS/MS processing this was set to a maximum of 2 charges with a precursor charge range of 1-4. For data obtained with the nanolc Orbitrap MS system, the precursor charge range was changed to 1-5 and the maximum intermediate scan count was set to The generated peak lists were uploaded to an in-house Mascot server (version ; Matrix Science Ltd., London, United Kingdom) and searched against the SwissProt database (dated ) and a decoy database. Mascot searches were performed using tryptic cleavage conditions allowing for 2 missed cleavages, carboxymethylation as a variable modification and the taxonomy was set to Homo Sapiens. For the QTOF data, the peptide and MS/MS tolerance were set to 0.1 Da, while for the orbitrap data these settings were 5 ppm and 0.9 Da, respectively. The search results were imported into Scaffold (Version 3.6.4, Proteome Software Inc., Portland, OR, USA) and searched with X! Tandem using a subset of the same SwissProt database. Protein identification criteria included at least 2 unique peptides for each protein hit, protein identification probabilities of 99% and peptide identification probabilities of 95%. With these settings, false discovery rates (FDR) of <0.1% were obtained. 4.3 RESULTS AND DISCUSSION Initial experiments Initially, parameters like solvent compatibility, centrifugation time, protein recovery, and filter bleeding were evaluated. The filter plates are compatible with all of the solvents needed, but the supplied polystyrene lids should be replaced by polyolefin lids that are compatible with organic solvents. The solution flow-through time solely depended on the volume loaded into the wells of the filter plate. As acknowledged by the supplier, differences in flow rate were observed depending on the well location in the filter plate. However, complete flow-through of 200 µl of any of the solutions used was achieved within 2 h at 4,000 rpm and 4 ºC. The recovery of HSA (at 1.0, 0.2 and 0.04 mg/ml in lysis buffer) was tested following centrifugation and resuspension in lysis buffer. Every 15 min, a sample was taken and analyzed with the BCA assay. Complete resuspension was achieved within 15 min with an average HSA recovery of 107 ± 11% (n = 3, see Figure 4.2. Prolonged exposure of the filter plates to a HSA solution in lysis buffer in the absence of centrifugal force did not lead to filter bleeding after two days at 4 C. 74

76 Recovery 140% 120% 100% 80% 60% 40% 20% 1 mg/ml 0.2 mg/ml 0.04 mg/ml 0% Time (min) Figure 4.2 Resuspension of HSA in the filter plates following centrifugation measured as the recovery at 15 min time intervals. Unique peptides a Without MeOH With MeOH b Without MeOH Proteins With MeOH c 1 st Elution 2 nd Elution d 1 st Elution 2 nd Elution Figure 4.3 Venn diagrams of unique peptides and proteins identified in the cell lysate digests with and without additional wash or elution steps prior to LC QTOF MS analysis. 75

77 4.3.2 Additional methanol washing and elution steps 4 The filter plate protocol leaves room for the incorporation of additional steps, if needed. For instance, certain cell types may require an organic solvent washing step for removal of cell membrane constituents. This level of flexibility in the protocol is illustrated by evaluating the effect of incorporating an additional methanol washing step into the protocol using HEK293T cells cultured and lysed in a Petri dish. Three cell lysate samples (100 µl, 35 µg of protein) were spiked with HSA (0.1, 0.01 and 0 mg/ml) and washed with 50% methanol directly after transfer to the filter plate followed by a centrifugation step. The results of these samples were compared to the results of an identical sample set prepared without the additional washing step. The digests (10 µl, 3.5 µg) were rapidly analyzed on the rapid resolution LC QTOF MS. The results of the data analysis are depicted in Figure 4.3 and show that after the additional washing step resulted in only a slightly lower number of identified unique peptides (240 vs. 275) and proteins (34 vs. 38) (see Figure 4.3a and 4.3b, respectively). The methanol washing step is compatible with the filter plate, but was not required for the cell type analyzed in this study and was omitted from the protocol. Alternatively, a second elution step with organic solvent can also be incorporated to determine whether complete elution of the proteolytic peptides was achieved in the first elution step. Therefore, following elution of the tryptic peptides from the cell lysate samples spiked with HSA (0.1 or 0.01 mg/ml) using 100 µl of 1% FA in 50 mm ABC, a second elution step was performed with 100 µl of 1% FA in 50% methanol. The elution fractions from all samples (n = 4) were collected in separate plates and analyzed with LC QTOF MS, see Figure 4.3c and 4.3d for the results. The total ion chromatograms in Figure 4.4 demonstrate that the majority of peptides indeed elute in the first fraction. From the 94 peptides identified in the second elution fractions, only 7 were not identified in the first elution fractions (a list of all identified proteins and peptides is available as supporting information, Table S1 upon request). As this did not lead to any additional protein identifications, the experiment most likely indicates a concentration-related effect rather than non-specific binding of peptides to the filter membrane Comparison of the filter plate protocol with the gel-filtration protocol for cellular proteomics The low number of identified peptides and proteins in the previous experiments is caused by combining a fast LC separation and a QTOF with a relatively low acquisition rate (2.8 s to achieve a survey spectrum and two MS/MS spectra). Therefore, the samples of the next experiment were reanalyzed with a nanolc Orbitrap MS system that was programmed to acquire a survey spectrum and ten MS/MS spectra in only 2.4 s. In this final comparison experiment, the optimized high-throughput filter plate protocol was compared to a column-based gel-filtration protocol [5] commonly used in our laboratory using HEK293T cells lysed in a 96-well plate (20 µg protein per well). For each sample from both protocols 0.25 μg (1 µl) of the total protein digest was analyzed twice by the nanolc Orbitrap MS system. Both protocols are depicted in Figure 4.1. The Venn diagrams in Figure 4.5 show that despite the smaller amount of protein analyzed (0.25 µg vs 3.5 µg), the results have significantly improved in comparison to the previous experiment due to the high 76

78 7 x Without methanol wash step, 1 st elution step Without methanol wash step, 2 nd elution step With methanol wash step, 1 st elution step With methanol wash step, 2 nd elution step Counts vs. Acquisition Time (min) Figure 4.4 Chromatograms obtained from LC MS analysis of the cell lysate digests prepared with or without an initial methanol wash step and with or without a second elution step using 1% FA in 50% methanol. 4 a Filter plate Unique peptides Gel-filtration b Filter plate Proteins Gel-filtration Figure 4.5 Venn diagrams of the identified unique peptides (a) and proteins (b) in cell lysate digests prepared with the filter plate or gel-filtration protocol and analyzed with nanolc Orbitrap MS. 77

79 efficiency of the nanolc Orbitrap MS system. Figure 4.5 also shows large differences in the number of identified proteins and peptides between the cell lysate samples prepared with either the filter plate or the gel-filtration protocol. With the filter plate protocol, the number of identified proteins is 4-fold higher (440 vs. 109) and the number of identified peptides is 6-fold higher (2083 vs. 347) than with the gel-filtration column protocol. (A list containing all identified peptides and proteins is available as supporting information, Table 2S upon request.) This large difference may partly be due to the risky freeze-drying step (possibly leading to non-specific protein adsorption) that is required due to the 4-fold dilution at the start of the gel-filtration protocol (from 200 µl to 750 µl for application and elution from the column) and to achieve the same final concentration level as in the filter plate protocol. Next to that, the multiple sample transfer steps may have resulted in protein loss, thereby causing a reduction in the overall sensitivity of this method. These results may also indicate that the gel-filtration column protocol is more suitable for less complex and more concentrated protein samples. 4 In contrast, the number of sample transfer steps is minimized in the filter plate protocol, which also allows for a 2.5-fold sample concentration (from 200 µl to 80 µl) and avoids the freeze-drying step. Therefore, this protocol is less prone to sample losses and more sensitive. The obtained results show its applicability for global cellular proteomics. Although the buffer-exchange steps via centrifugation can be quite lengthy ( h, depending on the sample volume) in comparison with gel-filtration (completed within 1 h including equilibration of the columns), the filter plate protocol is less laborious, reduces buffer consumption, requires less manual steps and is still completed within one day. It also allows for the simultaneous preparation of 96 samples, while for practical reasons this number is much more limited for the gel-filtration columns. 4.4 CONCLUSION To summarize, a high-throughput sample preparation method employing 96-well filter plates was developed for bottom-up cellular proteomics. This protocol allows for 1:1 transfer of cell lysates from 96-well plates, for instance from functional assays performed in the 96-well format. The 96-well filter plate sample preparation format was found to be superior to the individual treatment of samples in terms of sample throughput and sensitivity. As opposed to other media, the use of MWCO filter membranes contributes to the versatility of the 96-well plate protocol and is not biased towards any type of proteins or peptides. 78

80 ACKNOWLEDGEMENTS This research was performed within the framework of project D3-201 Towards novel translational safety biomarkers for adverse drug toxicity of the Dutch Top Institute Pharma. The authors greatly thank Marija Mladic for the culturing of the HEK293T cells. REFERENCES [1] Levin, Y., Schwarz, E., Wang, L., Leweke, F. M., Bahn, S., Label-free LC MS/MS quantitative proteomics for largescale biomarker discovery in complex samples. J. Sep. Sci. 2007, 30, (14), [2] Simpson, D. C., Ahn, S., Pasa-Tolic, L., Bogdanov, B., Mottaz, H.M., et al., Using size exclusion chromatography- RPLC and RPLC-CIEF as two-dimensional separation strategies for protein profiling. Electrophoresis. 2006, 27, (13) [3] Gundry, R.L., White, M.Y., Murray, C.I., Kane, L.A., Fu, Q., et al., Preparation of proteins and peptides for mass spectrometry analysis in a bottom-up proteomics workflow. Curr. Protoc. Mol. Biol. 2009, Chapter 10, Unit [4] Guo, X., Kristal, B.S., The use of underloaded C(18) solid-phase extraction plates increases reproducibility of analysis of tryptic peptides from unfractionated human plasma. Anal. Biochem. 2012, 426, (1), [5] Switzar, L., Giera, M., Lingeman, H., Irth, H., Niessen, W.M., Protein digestion optimization for characterization of drug protein adducts using response surface modeling. J. Chromatogr. A. 2011, 1218, (13), [6] Wisniewski, J.R., Zougman, A., Nagaraj, N., Mann, M., Universal sample preparation method for proteome analysis. Nat. Methods. 2009, 6, (5), [7] Katayama, H., Satoh, K., Takeuchi, M., Deguchi-Tawarada, M., Oda, Y., et al., Optimization of in-gel protein digestion system in combination with thin-gel separation and negative staining in 96-well plate format. Rapid. Commun. Mass Spectrom. 2003, 17, (10), [8] Pluskal, M.G., Bogdanova, A., Lopez, M., Gutierrez, S., Pitt, A.M., Multiwell in-gel protein digestion and microscale sample preparation for protein identification by mass spectrometry. Proteomics. 2002, 2, (2), [9] Nissum, M., Schneider, U., Kuhfuss, S., Obermaier, C., Wildgruber, R., et al., In-gel digestion of proteins using a solid-phase extraction microplate. Anal. Chem. 2004, 76, (7), [10] Chelius, D., Xiao, G., Nichols, A.C., Vizel, A., He, B., et al., Automated tryptic digestion procedure for HPLC/MS/MS peptide mapping of immunoglobulin gamma antibodies in pharmaceutics. J. Pharm. Biomed. Anal. 2008, 47, (2), [11] Hou, W., Ethier, M., Smith, J.C., Sheng, Y., Figeys, D., Multiplexed proteomic reactor for the processing of proteomic samples. Anal. Chem. 2007, 79, (1), [12] Meiring, H.D., van der Heeft, E., ten Hove, G.J., de Jong, A.P.J.M., Nanoscale LC MS(n): technical design and applications to peptide and protein analysis. J. Sep. Sci. 2002, 25,

81

82 Chapter 5 Identification and quantification of drug albumin adducts in serum samples from a drug exposure study in mice Linda Switzar, Lydia M. Kwast, Henk Lingeman, Martin Giera, Raymond H. Pieters and Wilfried M.A. Niessen Journal of Chromatography B, 2013, , 53-61

83 ABSTRACT The formation of drug protein adducts following the bioactivation of drugs to reactive metabolites has been linked to adverse drug reactions (ADRs) and is a major complication in drug discovery and development. Identification and quantification of drug protein adducts in vivo may lead to a better understanding of drug toxicity, but is challenging due to their low abundance in the complex biological samples. Human serum albumin (HSA) is a well-known target of reactive drug metabolites due to the free cysteine on position 34 and is often the first target to be investigated in covalent drug binding studies. Presented here is an optimized strategy for targeted analysis of low-level drug albumin adducts in serum. This strategy is based on selective extraction of albumin from serum through affinity chromatography, efficient sample treatment and clean-up using gel filtration chromatography followed by tryptic digestion and LC MS analysis. Quantification of the level of albumin modification was performed through a comparison of non-modified and drug-modified protein based on the relative peak area of the tryptic peptide containing the free cysteine residue. The analysis strategy was applied to serum samples resulting from a drug exposure experiment in mice, which was designed to study the effects of different acetaminophen (APAP) treatments on drug toxicity. APAP is bioactivated to N-acetyl-p-benzoquinoneimine (NAPQI) in both humans and mice and is known to bind to cysteine34 (Cys34) of HSA. Analysis of the mouse serum samples revealed the presence of extremely low-level NAPQI albumin adducts of approximately 0.2% of the total mouse serum albumin (MSA), regardless of the length of drug exposure. Due to the targeted nature of the strategy, the NAPQI-adduct formation on Cys34 could be confirmed while adducts to the second free cysteine on position 579 of MSA were not detected. 5 82

84 5.1 INTRODUCTION Drugs that are associated with adverse drug reactions (ADRs) and idiosyncratic drug reactions (IDRs) frequently possess the propensity for bioactivation and subsequent formation of drug protein adducts. [1] Although the exact mechanisms behind ADRs and IDRs still remain largely uncertain, drug protein adducts are suggested to be directly linked to selective organ toxicity and immune-mediated response. [2-4] Elucidation of these mechanisms will lead to a better understanding of drug toxicity and may be achieved by identification of the target proteins of bioactivated drugs and their involvement in biological processes, such as inflammatory responses. [1] However, the detection of drug protein adducts in vivo is extremely challenging since only minute amounts are present. Additionally, the low abundance of adducts in tissues and biofluids that possess a very wide dynamic range and a large excess of non-adducted proteins further complicates their detection. Therefore, sensitive and selective analytical methodologies are needed for the analysis of drug protein adducts in complex biological samples. Many target proteins of reactive drug metabolites have been identified from in vitro and in vivo animal experiments using global proteomics approaches, such as two-dimensional gel electrophoresis in combination with mass spectrometry (MS) detection [5]. In such approaches, relevant protein spots are selected for further treatment based on autoradiography, in cases where radiolabeled drugs are used [6-7], or immuno-selective staining with antibodies raised against specific drugs [8-9]. The selected protein spots are excised from the gel, in-gel digested and analyzed with liquid chromatography mass spectrometry (LC MS)[9] or matrix-assisted laser desorption ionization (MALDI)-MS[6-8]. Using immuno-blotting, 15 hepatic proteins, 12 of which are novel, were identified recently as likely targets of tienilic acid in rats [9], while a total of 64 protein targets (42 cytosolic and 24 microsomal) of radiolabeled thiobenzamide were identified in rat liver using phosphorimaging [6]. Identification of the modification site, e.g., via detection of an adducted peptide in a proteomics-based strategy, would unambiguously confirm adduct formation to the identified protein, but this is often not achieved due to the low level of covalent binding [10] and the small amount of protein present in the gel spot. Additionally, detection of the adducted peptide would provide a means for quantification of the level of adduct formation, which could have important clinical and toxicological relevance [5]. 5 In vitro experiments using trapping agents have shown that numerous reactive drug metabolites are reactive toward the thiol group in cysteines [11] and many global proteomics approaches have identified human serum albumin (HSA) as a protein target for covalent binding. [1] This is due to the fact that HSA possesses an unpaired cysteine on position 34 (Cys34) that is located on the surface of the protein and thus accessible to reactive drug metabolites, as can be seen in an image of the crystal structure (PDB ID: 2BXB [12], created with Protein workshop [13]) shown in Figure 5.1. Using this information, a targeted approach can be developed to investigate covalent binding to HSA. Such a targeted approach includes selective protein isolation, e.g., by affinity chromatography, prior to protein digestion and subsequent LC MS analysis. This approach would provide the required selectivity and sensitivity for studying low-abundant proteins and their interactions. [14-15] The increased sensitivity of targeted approaches may also be useful for monitoring of drug protein 83

85 adducts in clinical samples [16] where radiolabels or other distinct features for selective detection are not applicable. A limited number of examples of a targeted approach for the confirmation of in vivo adduct formation have been reported for drugs [17-18] and other xenobiotic compounds [10]. In these studies, it was found that HSA, isolated from patient samples, was adducted at Cys34 by nitrogen mustards [18] and N-acetyl-p-benzoquinoneimine (NAPQI), the reactive metabolite of acetaminophen (APAP) [17] Covalent binding studies are an integral part of drug candidate evaluation in the pharmaceutical industry and may determine the fate of a lead compound. However, there is a need for advanced proteomicsbased methodologies to study the relationship between drug protein adducts and toxicity. [19] The NAPQI albumin adduct is perhaps the most well-known example of drug protein adduct formation and is therefore often used as a model adduct. [20] Drug toxicity studies are generally performed in animal models, thus it is fortunate that the protein sequence and structure, i.e., disulfide bridges, of serum albumins are well conserved among various species. [21-22] Nevertheless, it is important to note that small differences in the protein sequence may lead to different binding effects. [23] Albumins of most species, including HSA and mouse serum albumin (MSA), contain the free cysteine on position 34. However, MSA represents a rather unique case because it also contains an additional free cysteine at position 579. Unfortunately, a crystal structure of MSA is not available, but, considering the fact that this cysteine is close to the C-terminus and the protein sequence is very similar to that of HSA, it is likely that this second free thiol in MSA is also accessible to reactive drug metabolites, see Figure 5.1 for the crystal structure of HSA Figure 5.1 Crystal structure of HSA (PDB ID: 2BXB [12], created with Protein workshop [13]) indicating the position of Cys34 and location of residue 579, which is a cysteine in MSA. 84

86 In the here presented research, an analytical strategy was developed and optimized for identification and quantification of low-abundant drug albumin adducts. The optimized strategy was subsequently applied to serum samples resulting from a drug exposure study in mice. In this study, the mice received a dose of APAP either for a single day or for seven consecutive days in order to study the kinetics of ADRs and determine whether there is a correlation with the formation of NAPQI MSA adducts. 5.2 MATERIALS AND METHODS Chemicals and materials Mouse serum albumin (MSA, 96%), bovine serum albumin (BSA, 98%), human serum albumin (HSA, 97-99%), acetaminophen (APAP, 99%), disodium hydrogen phosphate, sodium dihydrogen phosphate, potassium chloride, sodium chloride, Bradford reagent, guanidine HCl (GHCl), ammonium bicarbonate (ABC), HPLC standard peptide mixture, [Met 5 ]Enkephalin acetate ( 95%, internal standard), dithiothreitol (DTT), iodoacetic acid (IAA), N-acetyl-p-benzoquinone imine (NAPQI) and ethanol (96%) were obtained from Sigma Aldrich (Schnelldorf, Germany). A synthetic version of the HSA peptide ALVLIAFAQYLQQC 34 PFEDHVK was kindly provided to us by R. Ekkebus (NKI, Amsterdam, the Netherlands). The HiTrap blue HP albumin affinity columns, prepacked with Blue Sepharose High Performance, (1 ml column volume) and NAP-25 gel filtration columns, prepacked with Sephadex G-25 DNA Grade, (2.5 ml column volume) were purchased from GE Healthcare (Diegem, Belgium). Trypsin from bovine pancreas (EC ) was supplied by Roche (Almere, The Netherlands). LC MS analyses were performed with ULC MS grade acetonitrile (ACN, 99.95%) and formic acid (FA, 99%) from Biosolve (Valkenswaard, The Netherlands). Purified water was provided by a Millipore Milli-Q unit (Amsterdam, The Netherlands). For the mouse oral exposure experiments, APAP was dissolved in distilled water (Aqua B. Braun, Melsungen, Germany) Mouse study design 5 Four to six weeks old female C3H/HeN mice were purchased from Harlan (Venlo, The Netherlands). Mice were specific pathogen-free and maintained under barrier conditions in filter-topped macrolon cages with wood chip bedding at a mean temperature of 23 ± 2 C, 50-55% relative humidity and a 12 h light/dark cycle. Drinking water and standard laboratory food pellets were provided ad libitum. The experiments were conducted according to the guidelines of, and with permission from, the animal experiments committee of Utrecht University. The C3H/HeN mice (n = 8) received either a single dose or seven consecutive daily doses of 300 mg/kg APAP. This dose was chosen using the maximum tolerable dose as described in the datasheet of the compounds and as used in previous experiments. [24] In both experiments, the control group consisted of eight mice that received the vehicle. The mice used in this study were part of a larger group (n = 112) that showed little variation in weight (mean weight ± SD of ± g) h following the last oral dose, blood was drawn by cheek pouch puncture and both plasma and serum was collected 85

87 (Minicollect, Kremsmünster, Austria) for further analysis. Furthermore, a part of the liver was isolated, fixed in formalin and subsequently embedded in paraffin. Alanine aminotransferase (ALAT) and aspartate aminotransferase (ASAT) levels were analyzed in plasma samples on the AU400 Chemistry System (Beckman Coulter Nederland B.V., Woerden, The Netherlands) at NOTOX B.V. ( s-hertogenbosch, The Netherlands). Plasma enzyme levels of APAP treated animals were compared to the combined control values obtained from both series of kinetics experiments. Multiple comparisons of group means were analyzed using one-way ANOVA with Dunnett s as post test. A value of p < 0.05 was considered significantly different compared to the control values. Data was analyzed using Graphpad Prism version 5.00 for Windows (Graphpad Software, San Diego, CA, USA) Optimization of sample preparation protocol 5 The conditions of the albumin affinity chromatography, gel filtration chromatography and freeze-drying steps were optimized using standard serum albumin solutions with varying concentrations in various buffers. Initial experiments were performed with BSA or HSA; selected experiments were repeated with MSA. Recoveries were determined based on the Bradford assay quantification results, which were obtained as follows. A series of calibration solutions of mg/ml were prepared of the albumin standard used in the specific optimization experiment and in the appropriate buffer. The readout was performed in 96-well plates in which 10 µl of sample or calibration standard was added to 200 µl Bradford reagent, in duplicate. The plates were incubated at room temperature for min before readout in triplicate at 595 nm using a Victor Multilabel Counter plate reader from Perkin Elmer (Groningen, The Netherlands). The average absorbance of the triplicate measurements of the duplicate calibration samples was corrected for the blank and plotted against the concentration in mg/ml. The unknown protein concentrations of the samples were calculated using the linear regression equation obtained from the calibration curve. LC MS separation and detection conditions were optimized using a tryptic digest of standard MSA or NAPQI MSA samples, prepared similarly as described in the next sections Preparation of mouse serum and MSA standard samples Ideally, 50 μl of serum from each control (n = 8) and dosed (n = 8) mouse was used for analysis. In some cases, a smaller volume of serum was obtained and, therefore, all serum samples were weighed to determine the available volume. The 50 μl portions of mouse serum were adjusted to the composition of the HiTrap binding buffer by dilution with 950 µl of 20 mm sodium phosphate, ph 7.0. Additionally, a 10 µm reference standard of MSA in binding buffer (n = 8) was simultaneously prepared and received the same treatment as the serum samples. Albumin was extracted from the mouse serum using HiTrap blue HP columns. The MSA reference standards underwent the same treatment in order to determine the recovery of MSA from the affinity column. The column was equilibrated with 10 column volumes of binding buffer followed by application of the sample. 86

88 The unbound proteins were removed by washing the column with 10 column volumes of binding buffer. The bound MSA was eluted with 10 ml of elution buffer, consisting of 20 mm sodium phosphate, 2.0 M sodium chloride, ph 7.0. The first 3 ml of the elution step, containing the bulk of the bound MSA, was collected and used for further treatment. The total protein content was determined by Bradford assay as described in the previous section. The purified MSA was buffer exchanged to denaturation buffer (2.0 M GHCl, 50 mm ABC, ph 8.5) using NAP-25 gel filtration columns. An improved protocol was used to avoid further dilution of the sample while maintaining a high recovery. The columns were equilibrated with 25 ml of denaturation buffer before application of the 3 ml sample. The protein fraction was subsequently eluted by application of an equal volume (3 ml) of the same buffer thereby avoiding dilution of the sample. The samples were allowed to denature overnight at 4 C.The following morning, the disulfide bridges in the MSA samples were reduced by the addition of a 50-fold molar excess (50 total number of cysteines (36) MSA concentration) of DTT and incubation at 50 C for 30 min. After reduction, the samples were allowed to cool down to room temperature and the resulting thiol-groups were alkylated with a 75-fold molar excess of IAA (1.5-fold excess over DTT) at room temperature and in the dark for 30 min. Subsequently, a buffer exchange and desalting step was performed using the same NAP-25 protocol mentioned above, only now Milli-Q water was used for equilibration of the column and elution of the protein. The samples were concentrated by freeze-drying for 6.5 h using a Centrivap concentrator (Labconco, Kansas City, MO, USA), which was kept at room temperature and was connected to an automatic freeze dryer (The VirTis Company Inc, Gardiner, NY, USA). The dried protein samples were either stored at -20 C until further treatment or directly redissolved in 250 µl of 50 mm ABC buffer, ph 8.4, followed by digestion. Tryptic digestion was performed overnight using optimized conditions as described elsewhere. [25] In short, trypsin was added to the samples in a protein-to-enzyme ratio of 100:1 (w/w). The samples were subsequently incubated at 24 C overnight (~15 h) followed by the addition of FA to a final concentration of 0.1%. Finally, the internal standard was added to a final concentration of 31.5 µm and a final volume of 350 µl was obtained by the addition of digestion buffer. The samples were either analyzed immediately with LC MS or stored at -20 C In vitro preparation of NAPQI MSA adducts To assess whether the cysteine 579 (Cys579) residue in MSA is accessible to reactive drug metabolites and whether adduct formation can take place at this site, the NAPQI MSA adduct was prepared in vitro according to the following protocol. Synthetic NAPQI (1 mg) was dissolved in 80 µl of DMSO and 30 µl of this solution, representing a 50-fold molar excess, was added to 3 mg of MSA reconstituted in 2.97 ml of a 10 mm phosphate buffered saline (PBS) buffer, ph 7.4. The adduct formation reaction was left to proceed at room temperature for 1 h before buffer exchange to PBS buffer. Five sequential gel filtration steps were performed in order to completely remove the excess NAPQI and prevent adduct formation to other cysteines after reduction of the disulfide bridges. From then on, the purified NAPQI MSA received the same treatment as the mouse serum samples and was analyzed in triplicate. 87

89 5.2.6 LC MS analysis The digested MSA samples were analyzed with a 1200 series Rapid Resolution LC system coupled to a 6520 QTOF mass spectrometer (Agilent Technologies, Amstelveen, The Netherlands), that was controlled by the Agilent Masshunter Workstation Acquisition software (version B.02.00). The tryptic peptides were separated on an Agilent XDB-C18 column (50 mm 4.6 mm, 1.8 µm particles) that was protected by a C18 guard column (4 mm 2 mm) from Phenomenex (Utrecht, The Netherlands). The LC MS settings were based on a previously published method for the analysis of NAPQI HSA digests [25] and slightly modified for NAPQI MSA digests. In order to gain additional retention for the Cys34 and Cys579 peptides, the mobile phases were changed to 2.5% ACN, 0.1% FA in water for eluent A and 2.5% water, 0.1% FA in ACN for eluent B. In addition, the gradient method was extended to a total runtime of 48 min. Gradient elution was performed by holding the % B at 0% for the first 5 min, followed by a linear increase to 40% B in 23 min. The column was then washed at 100% B for 7 min and re-equilibrated for the next run at 0% B for 13 min. The flow rate was set to 0.6 ml/min and the temperature of the thermostated column compartment was maintained at 40 C. 5 Using an internal switch valve, the LC flow from 4-28 min was directed to the mass spectrometer, which was operated in 2 GHz, extended dynamic range mode. The electrospray ionization source was operated in positive mode (ESI+), the capillary voltage was set to 3500 V and nitrogen ( %) was used as the drying (350 C) and nebulizer gas at a flow rate of 12 L/min and a pressure of 60 psig, respectively. Profile data was acquired at a rate of 1.25 spectra/s in data-dependent mode where the most intense ion (m/z ) was selected for fragmentation and subsequently excluded from fragmentation for 0.2 min. Fragmentation spectra of selected ions were recorded over an m/z range from 50 to 2000, at a rate of 1.03 spectra/s using a fixed collision energy voltage of 20 V and nitrogen as the collision gas. An aqueous blank sample and the HPLC peptide standard mixture (containing 0.5 µg/ml of Gly-Tyr, Val-Tyr-Val, [Met 5 ]enkephalin acetate, Leu-enkephalin and angiotensin II acetate) were analyzed after every sample run to check the stability of the LC MS system throughout the sequence. These control samples were analyzed using a shorter method with a runtime of 22 min. During this method, the concentration B increased from 0-50% over the first 7 min, was held constant at 100% B for 5 min, and finally held at 0% B for 10 min. Full-spectrum MS data (m/z ) was collected during the first 7 min of the run at a rate of 1.03 spectra/s. The LC flow rate and ESI source conditions were the same as described above Data analysis Peak extraction (using a 20 ppm half-width m/z window) and integration was performed with the Agilent Masshunter Qualitative Analysis software (version B.02.00). The peak areas were normalized to the peak area of the internal standard. The level of adduct formation was determined by the ratio between the peak area of the NAPQI cys peptide divided by the total peak area of cys peptide (sum of all charge states of the carboxymethylated and NAPQI modified). 88

90 5.3 RESULTS AND DISCUSSION Method development and optimization Albumin is the most abundant serum protein, but only a very small percentage is modified by reactive drug metabolites. This means that in our mouse study the NAPQI MSA adduct is extremely low abundant and, adding to this challenge, it is in the presence of a large excess of non-modified MSA. Furthermore, the NAPQI modification represents a very small addition of only Da to the large MSA protein molecule of 65.9 kda and does not have a large effect on the protein characteristics. The difference between the two MSA species is, therefore, too small to achieve a separation using conventional techniques. For the analysis of NAPQI MSA adducts in serum, a sample preparation and analysis methodology was developed and optimized in order to deal with the low abundance of the NAPQI MSA adduct (Figure 5.2). The optimization of the various steps is discussed in some detail below. The albumin, both NAPQI and non modified, was extracted from the mouse serum by albumin affinity chromatography using HiTrap cartridges. If available, a starting sample volume of 50 µl mouse serum was used that was adjusted to the HiTrap binding conditions by dilution with binding buffer. To determine the optimal dilution volume, different sample volumes of ml were evaluated, but significant differences in the recovery of BSA were not detected (69 ± 8%, n = 8). An application volume of 1.0 ml (recovery 71 ± 11%, n = 4) was chosen for use in the final protocol. Regarding the elution volume, the bulk of HSA (> 90%) eluted within the first 3.0 ml from the HiTrap cartridge, thus this was considered as the optimum elution volume. Finally, a different buffer system based on GHCl and ABC at ph 8.5 was also investigated for HiTrap binding and elution, but this led to a decreased recovery of <50%. 5 Serum sample 50 µl Bradford assay Denaturation, reduction & alkylation Freeze-drying LC-MS analysis, identification & quantification HiTrap Albumin affinity chromatography NAP-25 gel filtration NAP-25 gel filtration Tryptic digestion Figure 5.2 Sample preparation and analysis strategy. 89

91 The recovery of BSA and HSA from the HiTrap columns using the standard Sodium phosphate binding and elution buffers at ph 7.0 was comparable, 87% and 90% (n = 4), respectively. However, the recovery decreased over time with repetitive use of the columns. Therefore, new HiTrap columns were used for the mouse serum samples. Unfortunately, the binding of MSA to the HiTrap columns is less efficient [26], resulting in a recovery of 59 ± 9% (n = 16). The low recovery could easily be compensated by increasing the initial starting volume of 25 µl to 50 µl serum, when available. Although this also leads to a higher concentration of non-modified MSA in the final sample, the final concentration of NAPQI MSA will theoretically be in the nanomolar range (after the freeze-drying step), assuming that ~0.1% of the total MSA is modified by NAPQI. This concentration of NAPQI MSA is well within the detectable range. The HiTrap albumin affinity step leads to a 60-fold dilution of the original 50 µl of serum to 3 ml of purified albumin. To avoid further dilution and sample losses, the standard NAP-25 gel filtration protocol (application of 2.5 ml of sample and elution with 3.5 ml buffer) was adapted for better alignment with the HiTrap elution step. Therefore, the total 3.0 ml of eluted sample from the HiTrap cartridges was applied to the gel filtration column and elution from this column was performed with 3.0 ml of buffer. Using a 1 mg/ml BSA solution in water, the recovery of this protocol was found to be above 90% and with two consecutive gel filtration steps the recovery was still above 85%. In order to counteract the 60-fold dilution and increase sensitivity, the effect of a freeze-drying step on protein stability and recovery was also investigated. After 6 h of freeze-drying and redissolving the dried protein in a volume of 250 µl, a more than 10-fold concentration factor and a recovery of 94 ± 6% (n = 6) were achieved. 5 The tryptic digestion and LC MS conditions were optimized previously for NAPQI HSA Cys34 adducts with respect to protein coverage and peak area of the modified Cys34 peptide [25]. These optimized conditions also provided better results in the identification of MSA adducts and, thus, were applied in the current study. However, the peptides of interest containing the possible adduct formation sites resulting from the tryptic digestion of (NAPQI )MSA, C 34 SYDEHAK and C 579 KDALA, are much shorter in length than the Cys34 peptide of HSA, ALVLIAFAQYLQQC 34 PFEDHVK. The Cys579 peptide still contains a lysine, but this is not considered as a tryptic cleavage site due to the cysteine in the p2 position and aspartic acid in p1, as is stated for the tryptic cleavage rules in the ExPASy database. [27] This part of the MSA sequence was also never detected in any other form. The relatively short carboxymethylated Cys34 and 579 peptides from MSA showed little retention and co-eluted with the internal standard. Therefore, some adjustments to the LC MS method were made for increased retention and separation of the peptides of interest. Additionally, several peptides were tested for use as an internal standard; [Met 5 ]Enkephalin (YGGFM) was selected due to its favorable retention time of 18.8 min (peak 5). With this optimized method, the C 34 (carboxymethyl)sydehak (peak 1, tr 4.9 min) and C 579 (carboxymethyl)kdala (peak 2, tr 11.4 min) could be separated and detected, see Figure 5.3. Furthermore, two peaks were observed for each NAPQI cys peptide adduct, which is most likely due to the formation of regioisomers [17], see peaks 3 and 4, a and b in Figure

92 HN O HN O S - Protein OH 1 OH S - Protein 2 3b 4b 5 4a 3a Counts vs. Acquisition Time (min) Figure 5.3 LC MS results from the analysis of a tryptic digest of NAPQI MSA. The extracted-ion chromatograms (EICs) of the doubly-charged ions with m/z (peak 1), C 34 (carboxymethyl)sydehak), m/z (peak 2), C 579 (carboxymethyl)kdala), m/z (peaks 3a and b), C 34 (NAPQI)SYDEHAK), m/z (peaks 4a and b), C 579 (NAPQI)KDALA) and of the singly-charged ion with m/z (peak 5), internal standard are shown in black. In the background, a total ion chromatogram (TIC) is shown in grey. For purpose of clarity, different scales were used for the y-axis. Proposed structures of the regioisomeric adducts are shown in the top left corner Adduct quantification Information about the level of albumin adduct formation present in the mouse serum is necessary in order to determine the effect of different drug exposure regimes in the two mouse studies. Ideally, quantification of the NAPQI MSA adduct in the samples would be done at the protein level. However, since the resolution between MSA and NAPQI MSA and/or the sensitivity were simply insufficient in MALDI-TOF MS and Orbitrap MS experiments (data not shown), quantification could only be done at the peptide level where the NAPQI modified peptides are easily separated from the non-modified species. In order to achieve reliable quantification, an internal standard must be used. [28] Given the large sequence homology between albumins, a tryptic Cys34 or Cys579 peptide from another albumin with a slightly different sequence would be a good choice. Unfortunately, MSA is the only commercially available albumin with a free Cys579 and yields a much shorter Cys34 peptide than other albumin species. Another possibility would be absolute adduct quantification using a synthetically prepared NAPQI Cys34 peptide, as applied by Damsten et al. [17]. Using a synthetic NAPQI Cys34 peptide, levels of 3-35 pmol/ml of NAPQI HSA adducts in patient serum were detected. However, this approach requires a labeled reference peptide to be synthesized for each adduct, which can be a tedious process. An extra limitation for the preparation of reference adducts is the limited availability of reactive metabolites since NAPQI is the only one that can be obtained commercially. 5 91

93 Another evaluated option was the use of a MSA reference standard for external quantification of the NAPQI MSA level in the mouse serum samples. This was accomplished by comparing the peak areas of five selected MSA peptides to the peak areas of the corresponding peptides in the tryptic digests of the purified mouse serum samples. This provides the total MSA concentration. Subsequently, a decrease in the peak areas of the non-adducted Cys34 and 579 peptides in both types of samples can be assumed to be a measure of the percentage of NAPQI modified MSA in the mouse serum samples. The five reference peptides were selected from the MSA digest based on the following criteria: 1) not to contain cysteines in order to prevent problems with incomplete reduction and alkylation, 2) be well distributed over the chromatogram and the protein sequence, 3) no missed cleavages in or next to peptide in order to prevent inaccuracies due to incomplete digestion, 4) limited coelution with other peptides. All peptides were normalized to the internal standard to correct for changes of MS signal intensity over time. Using this strategy, the MSA concentration in the in vitro generated NAPQI MSA samples could be determined accurately at 84 ± 14% of the actual concentration (n = 5 or 7). The adduct levels were predicted to be 40 ± 3% and 51 ± 13% for Cys34 and 579, respectively. Unfortunately, when this strategy was applied to the serum samples from the mouse study, it suffered from matrix effects most likely caused by the co-extraction of other serum proteins during the albumin affinity chromatography step and the biological variation. These phenomena result in high %RSD values making it problematic to accurately quantify adduct levels of 0.1%. 5 Finally, it was decided to perform a relative quantification of the NAPQI MSA adduct level within each sample without the use of any reference standard, thereby avoiding all disadvantages described above. This was done by determination of the relative ratio between the peak areas of all charge states of the NAPQI cys peptide and that of the total cys peptide peak area, which is the sum of the NAPQI cys and carboxymethylated-cys peptide. Hereby, it was assumed that the ionization efficiency of both forms of the cys peptides was similar. With this method, all samples could be subjected to quantification under the same conditions and low adduct levels could be determined Analysis of in vitro NAPQI MSA In contrast to HSA, MSA contains a second free cysteine at position 579. No information was available about the accessibility of this cysteine for small molecules. By in vitro incubations of MSA with synthetic NAPQI under non-denaturing conditions, it was shown that Cys579 can also be modified with NAPQI (synthetic labeling efficiency 79 ± 1.3%), even to a higher extent than Cys34 (55 ± 2.5%). If this adduct is also found in vivo, it could have implications for the extrapolation of drug toxicity effects in mice to human. For both the C 34 (NAPQI)SYDEHAK and C 579 (NAPQI)KDALA peptides, the formation of the later eluting regioisomer is strongly favored over the other. Sequence analysis of the MS/MS spectra of the ions m/z (tr 4.9 min, C 34 (carboxymethyl)sydehak) and m/z (tr 12.9 min, C 34 (NAPQI)SYDEHAK), shown in Figure 5.4 a and b, confirms the presence of the NAPQI adduct on the Cys34 position by the presence of an almost complete y-ion sequence. Characteristic ions of the doubly- (m/z ) and triply-charged (m/z ) C 34 (NAPQI)SYDEHAK peptide of the other regioisomer (tr 12.1 min) were observed, their intensity was insufficient for adequate MS/MS fragmentation. 92

94 x Product Ion (4.9 min) ( [z=2]) 2.2 K A H E D Y S C(carboxymethyl) 2.0 a MH x Product Ion (12.9 min) ( [z=2]*) b 1.0 KA H E D Y S C(NAPQI) MH x10 3 Counts vs. Mass-to-Charge (m/z) + Product Ion (11.4 min) (339.66[z=2]) 1.4 c C(carboxymethyl)K D A LA MH x Product Ion (15.0 min) (385.19[z=2]) C(NAPQI)K D A LA Counts vs. Mass-to-Charge (m/z) d MH + 5 Figure 5.4 MS/MS spectra of C 34 (carboxymethyl)sydehak (a), C 34 (NAPQI)SYDEHAK (b), C 579 (carboxymethyl)kdala (c) and C 579 (NAPQI)KDALA (d). 93

95 Due to the short length of the Cys579 peptide, the MS/MS spectra of the C 579 (carboxymethyl)kdala (tr 11.4 min, Figure 5.4 c) and C 579 (NAPQI)KDALA (tr 15.0 min, Figure 5.4 d) are more challenging to interpret, but the presence of the NAPQI cys adduct on position 579 could be confirmed from a partial b-ion sequence. The identification of the C 579 (NAPQI)KDALA peptides is further complicated by the adjacent elution of isobaric peptides (m/z , tr 14.1 and 15.2 min) with the same charge state. However, the MS/MS spectra of these ions contain a fragment ion with m/z characteristic of a C-terminal lysine (data not shown) and, thus, are distinctly different peptides Histological and liver enzyme analysis Hepatoxicity caused by oral exposure to APAP was determined by both histological analysis and plasma levels of the liver enzymes ALAT and ASAT. Blood flow in the liver runs from the portal area toward the central veins from which blood is subsequently transported to the inferior vena cava. Previous animal experiments showed that administration of APAP resulted in liver damage, indicated by the presence of necrotic areas around the central vein areas. [29-32] Histological examination of the mouse livers from the current study showed similar necrotic centrilobular areas in the livers from mice orally exposed to APAP, which could be identified by minor cell swelling and loss of hepatic structure at this site, see Figure 5.5. Surprisingly, no clear differences in liver damage and necrotic areas were observed between the groups that received a single or seven consecutive exposures to APAP. In addition to the histological analysis, serum levels of the liver enzymes ALAT and ASAT were determined, see Figure 5.6 for the resulting bar graphs. Serum biomarkers, such as ALAT and ASAT levels, are generally 5 Figure 5.6 Serum levels of liver enzymes from animals orally exposed to vehicle or APAP. Mice were orally exposed to a single or seven consecutive doses of vehicle or APAP. Within 24 hours of the last oral dosing, serum and plasma was collected and ALAT (a) and ASAT (b) levels in plasma were determined (each bar represents the mean ± SEM of 8-16 (controls) or 3-8 (APAP) animals per group). *** p<

96 a PV PV CV PV CV b CV PV 5 CV PV CV Figure 5.5 Histological representation of liver. Mice were exposed orally to either a single or seven consecutive doses of acetaminophen. Within 24 h of the last dose, a part of the liver was collected and paraffin embedded sections of the liver were stained with hematoxylin and eosin. Representative sections for controls (a) and APAP (b) treated animals are shown. PV: portal vein, CV: central vein (Magnification 4x) 95

97 used to indicate the presence of liver injury following drug exposure. [32-34] In our experiments, the level of ALAT in plasma was significantly increased in both the single and seven times exposed groups, which is indicative for APAP-induced liver damage. However, this effect was not observed for ASAT levels. Similar to the histological analysis, also no differences were observed between the single and multiple dosed groups in terms of liver enzymes. Several experiments have shown time dependent increases in liver enzyme levels following administration of APAP. However, there does not seem to be a clear relationship between ALAT en ASAT levels, as exemplified by several studies performed in animals [35-36] and humans [17, 37] Analysis of NAPQI MSA in mouse serum samples Finally, the optimized analysis and quantification strategy was applied to the serum samples from the mouse study. The one- and seven-day experiment each consisted of eight control mice and the same number of dosed mice, resulting in the collection of 16 serum samples for each experiment. Low-abundant NAPQI MSA adducts to Cys34 were detected in all of the serum samples collected from the dosed mice, of both the one-day and seven-day experiment, see Figure 5.7. In the extracted ion chromatogram (EIC) of triply-charged C 34 (NAPQI)SYDEHAK (m/z at 12.7 min, Figure 5.7a), a significant difference was observed between the control and dosed mouse serum sample. The other regioisomer of C 34 (NAPQI) SYDEHAK is also present at 12.1 min, albeit at a very low intensity. No difference was observed between control and dosed mice in the EICs of m/z , the expected m/z of the doubly-charged C 579 (NAPQI) KDALA (Figure 5.7b). Although the Cys579 residue could be modified by NAPQI in vitro, this adduct was not detected in vivo. Reasons for this could be a preference for modification of Cys34 by NAPQI in vivo, resulting 5 x EIC( ) Scan 3.2 * * Counts vs. Acquisition Time (min) a x EIC( ) Scan Counts vs. Acquisition Time (min) b Figure 5.7 LC MS results of a control (black trace) and APAP-dosed (grey trace) serum sample from the mouse study. EIC s of m/z (a) of the triply-charged C 34 (NAPQI)SYDEHAK peptide and m/z (b) representing the doubly-charged C 579 (NAPQI)KDALA peptide. 96

98 in a lower modification level of Cys579 and possibly insufficient abundance for detection, or a complete absence of Cys579 modification. The total MSA concentration in the final sample and in serum was calculated based on the results of the Bradford assay and the empirically established recoveries of MS after each of the sample preparation steps, see Table 1. Liver toxicity is the ADR associated with APAP overdose and changes in the serum albumin level could be a first indication of a disruption in liver function. However, the albumin levels in the APAP-dosed mice were not different from those of the control mice, nor were there significant differences detected between the one- and seven-day experiments. By application of the relative quantification strategy based on the tryptic Cys34 peptides, the NAPQI MSA adduct levels detected in the dosed mouse serum samples of both the one- and seven-day experiment were found to be 0.20% and 0.21% of the total MSA concentration, respectively, which represents a final sample concentration of NAPQI MSA in the nanomolar range (Table 5.1). As expected, the detected adduct levels are extremely low, but, due to the prolonged drug treatment, higher NAPQI MSA adduct levels were expected in the seven-day experiment. Surprisingly, the detected adduct levels were similar for both mouse experiments, which is in good agreement with the plasma levels of liver enzymes and histological liver analysis. A possible explanation of these results could be adaptation to the drug treatment after prolonged exposure. On the other hand, the liver function may have already been severely disrupted after a single high dose of APAP such that prolonged exposure on seven consecutive days did not lead to the formation of more NAPQI and subsequent NAPQI MSA adducts than in the one-day experiment. The average NAPQI MSA serum concentration was calculated to be 2.1 and 1.8 nmol/ml serum for the one- and seven-day experiment, respectively. These levels are 50 to 700-fold higher than previously detected in patients after APAP overdose. [17] This disagreement may have several causes, such as inter-species differences and the different methods used for quantification. Furthermore, relatively high standard deviations were observed for the NAPQI MSA adduct levels. These may partly be due to variation caused by the sample preparation and analysis method, but mainly results from the individual biological variation between the eight mice in a treatment group. Similar to humans, each mouse is genetically distinct and a mouse population may display the same variation in response as a human population 5 Table 5.1 Summary of quantification results. Calculated MSA values 1 day 7 days (average ± stdev, n = 8) APAP Control APAP Control Total mg MSA/mL serum 69.3 ± ± ± ± 5.0 µm MSA final sample 60.3 ± ± ± ± 5.5 % MSA modified by NAPQI 0.20 ± ± µm NAPQI MSA final sample 0.12 ± ± nmol NAPQI MSA/mL serum 2.1 ± ±

99 [38-39] This was also observed for the current experiment, as reflected by the physiological behavior of the mice as well as the detected NAPQI MSA levels and standard deviations. On the other hand, great care should always be taken when extrapolating results from in vitro experiments to in vivo studies and from animal models to the human situation. The difference in covalent drug binding sites in vitro and in vivo, and between HSA and MSA may not be of any influence on the observed ADRs of the two species, but, when studying the mechanisms behind ADRs, a difference in number and absence or presence of drug binding sites on target proteins involved in human signaling pathways may influence the outcome significantly. 5.4 CONCLUSION The developed analytical strategy was successfully applied to mouse serum samples resulting from a drug exposure study and achieved detection and quantification of low-abundant drug albumin adducts despite the presence of an excess of non-modified protein. As expected, the NAPQI MSA adduct levels were extremely low, approximately 0.2% of the total MSA, but, due to the optimized strategy, the NAPQI MSA concentration in the final samples was still in a detectable range. Additionally, the increased selectivity and sensitivity of the targeted methodology allows for confirmation and characterization of the NAPQI MSA adduct in terms of localization of the adduct formation site. Another advantage of the examined identification and quantification approach is its generic nature, which permits application to different drug albumin adducts. 5 ACKNOWLEDGEMENTS The authors acknowledge the technical support provided by Dr. M. Stitzinger and M.H.M. van Tuyl, M.Sc. and the technicians of NOTOX BV ( s-hertogenbosch, The Netherlands) for measuring liver enzymes. This research was performed within the framework of project D3-201 Towards novel translational safety biomarkers for adverse drug toxicity of the Dutch Top Institute Pharma. 98

100 REFERENCES [1] Zhou, S.F., Chan, E., Duan, W., Huang, M., Chen, Y.Z., Drug bioactivation, covalent binding to target proteins and toxicity relevance. Drug Metab. Rev. 2005, 37, (1), [2] Castell, J.V., Castell, M., Allergic hepatitis induced by drugs. Curr. Opin. Allergy. Clin. Immunol. 2006, 6, (4), [3] Ju, C., The role of haptic macrophages in regulation of idiosyncratic drug reactions. Toxicol. Pathol. 2009, 37, (1), [4] Masson, M.J., Peterson, R.A., Chung, C.J., Graf, M.L., Carpenter, L.D., et al., Lymphocyte loss and immunosuppression following acetaminophen-induced hepatotoxicity in mice as a potential mechanism of tolerance. Chem. Res. Toxicol. 2007, 20, (1), [5] Zhou, S., Separation and detection methods for covalent drug protein adducts. J. Chromatogr. B. 2003, 797, (1 2), [6] Ikehata, K., Duzhak, T.G., Galeva, N.A., Ji, T., Koen, Y.M., et al., Protein targets of reactive metabolites of thiobenzamide in rat liver in vivo. Chem. Res. Toxicol. 2008, 21, (7), [7] Qiu, Y., Benet, L.Z., Burlingame, A.L., Identification of the hepatic protein targets of reactive metabolites of acetaminophen in vivo in mice using two-dimensional gel electrophoresis and mass spectrometry. J. Biol. Chem. 1998, 273, (28), [8] Landin, J.S., Cohen, S.D., Khairallah, E.A., Identification of a 54-kDa mitochondrial acetaminophen-binding protein as aldehyde dehydrogenase. Toxicol. Appl. Pharmacol. 1996, 141, (1), [9] Methogo, R.M., Dansette, P.M., Klarskov, K., Identification of liver protein targets modified by tienilic acid metabolites using a two-dimensional Western blot-mass spectrometry approach. Int. J. Mass Spectrom. 2007, 268, (2-3), [10] Koen, Y.M., Yue, W., Galeva, N.A., Williams, T.D., Hanzlik, R.P., Site-specific arylation of rat glutathione s-transferase A1 and A2 by bromobenzene metabolites in vivo. Chem. Res. Toxicol. 2006, 19, (11), [11] Gan, J., Ruan, Q., He, B., Zhu, M., Shyu, W.C., et al., In vitro screening of 50 highly prescribed drugs for thiol adduct formation--comparison of potential for drug-induced toxicity and extent of adduct formation. Chem. Res. Toxicol. 2009, 22, (4), [12] Ghuman, J., Zunszain, P.A., Petitpas, I., Bhattacharya, A.A., Otagiri, M., et al., Structural basis of the drug-binding specificity of human serum albumin. J. Mol. Biol. 2005, 353, (1), [13] Moreland, J.L., Gramada, A., Buzko, O.V., Zhang, Q., Bourne, P.E., The Molecular Biology Toolkit (MBT): a modular platform for developing molecular visualization applications. BMC Bioinformatics. 2005, 6, 21. [14] Azarkan, M., Huet, J., Baeyens-Volant, D., Looze, Y., Vandenbussche, G., Affinity chromatography: A useful tool in proteomics studies. J. Chromatogr. B. 2007, 849, (1 2), [15] Lee, W.C., Lee, K.H., Applications of affinity chromatography in proteomics. Anal. Biochem. 2004, 324, (1), [16] Yang, X.X., Hu, Z.P., Chan, S.Y., Zhou, S.F., Monitoring drug protein interaction. Clin. Chim. Acta. 2006, 365, (1-2), [17] Damsten, M.C., Commandeur, J.N., Fidder, A., Hulst, A.G., Touw, D., et al., Liquid chromatography/tandem mass spectrometry detection of covalent binding of acetaminophen to human serum albumin. Drug Metab. Dispos. 2007, 35, (8), [18] Noort, D., Hulst, A.G., Jansen, R., Covalent binding of nitrogen mustards to the cysteine-34 residue in human serum albumin. Arch. Toxicol. 2002, 76, (2), [19] Evans, D.C., Watt, A.P., Nicoll-Griffith, D.A., Baillie, T.A., Drug protein adducts: an industry perspective on minimizing the potential for drug bioactivation in drug discovery and development. Chem. Res. Toxicol. 2004, 17, (1), [20] Hoos, J.S., Damsten, M.C., de Vlieger, J.S.B., Commandeur, J.N.M., Vermeulen, N.P.E., et al., Automated detection of covalent adducts to human serum albumin by immunoaffinity chromatography, online solution phase digestion and liquid chromatography mass spectrometry. J.Chromatogr. B. 2007, 859, (2), [21] Loo, J.A., Edmonds, C.G., Smith, R.D., Tandem mass spectrometry of very large molecules: serum albumin sequence information from multiply charged ions formed by electrospray ionization. Anal. Chem. 1991, 63, (21), [22] Otagiri, M., A molecular functional study on the interactions of drugs with plasma proteins. Drug Metab. Pharmacokinet. 2005, 20, (5), [23] Lin, J.H., Species similarities and differences in pharmacokinetics. Drug Metab. Dispos. 1995, 23, (10), [24] Kwast, L.M., Fiechter, D., Hassing, I., Bleumink, R., Boon, L., et al., Oral exposure to drugs with immune-adjuvant potential induces hypersensitivity responses to the reporter antigen TNP-OVA. Toxicol. Sci. 2011, 121, (2),

101 [25] Switzar, L., Giera, M., Lingeman, H., Irth, H., Niessen, W.M., Protein digestion optimization for characterization of drug protein adducts using response surface modeling. J. Chromatogr. A. 2011, 1218, (13), [26] Kelleher, P.C., Smith, C.J., Pannell, R., Chromatography of non-human albumins on Cibacron Blue agarose : Application to the separation of albumin from rat alpha-fetaprotein. J. Chromatogr. A. 1979, 173, (2), [27] Gasteiger, E., Gattiker, A., Hoogland, C., Ivanyi, I., Appel, R.D., et al., ExPASy: the proteomics server for in-depth protein knowledge and analysis. Nucleic Acids Res. 2003, 31, (13), [28] Bronsema, K.J., Bischoff, R., van de Merbel, N.C., Internal standards in the quantitative determination of protein biopharmaceuticals using liquid chromatography coupled to mass spectrometry. J. Chromatogr. B. 2012, , (0), [29] Huang, Q., Jin, X., Gaillard, E.T., Knight, B.L., Pack, F.D., et al., Gene expression profiling reveals multiple toxicity endpoints induced by hepatotoxicants. Mutat. Res. 2004, 549, (1-2), [30] Jollow, D.J., Mitchell, J.R., Potter, W.Z., Davis, D.C., Gillette, J.R., et al., Acetaminophen-induced hepatic necrosis. II. Role of covalent binding in vivo. J. Pharmacol. Exp. Ther. 1973, 187, (1), [31] Mitchell, J.R., Jollow, D.J., Potter, W.Z., Davis, D.C., Gillette, J.R., et al., Acetaminophen-induced hepatic necrosis. I. Role of drug metabolism. J. Pharmacol. Exp. Ther. 1973, 187, (1), [32] Singhal, R., Ganey, P.E., Roth, R.A., Complement activation in acetaminophen-induced liver injury in mice. J. Pharmacol. Exp. Ther. 2012, 341, (2), [33] Amacher, D.E., Serum transaminase elevations as indicators of hepatic injury following the administration of drugs. Regul. Toxicol. Pharmacol. 1998, 27, (2), [34] Ozer, J., Ratner, M., Shaw, M., Bailey, W., Schomaker, S., The current state of serum biomarkers of hepatotoxicity. Toxicology. 2008, 245, (3), [35] Yapar, K., Kart, A., Karapehlivan, M., Atakisi, O., Tunca, R., et al., Hepatoprotective effect of L-carnitine against acute acetaminophen toxicity in mice. Exp. Toxicol. Pathol. 2007, 59, (2), [36] Kim, Y.C., Lee, S.J., Temporal variation in hepatotoxicity and metabolism of acetaminophen in mice. Toxicology 1998, 128, (1), [37] Singer, A.J., Carracio, T.R., Mofenson, H.C., The temporal profile of increased transaminase levels in patients with acetaminophen-induced liver dysfunction. Ann. Emerg. Med. 1995, 26, (1), [38] Festing, M.F., Genetic factors in toxicology: implications for toxicological screening. Crit.Rev. Toxicol. 1987, 18, (1), [39] Lathe, R., The individuality of mice. Genes Brain Behav. 2004, 3, (6),

102 101 5

103

104 Chapter 6 Chapter 6 Summary, conclusion and perspectives

105 6 104

106 6.1 SUMMARY The occurrence of ADRs is one of the main reasons for the high attrition rates in the drug discovery and development pipeline and contributes to the high costs for the pharmaceutical industry to bring new drugs to the market. A comparison of the various reasons for drug attrition between 1991 and 2000 is depicted in Figure 6.1. [1] Next to efficacy and commercial reasons, toxicology issues are an important reason for attrition, with even a significant increase between 1991 and A more recent study of drug failures in phase II and phase III clinical trials performed in shows that drug safety continues to be a major issue in drug development accounting for 22% of the overall failure rate of drugs. [2] The assessment of the metabolic stability of drugs and the characterization of metabolites has been recognized as an important aspect in the production of safer drugs. [3] In these so-called MIST (Metabolites in Safety Testing) guidelines, the monitoring and toxicity assessment of (re)active drug metabolites throughout the whole drug development process is regulated. Characterization of reactive metabolites is often performed using trapping agents, such as glutathione, in order to gain a better understanding of the metabolic pathways behind drug (bio)activation. [4-5] However, from an industry point of view, the potential for metabolic activation of drugs should be minimized because this is seen as the first step towards drug toxicity. [6] As outlined in Chapter 1, there are several theories on the mechanism behind ADRs. One of the assumptions underlying this thesis is that the covalent binding of reactive metabolites to proteins plays an important role in ADRs. This is known as the hapten theory. [7] In current practice, covalent binding of reactive metabolites to liver proteins is only assessed on a global level using radiolabeled drugs and radioactivity 6 Figure 6.1 Reasons for attrition in 1991 and Reprinted by permission from Macmillan Publishers Ltd: Nature Reviews Drug Discovery [1],

107 assays. A conservative threshold value of 50 pmol drug equiv/mg total liver protein has been suggested to aid in the selection of suitable drug candidates [6], but less stringent requirements may apply for drugs for severe diseases that are difficult to treat or incurable [8]. However, studies have shown that not all drug protein adducts may induce an immune response, for instance, comparable covalent binding levels of acetaminophen (APAP) and its non-toxic regioisomer 3 -hydroxyacetanilide in vivo in mice showed that only APAP led to hepatotoxicity. [9] This finding may indicate that only adduct formation of to certain critical proteins plays a role in organ toxicity. However, the same regioisomer was found to be equally or more toxic than APAP in rat and human in an ex vivo study using precision-cut liver slices. [10] Therefore, a detailed investigation is needed to assess which drug protein adducts are formed and if they are related to ADRs. Current strategies for the identification of hepatic protein targets of reactive metabolites involve the use of a radiolabeled drug in animal experiments followed by analysis of the liver proteins with two-dimensional gel electrophoresis combined with radioactivity assays and mass spectrometry. [11] The list of identified protein targets continues to grow, but the identification of the specific reactive metabolite, protein target and modification site using such strategies is a significant challenge due to the failure in detection of the adducted peptides. [12] The aim of the research described in this thesis was to develop an advanced analytical methodology, based on proteomics approaches, to study drug protein adducts in vivo in biofluids and eventually tissues. The essential difference between a global strategy and our envisaged approach is that we wanted to approach the topic at protein level in order to obtain absolute proof of drug protein adduct formation. The aim was to use targeted proteomics approaches to establish the identity of the drug protein adducts at the (single) protein level and use this information to (semi-)quantitatively assess the level of adduct formation. The general workflow in a proteomics approach consists of a number of steps: (1) separation and/or fractionation of a complex protein sample, (2) denaturation, reduction and alkylation of the isolated protein (mixture), (3) enzymatic digestion of the proteins into peptides, (4) LC MS/MS analysis of the peptides, and (5) data processing using bioinformatics tools. We considered the conversion of the proteins into peptides to be one of the critical steps in the complete procedure and decided to optimize this step. In order to be able to adequately monitor this optimization, all sample preparation steps were optimized to obtain optimum results. 6 Additionally, the LC MS/MS analysis had to be optimized, especially considering the fact that new rapid-resolution LC using columns packed with (in those days) innovative 1.8 μm ID particles would be used. Under such conditions, it is important to find the best compromise between speed of analysis and information content. In fact, when considering information content in relation to LC MS MS performance, optimization can be directed at optimum response for a particular drug protein target or at the highest possible sequence coverage of the protein(s) analyzed. The former would be especially important in targeted analysis, whereas the latter would possibly enable the discovery of yet unknown drug protein adducts. The same line of thinking can also be applied to enzymatic digestion. In most bottom-up proteomics experiments, the emphasis lies on achieving high protein coverage for accurate protein identification, whereas detection and identification of the modification and the modification site, i.e., only a certain part of the protein sequence, is paramount for confirmation of drug protein adduct formation. 106

108 However, the in vivo concentrations of drug protein adducts are generally very low, as typically less than 1% of the protein is adducted. Thus, a well-designed and optimized analytical methodology is required to meet the challenge in the detection of low-abundant drug protein adducts. Chapter 3 describes the optimization of all phases of sample preparation and analysis specifically for drug albumin adducts with a main focus on enzymatic digestion. The effect of different enzymes and digestion conditions on the identification of drug albumin adducts in terms of protein coverage and detection of a specific peptide containing the modification was evaluated through the innovative application of a Design of Experiments. This approach allows for a more efficient and more accurate determination of the optimum digestion conditions through simultaneous optimization of multiple variables and responses, and avoids the disadvantages associated with the traditional one-variable-at-a-time approach. The digestion conditions were optimized with a model adduct of monochlorobimane to human serum albumin (MCB HSA), and the optimized conditions were applied to an adduct of N-acetyl-p-benzoquinoneimine (NAPQI), the reactive metabolite of APAP, to HSA (NAPQI HSA). In this study, it was shown that optimization of digestion conditions for a specific application is very meaningful because it leads to increased efficiency and sensitivity for the detection and identification of NAPQI HSA adducts. In particular for the less specific and less well-described enzyme thermolysin, both the protein coverage and sensitivity for the adduct modification site was improved several fold. The approach described above is termed bottom-up proteomics because it involves digestion of the proteins into peptides and subsequent identification of the peptides in order to establish the identity of the original protein, thus from the bottom up. Despite increasing popularity of other proteomics approaches, bottom-up proteomics remains the method of choice for the analysis of proteins (and their modifications). Protein digestion is the most crucial step in such an approach and can be performed in a variety of ways using proteolytic enzymes for enzymatic digestion or using chemicals for non-enzymatic digestion. Many different enzymes and reagents exist for this purpose, but traditional digestion protocols often apply trypsin, the gold standard in enzymatic digestion, and are performed during an overnight incubation. Protein digestion is often the bottleneck in terms of time consumption and, in recent years, much effort has been invested in the development of techniques for the acceleration of this process. Alternatively, increased throughput can also be achieved through incorporation of the digestion step into online systems, which allows for automation and reduces sample handling. These state-of-the-art protein digestion strategies are reviewed in Chapter 2 of this thesis and may reduce protein digestion time from hours to as short as seconds, which greatly reduces the time needed for sample preparation. As in vivo analysis of drug protein adducts was aimed at, another important aspect of sample preparation is the extraction of the target protein from a complex biological matrix. As stated above, less than 1% of a protein is modified by reactive drug metabolites, which means that the very similar, non-adducted protein is present in at least a hundred-fold excess. Selective purification methods for extraction of only the adducted proteins are not yet available, meaning that, following extraction of the target protein, a mixture of adducted (<1%) and non-adducted (>99%) protein is obtained. A sufficient amount of target protein needs to be extracted, otherwise the presence of the drug protein adduct may be masked by the excess of non-adducted protein. Chapter 5 describes the development of a dedicated sample preparation, digestion and analysis protocol for the in vivo identification and quantification of drug albumin adducts in serum. This protocol consists of a number of steps: (1) affinity chromatography for the extraction of albumin from 6 107

109 serum, (2) buffer exchange of the extracted albumin sample to denaturation buffer using a gel-filtration column, (3) disulfide reduction and cysteine alkylation to achieve complete unfolding of the extracted albumin, (4) buffer exchange of the unfolded albumin sample to pure water using a gel-filtration column, (5) freeze-drying to concentrate the sample and to bring the NAPQI albumin adduct within the detectable range, (6) enzymatic digestion using the optimized conditions from Chapter 3, (7) LC MS/MS analysis of the generated peptides, and (8) data analysis, both manually and with the aid of database search software, for identification of drug albumin adducts. For final application of the developed protocol to actual biological samples, collaboration was sought to one of the partners in the larger project group that performs animal experiments, studying (among other) immunoactivity of drug protein adducts in mice models. For our project, we received serum samples from a drug exposure study in which the mice received a high dose of APAP on either a single or seven consecutive days. Since the mouse is a small animal, the serum samples are only available in very limited amounts, which also added to the challenge. The mouse study was designed to study the kinetics in ADRs via the different treatment groups, thus, it was also pertinent to develop a quantification strategy to assess whether a relationship to the formation of NAPQI albumin adducts exists. Several options, such as the use of albumins of other species and synthetic NAPQI albumin adducts, were evaluated, but the best approach was found to be a relative quantification via comparison of the peptide peak areas obtained from the adducted and non-adducted albumin. One other factor that demanded consideration was the presence of two free cysteines on the surface of mouse serum albumin (MSA), which is different from human albumin where only one surface cysteine is present. In the end, the developed strategy allowed for successful identification of NAPQI MSA adducts in vivo in mouse serum for the first time. NAPQI modification was only detected on one of the free cysteines, which could be due to an in vivo preference for this modification site. On the other hand, despite the large dose of 300 mg APAP/kg that was administered to the mice, only extremely low levels of NAPQI albumin binding of 0.2% of the total albumin were detected, which may indicate that low-level NAPQI-adduct formation to the other modification site may have been below the detection level. Furthermore, similar adduct levels were detected in both treatment groups, which may suggest that a single high dose of APAP may already severely affect liver function and, thus, the production of both reactive drug metabolites as well as serum albumin. 6 As described above, sample preparation protocols for bottom-up proteomics can be very laborious and time-consuming, especially when a large number of samples need to be treated. Although the use of gel-filtration columns in the previous study was effective, the number of samples that could be prepared simultaneously with this individual sample clean-up procedure was limited for practical reasons. When large sample sets need to be analyzed for comparative studies or quantitation purposes, the sample preparation and analysis should ideally be performed in one batch. Chapter 4 describes the development of a high-throughput sample preparation approach based on 96-well filter plates that allows for the simultaneous preparation of 96 samples. The use of a molecular weight cut-off membrane has several advantages, including a reduction of the number of sample transfer steps, thereby reducing the potential for sample loss, and enabling simultaneous sample concentration. The latter also avoids the use of a freeze-drying step and increases the overall sensitivity. The particular protocol described in this chapter was developed for global cellular proteomics and performed significantly better than the gel-filtration protocol 108

110 in terms of the number of identified peptides and proteins. This difference in performance was already visible using the rapid-resolution LC QTOF system for analysis, but was far more evident when the filter plate methodology was combined with more advanced nanolc Orbitrap MS analysis. The final filter plate and LC MS analysis protocol resulted in the identification of more than 400 cellular proteins. This number may still be increased through the use of two-dimensional LC to improve peptide separation and detection. The developed methodology can also be used for targeted protein analysis and may easily be applied to other proteomic samples. 6.2 DISCUSSION AND PERSPECTIVES The project plan provided basically two general lines of research, one directed at the offline and online targeted analysis of specific drug protein adducts, e.g., NAPQI albumin, and another one at non-targeted analysis, enabling screening for (circulating or tissue) proteins that are susceptible to adduction. The proposed online targeted approach was envisaged based on previous research in our group, involving the development of an online system for the analysis of target proteins in complex matrices using immunoaffinity selection of the target protein, subsequent digestion of the protein in an immobilized enzyme reactor, collection of the resulting peptides on an SPE column, and finally LC MS/MS analysis of the peptides [13-14]. The project plan turned out to be too ambitious for application to drug protein adduct analysis in vivo, given the high analytical challenges that have to be met. The in vivo concentrations of drug protein adducts are generally very low, as typically less than 1% of the protein is adducted (Chapter 5). NAPQI HSA is the most well-known example of drug protein adduct formation and only results in ADRs after an overdose of APAP, which leads to increased levels of NAPQI resulting in depletion of GSH and detectable levels of NAPQI albumin adducts. [15] Most types of adverse reactions are considered to be dose-related, see Chapter 1, thus may also result in detectable drug protein adducts. However, some drugs may already result in immune-mediated ADRs after a single low dose or after extended use at low dose, which means that the level of drug protein binding may be even lower. In addition, the work in this thesis was mainly focused on the analysis of drug-adduct formation to albumin, which is the most abundant serum protein. Other protein targets of reactive drug metabolites will have lower abundance, which may also imply that the drug protein adducts will be even more difficult to detect, especially in global proteomics studies. Another factor that may influence the in vivo concentration of drug protein adducts is the generation of multiple reactive metabolites. Although APAP only produces one reactive metabolite, which can react with cysteines in two different ways, other drugs, such as diclofenac and clozapine, are bioactivated to multiple reactive species, which may lead to the formation of different adducts at the same protein. In addition, the multiple reactive metabolites may target more than one protein, which has also become evident from the in vivo covalent binding studies with radioactive labeled drugs. Both issues may result in a dilution effect, which leads to failure in the detection of drug protein adducts. [12] 6 109

111 Furthermore, if a small-molecule drug is covalently bound to the protein, only very minor changes in mass and/or physicochemical properties result. Thus, in order to analyze the drug protein adduct within a complex biological matrix containing an at least hundred-fold excess of the very similar, non-adducted protein requires highly specific sample pretreatment as well as extensively optimized steps in the analytical methodology. The online digestion systems are more suitable for the analysis of concentrated, purified proteins and this is difficult to achieve for drug protein adducts. Therefore, the focus was shifted more towards offline methodologies. In the end, we succeeded in performing semi-quantitative analysis of low-level APAP-mouse serum albumin adducts in vivo using an offline approach (Chapter 5). The strategy was extensively optimized, but may be further improved through the development and application of selective drug protein adduct extraction. Adduct formation to HSA by systemic electrophiles as well as drugs takes place on its single free cysteine residue on position 34. Therefore, a method for the purification of modified HSA using a thiol affinity resin has been suggested. [16] Only HSA that does not contain any modifications to this cysteine binds to the resin. The non-bound fraction, containing the modified HSA, can subsequently be processed and analyzed. This approach could be added to the developed strategy following HSA extraction from serum to enrich drug-hsa adducts prior to digestion. However, it is not selective for drug protein adducts because other cysteine modifications, such as oxidation (non-mercaptalbumin), will be co-enriched as well. Additionally, the applicability of this enrichment technique is limited when protein targets of reactive metabolites possess multiple free cysteines, such as glutathione-s-transferase p1 [17] and mouse serum albumin [18]. 6 Therefore, there still is a need for truly selective enrichment strategies for drug protein adducts. One possibility may be the use of drug protein adduct-specific antibodies, used either at the protein or the peptide level. Antibodies are raised in animal models, to obtain polyclonal antibodies, or in immortalized cell lines, to produce the more selective monoclonal antibodies. These antibodies could be raised against the intact drug protein adduct to enrich on the protein level, but may also be raised against signature peptides containing the adduct formation site in order to perform enrichment on the peptide level following digestion of the drug protein adduct. Antibody enrichment on the peptide level has already proven its usefulness in protein biomarker studies as part of the Stable Isotope Standards and Capture by Anti-Peptide Antibodies (SISCAPA) approach, in which also isotopically labeled, synthetic peptides are used for quantitation of the enriched target peptides. [19] Such an approach could potentially be very valuable in drug protein adduct analysis. The results obtained in Chapter 4 show that a sample preparation approach based on molecular weight cut-off (MWCO) membranes performed significantly better for global cellular protein profiling than the gel-filtration procedure, which was used in Chapter 5. Sample handling and transfer, and thus, sample loss is minimized with the filter plate methodology, which may be very beneficial for the global or targeted analysis of low-abundant proteins. Additionally, the study also showed that the type of LC MS instrument used for analysis greatly affected the results. The high resolution and efficiency of the nanolc Orbitrap MS system improved the number of identified peptides and proteins in comparison with the rapid-resolution LC QTOF system. The possibility of performing multiple MS/MS fragmentations in a short time period means that not only the highest abundant, but also less abundant peptide ions will be selected for fragmentation. 110

112 This feature, in combination with the increased sensitivity of nano-electrospray ionization (nano-esi), may lead to the additional identification of low-abundant peptides, which are often related to low-abundant modifications. The work in this thesis was focused mainly on the targeted analysis of drug albumin adducts, but the other envisaged research line was the development of a non-targeted screening methodology for the identification of potential target proteins. A first step in this direction was made by setting up an approach for the identification of proteins containing free surface cysteines that are accessible to reactive drug metabolites using fluorescent labeling. This approach consisted of (1) albumin depletion from human serum, (2) thiol-reactive fluorescent labeling of the free cysteine-containing proteins in the depleted serum, (3) size-exclusion chromatography (SEC) in combination with fluorescence detection for fractionation and collection of the fluorescent protein fractions, (4) sample preparation and digestion using spin filters and previously optimized conditions, (5) offline LC MALDI-TOF/TOF MS analysis of the digested fractions, and (6) data analysis (manual and using bioinformatics tools) for identification of labeled proteins. Using this approach, three labeled proteins were identified, but only one of the proteins, Ig alpha-1 chain C region, contained a fluorescent label on a cysteine. The other two identified proteins, Alpha-2-macroglobulin and Ig kappa chain C region, were found to be labeled on one or more histidines. The confidence in their identification was high and additional examination of the 3D-structures of the identified proteins, when available, showed that the identified labeling sites are on the surface of the protein and may, thus, Lu Figure 6.2 SEC fractionation of albumin-depleted, fluorescently-labeled serum. The blue line represents the fluorescence chromatogram measured at an excitation wavelength of 390 nm and an emission wavelength of 478nm, and the red line represents the UV chromatogram measured at 220 nm. The black dotted line represents the FLD chromatogram obtained from the analysis of MCB. The fluorescent fractions that were collected for analysis are labeled 1-5. Fluorescently labeled Ig alpha-1 chain C region, Alpha-2-macroglobulin, and Ig kappa chain C were detected in these fractions (L. Switzar et al., unpublished data). min 6 111

113 be accessible to the label or, possibly, a reactive drug metabolite. The fluorescent label used in this study was MCB, which is thiol-reactive. However, the large excess of the label and possibly other conditions, such as the long (4 h) reaction time, that were used may have led to the off-target labeling of histidine residues. Therefore, further optimization of the fluorescent labeling conditions is required to reduce this non-selective labeling. Furthermore, serum was chosen in this initial study, but, due to the complexity of this matrix and the wide dynamic range, only abundant labeled proteins were identified. The sensitivity for lower abundant proteins may be improved through depletion of the most abundant proteins, for instance, using a multiple affinity removal system [20], or a different fractionation technology, since the resolution of SEC is relatively low (see Figure 6.2). Additionally, another matrix, such as liver, may perhaps be a better option, considering the fact that reactive drug metabolites are formed in the liver. Although little is known about the mechanisms behind drug protein adduct formation, the short half-lives of the highly reactive metabolites may imply that adduct formation takes place in the liver, which may also suggest that higher adduct concentrations are present in this matrix. Finally, whereas the use of the offline LC MALDI-TOF MS system was already a significant improvement over the rapid-resolution LC QTOF system, further improvements may be achieved through the use of even more efficient MS instruments, such as the Orbitrap MS. 6 All of the methods described in this thesis have been based on bottom-up proteomics. However, recent developments in MS instrumentation now allow for the detection of intact proteins and identification of their post-translational or other modifications. From these so-called top-down proteomics experiments, the intact protein mass is obtained. Proteins up to 50 kda may also be fragmented to obtain sequence information and, thus, identification and localization of modifications may be achieved. Although the mass difference between a drug-adducted protein and its non-modified form is too small to achieve a separation by common chromatographic separation technologies, it is sufficient for mass spectrometric discrimination. In theory, low resolution instruments (R<5000) should already be able to distinguish both species of the protein. For example, the mass spectrometry discrimination of non-modified HSA (~66500 Da) and the NAPQI HSA drug protein adduct ( Da) requires a resolution (R=M/ΔM, full width measured at half-maximum) of ~500, based on the singly charged species. However, from our own experience, the large difference in concentration, the obtained peak widths and the presence of other, naturally occurring protein modifications (isoforms, post-translational modifications) result in baseline noise/crowding and obscure the low-abundant drug protein adduct. In addition, small mass modifications to proteins may also be obscured by overlapping isotopic distributions of intact proteins. [21] Figure 6.3 shows a comparison of the protein profiles obtained from the analysis of an intact MCB HSA adduct (mass shift of Da) using MALDI-TOF MS and ESI-QTOF MS. The typical resolution of a MALDI-TOF instrument operated in reflectron mode is for analytes with a mass of <5000 Da, but larger molecules, such as proteins, can only be analyzed in linear mode with a significantly lower resolution of ~1000. Figure 6.3a shows that the effective resolution (m/z / FWHM) obtained for HSA actually is <100, most likely due to its higher mass. This resolution is not sufficient for resolving the various forms of HSA, including MCB HSA, present in the sample. Since mass is measured as a function of charge, the production of multiply charged ions with ESI leads to multiple charge states of the protein that are detected in the lower m/z range where higher resolution can be obtained. In addition, the resolution of the MaXis 112

114 a Intens. [a.u.] % m/z Intens b m/z Intens. 4 x c m/z 6 Figure 6.3 Results of the analysis of intact MCB HSA; (a) MALDI-TOF MS spectrum obtained with a Bruker Ultraflex III instrument using sinapinic acid as matrix, (b) ESI-QTOF MS spectrum obtained with a Bruker MaXis QTOF instrument and (c) deconvoluted ESI-QTOF MS spectrum. The total protein concentration in this particular sample was 0.9 mg/ml and an estimated 10-20% of the HSA was modified by MCB at the cysteine 34 position (L. Switzar, unpublished data). 113

115 ESI-QTOF instrument used for this experiment is considerably higher ( ) and is sufficient for the analysis of intact HSA. Even with this high-resolution instrument, the overlapping charge envelopes could not be completely resolved (effective resolution ~2000), probably due to the presence of other HSA isoforms, see Figure 6.3b. However, deconvolution of the obtained data does show a distinction between non-modified HSA and MCB HSA and may even reveal the presence of a second MCB modification on HSA, see Figure 6.3c. Thus, high-resolution instruments or rather ultra-high resolution instruments, such as Fourier transform ion cyclotron resonance (FTICR) and Orbitrap MS, are required for analysis of intact drug protein adducts. However, sensitivity may become an issue, in particular with increasing protein mass and lower modification levels. For example, the MCB modification level of HSA in this sample was approximately 10-20%, but the in vivo modification will be considerably lower. Therefore, selective drug protein adduct purification methods will still be required for identification of drug protein adducts from in vivo samples using top-down proteomics. 6 Ion mobility MS (IMMS) presents another possibility for the detection and identification of drug protein adducts. In IMMS, proteins are separated based on size and shape (folding and conformation) rather than mass. In order to achieve separation in IMMS, covalent adduction of a reactive metabolite to a protein should induce a sufficient change in the shape of the protein (collision cross section) since the size may not change significantly due to the addition of a small metabolite to a relatively large protein. Many examples exist of the IMMS investigation of protein-ligand interactions, for instance, the interaction of the Bcl-2 related survival factor protein (Bcl-xL, ~26 kda) and the small-molecule inhibitor ABT-737 (~800 Da). [22] Binding of ABT-737 induces a conformational change in Bcl-xL and the Bcl-xL/ABT-737 complex could be separated from Bcl-xL after optimization of instrument parameters. Although this example resembles the situation of covalent modification of proteins by reactive drug metabolites, the metabolites are generally smaller than 800 Da. It will need to be investigated whether they induce a change in conformation and/or collisional cross section large enough to achieve successful ion mobility separation. However, commercially available IMMS instruments can separate isoforms of large proteins, such as minor glycoforms of IgG (148 kda), and have also shown to be able to achieve isotopic resolution for proteins up to 30 kda. [23] These IMMS instruments also possess the option for top-down sequencing of proteins through ETD fragmentation. If ETD fragmentation could take place following ion mobility separation of the drug protein adduct from its non-modified form, this would be an ideal instrument for confirmation and localization of adduct formation. However, the ETD cell is located in front of the ion mobility tube and only CID fragmentation can be performed following IMMS separation, which is not suitable for the fragmentation of large molecules. When dedicated detection and identification strategies for individual drug protein adducts become available, the covalent binding thresholds will need to be reevaluated. Currently, the threshold for global covalent binding is 50 pm [6], but when the identity of the specific drug protein adduct(s) is known and the relationship to ADRs can be established, this threshold will need to be redefined for each drug or drug protein adduct. Quantification of in vivo drug protein adduct formation may be performed using existing methods, such as spectral counting or the above suggested modified SISCAPA approach. Other possibilities may include the use of isotopically labeled drugs in animal studies, or synthetic, isotopically labeled reference standards of drug protein adducts for quantification in human samples from clinical trials. In the end, state-of-the-art LC MS based proteomics methods for detection, identification and 114

116 quantification of drug protein adducts may contribute to the establishment of the relationship of drug protein adducts to ADRs and the levels at which they may cause toxicity, which in turn may aid to reduce drug attrition rates and the development of safer drugs. REFERENCES [1] Kola, I., Landis, J., Can the pharmaceutical industry reduce attrition rates? Nat. Rev. Drug Discov. 2004, 3, (8), [2] Khanna, I., Drug discovery in pharmaceutical industry: productivity challenges and trends. Drug Discov. Today. 2012, 17, (19-20), [3] FDA, Guidance for Industry. Safety Testing of Drug Metabolites. 2008, GuidanceComplianceRegulatoryInformation/Guidances/ucm pdf [4] Dragovic, S., Boerma, J.S., van Bergen, L., Vermeulen, N.P.E., Commandeur, J.N.M., Role of Human Glutathione S-Transferases in the Inactivation of Reactive Metabolites of Clozapine. Chem. Res. Toxicol. 2010, 23, (9), [5] Dragovic, S., Gunness, P., Ingelman-Sundberg, M., Vermeulen, N.P.E., Commandeur, J.N.M., Characterization of Human Cytochrome P450s Involved in the Bioactivation of Clozapine. Drug Metab. Dispos. 2013, 41, (3), [6] Evans, D.C., Watt, A.P., Nicoll-Griffith, D.A., Baillie, T.A., Drug protein adducts: an industry perspective on minimizing the potential for drug bioactivation in drug discovery and development. Chem. Res. Toxicol. 2004, 17, (1), [7] Kevin Park, B., Coleman, J.W., Kitteringham, N.R., Drug disposition and drug hypersensitivity. Biochem. Pharmacol. 1987, 36, (5), [8] Baillie, T.A., Cayen, M.N., Fouda, H., Gerson, R.J., Green, J.D., et al., Drug metabolites in safety testing. Toxicol. Appl. Pharmacol. 2002, 182, (3), [9] Tirmenstein, M.A., Nelson, S.D., Subcellular binding and effects on calcium homeostasis produced by acetaminophen and a nonhepatotoxic regioisomer, 3 -hydroxyacetanilide, in mouse liver. J. Biol. Chem. 1989, 264, (17), [10] Hadi, M., Dragovic, S., Swelm, R., Herpers, B., Water, B., et al., AMAP, the alleged non-toxic isomer of acetaminophen, is toxic in rat and human liver. Arch.Toxicol. 2013, 87, (1), [11] Ikehata, K., Duzhak, T.G., Galeva, N.A., Ji, T., Koen, Y.M., et al., Protein targets of reactive metabolites of thiobenzamide in rat liver in vivo. Chem. Res. Toxicol. 2008, 21, (7), [12] Koen, Y.M., Yue, W., Galeva, N.A., Williams, T.D., Hanzlik, R.P., Site-specific arylation of rat glutathione s-transferase A1 and A2 by bromobenzene metabolites in vivo. Chem. Res. Toxicol. 2006, 19, (11), [13] Hoos, J.S., Damsten, M.C., de Vlieger, J.S., Commandeur, J.N., Vermeulen, N.P., et al., Automated detection of covalent adducts to human serum albumin by immunoaffinity chromatography, online solution phase digestion and liquid chromatography-mass spectrometry. J. Chromatogr. B. 2007, 859, (2), [14] Hoos, J.S., Sudergat, H., Hoelck, J.P., Stahl, M., de Vlieger, J.S., et al., Selective quantitative bioanalysis of proteins in biological fluids by online immunoaffinity chromatography-protein digestion-liquid chromatography-mass spectrometry. J. Chromatogr. B. 2006, 830, (2), [15] Damsten, M.C., Commandeur, J.N., Fidder, A., Hulst, A.G., Touw, D., et al., Liquid chromatography/tandem mass spectrometry detection of covalent binding of acetaminophen to human serum albumin. Drug. Metab. Dispos. 2007, 35, (8), [16] Funk, W.E., Li, H., Iavarone, A.T., Williams, E.R., Riby, J., et al., Enrichment of cysteinyl adducts of human serum albumin. Anal. Biochem. 2010, 400, (1), [17] Boerma, J.S., Dragovic, S., Vermeulen, N.P., Commandeur, J.N., Mass spectrometric characterization of protein adducts of multiple P450-dependent reactive intermediates of diclofenac to human glutathione-s-transferase P1-1. Chem. Res. Toxicol. 2012, 25, (11), [18] Switzar, L., Kwast, L.M., Lingeman, H., Giera, M., Pieters, R.H., et al., Identification and quantification of drug albumin adducts in serum samples from a drug exposure study in mice. J. Chromatogr. B. 2013, ,

117 [19] Anderson, N.L., Anderson, N.G., Haines, L.R., Hardie, D.B., Olafson, R.W., et al., Mass spectrometric quantitation of peptides and proteins using Stable Isotope Standards and Capture by Anti-Peptide Antibodies (SISCAPA). J. Proteome. Res. 2004, 3, (2), [20] Björhall, K., Miliotis, T., Davidsson, P., Comparison of different depletion strategies for improved resolution in proteomic analysis of human serum samples. Proteomics 2005, 5, (1), [21] Rhoads, T.W., Williams, J.R., Lopez, N.I., Morre, J.T., Bradford, C.S., et al., Using theoretical protein isotopic distributions to parse small-mass-difference post-translational modifications via mass spectrometry. J. Am. Soc. Mass Spectrom. 2013, 24, (1), [22] Atmanene, C., Petiot-Becard, S., Zeyer, D., Van Dorsselaer, A., Vivat Hannah, V., et al., Exploring key parameters to detect subtle ligand-induced protein conformational changes using traveling wave ion mobility mass spectrometry. Anal. Chem. 2012, 84, (11), [23] Waters, Enabling Greater Capability in the Characterization of Biomolecules with the Incorporation of Novel Off-Axis StepWave Ion Transfer Optics in the High Resolution SYNAPT G2-S System. 2011, com/webassets/cms/library/docs/ en.pdf 6 116

118 117 6

119

120 Appendices Nederlandse samenvatting List of abbreviations List of publications Curriculum Vitae Dankwoord

121 AP 120

122 NEDERLANDSE SAMENVATTING De ontdekking en ontwikkeling van medicijnen is een langdurig en kostbaar proces waarbij vele duizenden kandidaat-medicijnen worden onderzocht om uiteindelijk één nieuw medicijn op de markt te brengen. Eén van de hoofdredenen voor dit hoge uitvalspercentage en de hoge kosten voor de farmaceutische industrie zijn de mogelijke bijwerkingen (ADRs) van de kandidaat-medicijnen. Uit een vergelijking van de oorzaken voor de uitval van kandidaat-medicijnen tussen 1991 en 2000 blijkt dat, naast werkzaamheid en commerciële redenen, toxicologische kwesties in toenemende mate een reden zijn voor het hoge uitvalspercentage. [1] Ook een meer recente studie, uitgevoerd over de periode , laat zien dat de uitval van kandidaat-medicijnen tijdens fase II en fase III klinische studies de veiligheid van medicijnen een belangrijke rol blijft spelen in de ontwikkeling van nieuwe medicijnen en verantwoordelijk is voor 22% van alle uitval van kandidaat-medicijnen. [2] De beoordeling van de metabole stabiliteit van medicijnen en de karakterisering van medicijnmetabolieten is een belangrijk aspect in de productie van veiligere medicijnen. De zogenoemde MIST protocollen (metabolites in safety testing) [3] schrijven de controle en de veiligheidsbeoordeling van (re)actieve metabolieten voor tijdens het hele ontwikkelingsproces. Karakterisering van reactieve metabolieten wordt veelal uitgevoerd door deze te vangen met kleine moleculen, zoals glutathion, om een beter begrip te krijgen van de metabole routes die leiden tot de (bio)activering van medicijnen. [4-5] De farmaceutische industrie wil de kansen op metabolische activering van (kandidaat)medicijnen zo veel mogelijk beperken omdat dit als eerste stap richting toxiciteit wordt gezien. [6] In Hoofdstuk 1 worden de verschillende theorieën over het mechanisme achter de bijwerkingen van medicijnen besproken. Eén van de aannames die ten grondslag liggen aan dit proefschrift is dat de covalente binding van reactieve metabolieten aan proteïnen een belangrijke rol speelt bij het ontstaan van bijwerkingen. Dit staat bekend als de hapteen theorie. [7] In de huidige praktijk worden de covalentgebonden reactieve metabolieten aan leverproteïnen, de zogenoemde medicijn proteïneadducten, alleen globaal bepaald met behulp van radio-gelabelde medicijnen en radioactiviteitsmetingen. Een conservatieve drempelwaarde van 50 pmol medicijn equiv/mg totaal leverproteïne is voorgesteld om te helpen bij de selectie van geschikte kandidaat-medicijnen [6], maar minder strikte eisen worden gesteld aan medicijnen voor ernstige ziekten die moeilijk te behandelen of onbehandelbaar zijn [8]. Studies hebben echter aangetoond dat niet alle medicijn proteïneadducten een immuunreactie geven. Een in vivo studie in muizen heeft bijvoorbeeld aangetoond dat bij vergelijkbare covalente bindingsniveaus van paracetamol (APAP) en zijn niet-toxische regioisomeer 3 -hydroxyacetanilide alleen APAP resulteert in levertoxiciteit. [9] Deze observatie zou kunnen betekenen dat alleen adductvorming aan bepaalde kritieke proteïnen een rol speelt bij orgaantoxiciteit. Echter, in een andere, ex vivo studie uitgevoerd op precies-gesneden leverplakjes van rat en mens is aangetoond dat dezelfde regioisomeer leidt tot vergelijkbare of zelfs meer toxiciteit dan APAP. [10] Een gedetailleerd onderzoek is daarom nodig om vast te stellen welke medicijn proteïneadducten worden gevormd en of deze gerelateerd zijn aan AP 121

123 bijwerkingen. De huidige methoden voor identificatie van leverproteïnen die doelwit zijn van reactieve metabolieten gebruiken radio-gelabelde medicijnen in dierexperimenten gevolgd door de analyse van leverproteïnen met twee-dimensionale gelelektroforese gecombineerd met radioactiviteitsmetingen en massaspectrometrie. [11] De lijst met geïdentificeerde proteïnen blijft groeien, maar de identificatie van het specifieke metaboliet, het doelproteïne en de locatie van de modificatie met methoden zoals deze is zeer uitdagend omdat de peptiden die het adduct dragen meestal niet gedetecteerd worden. [12] Het doel van het onderzoek dat beschreven is in dit proefschrift was de ontwikkeling van geavanceerde analytische methoden, gebaseerd op proteomics -technieken, om medicijn proteïneadducten te bestuderen in in vivo biovloeistoffen en uiteindelijk in weefsels. Het essentiële verschil tussen een globale strategie en onze voorgestelde aanpak is dat wij het onderwerp op het proteïneniveau wilden benaderen om onomstotelijk bewijs van medicijn proteïneadductvorming te verkrijgen. Het doel was om doelgerichte proteomics-technieken te gebruiken om de identiteit van medicijn proteïneadducten vast te stellen op het (enkele) proteïneniveau en deze informatie te gebruiken om de mate van adductvorming (semi-)kwantitatief te bepalen. AP De proteomics-aanpak bestaat uit een aantal stappen: (1) scheiding en/of fractionering van een complex proteïnemonster, (2) denaturatie, reductie en alkylatie van (het mengsel van) geïsoleerde proteïnen, (3) enzymatische digestie van de proteïnen tot peptiden, (4) LC MS/MS-analyse van de peptiden, en (5) verwerking van de data met bioinformaticatechnieken. De omzetting van proteïnen tot peptiden is één van de kritieke stappen in de gehele procedure. Om dit proces adequaat te kunnen volgen en te optimaliseren zijn alle monstervoorbewerkingsstappen geoptimaliseerd. Daarnaast moest ook de LC MS/MS analyse geoptimaliseerd worden, vooral omdat nieuwe rapid-resolution LC-kolommen, gepakt met (in die dagen) innovatieve 1.8 µm ID deeltjes, zouden worden gebruikt. Hierbij is het belangrijk om het beste compromis te vinden tussen de analysesnelheid en de verkregen informatie. De optimalisatie kan gericht worden op een optimale gevoeligheid voor een bepaald medicijn proteïneadduct of op de hoogst mogelijke dekkingsgraad van de proteïnesequentie. Het eerste punt is bijzonder belangrijk voor de doelgerichte analyse van bekende medicijn proteïneadducten, terwijl het tweede punt juist belangrijk is voor de ontdekking van nog onbekende medicijn proteïneadducten. Dezelfde gedachtegang kan toegepast worden op enzymatische proteïnedigestie. In de meeste proteomics-experimenten ligt de nadruk op het behalen van een hoge dekkingsgraad van de proteïnesequentie voor een nauwkeurige proteïneidentificatie, waarbij detectie en identificatie van de modificatie en de locatie ervan, dat wil zeggen een bepaald deel van de proteïnesequentie, essentieel is voor vaststelling van medicijn proteïneadduct vorming. De in vivo concentraties van medicijn proteïneadducten zijn zeer laag omdat meestal minder dan 1% van een proteïne wordt gemodificeerd. Daarom is een goed-ontworpen en geoptimaliseerde analytische strategie nodig om de uitdaging in de detectie en identificatie van lage concentraties van medicijn proteïneadducten aan te gaan. Hoofdstuk 3 beschrijft de optimalisatie van alle fasen van monstervoorbewerking en analyse specifiek voor medicijn proteïneadducten, waarbij de focus voornamelijk ligt op de enzymatische digestie. Het effect van verschillende enzymen en digestiecondities op de identificatie van medicijn albumineadducten wat betreft de sequentiedekkingsgraad en de detectie van het gemodificeerde peptide is geëvalueerd 122

124 met de innovatieve toepassing van een Design of Experiments. Deze aanpak geeft een meer efficiënte en nauwkeuriger bepaling van de optimale digestiecondities door middel van gelijktijdige optimalisatie van meerdere variabelen en uitleesparameters. Het vermijdt de nadelen van de meer traditionele één-variabeleper-keer aanpak. De digestiecondities zijn geoptimaliseerd met een model-adduct van monochloorbimaan aan humaan serumalbumine (HSA), en de geoptimaliseerde condities zijn vervolgens toegepast op het adduct van N-acetyl-p-benzoquinoneimine (NAPQI), het reactieve metaboliet van APAP aan HSA (NAPQI HSA). In deze studie is aangetoond dat optimalisatie van de digestiecondities voor een specifiek doeleinde zeer zinvol is omdat het tot een verhoogde efficiëntie en gevoeligheid voor de identificatie van NAPQI HSA adducten leidt. Vooral voor het minder specifieke en minder goed gedocumenteerde enzym thermolysine was de sequentiedekkingsgraad en de gevoeligheid voor de locatie van de adductmodificatie sterk verbeterd. De hierboven beschreven aanpak wordt bottom-up proteomics genoemd omdat de proteïnen eerst worden gedigesteerd tot peptiden en vervolgens worden deze peptiden geïdentificeerd om de identiteit van het oorspronkelijke proteïne vast te stellen, dus van onder naar boven. Ondanks de toenemende populariteit van andere proteomics-methoden, heeft de bottom-up aanpak nog steeds de voorkeur voor de analyse van proteïnen (en hun modificaties). Proteïnedigestie is de meest cruciale stap in zulke methoden en kan op verschillende manieren gedaan worden met behulp van proteolytische enzymen of met behulp van chemicaliën voor non-enzymatische digestie. Hiervoor zijn veel verschillende enzymen en chemicaliën beschikbaar, maar klassieke digestieprotocollen maken gebruik van trypsine, de gouden standaard in enzymatische digestie, en worden uitgevoerd tijdens een overnacht digestie. Proteïnedigestie is vaak het knelpunt omdat het veel tijd kost. Recentelijk is veel aandacht besteed aan de versnelling van dit proces. Een verbetering van de monsterdoorvoer kan ook worden behaald door de digestie uit te voeren in een online systeem. Dat kan dan worden geautomatiseerd waardoor het aantal handmatige monstervoorbewerkingsstappen afneemt. Deze state-of-the-art digestietechnieken worden behandeld in Hoofdstuk 2 van dit proefschrift en kunnen de digestieduur reduceren van uren naar enkele seconden wat de duur van de monstervoorbewerking sterk verkort. Omdat in vivo analyse van medicijn proteïneadducten het doel is, is de extractie van het doelproteïne uit een complexe biologische matrix een belangrijk onderdeel van de monstervoorbewerking. Omdat minder dan 1% van een proteïne gemodificeerd wordt door reactieve metabolieten, is een meer dan honderdvoudige overmaat van een sterk overeenkomend, niet-gemodificeerd proteïne aanwezig. Selectieve opzuiveringsmethoden voor de extractie van alleen het proteïneadduct zijn nog niet beschikbaar. Dus wordt, na opzuivering van het doelproteïne, een mengsel verkregen van het medicijn proteïneadduct (<1%) en het niet-gemodificeerd proteïne (>99%). Daarom moet voldoende doelproteïne opgezuiverd worden, anders zal de aanwezigheid van het medicijn proteïneadduct worden gemaskeerd door de overmaat van het niet-gemodificeerde proteïne. Hoofdstuk 5 beschrijft de ontwikkeling van een protocol voor specifieke monstervoorbewerking, digestie en analyse ten behoeve van de in vivo identificatie en kwantificering van medicijn albumineadducten in serum. Dit protocol bestaat uit een aantal stappen: (1) affiniteitschromatografie voor de extractie van albumine uit serum, (2) bufferuitwisseling van het opgezuiverde albuminemonster naar denaturatiebuffer met een gelfiltratiekolom, (3) disulfide-reductie and cysteine-alkylering voor complete ontvouwing van het AP 123

125 opgezuiverde albumine, (4) bufferuitwisseling naar water met een gelfiltratiekolom, (5) vriesdrogen om het monster te concentreren en het NAPQI albumineadduct binnen de detectielimiet te brengen, (6) enzymatische digestie met behulp van de geoptimaliseerde condities van Hoofdstuk 3, (7) LC MS/MS analyse van de verkregen peptiden, en (8) data-analyse, zowel handmatig als softwarematig, voor de identificatie van medicijn albumineadducten. Voor de uiteindelijke toepassing van het ontwikkelde protocol op echte biologische monsters is een samenwerking gezocht met één van onze partners uit de projectgroep. Zij verrichten dierexperimenten voor het bestuderen van (onder andere) immunoactiviteit van medicijn proteïneadducten in muismodellen. Voor ons project hebben we serum monsters uit een medicijnblootstellingsonderzoek ontvangen waarin de muizen een hoge dosis APAP toegediend hebben gekregen op één of op zeven opeenvolgende dagen. Aangezien de muis een klein dier is, waren alleen kleine hoeveelheden van de serummonsters beschikbaar wat een extra uitdaging met zich meebrengt. De muizenstudie was ontworpen om de kinetiek van ADRs te bestuderen via de verschillende behandelgroepen. Dus was het ook noodzakelijk om een kwantificeringsmethode van de NAPQI albumine adducten te ontwikkelen. Verscheidene opties, zoals het gebruik van albuminen van andere diersoorten en synthetische NAPQI albumineadducten, zijn geëvalueerd, maar de meest betrouwbare aanpak bleek een relatieve kwantificering te zijn waarbij een vergelijking wordt gemaakt van de piekoppervlakken van specifieke peptiden voor het albumineadduct en voor het niet-gemodificeerd albumine. Een andere aandachtsfactor was de aanwezigheid van twee vrije cysteines in muizen serumalbumine (MSA), terwijl humaan albumine maar één vrije cysteine heeft. Uiteindelijk bleek de ontwikkelde methode succesvol en is voor het eerst het NAPQI MSA-adduct geïdentificeerd in vivo in muizenserum. De NAPQI modificatie werd alleen gedetecteerd aan één van de vrije cysteines, wat in zou kunnen houden dat er in vivo een voorkeur is voor modificatie van deze locatie. Ondanks een hoge toegediende dosis van 300 mg APAP/kg werden alleen extreem lage concentraties van NAPQI albuminebinding gedetecteerd van 0.2%. Dit zou kunnen betekenen dat de concentratie van het NAPQI albuminemodificatie op de andere locatie misschien onder de detectielimiet zat. Verder zijn vergelijkbare adductniveaus gedetecteerd in beide behandelgroepen, wat kan suggereren dat een enkele hoge dosis APAP al voldoende is om de leverfunctie sterk te beïnvloeden en dus ook de productie van zowel reactief metaboliet als serumalbumineadduct. AP De monstervoorbewerkingsprotocollen voor bottom-up proteomics kunnen heel uitgebreid en langdurig zijn, wat vooral nadelig is wanneer grote aantallen monsters behandeld moeten worden. Wanneer een grote groep monsters geanalyseerd moet worden voor vergelijkingsstudies of kwantificeringsdoeleinden, moet de monstervoorbewerking en analyse idealiter in één keer uitgevoerd worden. Hoofdstuk 4 beschrijft de ontwikkeling van een hoge-doorvoer monstervoorbewerkingsmethode gebaseerd op 96-wel-filterplaten waarmee 96 monsters gelijktijdig opgewerkt kunnen worden. Het gebruik van een membraan met een bepaalde poriegrootte heeft meerdere voordelen, zoals het terugbrengen van het aantal benodigde monsteroverbrengstappen, waarbij mogelijk monsterverlies wordt beperkt, en de mogelijkheid tot gelijktijdige concentrering van het monster. Daarmee kan ook een vriesdroogstap worden vermeden, wat de totale gevoeligheid verhoogt. Het protocol dat beschreven wordt in dit hoofdstuk is ontwikkeld voor globale celproteomics en presteerde beter dan het gelfiltratieprotocol wat betreft de aantallen geïdentificeerde peptiden en proteïnen. Het verschil in prestatie was al zichtbaar bij analyse van de 124

126 monsters met het rapid-resolution LC QTOF systeem, maar was nog duidelijker wanneer de filterplaatmethode gecombineerd werd met een meer-geavanceerd nanolc Orbitrap MS systeem. Het uiteindelijke protocol voor de filterplaatopwerking en LC MS analyse resulteerde in de identificatie van meer dan 400 cellulaire proteïnen. Dit aantal kan nog vergroot worden door de toepassing van tweedimensionale LC voor een verbeterde scheiding en detectie van peptiden. De ontwikkelde methodologie kan ook gebruikt worden voor doelgerichte proteomics en kan toegepast worden op andere typen proteïnemonsters. REFERENTIES [1] Kola, I., Landis, J., Can the pharmaceutical industry reduce attrition rates? Nat. Rev. Drug Discov. 2004, 3, (8), [2] Khanna, I., Drug discovery in pharmaceutical industry: productivity challenges and trends. Drug Discov. Today. 2012, 17, (19-20), [3] FDA, Guidance for Industry. Safety Testing of Drug Metabolites. 2008, GuidanceComplianceRegulatoryInformation/Guidances/ucm pdf [4] Dragovic, S., Boerma, J.S., van Bergen, L., Vermeulen, N.P.E., Commandeur, J.N.M., Role of Human Glutathione S-Transferases in the Inactivation of Reactive Metabolites of Clozapine. Chem. Res. Toxicol. 2010, 23, (9), [5] Dragovic, S., Gunness, P., Ingelman-Sundberg, M., Vermeulen, N.P.E., Commandeur, J.N.M., Characterization of Human Cytochrome P450s Involved in the Bioactivation of Clozapine. Drug Metab. Dispos. 2013, 41, (3), [6] Evans, D.C., Watt, A.P., Nicoll-Griffith, D.A., Baillie, T.A., Drug protein adducts: an industry perspective on minimizing the potential for drug bioactivation in drug discovery and development. Chem. Res. Toxicol. 2004, 17, (1), [7] Kevin Park, B., Coleman, J.W., Kitteringham, N.R., Drug disposition and drug hypersensitivity. Biochem. Pharmacol. 1987, 36, (5), [8] Baillie, T.A., Cayen, M.N., Fouda, H., Gerson, R.J., Green, J.D., et al., Drug metabolites in safety testing. Toxicol. Appl. Pharmacol. 2002, 182, (3), [9] Tirmenstein, M.A., Nelson, S.D., Subcellular binding and effects on calcium homeostasis produced by acetaminophen and a nonhepatotoxic regioisomer, 3 -hydroxyacetanilide, in mouse liver. J. Biol. Chem. 1989, 264, (17), [10] Hadi, M., Dragovic, S., Swelm, R., Herpers, B., Water, B., et al., AMAP, the alleged non-toxic isomer of acetaminophen, is toxic in rat and human liver. Arch.Toxicol. 2013, 87, (1), [11] Ikehata, K., Duzhak, T.G., Galeva, N.A., Ji, T., Koen, Y.M., et al., Protein targets of reactive metabolites of thiobenzamide in rat liver in vivo. Chem. Res. Toxicol. 2008, 21, (7), [12] Koen, Y.M., Yue, W., Galeva, N.A., Williams, T.D., Hanzlik, R.P., Site-specific arylation of rat glutathione s-transferase A1 and A2 by bromobenzene metabolites in vivo. Chem. Res. Toxicol. 2006, 19, (11), AP 125

127 AP 126

128 LIST OF ABBREVIATIONS ABC Ammonium bicarbonate ACN Acetonitrile ADR Adverse drug reaction ALAT Alanine aminotransferase ANOVA Analysis of variance APAP Acetaminophen ASAT Aspartate aminotransferase BCA Bicinchoninic acid BSA Bovine serum albumin CAD Collisional-activated dissociation CCD Central composite design CE Capillary electrophoresis CNBr Cyanogen bromide Cys34 Cysteine on position 34 Cys579 Cysteine on position 579 Da Dalton DOE Design of experiments DTT Dithiothreitol ECD Electron-capture dissociation EIC Extracted ion chromatogram ESI Electrospray ionization ETD Electron transfer dissociation FA Formic acid FASP Filter-assisted sample preparation FWHM Full width at half maximum FTICR Fourier transform ion cyclotron resonance GHCl Guanidine HCl GSH Glutathione HAc Acetic acid HCl Hydrochloric acid HDX Hydrogen/deuterium exchange HEK293T Human embryonic kidney cell line HPLC High performance liquid chromatography HSA Human serum albumin IAA Iodoacetic acid IdeS Immunoglobulin-degrading enzyme of Streptococcus pyogenes IDR Idiosyncratic drug reaction IgG Immunoglobulin G AP 127

129 IMER IMMS LC MS MALDI MCB MS MSA MS/MS MWCO m/z NAPQI NTCB OmpT OVAT PBS Pro4 PTM QTOF RRLC %RSD RSM SEC SD SISCAPA SPE SpeB TOF UHPLC Immobilized-enzyme reactor Ion mobility mass spectrometry Liquid chromatography mass spectrometry Matrix-assisted laser desorption ionization Monochlorobimane Mass spectrometry Mouse serum albumin Tandem mass spectrometry Molecular weight cut-off Mass-to-charge ratio N-acetyl-p-benzoquinone imine 2-Nitro-5-thiocyanobenzoate Outer membrane protease T One-variable-at-a-time Phosphate buffered saline Synthetic peptide H-Pro-Pro-Pro-Pro-OH Post-translational modification Quadrupole-time of flight Rapid resolution liquid chromatography Relative standard deviation Response surface methodology Size-exclusion chromatography Standard deviation Stable Isotope Standards and Capture by Anti-Peptide Antibodies Solid-phase extraction Streptococcal cysteine proteinase streptococcal exotoxin B Time-of-flight Ultra high perfomance liquid chromatography AP 128

130 LIST OF PUBLICATIONS Linda Switzar, Martin Giera, Henk Lingeman, Hubertus Irth and Wilfried M.A. Niessen, Protein digestion optimization for characterization of drug protein adducts using response surface modeling, Journal of Chromatography A, 2011, 1218, (13), Linda Switzar, Lydia M. Kwast, Henk Lingeman, Martin Giera, Raymond H. Pieters and Wilfried M.A. Niessen, Identification and quantification of drug albumin adducts in serum samples from a drug exposure study in mice, Journal of Chromatography B, 2013, , Linda Switzar, Martin Giera and Wilfried M.A. Niessen, Protein digestion: An overview of the available techniques and recent developments, Journal of Proteome Research, 2013, 12, (3), Linda Switzar*, Jordy A. van Angeren*, Martijn Pinkse, Jeroen Kool and Wilfried M.A. Niessen, A high-throughput sample preparation method for cellular proteomics using 96-well filter plates, accepted for publication in Proteomics. *These authors contributed equally AP 129

131 AP 130

132 CURRICULUM VITAE Linda Switzar was born on the 13th of May 1979 in Leidschendam, the Netherlands. In 1999, she finished the intermediate vocational education for interior design and worked in that field for 3 years as an all-round furniture upholsterer. After a change of heart, Linda enrolled at the Hogeschool Rotterdam in 2002 to study chemistry with a specialization in analytical-physical chemistry. She performed two internships, at Unilever R&D in Vlaardingen and at the Fire accelerants expertise area of the Dutch Forensics Institute in The Hague, both focused on the development of GC MS methods. After obtaining her Bachelor of Applied Science degree in 2006, Linda started with the chemistry master s programme at the VU University in Amsterdam, specializing in analytical sciences. During her major internship at the Analytical Chemistry group, Linda worked on the development of an analytical strategy for the identification of drug protein adducts in complex biological samples using LC MS and chemometrics. Realizing that she wanted to continue in this field of research, Linda seized the opportunity to start with a Ph.D. project on this topic in the same group, then called the BioMolecular Analysis group, after obtaining her Master of Science degree in In her Ph.D. research, she focused on the development and optimization of sample preparation techniques and analytical methodologies for the analysis of drug protein adducts. The results of this research are described in this thesis. Linda is currently working as a PostDoc at the Center of proteomics and metabolomics at the Leiden University Medical Center where she is developing mass spectrometry-based methods for the discovery of therapeutic serum biomarkers for exon-skipping genetic diseases. AP 131

133 AP 132

134 DANKWOORD Het lag niet geheel in de lijn der verwachting dat ik na mijn opleiding MBO interieur ontwerp ooit zou beginnen met een studie scheikunde en uiteindelijk zelfs zou gaan promoveren, maar zo is het verrasend genoeg toch gelopen. En wat is de tijd voorbij gevlogen! Om tot deze dag te komen heb ik al met al een lange weg afgelegd, die heel leerzaam en vooral ook erg leuk was en daar ben ik heel dankbaar voor! Ten eerste zou ik graag mijn promotoren en copromotor bedanken. Tijdens mijn master stage bij ACAS, later BMA, is mijn enthousiasme voor onderzoek verder gegroeid en werd mij duidelijk dat ik wilde gaan promoveren. Hubertus, ik heb je helaas niet lang meegemaakt in de groep omdat je je nu in de bredere zin inzet voor de analytische scheikunde en de exacte wetenschappen in het algemeen, maar veel dank voor het vertrouwen en de kans die je mij gegeven hebt. Wilfried, als Interim Head van de groep heb jij de laatste paar jaar een grote postieve invloed gehad op mij en mijn onderzoek. Helaas hebben we meer dan eens moeten constateren dat meneer Murphy hier misschien ook werkt, maar uiteindelijk is het goed gekomen. Ontzettend bedankt voor jouw enthousiasme, energie, inzet, organisatorisch vermogen en snelheid van manuscripten lezen en s beantwoorden! Henk, mijn eerste kennismaking met de VU was met jou voor inschrijving voor de master en sindsdien loop jij als een rode draad door mijn tijd op de VU. Bedankt dat je altijd zo n ongedwongen en open sfeer in de groep hebt gecreëerd. En ik vond het ontzettend fijn dat jouw deur altijd open stond voor advies en een goed gesprek of gewoon een beetje gezelligheid aan het eind van de dag! Martin, enorm bedankt voor de goede samenwerking en jouw input in mijn onderzoek. Dit heeft geleid tot mooie onderzoeksartikelen en een heel mooi review! Jeroen, een gezamenlijk project tijdens mijn laatste jaar heeft geleid tot een publicatie, bedankt daarvoor en voor de samenwerking van de afgelopen jaren! Maarten, bedankt voor jouw hulp tijdens mijn project en voor het delen van jouw ervaring en kennis! Mijn promotie tijd zou nooit zo leuk zijn geweest zonder mijn collega-promovendi en Postdocs! Ik heb de dynamiek van de academische wereld altijd heel erg leuk gevonden, waardoor je de kans hebt veel mensen te ontmoeten, al was het natuurlijk wel altijd jammer wanneer er iemand weg ging. Door de jaren heen heb ik veel kantoorgenoten gehad. Eerst deelde ik kamer O3.57 met Kim, Michele, en Dik. Bedankt voor de gezellige sfeer en dat jullie mij direct thuis hebben laten voelen! Daarna zat ik op de AiO kamer met Jon, Ansgar, Niels J., Filipe, Mark, en later ook met David en Dick-Paul. Bedankt voor alle discussies, adviezen en afleiding in de vorm van PhD-comics, darts en vele office-pranks! Daarnaast was ik ook regelmatig te vinden in het andere AiO kantoor(tje). Lygia, thanks for all the fun times at the VU and outside! I hope I will someday get the opportunity to visit you in Brazil. Niels M., bedankt voor alle bijzonder goede gesprekken die we gevoerd hebben! Ferry, jij hebt altijd wel een interessant verhaal te vertellen (elke keer weer ongelooflijk om te horen wat jou nu weer gevonden had op het internet!), wat ontzettend geholpen heeft om de dagelijkse sleur te doorbreken. Bedankt daarvoor! AP 133

135 Nadat de oude garde aan hun volgende uitdaging begonnen was, was de groep tijdelijk wat kleiner, tot er plotseling een vrouwelijke invasie plaatsvond! Marija, Reka and Dina, it was a pleasure to share an office with you and to finally have the oppertunity again for some girl-talk! Na wat reorganisatie in de M- en O-gangen, is de AiO kamer verhuisd naar N3.43 ( de balzaal ) waar ook Ingeborg en Jan-Hein hun nieuwe plek kregen. Ik vond het erg leuk om op de valreep ook nog met jullie een kantoor te delen! Natuurlijk kan een promovendus in deze groep niet zonder de massaspectrometrische kennis van Ben en in mijn begintijd heb ik ook veel samengewerkt met Marek, bedankt voor jullie hulp! De kerstdiners en groepsuitjes organiseerden we altijd samen met de spectroscopie groep en ook daar heb ik veel goede herinneringen aan, zowel aan de voorbereiding als aan de uitvoering. Bedankt Lineke, Silvia, Ivonne, Cecilia, Joost, Gert, Freek, Nel en Cees. Verder zijn er nog veel meer (tijdelijke) collega s geweest met wie ik ook met veel plezier heb samengewerkt. Gracias Leonor (and Carlos), you were in Amsterdam for only a short year, but it was an unforgettable experience! I hope we meet again in Madrid/Valencia/Amsterdam or elsewhere! Natuurlijk ook Gisèle, Bart, Willem, Rob, Erika, Petra, Zach, Anton, Jan Carel, Frank en alle anderen bedankt! Ook heb ik veel mensen ontmoet van de andere groepen op de VU, op de werkvloer en tijdens de vaak hilarische en legendarische borrels. Dank voor jullie gezelligheid Oscar, Mark, Dana, Maikel, Chimed, Chris, Stephanie, Sanja, Jan Simon, Jelle, Vanina, Jeroen en alle anderen! Het begeleiden van practica en studenten is ook een goede en leerzame ervaring geweest. Supawat, Vincent, Laure, Erinn, Salam, Daniel, Jordy, Eman, Rima en Kim, bedankt voor jullie inzet en bijdrage aan mijn onderzoek en proefschrift! Een analytisch chemicus kan niet zonder de kennis, ervaring en vindingrijkheid van de fijnmechanische werkplaats. De wanhoop was soms nabij, maar Dick, Klaas en Roald hadden altijd wel even tijd. Bedankt voor al jullie hulp! Mijn promotie project was onderdeel van een bijzonder groot TI Pharma consortium. Dank aan alle projectleden! Ik heb vooral veel geleerd van de samenwerking met wetenschappers uit verschillende expertisegebieden. In het bijzonder wil ik graag Lydia en Raymond bedanken voor de fijne samenwerking. We hadden er graag nog meer uitgehaald, maar toch heeft het geresulteerd in een mooie publicatie. AP Hier wil ik graag nogmaals mijn paranimfen, David en Dick-Paul, bedanken voor de gezellige tijd en hun hulp bij de laatste loodjes van mijn promotie! David (en Lisa), ook bedankt voor de gezellige boardgaming avondjes en ik hoop dat er nog veel zullen volgen! Dick-Paul, bedankt voor het mij op de hoogte houden van de laatste technische snufjes en het in ere houden van office prank-traditie! Toevalligerwijs zijn we nu alledrie ook weer collega s wat ik ontzettend leuk vind! 134

136 Verder wil ik nog graag alle familie en vrienden, die ik met name de laatste tijd veel te weinig aandacht heb kunnen geven, bedanken voor hun steun! Ik zou graag een paar mensen in het bijzonder willen noemen. HoSze, sinds het eerste jaar van de Hogeschool zijn we al vrienden en het is leuk dat we allebei ongeveer hetzelfde pad hebben gekozen, een master en daarna promoveren. Al is het wel in een iets andere richting, het is altijd fijn wanneer je de dagelijkse beslommeringen kunt bespreken met iemand die zich in dezelfde situatie bevindt. Bedankt voor je vriendschap! Simone en Erwin, onze vriendschap is al weer wat jaren eerder ontstaan, tijdens de opleiding MBO interieur ontwerp. Simone, het is grappig om te zien dat we hetzelfde startpunt hadden, maar allebei een rigoreuze carrièreswitch gemaakt hebben en nu allebei totaal iets anders doen! Wanneer ik jullie zie, is altijd weer gezellig en ik hoop dat er nog veel meer heerlijke etentjes en gezellige avonden zullen volgen! Bedankt voor jullie vriendschap! Fred, Clemence en Kimberley, mijn favoriete oom, tante en nichtje! Helaas is het er de laatste tijd niet veel van gekomen, maar het is altijd geweldig om bij jullie op bezoek te komen (en het dan eens een keer helemaal niet over wetenschap te hebben)! Bedankt voor jullie support en gezelligheid! Niet in de laatste plaats wil ik mijn moeder bedanken voor haar zorg en ondersteuning waardoor ik de kans had om de carrièreswitch naar de scheikunde te maken en verder te studeren. Soms maakte je wel eens zorgen of het niet een beetje teveel was, maar je was altijd geïnteresseerd en betrokken. Zonder jouw steun had dit proefschrift er misschien nooit geweest en daarom wil ik het graag opdragen aan jou. Linda AP 135

137