Identification of Adalimumab Epitopes recognized by Anti-Drug Antibodies (ADA) [306979]

[anonimizat] (ADA)

Master Thesis

vorgelegt von: [anonimizat]: Prof. Dr. Marcel Leist

Gutachter: Prof. Dr. Dr. Michael Przybylski

Konstanz, 2017

[anonimizat] – is affected by chemistry. – Linus Pauling, 1901-1994

i

Contents

Introduction 1

Immunogenicity of TNF-α inhibitors and clinical implications . . . . 1

Immunogenicity of TNF-α inhibitors and clinical implications . . . . 2

[anonimizat] . . . . . . . . . . . . . 3

Bioaffinity mass spectrometry for epitope elucidation and development

in protein and peptide structure determination . . . . . . . . . . . . . 4

1.4.1 Bioaffinity-mass spectromtry for epitope identification . . . . . 5

1.5 Aims of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

trometry 32

Matrix assisted laser desorption ionization mass spectrometry 48

Electrospray ionization mass spectrometry 49

Affinity based methods 50

Determination of affinity constants by SPR biosensor 50

[anonimizat] 53

[anonimizat] 56

Preparation of antibody affinity columns by the crosslinking approach 57

Epitope extraction of TNF-α 58

Epitope excision of TNF-α 58

Epitope extraction of Adalimumab 58

Software, structural modelling and amino acid sequences 59

Conclusions and outlook 60

References 61

iii

Thanks go to Prof. [anonimizat].

Thanks go to Prof. [anonimizat]-[anonimizat].

Thanks go to Prof. [anonimizat], Physiology, [anonimizat]. [anonimizat]-Pontoise, [anonimizat] , help and discussions during the thesis and providing Adalimumab drug and patients blood samples.

[anonimizat], Thermo-[anonimizat], for instrumental support and help with protein sequencing as well as providing chromatographic materials.

[anonimizat], [anonimizat].

iv

Adalimumab (ADA) is a fully humanized TNF-α blocking drug IgG antibody used in the treatment of juvenile idiopathic arthritis (JIA). [anonimizat]-[anonimizat]. [anonimizat] (SPR) [anonimizat]ry high affinities of (KD 10-9 M). In this work, the identification of an epitope of patient sera against Adali- mumab was performed by affinity mass spectrometry. The epitope was identified by proteolytic epitope extraction of digested Adalimumab. The epitope fractions were analysed by matrix assisted laser desorption ionization mass spectrometry (MALDI- MS) without further purification, and a 17 amino acid long epitope sequence was identified: 12-VQPGRSLRLSCAASGFTF-29.

Further, the epitope of Adalimumab located on the target protein TNF-α was for the first time studied under in vitro conditions, and identified by mass spectrometric epi- tope extraction and epitope excission. Two overlapping peptides were identified: 33- ANALLANGVELR-44 and 37-LANGVELRDQLVVPSEGLYLIY-59 by epitope ex-

traction of tryptic and chymotryptic peptide mixtures respectively. In addition two epitope peptides 7-TPSDKPVAHVVANPQAEGQLQWLNR-31 and the overlapping 7-TPSDKPVAHVVANPQAEGQLQWLNRR-32 were identified by epitope excision mass spectrometry in a conserved region of TNF-α.

v

Adalimumab (ADA) ist ein vollst¨andig humanisierter TNF-α blockierender IgG An- tik¨orper der in der Therapie von juveniler idiopathischer Arthritis verwendet wird. Immunogenit¨at gegen den Medikamenten-Antik¨orper wird h¨aufig in behandelten Pa- tienten festgestellt. Das Auftreten der anti-ADA Antik¨orper korreliert direkt mit Versagen der Therapie und vermindeterm Ansprechen der Patienten auf die Behand- lung aufgrund der schnellen Deaktivierung von Adalimumab. Die Bindung von anti- ADA-Antik¨orpern an Adalimumab wurde bereits durch Oberfl¨achenplasmonenreso- nanz Biosensor Messungen ermittelt: Die Interaktion von Adalimumab und anti- Adalimumab Antik¨orper ergab einen Dissoziationskonstanze KD von 10-9 M. Diese Ar- beit beschreibt eine erstmalige Epitop Identifizierung von Patientenserum gegen Adal- imumab mittels Affinit¨atsmassenspektrometrie. Das Epitop wurde durch proteolytis- che Epitop-Extraktion mit Adalimumab bestimmt. Epitop Fraktionen wurde mittels Matrix-unterstu¨tzte Laser-Desorption Ionisation Massenspektrometrie (MALDI-MS) analysiert und ein einziges 17 Aminosa¨uren langes Peptid der schweren Antik¨orper- kette identifiziert: 12-VQPGRSLRLSCAASGFTF-29.

Erstmals wurde das Epitop von Adalimumab auf dem Zielprotein, TNF-α, unter in vitro Bedingungen studiert und identifiziert mittels massenspektrometrischer Epitop- Extraktion und Epitop Excision. Zwei u¨berlappende Peptide wurden bestimmt durch Extraktion von tryptischen und chymotryptischen Peptid-Mischungen: 33-ANALLAN GVELR-44 und 37-LANGVELRDQLVVPSEGLY

LIY-59. Zwei weitere Peptide wurden im N-terminalen Bereich von TNF-α mittels Excision bestimmt: 7-TPSDKPVAHVVANPQAEGQLQWLNR-31 und 7-TPSDKPV AHVVANPQAEGQLQWLNRR-32. Die Epitope konnten einem konservierten Bere- ich auf TNF-α zugeordnet werden.

vi

ACN Acetontrile

ADA Adalimumab

ATZ amino acid anilinothialinone amino acid CDR Complementary determining region CRP C-reactive protein

DMP dimethyl pimelimidate

DTT Dithiothreitol

EDC N-(3-dimethylaminopropyl)-N-ethylcarbodiimid ELISA Enzyme linked immunosorbent assay

ESI Electrospray ionization

FC Constant region of antibodies FV Variable region of antibodies JIA Juvenile idiopathic arthritis IgG γ-Immunoglobulin

MALDI Matrix-assisted laser desorption-/ionization MS Mass spectrometry

NHS N-hydroxysuccinimide

RA Rheumatoid arthritis

SAW Surface acoustic wave

vii

SDS-PAGE Sodiumdodecylsulfat-polyacrylamid-gelelectrophoresis SPR Surface plasmon resonance

PBS Phosphate buffered saline

pH Negative logarithm of H33O+-iones concentration PTH amino acid phenylthiohydantoin-amino acid TEMED Tetramethylethylenediamine

TCEP Tris(2-carboxyethyl)phosphine TNF Tumor necrosis factor

ToF Time of flight

UV Ultraviolet

viii

Introduction

Immunogenicity of TNF-α inhibitors and clinical impli- cations

The pro-inflammatory cytokine Tumor necrosis factor (TNF)-α is a ligand promoting inflammatory response both at sites of injury and in the central nervous system (CNS) [1]. It is synthesized as a type-2 transmembrane protein, which is then cleaved to the soluble form. Both the transmembrane and soluble protein exhibit and mediate inflammatory effects. Figure 1 gives an overview of TNFα mediated proinflamma- tory pathways. The direct cellular effects of TNF-α include expression of cytokines, proliferation of cells and inhibition of apoptosis. This leads to fever and inflamma- tory reactions, production of acute phase proteins (CRP) and positive cell taxis on macrophages and neutrophiles [2].

Figure 1: Cellular response to activation of TNF-R1 and TNF-R2 through TNF-α. Both responses lead to activation of cellular signal cascades including (i) caspase mediated pathways leading to activation of Caspace 3 and following apoptosis and (ii) activation of the MEKK pathway which up-or down-regulates expression of genes linked to inflammatory respons es. This either induces expression of further pro-inflammatory cytokines and or inhibits or direct cell death (apoptosis).

Rheumatoid arthritis (RA) is an autoimmune disease causing inflammation of joints, leading to functional loss and eventually destruction of joints. Development of TNF-α blocking protein therapeutics allow better control and decreased rate of joint damage accompanied by improved physical conditions [3, 4]. Three protein biosimilars, the TNF-α receptor antagonist and two monoclonal antibodies Infliximab and Adalimumab have succeeded in the treatment of RA patients [5]. Alternative treatments such as with methotrexate have been replaced with these drug antibodies due to side effects of general immune supression after methotrexate treatment [6, 7].

Immunogenicity of TNF-α inhibitors and clinical impli- cations

All three drug-biosimilars used for the treatment of Rheumatoid arthritis in patients are considerable immunogenic [8, 9]. However, only Infliximab and Adalimumab show success in treatment of RA as well as also Crohns disease. Infliximab is a hu- man/mouse chimera antibody, with the variable region composed of anti-TNF-α CDR sequences. Such foreign, non-human sequences are readily recognized by the immune system with strong T-cell response and following high antibody production levels. As such, the foreign amino acid sequence of Infliximab is detected and presented to T-cells by dendritic cells which in turn stimulate B-cell differentiation and anti- drug antibody production. Moreover, even the fully humanized Adalimumab and Etanercept, which is a dimeric fusion protein of Human-TNFR2 and IgG1, stimulate anti-drug antibody formation.

Levels of anti-Infliximab antibodies went as high as 60% in Crohns diseased patients while antibodies were detectable in 10% of patients with RA. Adalimumab appears to be less immunogenic. In 3 different clinical trails, 6-12% of treated patients developed detectable anti-Adalimumab antibodies, while less than 1% of patients treated with both the immunospressive methotrexate and adalimumab together formed antibod- ies. Methotrexate is a powerful immune system supressant by reducing T-cell response and therefore reducing the patients immune response during treatment. Etanercept on the other hand seems to be the least immunogenic treatment option with only 2% of patients being anti-Etanercept positive [8, 9].

The resulting clinical implications for patients with a positive immune response to the treatments are severe: Possible results include: increased reactions to the infusion, shorter duration of pharmaceutical response and increased dosing frequency together with application of immunosupressive agents including methotrexate.

Studies on the pharmacokinetics of Adalimumab showed that the maximum concen- tration of Adalimumab was reached after 9.1 days, assuming a first order absorption rate of 0.28 day-1 calculated using the pharmacokinetic/pharmacodynamic (PK-PD) model [10]. In this study, three patients suspected to have anti-drug antibody forma- tion showed much higher clearance rates with a 50% increased clearance rate compared to normal patients. This resulted in lowered Adalimumab concentrations in blood. These results demonstrate the severity of the anti-drug antibody formation, making treatment of RA 50% less effective. Further, detection of anti-Adalimumab antibod- ies has proven difficult in patients, as normal ELISA assays are not applicable [11]. Therefore, it can be assumed that often antibody formation is underestimated. The

first successful and reliable test for anti-Adalimumab antibodies was established by Real-Fernandez et al. by an assay using a surface plasmon resonance biosensor [12]. However, the instrumentation is very expensive and trained personal is required for operation. Therefore broad screening for diagnostics was not performed.

To date there has been no approach published for treatment of antibody formation except switching medication or application of immunosupressiva. Identifying an epi- tope peptide could help to develop both treatment possibilities as well as easier and faster anti-Adalimumab diagnostic assays.

Biochemistry of antigen-antibody interactions

The human immune system consists of a network of cells, tissues and organs through- out the human body, defending the body against pathogens such as bacteria, fungi or viruses, as well as harmful molecules. The immune system can be differentiated into two major sub-classes: (1) The innate immune system which is responsible for unspecific and generic defence mechanisms against pathogens and the (2) adaptive immunity, during which specific pathogens are recognized resulting in an antigen specific response. This pathway creates the antigen specific immunological memory which is stored in certain cell types allowing faster elimination of pathogens or threats to the host in case of other encounters.The adaptive immune system is triggered if a threshold of ”non-self”-antigens in the body is reached leading to the B-cell humoral response and T-cell mediated immune response (fig. 2) [13].

B-memory cell

B-cell

Stimulation of B cell

Leading to B cell proliferation and specialization

Antigen presentation as MH- Complexes on cell surface

Recognition by T-cell

Production and secretion of antibodies

Figure 2: Adaptive immune response leading to the formation of antibodies via specialized B-cells and the role of T-cells and macrophages during the process. Upon stimulation of macrophages by phagocytosis of an antigen, the cells presented the antigens on their surface in complex with MHC receptors. This in turn leads to stimulation of both T-cells and B-cells, the later of which differentiate into B-memory and B-plasma cells if additional stimulation through T-cell secreted cytokines is sufficient. Diagram adapted from Nossal, 2003 [14].

Dendritic cells register infections by recognizing antigens with pattern recognition receptors (PRR) and express them as major histocompatibility complex II (MHC II) and peptide complexes on their cell surface (antigen representation). This step is cru- cial as here antigens (epitopes) are generated by digestion and lysis of foreign proteins or other biomolecules generating the structures which later antibodies will recognize

– their epitopes. B- and T-cells with antigen-complementary receptors recognize their specific antigen resulting in the specialization of each cell line. T-cells differentiate into various specialized cell lines such as T-helper cells and memory cells. B-cells specialize to plasma cells secreting soluble antibodies which recognize the antigens and start immunological chain reactions [15]. Later responses of the immune system rely on the capacity of the antibodies to react and bind to their corresponding epi- tope with segments of their complementary determining region (CDR) – known as the antibody paratope. In depth knowledge of epitopes is crucial for developement of epitope vaccines, characterization of cross-reactivities of antibodies and active sites on proteins such as cytochrome C and study of protein topography and conformation. There are a variety of methods and approaches for epitope identification as discussed in the next chapter.

Bioaffinity mass spectrometry for epitope elucidation and development in protein and peptide structure determi- nation

From the development of the first mass spectrometers at the turn of the 19th cen- tury, then called mass spectrographs, mass spectrometry has developed into a billion dollar industry nowadays with applications in many different scientific fields. How- ever, it was the development of soft ionization techniques in the late 1980s, mainly electrospray ionization (ESI) and matrix-assisted laser desorption (MALDI), which allowed major breakthroughs in the analysis of larger biomacromolecules with over 100 kDa in size. Both techniques allowed to surmount difficulties such as high sample consumption and low sensitivity for larger ions encountered with comparable ioniza- tion techniques like fast atom bombardment (FAB) or plasma desorption ionization (PDI). Due to the varied applicability of both ESI and MALDI in combination with other analytical tools, ESI and MALDI-MS are currently applied in various ”-omics” research areas for the study of proteomes, glycomes, metabolomes and lipidomes.

Due to these new developments, mass spectrometry became an established tool in the study not only of proteins but also of protein interactions. Especially electro- spray ionization, which allows the study of gas-phase dissociation of biomacromolecu- lar complexes with noncovalent interaction partners, became an increasingly powerful

method. The advantage that ESI can be performed in solution with physiological or near physiological pH and conditions allows the in vitro study of noncovalent complexes. As such, complexes such as antigen-antibody, enzyme-substrate or DNA- protein complexes can be directly studied. Further, as the charge distribution is directly proportional to the surface area of the ionized molecule, the m/z values in- dicate whether the protein is denatured or not [16].

A method that gained attention over the past years is the combination of hydrogen- deuterium exchange and mass spectrometry (HDX-MS). The protein or complex is incubated in a deuterated solvent allowing the exchange of hydrogen atoms along the peptide backbone situated at the surface of the complex. Hence the rate of the exchange is dependent on surface accessibility: Antigens bound by antibodies will be shielded through the complex formation and after proteolytic digestion (with pepsin) allows identification of epitope peptides [17].The method however needs to be carried out fast, as otherwise the deuterium exchange is reversed limiting its analytical power. Further approaches for identification of antibody epitopes include x-ray crystallogra- phy and NMR, immunoanalytical characterizations or epitope mapping by study of peptide arrays [18, 19, 20, 21].

1.4.1 Bioaffinity-mass spectromtry for epitope identification

A direct approach for epitope identification and characterization by bioaffinity mass spectrometry was initially presented by Suckau et al., whereby the limited proteolysis of a shielded antigen-antibody complex was utilized [22]. This methods combines two highly important methods: protein antigen proteolysis and mass spectrometric peptide mapping The methodology, epitope excision uses the shielding effect of the immune complex whereby only exposed and therefore accessible peptide bonds are cleaved through proteases (fig. 3b.). The epitope itself is not cleaved by the protease and after elution at acidic conditions can by identified by mass spectrometry. The complementary method called epitope extraction uses the analogous approach: The protein is first digested with proteases followed by formation of the immune complex and subsequent identification of the antigen peptide (epitope) by MS (fig. 3a.). The two methods are complementary to each other [23, 24].

Several different variations of this method have been developed Antibodies immo- bilized on beads such as sepharose being the most common approach. However, Macht et al. also presented the use of the size exclusion principle for epitope iden- tification for separation of antigens from antibodies after acidic dissociation of the immune complex [23]. This approach was later further developed by Al-Majdoub et al. by combination of epitope extraction using protein-G immobilized antibodies with

size exclusion chromatography. Upon elution of intact antibody antigen complexes from protein-G beads, intact immune complexes where subjected to size exclusion chromatography followed by mass spectrometric analysis [25].

Figure 3: Affinity mass spectrometric methids: Epitope extraction a.) and epitope excision b.). During epitope extraction, the protein is initially digested by proteases followed by extraction of the epitope from the proteolytic mixture using an immobilized antibody. The epitope can be analysed after washing away unbound peptides and elution of the epitope a.). For epitope excision, the immune complex is formed initially followed by proteolysis and washing/elution steps b.). Epitope peptides are analysed by mass spectrometry. Diagram adapted from Suckau et al. [22].

This approach has been successfully used for the identification and characteriza- tion of plaque specific amyloid-β antibodies [26, 27] as well as low affinity systems such as carbohydrate-recognising epitopes by carbohydrate recognition domain exci- sion (CREDEX) [28, 29]. Epitopes of high medical importance such as the Troponin T binding antibody designed for myocardial infarct diagnostic have also been charac- terized by this approach [23, 30].

Epitope extraction and excision are usually performed by packing the matrix cova- lently immobilized affinity ligand in a column. More recent methods report the use of combinations of biosensor and mass spectrometry to achieve the same goal: The combination of both methodologies is attractive due to the power of both analytcal methods: Kinetic information of the ligand interaction can be gained from the biosen- sor experiment, but not structural information is obtained. This crucial information can be gained by mass spectrometry which not only allows identification but also structural characterization of epitope peptides.

The online combination of surface acoustic wave (SAW) biosensor with electrospray mass spectrometry (ESI-MS) for the first time allowed the simultaneous characteriza- tion of the affinity constant (kinetic analysis) and mass spectrometric identifaction of the epitope (structural analysis) [31, 32]. Electrospray is especially suitable for direct online coupling due to use of a liquid flow for analyte introduction to the mass spec- trometer. There have been reports of biosensor and matrix assisted laser desorption ionization (MALDI) MS combinations but those methods cannot be described as di- rect coupling or online method as the biosensor chip needs to be transferred manually to the MALDI instrument. Consequently, this approach has proven less reliable and less successful [33]. The online combination of SAW and MS was used to study affin- ity systems such as amyloid-β and α-synuclein interaction [34], amyloid-β antibody interaction and identification and quantification of the interaction of 3-nitrotyrosine- modified peptides with nitrotyrosine-specific antibodies [35].

The online combination of both biosensor and mass spectrometry proves very power- ful allowing automated mass spectrometric epitope identification and characterization yet room for further development of the system is present.

The methods described above for epitope identification are the basis for the work carried out in this thesis. Primary structure characterization by proteolytic peptide mapping of antibodies proves difficult due to their proteolytic stability. This stability allows at the same time the use of antibody ligands during epitope excision, how- ever, proves to be challenging for identification of an epitope located on an antibody. Therefore, thorough method development was required at the beginning of this thesis.

Aims of this thesis

Adalimumab is an important antibody medication used in the treatment of a variety of diseases from rheumatoid arthritis to Crohns disease. Understanding of the underlying epitope that leads to the formation of anti-drug antibodies and the mechanism of drug- deactivation through such antibodies can lead to improved drug design to development of future treatment approaches. Although both epitope excision and extraction are established methods, careful adjustment to the studied proteins is required.

Using the tools of bioaffinity mass spectrometry, the goals of this thesis were as follows:

Primary structural characterization of Adalimumab

Epitope analysis requires careful characterization of proteolytic digests and anal- ysis of generated peptide fragments. Workflow development for sample prepa- ration (i) followed by peptide mapping of (ii) tryptic and (iii) chymotryptic digests and mass spectrometric analysis.

Identification of anti-Adalimumab epitope

Epitope identification was carried out with bioaffinity mass spectrometry: Prepa- ration of affinity columns using enrichments of two different patient sera (i) fol- lowed by epitope extraction experiments (ii).

Primary structural characterization of TNF-α

This step is to be carried out analogous to point 1. The sample preparation needs to be optimized for this structurally different protein.

Affinity characterization of Adalimumab and TNF-αinteraction and identification of Adalimumab epitope on TNF-α

Characterization of affinity interaction between Adalimumab and TNF-α using SPR biosensor (i). Epitope extraction experiments with immobilized Adali- mumab (ii). Epitope excision by incubation of Adalimumab with TNF-α fol- lowed by proteolytic digestion.

Results and Discussion

Epitope identification is not a standardized procedure, as every protein and peptide behaves differently during each analytical step. The general approach for epitope analysis is displayed in figure 4. The protein is digested and analyzed followed by epitope identification by epitope extraction or excision [36, 24, 37]. The proteins and peptides are then subjected to SPR and MS measurement for structural characteri- zation and kinetics evaluation of the affinity pair.

Figure 4: Approach for epitope identification employed during this work for identification of anti- Adalimumab epitopes.

Primary structural analysis of therapeutic antibody Adal- imumab by mass spectrometry

Peptide mapping is defined as characterization of enzymatically generated peptide fragments by mass spectrometry, which is a precondition for safe identification of epi- tope determination. A work-flow of Adalimumab sample preparation was established ensuring reproducibility of analytical experiments with simultaneously ease of appli- cation.

Figure 5 displays the overview of the sample preparation and work-flow employed during this work for the complete primary structure analysis by mass spectrometry. As antibodies have a highly stable 3D structure for both chemical and enzymatic degradation, samples for primary structure determination were firstly reduced and alkylated followed by 1-dimensional gelelectrophoresis (fig. 5a). Through this, chem- ical separation of the four polypeptide chains (two heavy chains and two light chains) connected by disulfide bridges was achieved allowing separate analysis for primary structure determination. This step was carried out in order to reduce sample com- plexity for mass spectrometric analysis of protein digestion mixtures. The separated heavy and light chains were then processed separately by ingel-digestion and desalting by ZipTip procedure prior to MALDI-ToF MS (Fig. 5b).

a.

Figure 5: Established workflow of Adalimumab sample preparation for mass spectrometric analysis. The heavy and light chain were firstly separated by 1D SDS-PAGE a.); followed by in-gel digestion and MALDI-ToF analysis.

Reduction was carried out using both DTT and TCEP, for comparison of both reagents. TCEP and DTT both have similar redox potentials (E TCEP, pH = 2.5 = – 290 mV, E DTT, pH = 7.5 = -330 mV) at their respective working pH values [38, 39]. However, TCEP is considered an MS friendly reagent which is not required to be removed prior to MS analysis. Further, it does not react readily with sulfhydryl groups during the reduction reaction and is therefore generally considered the su- perior reducing agent [40]. Yet both reagents were tested during optimization for sample preparation. The reaction with DTT was performed for 1 h at 60◦C while the reduction with TCEP was performed for 10 min at 98◦C, after which the protein was alkylated with iodoacetamide. 20 µg of protein from each reduction reaction was loaded onto a 15% resolving polyacrylamide gel. Electrophoresed gels were treated with Coomassie staining, showing two distinct bands on the gel (fig. 6) indicating by the reduction of the disulfide bonds connecting the different protein chains of the

quaternary structures of the immunoglobuline in the FC area. The heavy chain was detected at around 50 kDa while the light chain was detected as a double band at around 20 to 25 kDa according to the molecular weight marker. For both samples, a very faint band between 100 and 200 kDa was detected, corresponding to non-reduced antibody. No band was detected at 75 kDa, which would correspond to a dimer of light and heavy chain, hence the reduction reaction ocurred nearly 100% quantita- tive. There was no significant difference between the yield of the reaction between TCEP or DTT. However, as TCEP was provided as hydrochloride salt, an additional neutralization step prior to the alkylation reaction was required to readjust the pH to 8. Therefore, DTT was found to be the better reduction agent for this work and was used for all further reduction reactions.

Figure 6: SDS-PAGE of reduced and alkylated Adalimumab. The heavy chain was found to run at about 50 kDa, while the light chain was found to run as a double band at 25 kDa according to the molecular weight marker (PR). Further, the gel showed no comparable difference in the yield of the reduction reaction performed either with DTT or TCEP.

An aliquote of the reduced and alkylated sample was also taken for analysis by MALDI-ToF MS to control the yield of the alkylation reaction prior to performing proteolytic digestion experiments. 0.5 µl sample was mixed with 1 µl MALDI ma- trix mixture, loaded on a MALDI target and measured using the pre-programmed LP-ProtMix method (Bruker). The spectrum was acquired in linear ToF-mode (Fig. 3). The spectrum shows the singly and doubly charged ions corresponding to the molecular weights of the fully alkylated light chain of Adalimumab. The heavy chain was found to not ionized as efficiently, possibly due to the strong glycosylation on the FC region of the heavy chain, compared to the non glycosylated light chain. Strongly glycosylated proteins are known to sometimes ionize less effectively leading to ion suppression effects.

The light chain was detected at an m/z value of 23683.02 corresponding to the fully alkylated protein: Alkylation with iodoacetamide modifies the sulfhydryl-groups of cysteine side chains by introduction of carboxyamidomethyl-groups, thus preventing the re-formation of disulfide bridges. This leads to an increase of the protein or peptide mass by 57.02 Da per modified cysteine residue. The Adalimumab light chain with non alkylated cysteine residues has a molecular weight of MTheoretical = 23392.57 Da. The experimental m/z value of 23683.02 corresponds to a mass increase of 290.45 Da. This coincides to the introduction of exactly five carboxamidomethyl-modifications at the five cysteine residues (Cys23, Cys-88, Cys-134, Cys-194, Cys-214) located on the light chain of Adalimumab. Therefore, the alkylation reaction occurred quantitative. Although the heavy chain could not be detected directly, it can be safely assumed that if the light chain was fully alkylated during the reaction, then the heavy chain cysteine residues should also be fully modified with carboxyamidomethyl-groups.

Figure 7: MALDI-ToF spectrum of fully reduced and alkylated Adalimumab. Only singly and doubly charged signals for the completely alkylated light chain were detected. The heavy chain was found to ionize less due to the strong glycosylation on the constant region of the heavy chain. This leads to ion surpression effects causing the signal of the heavy chain to be much lower than the light chain signal. The signal was so low that it could not be detected.

Adalimumab heavy chain primary structure characterization by proteolytic mass spectrometry

For identification of epitope peptides of Adalimumab recognized by anti-drug anti- bodies, detailed analysis of all possible proteolytic peptides to be generated for the epitope extraction experiment is important. This approach for proteolytic protein identification and characterization is defined as bottom-up approach, complementary to the top-down approach by gas-phase fragmentation during MS/MS experiments [41]. Analysis was carried out by ingel digestion of Adalimumab heavy and light chain separately followed by MALDI-ToF mass spectrometry.

Adalimumab is an IgG1 class antibody, produced as fully humanized drug via phage display. The immunoglobuline consists of four polypeptide chain, two heavy chains and two light chains. Both chains consist of a variable and a constant region, where large segments of the Fab domain are made up of the variable segments FV which bind the antigen via their CDR loops. The loops of the constant region FC are stabilized by glycosylations [42, 43].

Figure 8: Schematic representation of an IgG1 class antibody. Different segments are marked and labelled accordingly: IgG class antibodies are generally diveded into two segments: the variable region FV and the constant region FC. The variable region comprised of both heavy (blue) and light chains(red) makes up the complementary determining region (CDR) with the CVH region of the heavy chain and the VL of the light chain. The Fab fragment consists of the variable region plus the first constant regions of both heavy and light chain (CH1 and CL respectively).

Ingel digestion was performed after washing and rehydration of gel pieces with digestion buffer. Each band of heavy and light chain gel pieces of Adalimumab were incubated with both chymotrypsin and trypsin overnight at 37◦C for prote- olytic degradation. Trypsin cleaves peptide bonds after basic amino acids lysine (K) and arginine (R) while chymotrypsin is considered a complementary protease cleav- ing after aromatic residues tyrosine (Y), tryptophane (W), phenylalanine (F) and the non-aromatic amino acid lysine (L). Following desalting by ZipTip procedure, the

digestion mixtures were applied to a MALDI target and measured with both linear and reflectron ToF mode (fig. 10). The FV sequence (1-124) was suspected to hold the anti-drug-antibody epitope from pediatric patient sera as the FC sequence is con- served across all human IgG1 immunoglobulins. Antibodies with FC epitopes will consequently bind to all human IgG1 antibodies and not just Adalimumab. Hence the FV sequence is discussed here in more detail; for an overview of the complete peptide mapping see chapter 2.1.2. MALDI-ToF analysis yielded a sequence coverage of 100% of the FV domain from peptide fragments generated during tryptic digestion. Figure 9 shows the nine in-silico tryptic cleavage sites located on the variable region. Blue arrows show the cleavage sites identified , while red arrows show missed cleavage sites. However, Lys-43 is preceded by Pro-41, which will reduce enzymatic cleavage considerably. This matches the literature cleavage specificity of trypsin whereby this protease will cleave C-terminal peptides bonds after lysine and arginine but not after K and R which are proceeded by proline. The second missed cleavage was Arg-67.

Figure 9: Sequence of the variable region of the heavy chain of Adalimumab. Tryptic cleavage sites are marked by arrows: blue arrows indicate postively identified cleavage sites, read arrows mark missed cleavage sites. The identified peptide sequence is underlined.

Table 1 lists all peptide fragments identified my mass spectrometric analysis. The glycosylation on the FC region of Adalimumab at Asn-301 was found by the detec- tion of two ions: one ion corresponding to the biantennary G0F glycosylated peptide 297-305 as well as two non-glycosylated ions 293-305 and 297-305 were found (struc- ture displayed in fig. 11). During the energy intense ionization process of MALDI- MS, more unstable covalent bonds such as the linkage of Asn-N-acetylglucosamine are known to fragment. Therefore both the glycosylated peptide as well as non- glycosylated peptides 297-305 with a mass difference of 1868.58 Da coinciding to a G0F glycosylation modification with loss of two terminal N-acetylgalactosamine were detected. This glycosylation of Adalimumab was found to be the most abundant type of glycosylation on this biosimilar drug antibody [44].

N-Acetylglucosamine

Mannose Fucose

Asn301-Ser302-Thr303

Figure 11: Glycosylation structure of Adalimumab as identified by MALDI-ToF. The glycan corre- sponds to the biantennary oligosaccharide with a G0F structure: The N-linked acetylgalactosamine is attached to N-301 on the Adalimumab heavy chain. To this sugar residue another acetylgalac- tosamine linked to three mannose residues which branch of are connected. The biantennary structure is capped by two further N-acetylgalactosamine. One Fucose residue is attached to the first acetyl- galactosamine. This glycosylation is common among human IgG1 antibodies [45]

Table 1: Adalimumab heavy chain tryptic peptides identified by MALDI-ToF and list of experimental and theoretical m/z values

The spectrum acquired of chymotryptic digestion mixture is displayed in figure 13, with all in silico cleavage sites located on the FV domain indicated in figure 12. Inspection of the sequence shows much more possible chymotryptic cleavage sites lo- cated in this section of the protein compared to tryptic cleavage sites. Therefore, the chymotryptic digestion yields a lot more peptide fragments by digestion of the variable domain, making chymotrypsin the more suitable enzyme for epitope analysis. Shorter and more peptide fragments make the identification of the minimal epitope sequence easier. Figure 12 shows that also chymotrypsin did not cleave all predicted peptide bonds, similarly to trypsin. The digestion with chymotrypsin also yielded a glycosylated peptide (301-317) corresponding to the G0F modification already ob- served in the tryptic digestion mixture, however, the non glycosylated peptide was not observed. A list of all peptides identified by MALDI-ToF is listed in table 2.

Figure 12: Sequence of the variable region of the heavy chain of Adalimumab. Chymotryptic cleavage sites are marked by arrows: blue arrows indicate cleavage sites resulting from peptide fragments identified in the chymotryptic mixture by MALDI-MS, while red arrows indicate missed cleavage sites. The identified peptide sequence is underlined.

Figure 13: MALDI-ToF spectrum of chymotryptic mixture of Adalimumab heavy chain. Signals mark with black dots indicate unidentified signals. The glycosylated peptide 301-317 is marked by a red dot.

Table 2: Adalimumab heavy chain chymotryptic peptides identified by MALDI-ToF and list of ex- perimental and theoretical m/z values

Overview of of Adalimumab heavy chain mass spectrometric map- ping and sequence analysis

Figure 14: Comparison of sequence coverage from tryptic and chymotryptic digestion analysis. Tryp- tic peptide fragments are underlined blue, chymotryptic fragments are underlined in red. Cleavage sites are indicated by vertical lines.

Figure 14 summarizes the results of the primary structural analysis by proteolytic peptide mapping, comparing the results of both tryptic and chymotryptic digestion. The tryptic peptides are underlined in blue, showing that these fragments cover the constant domain of the Adalimumab heavy chain much better compared to the chy- motryptic fragments (underlined in red). However, the chymotryptic peptides map the variable region with much more overlapping peptide fragments. This is mainly due to more available cleavage sites (twenty five chymotryptic cleavage sites compared to nine tryptic cleavage sites) across the FV sequence. Further, the FV region is much more hydrophobic than the constant domain of the heavy chain, making it more ac- cessible to chymotryptic cleavage than tryptic cleavage. Additionally, a lot of peaks in the tryptic mass spectrometric peptide mapping could not be assigned to tryptic peptides, indicating unspecific cleavage adducts.

The variable domain was 100% covered by overlapping peptide fragments. The total sequence coverage of the heavy chain was calculated to be 96%. The constant region is covered much better by tryptic peptides. The glycosylation located at Asp-301, which is the only location of the N-glycosylation consensus sequence N-Xxx-S-T, was confirmed as G0F glycosylation during analysis of the tryptic peptide mixture as de- scribed earlier already by Rodrigues et al.[44]. Missing sequence parts covered neither by trypsin or chymotrypsin include 218-221 and 279-291. However neither sequence are of particular interest to this work as the epitope is suspected on the variable sub-domain as mentioned above.

Adalimumab light chain primary structure characterization by pro- teolytic mass spectrometry

Structure analysis of the light chain was carried analogous to the structure elucidation of the heavy chain. Ingel digestion was performed using chymotrypsin overnight fol- lowed by the peptide extraction and sample desalting. Figure 16 displays the MALDI- ToF spectrum acquired from the chymotryptic peptide mixture which is summarized in figure 15 and table 3. The variable domain of the light chain (1-121) is completely covered by the peptide fragments. 20 in-silico chymotryptic fragments were identified within the variable domain sequence, of which five missed cleavages were found by MS analysis. Sequence coverage for the FV region was 96%, the coverage of the total polypeptide chain was calculated to be 87%.

Figure 15: Sequence of the variable region of the light chain of Adalimumab. Chymotryptic cleavage sites are marked by arrows: blue arrows indicate positively identified cleavage sites, read arrows mark missed cleavage sites. Sequence segments covered by peptide fragments are underlined in blue.

Figure 16: MALDI-ToF spectrum of chymotryptic mixture of Adalimumab light chain.

Table 3: Adalimumab light chain chymotryptic peptides identified by MALDI-ToF and list of exper- imental and theoretical m/z values

N-terminal sequence by Edman sequencing

Confirmation of the N-terminal heavy and light chain was performed by automated Edman sequencing . This chemical identification has the advantage of differentiating between isobaric amino acids leucine and isoleucine which is not possible by mass spec- trometry without powerfull MS/MS techniques. Further, knowledge the N-terminal sequences confirms the starting point of proteins and peptides derived by biotechno- logical methods such as phage displayed Adalimumab: During protein translation, the mRNA translates the proteins from N-terminus to C-terminus. Therefore com- plete proteins and peptides should show a complete and correct N-terminal sequence.

The heavy and light chain were analysed simultaneously, producing two peaks during each chromatographic separation. Below the first three cycles and the last cycle of the first 15 N-terminal residues are displayed. Further chromatograms are listed in the supplementary materials.

min

N-terminal sequence heavy chain: E – N-terminal sequence light chain: D –

Figure 17: Chromatogram recorded at 269 nm of PTH amino acid derivatives. The first two amino acids identified were D and E. Identified residues assigned to the different chains are listed below the chromatogram.

N-terminal sequence heavy chain: E – V N-terminal sequence light chain: D – I

Figure 18: Chromatogram recorded at 269 nm of PTH amino acid derivatives. Residues number two cleaved of the heavy and light chain. Identified residues assigned to the different chains are listed below the chromatogram.

min

N-terminal sequence heavy chain: E – V – Q N-terminal sequence light chain: D – I -Q

Figure 19: Chromatogram recorded at 269 nm of PTH amino acid derivatives. Residues number three cleaved of the heavy and light chain. In both cases the amino acid identified was glutamine explaining the intense peak at 5.50 min. Identified residues assigned to the different chains are listed below the chromatogram.

min

N-terminal sequence heavy chain: E – V – Q – L – V – E – S – G – G – G – L – V – Q – P – G – R N-terminal sequence light chain: D – I -Q – M – T – Q – S – P – S – S – L – S – A – S – V

Figure 20: Chromatogram recorded at 269 nm of PTH amino acid derivatives. Fifteenth cycle of Edman degradation yielding still acceptable identification of amino acids from both light and heavy chain simultaneously. Assembled amino acid sequence up to fifteenth residue assigned to the different chains are listed below the chromatogram.

The sequence elucidated by Edman sequencing consisting of the first fifteenth residues each was consistent with results obtained by mass spectrometry. Further cycles up to 20 degradation steps were carried out, however the yields showed signifi- cant reduction and chromatograms could not be interpreted safely. As the PTH-amino acids are retained so well on the column, PTH derivates are washed out slowly dur- ing each cycle, increasing the background noise allowing no simply identification of the actual amino acid. By calculation of repetitive yields and expected yields per PTH-amino acid, differentiation between a new peak and a wash out peak is possible, allowing identification of peaks: Through integration of peaks, the total amount of each PTH-amino acid is possible. By comparison of the peak area of the previous cycle, the new peaks can be differentiated from wash out peaks. Separation of heavy and light chain by SDS-PAGE followed up by electro-blotting which can be used directly for sequencing was attempted but not successful.

Identification of an anti-Adalimumab-antibody specific epitope using mass spectrometry

Pediatric patients with juvenile idiopathic arthritis are treated with Adalimumab to block TNF-α, thus lessening the inflammatory response of the arthritic disease and improving the living conditions of the patient. These patients are known to develop anti-drug antibodies (ADA) against the therapeutic Adalimumab antibody, despite the fact that Adalimumab is developed as a fully humanized immunoglobuline. Con- sequently the treatment is rendered ineffective and needs to be stopped.

To identify the epitope sequence recognized by the anti-drug antibodies present in sera of treated patients, epitope extraction combined with mass spectrometric anal- ysis was employed. Antibody immobilization was performed using a POROS protein G prepacked sensor cartridge, to which the patient IgG enriched serum fraction was immobilized by DMP crosslinking. The patient sera were collected after the acute phase of the anti-drug antibody formation phase, where mainly IgG-type antibodies are being formed. The serum affinity column was tested by applying 10 µg of Adal- imumab and monitoring flow through and elution at 280 nm. The chromatogram is shown in figure 21. Adalimumab was found to elute after 32 min after switching to elution, showing that the immobilized antibodies from patient serum are able to bind and capture the protein.

Figure 21: Affinity chromatography of Adalimumab with immobilized IgG fraction of patient serum. 10 µg of Adalimumab were applied to the column at a flow rate of 50 µl min-1. During washing for the first 20 min, no elution was observed.After 20 min, buffer was switched to elution buffer with pH 2.5, eluting the affinity captured Adalimumab after 32 min.

For epitope extraction, a total of 100 µg of Adalimumab chymotryptic digest of heavy and light chain combined was applied to the serum affinity column. Non- specifically bound peptide fragments were washed away (supernatent fraction) and elution performed at pH 2. The MALDI-ToF spectrum (figure 22) of the elution fraction reveals only one peptide corresponding to Adalimumab heavy chain 12-29. The epitope sequence is located before and at the beginning of CDR-H1, one binding region of Adalimumab to TNF-α. Table 4 lists the identified epitope sequence with corresponding mass.

In order to confirm the epitope sequence, the extraction experiment was repeated us- ing IgG enrichment from a second patients serum. The affinity column was prepared by the same procedure as before. Figure 23 shows the elution fraction, yielding the same single peptide of Adalimumab heavy chain 12-29. This indicates the possibility of a single uniform epitope recognized by antibodies formed in different Adalimumab treated patients. Therefore the anti-drug antibodies formed in patients upon an im- mune reaction are generated against the same peptide fragment of Adalimumab in all treated patients. In order to confirm this hypothesis, a larger group of patients would be required to be tested. The 12-29 peptide could not be retained non-specifically through protein-G used for IgG sera immobilization as protein G binds specifically to the FC domain. Confirming the location of the elucidated epitope at CDR-H1 should be performed by repeating the experiments using a tryptic peptide mixture as well as by performing proteolytic epitope excision. During epitope excision, the shielding effect of the antigen-antibody complex is utilized preventing cleavage of epi- tope segments due to steric blockage of the protease. Thereby discontinuous epitopes consisting of two or more epitope sequences not located discontinuously on the pri- mary amino acid structure the can be determined. During epitope extraction, such conformational epitopes are not always elucidated as the immune complex is formed after digestion.

In order to elucidate the functionality of this epitope, a competitive binding assay by SPR biosensor with immobilized Adalimumab and a mixture of anti-Adalimumab antibodies, TNF-α and the epitope peptide should be performed. In case the epitope is functional, it will bind the anti-Adalimumab antibodies blocking their CDRs and Adalimumab will interact with TNF-α and give rise to an SPR signal. This will clarify the functionality of this epitope.

Table 4: Summary of mass spectrometric epitope extraction data for Adalimumab

26CDR-135 50CDR-263

1EVQLVESGGGLVQPGRSLRLSCAASGFTFDDYAMHWVRQAPGKGLEWVSAITWNSGHIDYADSVEGRFTI70

Figure 22: Mass spectrometric epitope extraction of Adalimumab using chymotrypsin. Patient sera 02b was used for affinity column preparation. Non-identified signals are marked by black dots. The enlarged section shows the isotopic distribution pattern of the peak. The sequence of Adalimumab 1-70 with CDR-H1 and CDR-H2 (marked in blue) is shown above the spectrum.

12VQPGRSLRLSCAASGFTF29

[12-29]+

1954.60

+

[M+H]

= 1955.23

[M+H]

+calc

= 1954.60

exp

Figure 23: Mass spectrometric epitope extraction of Adalimumab using chymotrypsin.The enlarged section shows the isotopic distribution pattern of the peak. Patient sera 09b was used for affinity column preparation. The affinity column was prepared using serum from a second patient.

EVQLVESGGG LVQPGGSLRL SCAASGFTFD DYAMHWVRQA PGKGLEWVSA ITWN–SGHI 60 EVKLEESGGG LVQPGRSMKL SCVASGFIFS NHWMNWVRQA PEKGLEWVAE IRSKSINSAT 60

DYADSVEGRF TISRDNAKNS LYLQMNSLRA EDTAVYYCAK VSYLSTASSL DYWGQGTLVT 120 HYAESVKGRF TISRDDSKSA VYLQMTDLRT EDTGVYYCSR NYYG–STY DYWGQGTTLT 120

VSS 123

VSS 123

DIQMTQSPSS LSASVGDRVT ITCRASQGIR NYLAWYQQKP GKAPKLLIYA ASTLQSGVPS 60 DILLTQSPAI LSVSPGERVS FSCRASQFVG SSIHWYQQRT NGSPRLLIKY ASESMSGIPS 60

RFSGSGSGTD FTLTISSLQP EDVATYYCQR YNRAPYTFGQ GTKVEIKR 108 RFSGSGSGTD FTLSINTVES EDIADYYCQQ SHSWPFTFGS GTNLEVKR 108

Figure 24: Sequence alignment of Adalimumab and Infliximab fab fragments. The heavy chain (HC) and light chain (LC) of each antibody were aligned, with differences in amino acid sequences highlighted in blue. The anti-Adalimumab epitope on Adalimumab is marked in red.

Figure 25: Crystal structure of Adalimumab fab (blue and green) fragment in complex with TNF- α(orange). The identified epitope (Adalimumab heavy chain 12-29) is marked in red. From the crystal structure it is clear that the anti-Adalimumab antibody epitope is not located directly at the interaction site of Adalimumab and TNF-α. The epitope was expected here as the anti-Adalimumab antibodies are known to block Adalimumab-TNF-α interaction [46]

The first generation TNF-α blocking antibody Infliximab primary amino acid se- quence is compared with Adalimumab in figure 24. The epitope region sequence of anti-Adalimumab antibodies (marked in red) shows that four amino acids are differ- ent between the two antibodies explaining why anti-Adalimumab antibodies do not bind Infliximab. Patient treated with Adalimumab are switched to Infliximab suc- cessfully after observed formation of anti-drug antibodies. This shows that patients anti-Adalimumab antibodies do not bind Infliximab, and treatment is conducted suc- cessfully. Further this implies the importance of residues Gly-16, Leu-18, Arg-19 and Thr-28 for Adalimumab/anti-Adalimumab interaction.

Figure 25 shows the crystal structure of the Adalimumab fab complex with TNF- α. The anti-Adalimumab antibody epitope segment on Adalimumab CVH 12-29 is marked in red, showing a part of an anti-β-parallel loop at the tip of CDR-H1 iden- tified as binding site. This binding site is not directly located at the same site as TNF-α binding (Adalimumab paratope), however directly proximal to the interaction segment at CDR-H1. The anti-Adalimumab epitope was expected to overlap with the TNF-α paratope on Adalimumab. Hu et. al. 2014 elucidated the Adalimumab/TNF- α epitope and paratope interaction by crystal structure of the fab/protein complex [47]. showing that CDR-H1 only interacted with TNF-αat Asn-31 and Trp-33. Other segments such as CDR-H2 contributed much more towards the interaction. Hence the anti-Adalimumab epitope seems to not overlap with the TNF-α paratope.

Therefore, elucidation of the Adalimumab epitope/paratope became a major goal of this thesis, as previous data acquired from crystal structural studies seemed to conflict with the data collected during this thesis. However, it has to be kept in mind that crystal structures do not represent in-vivo conditions as changes to protein structures can occur during the crystallization procedure.

KD Determination of TNF-α and Adalimumab interac- tion

Strong affinity interaction of high importance for epitope analysis by proteolytic ex- traction and excision as the formation of the immune complex between antibody and antigen needs to be stable and not dissociate easily. The lower the affinity constant, KD is, the stronger the affinity interaction between the two binding partners is. The KD value is directly equal to the association constant kon divided dissociation con- stant koff of the interacting molecules [48, 49, 50].

The affinity constant between TNF-alpha and Adalimumab was measured by surface

plasmon resosnance (SPR) biosensor with immobilized Adalimumab and a dilution series of human recombinant TNF-α (soluble fraction, 77-233). The sensorgram was recorded over 10 minutes for each association and dissociation phase followed by re- generation steps after each sample injection. The data was evaluated using a one to one Langmuir binding model with both kon and koff constants set to global fitting types. The affinity constant was determined to be 6.16 x 10-10 M (see table 5). The lower the affinity constant, the stronger the affinity interaction exhibited by the affin- ity pair. The χ2 error was calculated to be 0.83. This error calculation known as Pearsons χ2 error test represents the statistical goodness of the fitted data. For this measurement, the value of 0.83 presents a small difference between the fitted data (black lines in fig. 26) and the acquired data (coloured lines) [51].

The affinity between Adalimumab and TNF-α is very high. By comparison the affin- ity between the other two widely used TNF-α blocking biosimilars Infliximab and Etanercept, the KD value measured also by SPR was 5.9 x 10-9 M and 2.0 x 10-9 M respectively. The values are 10 orders of magnitude higher expressing the lower affin- ity towards TNF-α [52]. Further the same study found that the affinity of all three biosimilars was lower towards membrane bound TNF-α compared to soluble TNF. However, these results could not be linked towards any differences in immunogenic- ity or clinical efficacy for treatment as described earlier. Only Adalimumab and Infliximab show efficacy in treatment of Crohns disease while Etanercept did not. Adalimumab (6-12%) and Etanercept (2%) show much lower immunogenicity than Infliximab (60%) [47].

kon

KD =

k

off

Table 5: Summary of KD value calculated for the interaction of immobilized Adalimumab with TNF-

α from kon and koff values

kon koff KD χ2 error 5.90 x 105 3.63 x 10-4 6.16 x 10-10 0.83

Time (min)

Figure 26: SPR sensorgram of interaction between human recombinant TNF-α and immobilized Adalimumab at various concentrations. TNF-α was injected over Adalimumab with increasing con- centrations. The recorded sensorgrams were evaluated according to a one to one binding model.

Primary structure characterization of TNF-α by bottom- up mass spectrometry

For further insights into the pharmaceutical mechanism of the anti-Adalimumab an- tibodies, the interaction between Adalimumab and its normal antigen, TNF-α, was further investigated. To achieve this, TNF-α was firstly structurally characterized by mass spectrometric peptide mapping prior to epitope analysis.

TNF-α is a 17.35 kDa large cytokine consisting of 233 amino acids, which form upon protein translation a membrane bound homotrimer [53, 54]. This trimer is then cleaved by the TNF-α converting enzyme into the active soluble homotrimer, each subunit consisting of 157 amino acids. By binding of either TNF-R1 or TNF-R2, TNF-α induces the classic inflammatory symptoms including heating, pain, redness and swelling of affected areas. Further reactions include cell death, migration of neu- trophiles towards the affected area and phagocytosis by macrophages. Adalimumab minimizes such reactions by binding and blocking TNF-α, making it a universal drug used in treatment of inflammatory diseases including arthritis. Structurally, TNF- α monomers consist of a jelly roll β-sheet sandwhich with anti-parallel β strands.

The primary amino acid structure together with predicted tryptic cleavage sites is presented in figure 14. The MALDI-ToF spectrum is shown in figure 15, with blue ar- rows indicating in the sequence cleavage sites corresponding to the peptide fragments. Six missed cleavage sites were observed, one of which is followed by a C-terminal pro- line residue 10K-P12. The total sequence coverage for the tryptic peptide mapping was calculated as 95%.

1 10 20 30 40 50 60

VRSSSRTPSD KPVAHVVANP QAEGQLQWLN RRANALLANG VELRDNQLVV PSEGLYLIYS

61 70 80 90 100 110 120

QVLFKGQGCP STHVLLTHTI SRIAVSYQTK VNLLSAIKSP CQRETPEGAE AKPWYEPIYL GGVFQLEKGD RLSAEINRPD YLDFAESGQV YFGIIAL

Figure 27: Mass spectrometric peptide mapping of tryptic peptide mixture. Tryptic cleavage sites are marked by arrows: blue arrows indicate cleavage sites resulting from peptide fragments identified in the peptide mixture by MALDI-MS, while red arrows indicate missed cleavage sites. The protein sequence covered by the peptide fragments is underlined in blue.

[7-32]1+

2913.43

Figure 28: MALDI-ToF spectrum of tryptic mixture of TNF-α. Signals marked with black dots indicate unidentified signals.

Table 6: TNF-α tryptic peptides identified by MALDI-ToF and list of experimental and theoretical m/z values

The peptide mapping was repeated using the protease chymotrypsin (fig. 29 and 30). Missed cleavage sites where found in areas of highly hydrophobic sequence parts, or with two or more possible residues for cleavage such as 54L-Y-L-I-Y60. The total sequence coverage was 100%.

1 10 20 40 50 60

VRSSSRTPSD KPVAHVVANP QAEGQLQWLN RRANALLANG VELRDNQLVV PSEGLYLIYS

61 70 80 90 100 110 120

QVLFKGQGCP STHVLLTHTI SRIAVSYQTK VNLLSAIKSP CQRETPEGAE AKPWYEPIYL GGVFQLEKGD RLSAEINRPD YLDFAESGQV YFGIIAL

Figure 29: Mass spectrometric peptide mapping of chymotryptic peptide mixture. Chymotryptic cleavage sites are marked by arrows: blue arrows indicate cleavage sites resulting from peptide fragments identified in the peptide mixture by MALDI-MS, while red arrows indicate missed cleavage sites. The protein sequence covered by the peptide fragments is underlined in blue.

[88-115]1+

3204.90

Figure 30: MALDI-ToF spectrum of chymotryptic mixture of TNF-α . Signals mark with black dots indicate unidentified signals.

Table 7: TNF-α chymotryptic peptides identified by MALDI-ToF and list of experimental and theoretical m/z values

TNF-α epitope recognition by Adalimumab

Epitope extraction was combined with mass spectrometry for elucidation of the recog- nition pattern of Adalimumab. Adalimumab was covalently immobilized on CNBr- activated Sepharose beads and packed in a microcolumn. Tryptic and chymotryptic digests were incubated for 2 h before removal of the proteolytical fragments and dis- sociation of the immune complex by addition of 0.1% TFA.

The MALDI-ToF spectrum of the elution fraction is shown in figure 26 (epitope ex- traction with tryptic mixture) and figure 27 (epitope extraction with chymotryptic mixture). Each spectrum yielded one single peptide segment corresponding to TNF- α 33-44 (tryptic) and TNF-α 37-59 (chymotryptic). The data is summarized in table 7, showing an overlapping sequence 37L-A-N-G-V-E-L-R44.

Epitope excision was performed by incubation of immobilized Adalimumab with TNF- αfollowed by proteolytic digestion using trypsin on the affinity column. The beads were washed subsequently and the washing fractions checked until no further peptides were eluted. The antigen-antibody complex was dissociated under acidic conditions with TFA and the elution fractions combined for mass spectrometric analysis. The MALDI-ToF analysis yielded two N-terminal peptides: TNF-α 7-31 and 7-32 (fig. 33).

Figure 31: Mass spectrometric epitope exctraction of TNF-α using trypsin. Non-identified signals are marked by black dots. The sequence of TNF-α 20-70 is displayed above, with epitope sequences of tryptic and chymotryptic mixtures underlined in blue. The enlarged section shows the isotopic distribution pattern of the peak.

37LANGVELRDQLVVPSEGLYLIY59

[37-59]2+

1288.04

Figure 32: Mass spectrometric epitope exctraction of TNF-α using chymotrypsin. The enlarged section shows the isotopic distribution pattern of the peak.

7TPSDKPVAHVVANPQAEGQLQWLNR31

[7-31]1+

2757.44

Figure 33: Epitope excision of TNF-α with trypsin as identified by MALDI-ToF.Isotopic resolution was not achieved for this measurement.

Table 8: Summary of mass spectrometric epitope extraction data for TNF-α

In conclusion, the epitope of Adalimumab on TNF-α is composed of at least two discontinous segments on the surface of TNF-α as identified by both epitope excision and extraction: The N-terminal segment 7-32 and the loop connecting two beta sheets between 37 to 44 (see fig. 35).

TNF-α and the complementary cytokine lymphotoxine-α, also known as TNF-β, have four known either identical or conserved sequences: 11-13, 37-42, 49-57 and 155-157 (see fig. 34). These sequence are conserved among human, mouse and horse and either identical or similar between TNF-α and lymphotoxin. Both TNF-αand lym- photoxin bind to the TNF-receptors TNF-R1 and TNF-R1, making similar sequences very likely for receptor binding locations [53]. Noteworthy are the sequence similari- ties of the proposed receptor binding region, also known as the tip protein of TNF-α, the N-terminal 11K-P-V13 and the C-terminal 155I-A-L157 of TNF-α which are identical with lymphotoxin-α. Further, the identical residue side chain of Tyr-56 projects into a cavity lined by polar and small hydrophobic side chains and is the only residue that deemed important for receptor binding and was also identified within one epitope se- quence identified by mass spectrometric epitope extraction. This residue was proven functionally important for the correct folding of the tip segment which is involved in the receptor binding through mutation of the tyrosine residue. Upon substitution of Tyr-56, binding with the receptor could no longer be observed suggesting either loss of conformation or direct involvement of this residue in receptor binding. However, all other residues suggested of importance for receptor binding are located at the C- and N-terminal segments at the base of the protein (segments 11-13 and 155-157, marked

in magenta in fig. 35) and in close proximity to the proteolytic epitope segments 33-44 and 37-59. Therefore Adalimumab inhibits binding of TNF-α to its natural tar- get receptor TNF-R1/R2 by blocking the binding sites on TNF-α directly or through steric hindrance. In case of a directly blocked TNF-receptor binding site, the discon- tinuous epitope as identified in this work, is expected. Both the N-terminal conserved sequence 11-13 and the conserved region 37-42 were identified during epitope excision and epitope exctraction respectively.

Then further peptides situated at the C-terminus could be identified by mass spec-

––––––VRSSSRTPSD 10

LPGVGLTPSAAQTARQHPKMHLAHSTL 33

KPVAHVVANPQAEGQLQWLNRRANALLANGVELRDNQLVVPSEGLYLIYSQVLFKGQG– 68 KP AH P L W A L G L N L VP G Y YSQV F G KPAAHLIGDPSKQNSLLWRANTDRAFLQDGFSLSNNSLLVPTSGIYFVYSQVVFSGKAYS 87

–CPSTHVLLTHTISRIAVSYQTKVNLLSAIKSPCQRETPEGAEAKPWYEPIYLGGVFQL 126 S L H Y V LLS Q G PW Y G FQL PKATSSPLYLAHEVQLFSSQYPFHVPLLSS––QKMVYPGLQ-EPWLHSMYHGAAFQL 141

EKGDRLSAEINRPDYLDFAESGQVYFGIIAL 157 GD LS L S V FG AL

TQGDQLSTHTDGIPHLVLSPS-TVFFGAFAL 171

Figure 34: Sequence alignment of primary amino acid sequence of human TNF-α (TNF) and lymphotoxin-α (LYMPH). Conserved amino acids are repeated in the middle. Conserved TNFα se- quences are marked in blue, while the overlapping epitope sequence is coloured in red. Receptor binding sequences 11-13 and 155-157 of TNFα are conserved and marked in blue.

trometry. An alternative would be that Adalimumab blocks the binding of TNF-α to its receptor by binding in close proximity to the receptor binding site. This steric hindrance would thereby make binding at the receptor pocket not possible. Compet- itive binding assays carried out previously showed that Adalimumab mode of action prevents TNF-α binding to its receptor, hence allowing only the two conclusions described above.

The aforementioned crystal structure analysis of Adalimumab fab and TNF-α struc- ture contradicts the findings of previous crystal structural studies and substitution experiments as well as the epitope data raised during this work. Eck et. al. 1989 showed that the previously described four conserved polypeptide sequences are impor- tant for receptor binding not only by discussion of structural features but also through substitution or elimination of amino acids important for the structural integrity of the TNF-α tip segment at the N- and C-terminus. Amino acid side chain residues proposed by Hu et al. to be of importance for TNF-receptor binding do neither con- cur with these findings (see table 9) nor with the epitope data. It has to be noted that none of the residues proposed by Hu et al. is conserved [47, 53]. Further, figure 36 shows the residues identified by Eck et al. marked red in the crystal structure. According to this crystal structure, the receptor binding sites of TNF-α are located far away from the Adalimumab binding site. However, it was shown previously by competitive binding assays, that TNF-α blocking antibodies such as Adalimumab and Infliximab exhibit their pharmaceutical function by blocking the binding site of TNF- α to its receptor [55, 5, 52]. This property has also been reviewed [56]. Consequently, the crystal structure data is neither conclusive with the pharmaceutical functionality of the drug-antibodies nor the epitope data.

Epitope excision b. Epitope exctraction

Figure 35: Ribbon diagram of the TNF-α crystal structure. The epitope segments identified by both epitope excision TNF-α 7-32 (a) and epitope extraction TNF-α33-59 (b) are marked in red. The conserved receptor binding segments TNF-α 11-13, 37-42 and 49-57 are highlighted in green and overlap with the epitope segments. The C-terminal receptor binding segment 155-157 was neither identified during epitope extraction or epitope excision. Either Adalimumab does not recognize this part on TNF-αor due to the low binding of individual segments of discontinuous epitopes was not found during the experiments.

Table 9: Comparison of residues important or directly involved in receptor binding of TNF-α as proposed in different publications [47, 53]. The epitope identification data matches with the findings of and predictions of Eck et al., but the residues identified of the Adalimumab fab and TNF-α crystal structure only match in a few cases

Eck et al.

Epitope data

Figure 36: Crystal structure of Adalimumab fab region in complex with TNF-α. The predicted receptor binding sites (a) by Eck et al. are highlighted in red as well as the overlapping epitope segments identified during this work (b). It is clearly visible, that this crystal structure model shows both receptor binding sites and epitope segments of the opposite site of TNF-α. Therefore the structure does not display the Adalimumab and TNF interaction under in vitro conditions.

Experimental Part

Materials and Reagents

All commercially available reagents used in this work were of either analytical grade or above if not stated otherwise. All buffers were filtered through a 0.2 µm filter and degassed by sonication under vacuum prior to each use. Stock solutions of antibodies and proteases were aliquoted and stored at -20◦C to avoid freeze-thaw cycles.

Proteolytic enzymes and reagents used for proteolysis.

TPCK treated sequencing grade trypsin and sequencing grade α-chymotrypsin were purchased from Promega Cooperation (USA). Human recombinant TNF-α soluble fraction (77-233) was purchased from Abcam.

Reagents for Edman sequencing.

All reagents and solvents were of sequencing grade and purchased from Merck (Darm- stadt, Germany). All reagents were stored at 4◦C prior to use. Opened bottles were stored under inert nitrogen atmosphere. Neat reagents were stored over a molecular sieve type 3A.

MALDI matrices and other mass spectrometry substances

α-cyano-4-hydroxycinnamic acid (CHCC) was purchased from Sigma-Aldrich. 2,5- dihydroxybenzoic acid (DHB) was purchased from Bruker Daltonics. MS calibration solution (NaI 2 µg/µl and CsI 50 ng/µl in 50/50 isopropanol/water) was purchased from Sigma-Aldrich.

Reagents and solvents for SPR measurements

Gold coated quartz chips for the SPR biosensor were purchased from Reichert Tech- nologies. 1-ethyl-3-(3-dimethylaminopropyl)carbodiimid (EDC) and N-hydroxysuccinimide (NHS) and ethanolamine were purchased from Sigma-Aldrich and of synthesis grade.

16-mercaptohexadecanoic acid was purchased from Sigma-Aldrich.

Materials and reagents for affinity columns

CNBR-activated Sepharose 4B was purchased from GE Healthcare. Pre-packed pro- tein G sensor cartridges, PEEK columns and POROS material were from Applied Biosystems, Thermofisher Scientific (Darmstadt). Column frits with 5 µm pore size were purchased from MoBiTec (Goettingen, Germany). Dimethyl pimelimidate (DMP) was purchased from Sigma-Aldrich.

Other Reagents

Coomassie Brilliant Blue R-250, tetramethylethylenediamine (TEMED) and glycine were purchased from Sigma-Aldrich. Rotiphorese Gel 30 was purchased from Carl Roth (Karlsruhe, Germany).

Proteolytic digestion of TNF-α

The endoproteases used in the present work for digestion of proteins display different specificities for the cleavage of peptide bonds. The serine-type endoproteases such as trypsin, Lys-C and chymotrypsin cleave peptide bonds in their reactive centre via a mechanism involving three specific amino acid residues known as the catalytic triad (His-57, Asp-102 and Ser-195). After processing of proteolytic products, peptide mix- tures can be analyzed with different methods (e.g. MS; LCMS, Edman Sequencing) to obtain structural information.

Prior to digestion, TNF-α was reduced and alkylated: 20 ug of human recombinant TNF-α (77-233 soluble fraction, Abcam) was dissolved in 25 µl of NH4HCO3 buffer (50 mM, pH 8.0) and reduced by adding 1.5 µl of reduction buffer (100 mM DTT in MilliQ). The sample was then incubated at 60◦C for 45-60 min. After the sample had cooled down to room temperature, 3 µl of freshly prepared alkylation buffer (100 mM iodoacetamide, in 50 mM NH4HCO3, pH 8.0) was added and the sample incubated in the dark for 45 min. The alkylation reaction was stopped by addition of 3 µl of reduction buffer. Freshly prepared trypsin or chymotrypsin solution was added to a final enzyme to substrate ratio of 1:10 and the sample digested for 2 h (chymotrypsin) or 3 h (trypsin) at 37◦C. Afterwards, a 1 µl aliquot was taken and analysed by MALDI-ToF MS to control the yield of the digestion. If the digestion procedure was found to be complete, the lysis reaction was stopped by addition of 5 µl 0.1% TFA.

Proteolytic digestion of Adalimumab

Antibodies are highly stable biomolecules with long life times in human blood. Adal- imumab is known to be stable for up to seven days at room temperature in clotted blood samples. Although this stability is highly desirable for treatment purposes, it makes proteolytic digestion challenging. For this reason a special protocol was devel- oped.

Adalimumab (HumiraQR , Abbvie) was supplied by Professor Rovero from the Depart-

ment of Neurosciences, Psychology, Drug Research and Child Health – Section of Pharmaceutical Sciences and Nutraceutics of the University of Florence (Italy).

In-gel digestion

For peptide mapping of Adalimumab, separation of the heavy and light chain prior to digestion was achieved by SDS-PAGE after reduction and alkylation. 800 µg of Adalimumab were dissolved in 88 µl of 50 mM NH4HCO3 (pH 8.0) with 2 M guanidine HCl to facilitate denaturation of any three-dimensional structures. The sample was then incubated at 95◦C for 15 minutes followed by reduction at 60◦C through addition of 7.2 µl of 1 M DTT (in MilliQ) for 1 h. The denatured and reduced sample was then allowed to cool to room temperature and 10 µl of alkylation buffer (see above) was added. The alkylation step was carried out in the dark for 1 h after which the reaction was quenched by adding 10 µl of 1 M DTT. To remove excess salt before gelelectrophoresis, the sample was desalted using Amicon Ultra 3K cutoff centrifuge filters. The sample was washed three times with 50 mM NH4HCO3 and diluted to a final concentration of 200 µg/µl.

SDS-PAGE (polyacrylamide-gel electrophoresis) was carried out according to the method of Laemmli [57]. Sodium dodecyl sulfate (SDS) is an ionic detergent, that intercalates into hydrophobic parts of proteins and disrupts compact folded struc- tures, causing an overall negative charge on proteins. Therefore proteins to be elec- trophoresed migrate towards the anode, allowing separation of proteins and peptides by size in the acrylamide gel. Gels were cast using a MiniProtean kit (Bio-Rad Lab- oratories, Hercules, USA). For one-dimensional SDS-PAGE, 10 µl of reduced and alkylated sample was mixed with 10 µl of 5x sample buffer. The samples were then electrophoresed in a 15% resolving gel (10% stacking gel) at 30 mA for 2-3 h. The gels were then washed with MilliQ water and stained using using the Coomassie staining procedure. The dye Coomassie Brilliant BlueR-250 binds stoichiometrically to pro- teins, allowing for detection of proteins as blue bands (or at 595 nm). Gels were first shacken with Coomassie staining solution (50% methanol, 10% acetic acid, 0.05% Coomassie Brilliant Blue R250) for 5-10 min at room temperature. For higher con- trast, gels were destained (50% methanol, 10% acetic acid in water) by shacking for 30 to 60 min. Bands of the heavy and light chain were excised from the gel and cut into small pieces (less than 1 mm), heavy and light chain pieces pooled together and washed with MilliQ water for 30 min two times while shaking. The gel pieces were then destained by incubation with 50% acetonitrile, 50% 50 mM NH4HCO3 buffer and shrunk by incubation with neat acetonitrile for 30 minutes followed by drying the pieces in a vacuum centrifuge. For digestion, the gel pieces were rehydrated with digestion buffer and incubated with either trypsin or chymotrypsin (10 µg each)

at 37◦C overnight. Peptides were extracted three times with 50% acetonitrile, 50%

50 mM NH4HCO3 buffer, 0.1% TFA and once with neat acetonitrile. The samples containing the extracted peptides stored at -20◦C.

In-solution digestion

Preparation of larger amounts of digested protein for epitope analysis, in-solution di- gestion was performed. Protein denaturation, as well as reduction and alkylation was carried out as described above. Aliquotes were taken after the completed alkylation reaction and verified by MALDI-TOF MS for quantitative reduction and alkylation reaction.

The sample was then diluted with 50 mM NH4HCO3 buffer (pH 8.0) until the guanidine- HCl concentration was below 1 M. Digestion was initiated by addition of 1:10 protease to protein ratio and samples incubated overnight at 37◦C while shaking. The prote- olysis was stopped by addition of 10 µl 0.1% TFA and samples either processed im- mediately or stored at -20◦C. For direct mass spectrometric analysis, samples where cleaned-up by ZipTip procedure using C18 ZipTip pipette tips (Merck Millipore, Darmstadt). The stepwise procedure consists of a. – wetting of the reverse-phase matrix pipette tip; b. – preequilibration and the pipette tip and binding of proteins and peptides to the matrix; c. – desalting and d. – elution of retained analyte.

Edman sequencing of Adalimumab

Determination of the initial N-terminal amino acid structure of Adalimumab was performed by automated Edman sequencing on an Applied Biosystems Model 494 Procise Sequencer (connected to a Microgradient System and a 785A Programmable Absorbance Detector).

Edman degradation is performed as a three step chemical degradation process (Fig. 1), by which the first N-terminal amino acid of a polypeptide chain is chemically cleaved as anilinothiozolinone derivate (ATZ-amino acid). The ATZ amino acid is selectively extracted after cleavage using ethyl acetate and butyl chloride and is con- verted to the more stable phenylthiodantion amino acid (PTH-amino acid) by treat- ment with aqueous TFA under heating at 64◦C for 9 min. Oxidation of PTH derivates during conversion is minimized by addition of 0.01% DTT. PTH amino acids absorb strongly at 269 nm, allowing for separation and detection using RP-HPLC.

For separation, an Applied Biosystems PTH-C18-column (220×2.1 mm, 5 µm) was used. The solvent system for the separation employed was 3.5% THF in water (sol- vent A) and 12% isopropanol in acetonitrile (solvent B). The separation gradient was run from 0% to 50% solvent B over 22 min at a flow rate of 325 µl min-1. Before the analysis was run, the column was pre-equilibrated by running the gradient at least two times over the column. By comparison of the retention time of every Edman degradation product with the retention time of the PTH derivative standards (see chromatogram fig. 2), identification of the N-terminal single amino acid sequence is possible.

Figure 37: Edman degradation of polypeptide chains and conversion of degradation derivates to PTH-amino acids.

Figure 38: Chromatogram of PTH-amino acid standard separation by RP-HPLC at 269 nm.

For sequencing, Adalimumab was buffer exchanged into 50 mM NH4HCO3 and applied as 5 µl aliquots (containing 10-20 pmol of protein) on a precycled glass fibre filter (Applied Biosystems) followed by drying with a stream of argon gas. The filters were then introduced into the corresponding sequencing glass cartridge. Analysis of the amino acid sequence was determined using the pulsed liquid method. The pulsed liquid method injects a small aliquot of liquid TFA onto the glass fibre filter for cleavage of the phenylisothiocyanate-protein after the coupling step. This method allows for faster run times and provides better repetitive yields compared to the gas- phase method which is better suitable for smaller proteins and peptide due to reduced wash-out of analyte from the filter.

Mass spectrometric methods

Mass spectrometry became the established method in the field of analytical pro- teomics. Due to low analyte consumption paired with fast analysis times, mass spec- trometry was developed more and more ever since its first discovery at the end of the 19th century. The major breakthrough for biopolymer analysis, such as proteins and polypeptides, is represented by the development of the soft ionization techniques: Ma- trix assisted laser desorption ionization (MALDI) and electrospray ionization (ESI). Both ionization types are usually coupled to different types of mass analsers and detec- tors, the most frequent types being time of flight (ToF), ion traps (IT) or quadrupole analyser. However, also more powerful analysers such as fourier transform ion cy- clotron resonance (FT-ICR) and orbitrap are coupled frequently, improving the ana- lytical capabilities through increased sensitivity and resolution.

The general schematics for a mass spectrometer is represented in figure 3; consist- ing of an ion source, a mass analyzer and a mass detector. The analyte is introduced into the mass spectrometer via the ion source, where ionization of the sample occurs, transferring the ions into the gas phase. The potential difference between U1 and U2 attracts the ions towards the interior of the MS. Ion optices then guide the ionized molecules into the mass analyzer where mass determination occurs followed by detec- tion of the ions in the mass detector (e.g. electron multiplier). The electronic signals are then processed and represented in a mass spectrum.

U1

Rough vacuum High vacuum

Source

Ion Optics

Mass Analyzer

Detector

Figure 39: Schematic representation of a basic mass spectrometer configuration. The sample is introduced via the source, transferred by ion optics (rough fore-vacuum) into the mass analyser (high vacuum) and afterwards to the detector for mass detection.

Matrix assisted laser desorption ionization mass spectrometry

MALDI-mass spectrometry was the preferred method for mass spectrometric analy- sis during this thesis due to its high sensitivity, broad mass range coverage and low sample consumption. For analysis with MALDI-MS, samples are mixed with an or- ganic acidic matrix and applied to a metal target plate. The matrix is required for better absorption of the laser radiation, in order to accumulate the energy from the laser more efficiently on the analyte, and therefore improve the ionization process. Ionization occurs through bombarding the analyte with short, pulsed lasers. During the laser strike, the matrix is rapidly heated, as the matrix molecules re-emitt the absorbed energy from laser, causing sublimation of the crystallized sample. Whether the actual ionization of the analyte occurs through proton transfer in the acidic ma- trix or by gas-phase proton transfer in the expanding gas plume upon heating through the laser is not yet fully understood. MALDI mass spectrometry usually leads to the formation of singly or doubly charged ions.

MALDI-ToF analysis was carried out on a Bruker Autoflex III (Bruker Daltonics, Bremen, Germany) equipped with a Nd:YAG laser source (4000 resolving power at 3200 Da in linear mode and 11000 resolving power below 3500 Da in reflectron mode). For acquisition of peptide spectra, α cyano-4-hydroxycinnamic acid (CHCC) was used as organic acid, while for acquisition of protein spectra, 2,3-dihydroxybenzoic acid (DHB) was used. CHCC matrix was prepared as a saturated solution in methanol and prepared fresh prior to every analysis by mixing one third of saturated matrix solution with one third 80% acetontrile , 0.1% TFA and one third 0.1% TFA. DHB matrix solution was prepared as a saturated solution of DHB in 70% acetonitrile. Aliquots of 0.5 µl sample solution were spotted with 1 µl of matrix solution mix on the metal target plate and allowed to dry at room temperature. Spectra were acquired with a set laser intensity between 50-70% and 1000 laser shots at 50 Hz per spectrum.

Electrospray ionization mass spectrometry

ESI is an atmospheric pressure ionization method, which produces intact ionized gas phase molecules from an analyte dissolved in solution. The sample is usually intro- duced into the ion source through infusion with syringes or a liquid chromatography system providing flow rates in the µl min-1 range. The ESI source consists of a cap- illary positioned in front of the mass spectrometer inlet. Between the capillary and the inlet, a potential difference in the low kV range is applied, which causes the in- jected analyte solution to form the Taylor cone. At the apex of this cone, the liquid stream breaks into smaller droplets when the Rayleigh limit is reached, meaning that the electrostatic repulsion of the particles exceeds the surface tension of the liquid stream. The droplets, with a diameter in the micrometer range, are attracted to the capillary entrance of the MS. On their flight path, the charge density on the surface of the micro-droplets increases further until the Rayleigh limit is reached, inducing Coulomb fission, splitting them into smaller, stabler droplets. The formation of gas- phase ions from solvated ions is currently still discussed, but two main theories have been proposed: According to the ion evaporation model, the solvated ions desorb from the charge droplets in the electric field and then enter the MS via the entrance capil- lary. The charge residue model proposes that Coulomb fission causes small droplets with only one solvated ion. The charge of the droplet is then transferred from the solvent to the molecule and the remaining solvent evaporates. By heating the ion source or addition of a dry gas flow as well as nebulizer gas, droplet formation and solvent evaporation may be accelerated. As a result, ESI produces mainly multiply charged ions, which can be further analysed in the MS.

ESI-MS analysis was carried out using a Waters Micromass Quattro Ultima (Wa- ters, Milford, USA) instrument. The mass spectra were acquired in the positive ion mode, scanning between 100-3000 m/z. Source parameters were tuned prior to each run to increase detection depending on the analyte and the experimental conditions. The ion source paramters were: nebulizing gas pressure , drying gas flow maximum, capillary voltage -3.4 kV, end plate offset 500 V and source temperature 150◦C. Ex- ternal calibration was carried out using an NaI/CsI ESI-tuning calibration standard.

Affinity based methods

Determination of affinity constants by SPR biosensor

The detection principle for an SPR biosensor is depicted in figure 4. This type of biosensor belongs to the class of optical biosensor, which measures the interaction of light with surface-adsorbed or immobilized molecules, allowing quantification and analysis of biomolecular interactions (such as antigen-antibody interaction) with real- time monitoring of changes in the surface-chemistry. In BIA-measurements, one in- teracting partner is immobilized on the chip while the other is passed over the surface in solution. Covalently immobilized affinity ligands on the sensing surface are used to capture the target protein or peptide from solution, thus altering the composition on the surface. In prism configured SPR biosensors (Kretschmann-arrangement), a polarized light beam undergoing total internal reflection at a solution-metal-prism interface causes optical excitation of metal electrons, thus causing propagation of these electrons parallel to the surface. These moving electrons called plasmons (elec- tromagnetic waves), can give rise to surface plasmon resonance (SPR). SPR occurs when the momentum of the incoming photons matches the momentum of the surface plasmons at a specific angle of the incident light, resulting in an electron vacancy among the metal atoms and energy transfer. The momentum of the plasmons, and thus the electric field, is highly sensitive to the composition of the medium above the metal surface. Changes in biomolecule configuration on the sensing surface, imposed by affinity interactions between molecules in solution and the modified metal surface, therefore require a change in the angle of the incoming light in order for SPR to occur. Consequently, the angle-intensity spectrum changes with physiochemical composition of the surface of gold-quartz chip.

Measurements of bioaffinity were performed with a two channel SR7500DC sur- face plasmon resonance (SPR) biosensor (Reichert Technologies Life Sciences, Buffalo, USA). KD determination was performed using either a commercial activated-dextran gold chip (Reichert Technologies, Buffalo, USA) or a freshly prepared self-assembled monolayer (SAM) chip. The SAM chips were prepared by immersing a plain gold chip in 10 mM 16-mercaptohexadecanoic acid in chloroform overnight at room tempera- ture while shaking. The chips were then rinsed thoroughly with ethanol and water and dried using a stream of air. The chips used for bioaffinity measurement were inserted into the instrument and the channels rinsed with running buffer (PBST, 10 mM Na2HPO4, 10 mM NaCl, 0.2% TWEEN, pH 7.5) at 250 µl min-1 to remove air and unspecifically adsorbed contaminants from the surface. For all following steps, the flow rate was reduced to 25 µl min-1. All solvents and buffers used were freshly prepared each time and degassed with sonication under vacuum. Carboxyl groups of the surface were activated using EDC/NHS chemistry: Activation with 1-ethyl-3-

Figure 40: Kretschmann configured surface plasmon resonance biosensor: The change in angle of a reflected, polarized light beam is detected by an optical detector, which is recorded by the biosensor instrument.

(3-dimethylaminopropyl)carbodiimid (EDC) and N-hydroxysuccinimide (NHS) was performed by injecting 250 µl of a 1:1 mixture of 250 mM EDC and 100 mM NHS for 10 min over the surface of the chip. Antibody immobilization was performed by injecting 33 µg of antibody dissolved in 10 mM NaOAc buffer (pH 5.2) over the left channel of the chip for 7 min, yielding the analyte channel, while the right channel was set as reference channel without any immobilized ligand. Unreacted succinate ester groups were then capped on both channels with 1 M ethanolamine (pH 8.5).

Y Y Y Y

Equilibrium

Y Y Y Y Y Y Y Y

Association

Dissociation

Y Y Y Y

Regeneration

Figure 41: Workflow of SPR measurements for determination of affinity interactions.

Figure (4) describes a typical cycle for an SPR experiment carried out during this work. Upon analyte injection, a. – the sample starts associating with the covalently immobilized affinity ligand on the gold chip until an equilibrium point is reached. At this point, the association and dissociation rates occur equally fast. The kon value is recorded during this phase. Afterwards, b. – the sample starts dissociating from the affinity ligand (measurement of koff and the chip surface can be regenerated – c, prior to the next sample injection – d.

For measurement of association and dissociation constants, a series of increasing sam- ple concentrations in PBST (10 mM, pH 7.5) were injected over the chip with the immobilized affinity ligand at 22◦C. Association and dissociation times were opti- mized for every different affinity pair. Regeneration of the chip surface was achieved by injection of 50 mM glycine (pH 2.5) for 30 sec and two times 1 mM NaOH (pH

9) for 30 sec. After every regeneration step, a buffer blank was injected. Recorded sensorgrams were evaluated using TraceDrawer Data Analysis software (Ridgeview In- struments AB,Vange , Sweden).) according to a 1:1 or 1:2 Langmuir binding model. The reference channel was substracted from the measurement channel during calcu- lation of the KD values, removing signals from non-specific interactions.

Online SPR-MS combination for epitope determination

Bioaffinity analysis with biosensors such as surface acoustic wave (SAW) or surface plasmon resonance (SPR) are well established methodologies for quantification of biomolecular interactions. However, the lack of structural information, which are not accessible with this method alone, is the main limiting factor for such analysis. In this work, the transfer of all required steps for an epitope identification experiment into a fully automated procedure is described, incooperating for the first time a fully automated digestion step. By combining an automated proteolytic digestion system with an online switching and desalting interface, enabling sample redirection either to an SPR-biosensor or ESI-MS, a fully automated, online system for epitope identi- fication was developed.

Affinity column

Epitope extraction

SPR-Biosensor

KD determination

ESI-MS

Structure Identification

Figure 42: Online SPR-MS system with integrated automated digestion system.

Instrumentation and microfluidic connections

Figure 6 represents a schematics of the instrumental combination and figure 7 dis- plays the details of the microfluidic interface connecting the instruments. Automated proteolytic digestion was performed using a DigestPro96 (Intavis AG, K¨oln) system, capable of performing both fully automated in-gel-digestion and in-solution digestion of proteins and peptides. The microfluidic system of the digestion robot was con- nected to one of two Rheodyne six-port, two position valves (Rheodyne LLC, USA) for transfer of the samples to the the online switching and desalting interface. The pipet robot of the DigestPro (Gilson Inc, USA) was reprogrammed with the Digest- Pro control software (Intavis AG, Ko¨ln) using ”Pipet ”, ”Aspirate” and ”Dispense” commands to allow autosampler functionality of the pipet robot. Samples for injec- tion onto either the SPR sensor chip or the switching interface were drawn up into a sample loop with an ”Aspirate” command followed by injection via a ”Dispense” command. The sequence of command steps was saved as a complete program and executed for experiments. An HPLC pump was connected to the Rheodyne valves providing easy buffer exchange and precise control of flow rates. The SPR biosensor (SR7500DC SPR biosensor, Reichert Technologies, USA) was connected to the first Rheodyne valve, for determination and quantification of affinity interactions, while the ESI-MS system (Micromass Quattro Ultima, Waters Coperation, USA) was con- nected to the second valve, after the desalting interface. The affinity column for epitope extraction was connected at the first valve. Online desalting of samples was performed using an OptiGuard C18 column (1.5 cm x 1 mm, 40 µm particle size). Connections were made with PEEKTM tubing (inner diameter: 0.25 mm).

Experimental details for online epitope extraction

Prior to each experimental step, all microfluidic connections were purged with the corresponding buffer or solvent to ensure defined experimental conditions. Sample in- jection of the analyte was performed by the reprogrammed DigestPro96 system. The Rheodyne valve was switched to injection position (valve position 2, fig. 7) and the analyte was directly injected onto the SPR chip for KD measurement. SPR experi- mental conditions were set as described above (chapter 3.6.1 Determination of affinity constants by SPR biosensor ). At the same time, the samples were automatically re- duced, alkylated and digested using sequencing grade proteases in the DigestPro96. All reagents for proteolytic digestion were prepared according the manufacturers in- structions. Digestions were carried out either in PBST buffer (10 mM, pH 7.5) or ammonium bicarbonate buffer (50 mM, pH 7.8). 20 µg of protein or peptide were digested using trypsin. After completion of the digestion, the digested samples were

injected via the first Rheodyne valve onto an affinity column with immobilized anti- bodies for epitope extraction (valve position 1, fig. 7)(for details of affinity column preparation see chapter 3.6.3 Preparation of antibody affinity columns using CNBr- activated Sepharose). After washing away unspecifically retained peptides, the affinity captured epitope peptide was eluted from the column at pH 2 to a C18 guard column for desalting (valve position 2). The guard column was washed with MilliQ and 5% acetonitrile to remove ESI-MS incompatible salts from the analyte (valve position 2) and the flow through directed to the waste. Elution from the RP-guard column was performed with 75% acetonitrile and the eluted peptides directed into the ESI-source of the mass spectrometer (valve position 1). Simultaneously, the affinity column was re-equilibrated with PBST buffer (valve position 1). After acquisition of MS spectra, the C18 column was washed extensively with acetonitrile before injection of the next sample.

The full system was set-up during the work of this thesis and theoretical parts eluci- dated, however, due to the limited time frame of this scientific work, the system was not yet tested in practise.

DigestPro

ESI-MS

HPLC

Biosensor Waste

DigestPro

ESI-MS

HPL

Biosensor Waste

Figure 43: Online SPR-MS system with integrated automated digestion system.

Preparation of antibody affinity columns using CNBr-activated Sepharose Antibody Immobilization

Antibodies were coupled to to Sepharose beads via the CNBr-coupling strategy. For this, CNBr-activated Sepharose was reacted directly with antibodies, resulting in the formation of a stable covalent amide bond between accessible amino-groups (e.g. from Lysine residues) and the CNBr functional group of the Sepharose material.

CNBr-activated Sepharose was preactivated by washing with 1 mM HCl (500 µl) after the resin was allowed to swell for 15 min at room temperature. Afterwards, the resin beads were centrifuged down at 1000 x g for 4 min and the supernatent dis- carded, followed by vortexing once more with 500 µl of 1 mM HCl and centrifugation. Afterwards, the beads were equilibrated in coupling buffer (100 mM NaHCO3, 500 mM NaCL, pH 8.3) by repeated centrifugation and vortexing steps as above. The equilibrated beads were then incubated with antibody diluted in coupling buffer (66.5 mg beads with 250 µg antibody) at room temperature for 2-3 h while shaking vigor- ously. The supernatent was kept after centrifugation of the beads in order to calculate the total immobilized amount of antibodies: The OD280 of the total supernatent was measured and using an IgG dilution series, through interpolation, the immobilized amount of antibodies was determined.

Blocking of unreacted CNBr-groups was performed by incubation with 1 M ethanolamine- HCl (pH 8.5) for 1.5 to 2 h at room temperature with shaking. The beads were washed once more with 500 µl coupling buffer, followed by a series of washing steps in alter- nating order: three times with high pH washing buffer (100 mM NaHCO3, pH 8.0) and three times with low pH washing buffer 100 mM NaOAc, pH 4.0. The material was immediately processed for packing in a column.

Column Packing

Sepharose material with immobilized antibodies was loaded into a POROS Prep Self Packing Device connected to POROS Peek Cartridge, whose total volume of 100 µl was reduced by addition of Peek filter material inside the cartridge to a final volume of 50 µl. The packing device was filled with running buffer (10 mM Na2HPO4, 150 mM NaCl, pH 7.5), which was also used for packing. The packing device and cartridge were connected to a Waters Alliance HPLC system and the column packed with an initial flow rate of 20 µl/min. The flow rate was kept constant, until no further increase in pressure was observed. After the column pressure had equilibrated, the flow rate was increased in increments of 20 µl/min until a final flow rate of 140 µl/min

was reached (final pressure: 2.4 bar).

The packed column was stored short term in running buffer and long term in running buffer with 0.02 % sodium azide at 4◦C.

Preparation of antibody affinity columns by the crosslinking ap- proach

Protein G is an immunoglobine-binding protein expressed by streptococcal bacteria, that binds specifically to the FV parts of IgG class antibodies with very high affinity. By crosslinking with a bi-functional reagent such as dimethyl pimelimidate (DMP), stable affinity columns can be generated. DMP is a homobifunctional crosslinking reagent, containing two imidate groups which react readily during a substitution re- action with free amino groups of lysine residues, resulting in the formation of a stable covalent linkage between the FC part of IgG antibodies and protein G. Thus, using this approach, it is possible to perform an enrichment of IgG fractions from whole patient serum and preparing an affinity column simultaneously.

A POROS G sensor cartridge containing covalently immobilized protein G was equilibrated with running buffer (10 mM NaHPO4, 150 mM NaCl, pH 7.5) was equi- librated for 10 minute at a flow constant flow rate of 1 ml min -1. Sera from pediatric patients were provided by the department of pediatric rheumatology, AOU Meyer, Florence, Italy and stored refridgerated until use. Prior to application to the protein G cartridge, sera samples were centrifuged at 4000 x g for 10 min and the supernatant diluted 1:100 with running buffer. For affinity capture of IgG antibodies, 2.5 ml of diluted serum was injected in increments of 500 µl to the cartridge at a flow rate of 0.5 ml min-1. Flow through was monitored at 280 nm using a UV/VIS detector (Dionex) and collected. Each flow through fraction was injected once more over the sensor cartridge. In order to blockfree protein G still remaining in the cartridge, 500 µl of 1:10 diluted human IgG fraction (Sigma-Aldrich) was injected over the column three times, until the flow through UV response yielded the same signal intensity. In order to permanently crosslink the captured IgG antibodies, 117 mg dimethyl pimelimidate (DMP) dissolved in 14 ml triethanolamine (100 mM, pH 8.5) was injected in incre- ments of 2 ml over the cartridge, at a flow rate of 0.5 ml min-1. The column was then washed for 5 min to remove any remaining crosslinking reagent. Reactive imidate groups of DMP were quenched by injecting 4 ml of ethanolamine (100 mM, pH 9.0) at a flow rate of 0.5 ml min-1. For conditioning the column, both running and elution buffer (100 mM NaCl-HCl, pH 2.5) were run over the column for 5 minutes each, at a flow rate of 1 ml min-1.

Epitope extraction of TNF-α

For the identification of the Adalimumab binding site on TNF-α through prote- olytic extraction mass spectrometry, 10 µg of protein was dissolved in 25 µl of 50 mM NH4HCO3 (pH 8.0) buffer and digested as described earlier (see chapter 3.2 Proteolytic digestion of TNF-α) after reduction and alkylation. Proteases (trypsin or chymotrypsin) were added to yield an enzyme to substrate ratio of 1:10 and the sample incubated for 2-3 h at 37 ◦C. An aliquot of 1 µl was analyzed my MALDI- ToF to check the yield of the digestion. The digestion mixture was then diluted with 100 µl Na2HPO4 buffer and loaded onto an affinity micro-column with immobi- lized Adalimumab (see chapter 3.6.3 Preparation of antibody affinity columns using CNBr-activated Sepharose). The column was closed and incubated overnight at 37◦C while shaking. The supernatent was collected and dried for subsequent MALDI-MS analysis. The column was washed 10x with PBS buffer (500 µl each) and 5x with MilliQ water for on-column desalting. Elution of affinity bound fractions was per- formed with 3x 0.1% TFA (500 µl each). The fractions were collected, dried using a vacuum centrifuge and resolubilized in 60% acetonitrile, 0.1% TFA in water for direct MALDI-analysis or in 5% acetonitrile, 0.1% TFA in water for desalting using C18 ZipTips.

Epitope excision of TNF-α

The epitope of Adalimumab on TNF-α was further identified by proteolytic epitope excision: The immobilized Adalimumab was incubated with 30 µg of TNF-α in 10 mM PBS buffer (pH 7.5) under while shaking for 2 h to allows the formation of the immune complex. Afterwards, 3 µg of sequencing grade Trypsin were added to the mixture and the column incubated under shaking at room temperature for 1 h to allow digestion of unshielded protein segments. The column was washed afterwards with 3 ml of PBS buffer and the last washing fraction was checked by MALDI-ToF MS. If no further peptides were found in the washing fraction, the peptides retained by the immune complex were dissociated by addition of 0.1% TFA in MilliQ water and eluted three times with 500 µl. Elution fractions were treated as for epitope extraction (see above) and analyzed by MALDI-ToF MS.

Epitope extraction of Adalimumab

Epitope determination of the anti-Adalimumab antibodies was performed by prote- olytic extraction. Adalimumab was prepared by digestion as described previously. Removal of trypsin from the digestion mixture was performed using Amicon Ultra 10K cutoff filters. The flow through containing the proteolytic peptide mixture was collected and the pH adjusted to 7-8 using NaOH. For epitope extraction, the sample

was diluted using PBS buffer (10 mM Na2HPO4, 150 mM NaCl, pH 7.5) to a final peptide concentration of 1 µg µl-1. 100 µl of this mixture were injected at a flow rate of 50 µl min-1 over the protein G immobilized patient serum in running buffer. Flow through was monitored at 220 nm and the column was washed until absorption reached again the baseline. The supernatent fraction was collected and lyophilized for later analysis. Elution was carried out by switching the running buffer to 150 mM NaCl (pH 2.5) at 100 µl min-1 flow rate. Elution fractions were also collected and either analysed directly by MALDI-ToF MS or concentrated first and desalted with C18 ZipTips prior to MS analysis.

3.7 Software, structural modelling and amino acid sequences

In-silico digests and analysis of proteolytic fragments was carried out using GPMAW (Lighthouse) and mMass. Amino acid sequences of Adalimumab were acquired from drugbank (accession number: DB00051). Amino acid sequence of TNF-α was ac- quired from UniProt (accession number: P01375) as well as lymphotoxin-α (accession number: P01374). Infliximab fab sequences were taken from Shi Hu et al. J. Biol. Chem. 2013;288:27059-27067. Structural models of proteins were generated using PyMOL. Crystal structures for modelling were downloaded from RCSB protein data bank.

Conclusions and outlook

In this work the identification of an epitope sequence on the therapeutic drug anti- body Adalimumab as recognized by treated patient serum is presented. The epitope sequence 12VQPGRSLRLSCAASGFTF29 located on the heavy chain of Adalimumab was identified by bioaffinity mass spectrometry using proteolytic epitope extraction. The epitope recognized by the patient sera was confirmed for two different patients suggesting a uniform epitope.

Further tests including epitope excision mass spectrometry and sampling further pa- tients need to be conducted. These results should confirm the findings of these exper- iments. The information obtained could be used for drug design not only for re-design of another Adalimumab analogous biosimilar drug antibody but information extrapo- lated for the inhibition mechanisms of anti-drug antibodies. As there are many other drug biosimilars on the market, the majority of which are antibodies, knowledge about immunogenicity will become of great importance. SPR binding assays and competi- tive binding assays using TNF-α, Adalimumab and anti-Adalimumab antibodies will be used for confirmation of Adalimumab deaktivation mechanisms. Synthetic epi- tope peptides can be used for characterization and confirmation of the epitope by SPR. Further, analysis of the Adalimumab paratope and comparison with the anti- Adalimumab antibody epitope could allow design of fab drug design and alternative treatment strategies. Apheresis techniques or treatment with epitope peptides for fil- tering and deactivating anti-Adalimumab antibodies in patients blood seem unlikely strategies at this stage as two other drugs, Infliximab and Etanercept are also avail- able for treatment of rheumatoid arthritis.

As the interaction between Adalimumab and the target protein TNF-α are only known from crystal structure data, representing non-in vivo conditions, epitope extraction was also performed for TNF-α under in vitro conditions. Two peptides were identified: 33ANALLANGVELR44 and 37LANGVELRDQLVVPSEGLYLIY59. Epitope excision

revealed the N-terminal peptide 7TPSDKPVAHVVANPQAEGQLQWLNRR32. The epitope was located at the predicted receptor binding sites explaining the biochemi- cal mechanism of Adalimumab: Adalimumab binds to the TNF-receptor binding site blocking the receptor interaction and thus reducing inflammatory responses in pa- tients.

Confirming the epitope will be performed by SPR kinetic analysis of candidate epi- tope peptides. To complete the study of the Adalimumab – TNF-α interaction, the Adalimumab paratope should be identified. This will be carried out by immobilizing TNF-α and digestion of Adalimumab.

References

G. Feuerstein, T. Liu, and F. Barone, “Cytokines, inflammation, and brain in- jury: role of tumor necrosis factor-alpha.,” Cerebrovascular and brain metabolism reviews, vol. 6, no. 4, pp. 341–360, 1993.

T. Hehlgans and K. Pfeffer, “The intriguing biology of the tumour necrosis fac- tor/tumour necrosis factor receptor superfamily: players, rules and the games,” Immunology, vol. 115, no. 1, pp. 1–20, 2005.

D. Hu¨rlimann, A. Forster, G. Noll, F. Enseleit, R. Chenevard, O. Distler,

M. B´echir, L. E. Spieker, M. Neidhart, B. A. Michel, et al., “Anti–tumor necro- sis factor-α treatment improves endothelial function in patients with rheumatoid arthritis,” Circulation, vol. 106, no. 17, pp. 2184–2187, 2002.

G. S. Firestein, “Evolving concepts of rheumatoid arthritis,” Nature, vol. 423, no. 6937, pp. 356–361, 2003.

M. E. Weinblatt, E. C. Keystone, D. E. Furst, L. W. Moreland, M. H. Weisman,

C. A. Birbara, L. A. Teoh, S. A. Fischkoff, and E. K. Chartash, “Adalimumab, a fully human anti–tumor necrosis factor α monoclonal antibody, for the treatment of rheumatoid arthritis in patients taking concomitant methotrexate: the armada trial,” Arthritis & Rheumatism, vol. 48, no. 1, pp. 35–45, 2003.

Y. Komaki, A. Yamada, F. Komaki, P. Kudaravalli, D. Micic, A. Ido, and

Sakuraba, “Efficacy, safety and pharmacokinetics of biosimilars of anti-tumor necrosis factor-α agents in rheumatic diseases; a systematic review and meta- analysis,” Journal of Autoimmunity, 2017.

Y. Berkun, D. Levartovsky, A. Rubinow, H. Orbach, S. Aamar, T. Grenader,

A. Atta, D. Mevorach, G. Friedman, and A. Ben-Yehuda, “Methotrexate re- lated adverse effects in patients with rheumatoid arthritis are associated with the a1298c polymorphism of the mthfr gene,” Annals of the rheumatic diseases, vol. 63, no. 10, pp. 1227–1231, 2004.

P. J. Anderson, “Tumor necrosis factor inhibitors: clinical implications of their different immunogenicity profiles,” in Seminars in arthritis and rheumatism, vol. 34, pp. 19–22, Elsevier, 2005.

F. B. Vincent, E. F. Morand, K. Murphy, F. Mackay, X. Mariette, and C. Mar- celli, “Antidrug antibodies (adab) to tumour necrosis factor (tnf)-specific neutral- ising agents in chronic inflammatory diseases: a real issue, a clinical perspective,” Annals of the rheumatic diseases, vol. 72, no. 2, pp. 165–178, 2013.

D. Ternant, E. Ducourau, P. Fuzibet, C. Vignault, H. Watier, T. Lequerr´e,

X. Le Lo¨et, O. Vittecoq, P. Goupille, D. Mulleman, et al., “Pharmacokinetics and concentration–effect relationship of adalimumab in rheumatoid arthritis,” British journal of clinical pharmacology, vol. 79, no. 2, pp. 286–297, 2015.

P. A. van Schouwenburg, S. Kruithof, G. Wolbink, D. Wouters, and T. Rispens, “Using monoclonal antibodies as an international standard for the measurement of anti-adalimumab antibodies,” Journal of pharmaceutical and biomedical anal- ysis, vol. 120, pp. 198–201, 2016.

F. Real-Fern´andez, R. Cimaz, G. Rossi, G. Simonini, T. Giani, I. Pagnini, A. M. Papini, and P. Rovero, “Surface plasmon resonance-based methodology for anti- adalimumab antibody identification and kinetic characterization,” Analytical and bioanalytical chemistry, vol. 407, no. 24, pp. 7477–7485, 2015.

A. K. Abbas, A. H. Lichtman, and S. Pillai, Cellular and molecular immunology. Elsevier Health Sciences, 2011.

G. J. Nossal, “The double helix and immunology,” Nature, vol. 421, no. 6921, pp. 440–444, 2003.

J. R. Rodgers and R. R. Rich, “Molecular biology and immunology: an intro- duction,” Journal of allergy and clinical immunology, vol. 88, no. 4, pp. 535–551, 1991.

M. Przybylski and M. O. Glocker, “Electrospray mass spectrometry of biomacro- molecular complexes with noncovalent interactions—new analytical perspectives for supramolecular chemistry and molecular recognition processes,” Angewandte Chemie International Edition in English, vol. 35, no. 8, pp. 806–826, 1996.

S. J. Coales, S. J. Tuske, J. C. Tomasso, and Y. Hamuro, “Epitope mapping by amide hydrogen/deuterium exchange coupled with immobilization of anti- body, on-line proteolysis, liquid chromatography and mass spectrometry,” Rapid Communications in Mass Spectrometry, vol. 23, no. 5, pp. 639–647, 2009.

U. Reineke and M. Schutkowski, Epitope mapping protocols, vol. 1. Springer, 2009.

S. Gerstenbruch, C. L. Brooks, P. Kosma, L. Brade, C. R. MacKenzie, S. V. Evans, H. Brade, and S. Mu¨ller-Loennies, “Analysis of cross-reactive and spe- cific anti-carbohydrate antibodies against lipopolysaccharide from chlamydophila psittaci,” Glycobiology, vol. 20, no. 4, pp. 461–472, 2010.

B. E. Bol´ıvar and J. T. Welch, “Studies of the binding of modest modulators of the human enzyme, sirtuin 6, by std nmr,” ChemBioChem, 2017.

S. Dey, A. Nandy, S. C. Basak, P. Nandy, and S. Das, “A bioinformatics approach to designing a zika virus vaccine,” Computational Biology and Chemistry, vol. 68, pp. 143–152, 2017.

D. Suckau, J. K¨ohl, G. Karwath, K. Schneider, M. Casaretto, D. Bitter- Suermann, and M. Przybylski, “Molecular epitope identification by limited pro- teolysis of an immobilized antigen-antibody complex and mass spectrometric peptide mapping,” Proceedings of the National Academy of Sciences, vol. 87, no. 24, pp. 9848–9852, 1990.

M. Macht, W. Fiedler, K. Ku¨rzinger, and M. Przybylski, “Mass spectrometric mapping of protein epitope structures of myocardial infarct markers myoglobin and troponin t,” Biochemistry, vol. 35, no. 49, pp. 15633–15639, 1996.

W. Fiedler, C. Borchers, M. Macht, S.-O. Deininger, and M. Przybylski, “Molec- ular characterization of a conformational epitope of hen egg white lysozyme by differential chemical modification of immune complexes and mass spectrometric peptide mapping,” Bioconjugate chemistry, vol. 9, no. 2, pp. 236–241, 1998.

M. Al-Majdoub, K. F. Opuni, Y. Yefremova, C. Koy, P. Lorenz, R. F. El-Kased, H.-J. Thiesen, and M. O. Glocker, “A novel strategy for the rapid preparation and isolation of intact immune complexes from peptide mixtures,” Journal of Molecular Recognition, vol. 27, no. 9, pp. 566–574, 2014.

J. McLaurin, R. Cecal, M. Kierstead, X. Tian, A. L. Phinney, M. Manea,

French, M. H. Lambermon, A. A. Darabie, M. E. Brown, et al., “Therapeu- tically effective antibodies against amyloid-β peptide target amyloid-β residues 4–10 and inhibit cytotoxicity and fibrillogenesis,” Nature medicine, vol. 8, no. 11, pp. 1263–1269, 2002.

R. Stefanescu, R. E. Iacob, E. N. Damoc, A. Marquardt, E. Amstalden,

M. Manea, I. Perdivara, M. Maftei, G. Paraschiv, and M. Przybylski, “Mass spectrometric approaches for elucidation of antigen–antibody recognition struc- tures in molecular immunology,” European Journal of Mass Spectrometry, vol. 13, no. 1, pp. 69–75, 2007.

A. Moise, S. Andr´e, F. Eggers, M. Krzeminski, M. Przybylski, and H.-J. Gabius, “Toward bioinspired galectin mimetics: identification of ligand-contacting pep- tides by proteolytic-excision mass spectrometry,” Journal of the American Chem- ical Society, vol. 133, no. 38, pp. 14844–14847, 2011.

R. Stefanescu, R. Born, A. Moise, B. Ernst, and M. Przybylski, “Epitope struc- ture of the carbohydrate recognition domain of asialoglycoprotein receptor to a monoclonal antibody revealed by high-resolution proteolytic excision mass spec- trometry,” Journal of the American Society for Mass Spectrometry, vol. 22, no. 1, pp. 148–157, 2011.

A. Wu, R. Valdes, F. S. Apple, T. Gornet, M. A. Stone, S. Mayfield-Stokes, A. M. Ingersoll-Stroubos, and B. Wiler, “Cardiac troponin-t immunoassay for diagnosis of acute myocardial infarction.,” Clinical chemistry, vol. 40, no. 6, pp. 900–907, 1994.

M. Dragusanu, B.-A. Petre, S. Slamnoiu, C. Vlad, T. Tu, and M. Przybylski, “On-line bioaffinity-electrospray mass spectrometry for simultaneous detection, identification, and quantification of protein–ligand interactions,” Journal of the American Society for Mass Spectrometry, vol. 21, no. 10, pp. 1643–1648, 2010.

M. D´ıaz-lobo, S. Guardiola, A. Moise, M. Gay, M. Vilanova, S. Slamnoiu, E. Gi- ralt, M. Przybylski, and M. Vilaseca, “On-line saw-biosensor-mass spectrometry as a powerful tool for studying biological complexes,” The Febs Journal, vol. 282, p. 253, 2015.

G. Treitz, T. Gronewold, E. Quandt, and M. Zabe-Ku¨hn, “Combination of a saw-biosensor with maldi mass spectrometric analysis,” Biosensors and Bioelec- tronics, vol. 23, no. 10, pp. 1496–1502, 2008.

S. Slamnoiu, C. Vlad, M. Stumbaum, A. Moise, K. Lindner, N. Engel, M. Vi- lanova, M. Diaz, C. Karreman, M. Leist, et al., “Identification and affinity- quantification of ß-amyloid and α-synuclein polypeptides using on-line saw- biosensor-mass spectrometry,” Journal of The American Society for Mass Spec- trometry, vol. 25, no. 8, pp. 1472–1481, 2014.

M. Dragu¸sanu, B.-A. Petre, and M. Przybylski, “Epitope motif of an anti- nitrotyrosine antibody specific for tyrosine-nitrated peptides revealed by a combi- nation of affinity approaches and mass spectrometry,” Journal of Peptide Science, vol. 17, no. 3, pp. 184–191, 2011.

Y. Zhao and B. T. Chait, “Protein epitope mapping by mass spectrometry,”

Analytical chemistry, vol. 66, no. 21, pp. 3723–3726, 1994.

M. Macht, A. Marquardt, S.-O. Deininger, E. Damoc, M. Kohlmann, and

M. Przybylski, “” affinity-proteomics”: direct protein identification from bio- logical material using mass spectrometric epitope mapping,” Analytical and bio- analytical chemistry, vol. 378, no. 4, pp. 1102–1111, 2004.

W. W. Cleland, “Dithiothreitol, a new protective reagent for sh groups,” Bio- chemistry, vol. 3, no. 4, pp. 480–482, 1964.

J. A. Burns, J. C. Butler, J. Moran, and G. M. Whitesides, “Selective reduc- tion of disulfides by tris (2-carboxyethyl) phosphine,” The Journal of Organic Chemistry, vol. 56, no. 8, pp. 2648–2650, 1991.

E. B. Getz, M. Xiao, T. Chakrabarty, R. Cooke, and P. R. Selvin, “A com- parison between the sulfhydryl reductants tris (2-carboxyethyl) phosphine and dithiothreitol for use in protein biochemistry,” Analytical biochemistry, vol. 273, no. 1, pp. 73–80, 1999.

N. L. Kelleher, H. Y. Lin, G. A. Valaskovic, D. J. Aaserud, E. K. Fridriksson, and F. W. McLafferty, “Top down versus bottom up protein characterization by tandem high-resolution mass spectrometry,” Journal of the American Chemical Society, vol. 121, no. 4, pp. 806–812, 1999.

G. M. Edelman, “Antibody structure and molecular immunology,” Annals of the New York Academy of Sciences, vol. 190, no. 1, pp. 5–25, 1971.

G. M. Edelman, “Covalent structure of a human γg-immunoglobulin. xi. func- tional implications,” Biochemistry, vol. 9, no. 16, pp. 3197–3205, 1970.

L. R. Rodrigues, J. A. Teixeira, H. C. van der Mei, and R. Oliveira, “Physico- chemical and functional characterization of a biosurfactant produced by lactococ- cus lactis 53,” Colloids and Surfaces B: Biointerfaces, vol. 49, no. 1, pp. 79–86, 2006.

M. Wuhrer, J. C. Stam, F. E. van de Geijn, C. A. Koeleman, C. T. Verrips,

R. J. Dolhain, C. H. Hokke, and A. M. Deelder, “Glycosylation profiling of immunoglobulin g (igg) subclasses from human serum,” Proteomics, vol. 7, no. 22, pp. 4070–4081, 2007.

K. Van Schie, M. Hart, E. de Groot, S. Kruithof, L. Aarden, G. Wolbink, and T. Rispens, “The antibody response against human and chimeric anti-tnf therapeutic antibodies primarily targets the tnf binding region,” Annals of the rheumatic diseases, vol. 74, no. 1, pp. 311–314, 2015.

S. Hu, S. Liang, H. Guo, D. Zhang, H. Li, X. Wang, W. Yang, W. Qian, S. Hou,

H. Wang, et al., “Comparison of the inhibition mechanisms of adalimumab and infliximab in treating tumor necrosis factor-associated diseases from a molecular view.,” The Journal of biological chemistry, vol. 289, no. 31, pp. 21296–21297, 2014.

D. Altschuh, M. C. Dubs, E. Weiss, G. Zeder-Lutz, and M. H. Van Regenmortel, “Determination of kinetic constants for the interaction between a monoclonal antibody and peptides using surface plasmon resonance,” Biochemistry, vol. 31, no. 27, pp. 6298–6304, 1992.

D. J. Oshannessy, M. Brighamburke, K. Soneson, P. Hensley, and I. Brooks, “De- termination of rate and equilibrium binding constants for macromolecular inter- actions using surface plasmon resonance: use of nonlinear least squares analysis methods,” Analytical biochemistry, vol. 212, no. 2, pp. 457–468, 1993.

A. W. Drake, D. G. Myszka, and S. L. Klakamp, “Characterizing high-affinity antigen/antibody complexes by kinetic-and equilibrium-based methods,” Ana- lytical biochemistry, vol. 328, no. 1, pp. 35–43, 2004.

M. J. Slakter, “A comparison of the pearson chi-square and kolmogorov goodness- of-fit tests with respect to validity,” Journal of the American Statistical Associ- ation, vol. 60, no. 311, pp. 854–858, 1965.

Z. Kaymakcalan, P. Sakorafas, S. Bose, S. Scesney, L. Xiong, D. K. Hanzatian,

J. Salfeld, and E. H. Sasso, “Comparisons of affinities, avidities, and complement activation of adalimumab, infliximab, and etanercept in binding to soluble and membrane tumor necrosis factor,” Clinical Immunology, vol. 131, no. 2, pp. 308– 316, 2009.

M. J. Eck and S. R. Sprang, “The structure of tumor necrosis factor-alpha at 2.6 a resolution. implications for receptor binding.,” Journal of Biological Chemistry, vol. 264, no. 29, pp. 17595–17605, 1989.

H. T. Idriss and J. H. Naismith, “Tnfalpha and the tnf receptor superfamily: Structure-function relationship (s),” Microscopy research and technique, vol. 50, no. 3, pp. 184–195, 2000.

B. Scallon, A. Cai, N. Solowski, A. Rosenberg, X.-Y. Song, D. Shealy, and

C. Wagner, “Binding and functional comparisons of two types of tumor necro- sis factor antagonists,” Journal of Pharmacology and Experimental Therapeutics, vol. 301, no. 2, pp. 418–426, 2002.

D. Tracey, L. Klareskog, E. H. Sasso, J. G. Salfeld, and P. P. Tak, “Tumor necrosis factor antagonist mechanisms of action: a comprehensive review,” Pharmacology & therapeutics, vol. 117, no. 2, pp. 244–279, 2008.

U. Laemmli, “Reagent and gel preparation for sds-page slab gel,” Nature, vol. 227, p. 680, 1970.