Evaluation of Techniwues for the Assessment of Protein Self-Interaction

(1)

Evaluation of Techniques for the

Assessment of Protein Self-Interaction

Final Report version 1.0

(2)

Evaluation of Techniques for the

Assessment of Protein Self-Interaction

Final Report version 1.0

Eef van den Elzen

Project Information

Title: Evaluation of techniques for the assessment of protein self-interaction Product: Final Report

Version: 1.0

Date: 10 June 2016

Period: 4.3 - 4.4 1 February 2016 - 17 June 2016

Study: Chemistry

Major: Forensic Chemical Research

School: Avans Hogeschool, ‘s-Herthogenbosch Company: Synthon Biopharmaceuticals BV, Nijmegen Contact Information

Supervisor (Synthon): Karin Geuijen E-mail: Karin.Geuijen@synthon.com

Supervisor (Avans): Linda Silvertand E-mail:

lhh.silvertand@student.avans.nl

Student: Eef van den Elzen E-mail: emw.vandenelzen@student.avans.nl

(3)

1.1 Abstract

This report describes the method development and method evaluation of techniques to determine the amount of protein self-interaction. In the early stage development of monoclonal antibodies, a high-throughput screening of physical properties of new entities is desired. The formulation buffer in which the antibody is present has a great influence on the degree of interaction. In case of a high tendency of self-interaction, the protein can form aggregates which can cause immunogenicity. The goal was to develop a high-throughput method to measure self-interaction that uses little quantities of antibodies and that is able to rank different formulation buffers and conditions in case of protein self-interaction.

The most used method to determine the amount of protein self-interaction is Self-Interaction Chromatography (SIC). This method is then used as a reference during the evaluation of the relatively new high-throughput techniques. The evaluated techniques were Surface Plasmon Resonance (SPR) and Bio-layer Interferometry (BLI). In both techniques the antibody of interest is immobilized to a sensor surface and then the same antibody is injected as sample to measure protein-protein interactions.

For both SPR and BLI experiments non-specific binding had to be reduced to a minimum. Additives like BSA, tween 80 added because but they could not prevent non-specific binding. When casein was added the total amount of binding decreased. Another known additive to prevent non-specific binding is sodium chloride. When NaCl (up to 150mM) was added, the total amount of binding to the sensor surface decreased. It is not known if specific binding was present before addition of NaCl which also decreased or if only non-specific interactions were measured. It is possible that no self-interaction is observed because the used protein is fully developed. For the future it is recommended to use less stable proteins when self-interaction is measured. These less stable proteins have to be tested in combination with NaCl or casein to determine if specific non-specific binding is reduced enough to only measure specific binding. The specific binding between proteins or the non-specific binding between the proteins and the sensor surface can be caused by different interactions like hydrophobic interaction, electrostatic interaction, hydrogen bridges and van der Waal forces.

Sodium chloride is not preferred as additive because it is also a factor in the buffer screening process. But by adding NaCl up to 1M it can be determined of non-specific binding is caused by hydrophobic interactions or by electrostatic interactions. When it is known what kind of interactions are measured the analysis can be further developed with different additives or sensors. Also other salts like ammonium sulfate can be added to create more stressed conditions for the proteins so more self-interaction will take place.

To overcome non-specific binding other immobilization setups were tried like quenching of the sensors with ethylenediamine in BLI experiments. Also other sensors were tried with a different matrix. In SPR experiments planar sensors were used that have a more 2-dimensional surface structure compared to the gel-like sensors which have a 3D surface structure. The idea was that this may reduce the non-specific interactions because of the different matrix structure. In BLI experiments streptavidin sensors were used, but all showed non-specific binding. Since the issue with non-specific binding has not been fully overcome, no good comparison between techniques can be made. When less developed proteins are analysed there was significant specific signal measured, but there is still non-specific interaction present. Thus this technique has potential but needs further optimization. BLI has a higher-throughput and flexibility compared to multiplexed SPR and therefore it is recommended to further develop this technique for application of the self-interaction measurements.

(4)

1.2 Samenvatting

Dit afstudeerverslag gaat over de vergelijking van de standaard techniek voor het bepalen van eiwit zelfinteracties met twee nieuwe technieken. Voor deze twee nieuwe technieken is ook een methodeontwikkeling uitgevoerd. Tijdens de ontwikkeling van nieuwe monoklonale antilichamen is het gewenst om een snelle beoordeling te kunnen maken van fysische eigenschappen van het eiwit. Een van deze eigenschappen is zelfinteractie; bij hoge mate van zelfinteractie bestaat er een kans op aggregatie wat kan leiden tot immunogeniciteit. De standaard techniek voor het meten van de mate van zelfinteractie is zelfinteractie chromatografie. Voor deze zelfinteractie chromatografie is veel materiaal nodig dat en de throughput is laag. Het doel was om deze techniek te vergelijken met nieuwe technieken met een hogere throughput zoals Surface Plasmon Resonance (SPR) en Bio-layer Interferometrie (BLI). Deze nieuwe technieken moeten buffers kunnen indelen op basis van zelfinteractie en antilichamen moeten in een vroeg stadium van de ontwikkeling met elkaar vergeleken kunnen worden op basis van de mate van zelfinteractie. Voor SPR en BLI geldt dat er wordt begonnen met het immobiliseren van het antilichaam op het sensoroppervlak. Vervolgens wordt interactie gemeten met hetzelfde antilichaam in het monster. Een ongewenst effect tijdens analyse van zelfinteractie is aspecifieke binding aan het sensoroppervlak of de matrix. Verschillende additieven zoals BSA en tween 80 bleken niet geschikt om aspecifieke binding te voorkomen. Door toevoeging van caseine wordt de hoeveelheid binding verminderd zowel op de specifieke als op de referentie sensor. Ook is zout toegevoegd om aspecifieke binding te voorkomen. Een zout zoals natriumchloride zou elektrostatische interacties tussen het geladen eiwit en het geladen sensoroppervlak moeten voorkomen. Door toevoeging van NaCl neemt de aspecifieke binding aan het sensor oppervlak af. Het is niet bekend of er geen specifieke binding was of dat deze ook gereduceerd wordt door de toevoeging van NaCl. Het is mogelijk dat er geen specifieke binding plaats vindt omdat het gebruikte eiwit volledig uit ontwikkeld is. In het vervolg wordt het aangeraden om minder stabiele eiwitten te gebruiken met een hogere mate van zelfinteractie. Deze eiwitten moeten geanalyseerd worden in combinatie met NaCl en caseine om te zien of specifieke binding gemeten kan worden. De specifieke zelfinteractie tussen eiwitten of de aspecifieke interactie tussen de eiwitten en het sensoroppervlak kan veroorzaakt worden door hydrofobe interactie, elektrostatische interactie, waterstof bruggen en vanderwaals krachten.

Natrium chloride is in dit geval niet gewenst als additief omdat het een factor die in het ideale geval gescreend wordt tijdens de ontwikkeling van nieuwe buffers. Maar om te bepalen of aspecifieke binding elektrostatische of hydrofobe interacties zijn kan de concentratie NaCl verhoogd worden tot 1M. Als de binding toe neemt bij een concentratie hoger dan 200mM dan wordt de binding veroorzaakt door hydrofobe interacties. Als de binding dan niet toe neemt wordt de binding bij veroorzaakt door elektrostatische interactie. Door te bepalen welke soort interactie gemeten wordt kunnen additieven en sensoren aangepast worden om de aspecifieke binding te verminderen.

Naast additieven is ook geprobeerd om de set-up van de immobilisatie te veranderen. Zo is bijvoorbeeld geprobeerd om het blokkeren van de sensor na immobilisatie uit te voeren met ethyleendiamine. Hierdoor zou de lading van het sensor oppervlak lager moeten zijn en op deze manier zou er minder aspecifieke binding moeten zijn. Ook zijn andere sensor oppervlakken gebruikt zoals een sensor met een vlakkere matrix bij SPR experimenten en een streptavidine sensor bij BLI experimenten. Alle set-ups en sensor oppervlakken lieten aspecifieke binding aan het oppervlak zien. Als met BLI eiwitten vergeleken worden die minder ver ontwikkeld zijn kan er specifiek signaal gemeten worden, maar er blijft ook aspecifieke binding aanwezig. Dus BLI is flexibeler en heeft heen hogere throughput dan multiplexed SPR maar er is nog verdere ontwikkeling nodig.

(5)

2 Introduction

This report is written to conclude the graduation internship that is part of the study chemistry, with as major Forensic Chemical Research, at the Avans Hogeschool located in ‘s-Hertogenbosch. The graduation internship of twenty weeks is completed at Synthon Biopharmaceuticals BV located in Nijmegen under supervision of Karin Geuijen.

The project was focused around setting up an optimal workflow for the high-throughput screening of protein self-interaction. The project was chosen because of the possibility to work with Surface Plasmon Resonance and new techniques such as Bio-Layer Interferometry.

The necessity of this evaluation is outlined by Wu et al.[ CITATION WuJ15 \l 1033 ]: “There is a significant

unmet need for methods that can evaluate self-association propensities of concentrated mAbs at the earliest stages in antibody discovery to avoid downstream issues”.

Because of this necessity the goal of this project was to develop an optimal workflow for the assessment of protein self-interaction. This was achieved by evaluating different high-throughput techniques and comparing them to results obtained from Self-Interaction Chromatography measurements as a reference technique. The optimal workflow needs to be able to screen a wide range of formulation buffers in a shortest possible time span and consuming the least amount of protein as possible. The first part of this report consists of a theoretical background of the project and the techniques used during this project. After that the used materials and methods are described. The results of these measurements are enclosed in the following chapter. Lastly the results are being discussed, the conclusions are drawn and recommendations are given.

3 Background

In this chapter the background information about the project is provided. Techniques such as Self-Interaction Chromatography, Bio-Layer Interferometry and Surface Plasmon Resonance are introduced. This chapter starts with information about the used proteins and why it is important to be able to determine the amount of self-interaction.

3.1 Monoclonal Antibodies

Monoclonal antibodies (mAbs) are a growing group of biopharmaceuticals that are highly specific and effective for the treatment of, for example, infectious and oncological diseases [ CITATION Pat10 \l 1033 ]. The antibody binds to tumour associated antigens with the Fab part; these antigens are relatively more expressed on malignant cells than cells of normal tissue. The monoclonal antibodies that are used as biopharmaceuticals in cancer treatment can target cells in two different mechanisms. The first mechanism inhibits cell division; this will stop the tumor growth, but it will not kill the cancer. After binding of the mAb to the cancer cell another part of the mAb, the so-called FC part interacts with a natural killer cell. When the natural killer cell interacts with the mAb, cell lysis is induced. This mechanism is shown in figure 1.

The second mechanism is based on a natural defence mechanism of the cell, which will eventually kill cancer cells. The mAb is internalized by the cancer cell and upon internalization the mAb is not able to interact with the natural killer cell, therefore cell lysis cannot be induced. However, in an Antibody Drug Conjugate (ADC) a cytotoxic drug is linked to an antibody with a specific target. Cells internalize the ADC either by phagocytosis, pinocytosis or receptor-mediated endocytosis. The internalization is followed by intracellular denaturation and catabolism [ CITATION Kla01 \l 1033 ]. The drug is then released in the cell and this induces apoptosis of the cell. This mechanism is shown in figure 2 [ CITATION Placeholder1 \l 1033 ] [ CITATION Ham10 \l 1033 ].

Figure 1: Mechanism of mAbs. The antibody binds to the antigen that is expressed on the tumour. Then the natural killer cell binds to the antibody and cell lysis is induced [ CITATION Ham10 \l 1033 ].

Figure 2: Mechanism of ADC. The ADC binds to the antigen expressed on the tumour. The ADC gets internalized and this is followed by denaturation of the ADC. The cytotoxic drug is released in the cell which induces apoptosis [ CITATION Men16 \l 1033 ].

(9)

3.2 Protein Stability

Protein stability is an important factor in the development of complex macromolecules such as monoclonal antibodies. Proteins are involved in a lot of biological processes and if they are not stable the function of the protein can change. Conformational, chemical and physical degradation can result in loss of activity and increased safety risks such as immunogenicity. A thermodynamically stable protein will be in its folded position. The folding and unfolding of a protein is reversible when the protein has not reached its melting point. When a protein is in its unfolded state it can undergo aggregation, this is caused by colloidal instability. Aggregation due to colloidal instability causes the protein to lose its functional domains and thereby lowering its bioactivity. Irreversible aggregation can be kinetically overcome by forcing the equilibrium in the formulation buffer to the folded state [ CITATION Uhl14 \l 1033 ]. The equilibrium can be shifted by changes in buffer, protein and salt concentration and it can be influenced by the pH and the buffer type. In its folded state a protein can still undergo a range of interactions with similar proteins that are also in their folded state. This is called self-interaction and it is caused by colloidal instability. Self-interactions can be caused by the following Self-interactions: hydrophobic Self-interactions, hydrogen bridges, van der Waals forces and electrostatic interactions[ CITATION Esf13 \l 1033 ][ CITATION LeB09 \l 1033 ] [ CITATION Pat96 \l 1033 ] [ CITATION Qui13 \l 1033 ].

Proteins have a poor bioavailability by most application routes[ CITATION LeB10 \l 1033 ], and because of this mAbs are frequently administered in high doses in small volumes (up to a few hundred mg/dose). Monoclonal antibodies are highly stable because they unfold slowly, but in high concentrations they can aggregate. Also other extreme conditions like freeze-thaw cycles, low pH and elevated temperature can cause mAbs to aggregate [ CITATION Esf13 \l 1033 ].

Due to self-interactions at high protein concentrations, aggregation may occur which can lead to high viscosity and low solubility. This limits the chances of an antibody to become a successful drug. Aggregation of the protein can cause immunogenicity which is a severe reaction of the immune system against the antibody. This is why it is necessary to see if an antibody has the tendency to self-interact in an early stage of the development. The methods that were used to measure self-interaction of proteins in this report include self-interaction chromatography (SIC), surface plasmon resonance (SPR) and bio-layer interferometry (BLI) and these are described below[ CITATION Esf13 \l 1033 ][ CITATION Uhl14 \l 1033 ] [ CITATION Sun \l 1033 ].

Figure 3: Schematic overview of protein stability. Self-interaction is influenced by colloidal stability. When the folded proteins self-interact, protein oligomers are formed. Oligomers cause less bioactivity of the antibody and can cause immunogenicity.

(10)

3.3 Surface Plasmon Resonance

Surface Plasmon Resonance (SPR) is a technique that is used to measure biomolecular interactions in real-time. A big advantage is that molecular interactions such as protein interactions and binding kinetics can be determined label-free. The ligand is immobilized to the sensor surface and then the injected sample passes over the sensor. If the proteins in the sample are the same as the immobilized protein, protein self-interaction can be measured. This self-self-interaction can be caused by electrostatic self-interaction, van der Waals forces, hydrogen bridges and hydrophobic effects [ CITATION Esf13 \l 1033 ][ CITATION Pat96 \l 1033 ] [ CITATION Qui13 \l 1033 ].

The sensor consists of a prism with a gold film of ±50nm thick on top. At the surface of the sensor free electrons are present. Under influence of an electric field the free electrons resonate. This vibration wave of free electrons (plasmons) on the surface of the gold film is called surface plasmon resonance. Polarized light goes through a prism and is reflected by the gold film. The wave of electrons becomes stronger if the energy of the light is absorbed. The SPR-angle is the angle at which the light loses the most intensity. This angle depends on the refractive index of the gold film and the prism. The refractive index is changed if large molecules like proteins bind to the proteins that are immobilized at the surface. In this case the refractive index changes as self-interaction between similar proteins takes place. The change of the SPR-angle is called SPR-shift[ CITATION SPR161 \l 1033 ].

Real-time association and dissociation of the protein to the immobilized ligand is shown in a sensorgram. In a sensorgram the SPR-shift is plotted against the time. A high tendency of protein self-interaction results in a high SPR-shift. For the binding of an antibody to a target high binding constants are desired. But for self-interaction experiments low binding constants are desired because protein self-self-interaction can cause aggregates and thereby cause immunogenicity [ CITATION Qui13 \l 1033 ].

For the immobilization of the protein to the sensor surface a Continuous Flow Microspotter (CFM) is used. The CFM is a spotting device which can immobilize 48 different proteins at the same time to the sensor surface. It is also possible to immobilize 48 times the same IgG to the sensor surface. The ligand solution cycles through the micro-channels of the CFM. The proteins stay in a liquid environment, in contrast to pin or non-contact printing, which makes it possible to print sensitive proteins [ CITATION Was15 \l 1033 ].

Figure 4: Principle of SPR. When the analyte binds to the immobilized ligand on the sensor surface, the SPR-angle changes. The SPR-angle is the angle at which the incoming light is absorbed the most by surface plasmons. [ CITATION van15 \l 1033 ].

(11)

The Easy2Spot sensors that were used in this project do not need to be activated before immobilization; these were pre-activated sensors. Other sensors have to be activated; in this case the carboxylic acid group of the sensor reacts with a NHS group in the presence of a carbodiimide like EDC (Figure 8). The sulfo-NHS ester that is formed can react with the primary amine of a protein to form a stable amide bond. Sulfo-NHS is preferred over Sulfo-NHS because the surface becomes less negatively charged when sulfo-Sulfo-NHS is used. The used protein is positively charged, thus when the sensor surface is less negatively charged there is less electrostatic interaction that can cause non-specific binding of the analyte to the sensor. After immobilization the sensor is quenched in the IBIS MX96 with ethanolamine.

The IBIS MX96 is a SPR instrument which can be used to determine several biomolecular interactions in a multiplexed manner. After quenching, sample is transported from the 96-well plate to the flow cell. In the flow cell, interaction is measured between the IgGs in the sample and the immobilized IgGs. With this technique many different sample conditions can be tested for the tendency of protein self-interaction in a short period of time[ CITATION SPR161 \l 1033 ][ CITATION Geu \l 1033 ].

3.3.1 Sensor surfaces

During the optimization process of an assay to determine biomolecular interactions with surface plasmon resonance the type of sensor surface has to be taken in to account. Different matrices that are coupled to the golden surface are commercially available. The most used surface is a carboxymethylated (CM) matrix. This carboxymethylated matrix comes in different degrees of carboxymethylation. A sensor with a high degree of carboxymethylation has a bigger capacity for the immobilization of a ligand. If an assay shows much non-specific binding a lower degree of carboxymethylation matrix or a more planar surface could be tried to reduce non-specific binding. Due to the lower degree of carboxymethylation the surface is less negatively charged and positively charged analytes are less attracted to the surface. If small molecules are analyzed it is better to use a sensor surface with a high degree of carboxymethylation. The length of the spacer also has an influence on the amount of non-specific binding. A short spacer is desired for higher sensitivity, but with a short spacer it is possible that there is steric hindrance when the analyte binds to the ligand. When steric hindrance occurs there is more ligand that has no analyte bound to it thus this ligand is susceptible for non-specific binding of other molecules in the sample. For the immobilization of the ligand to the sensor surface it has to bind with a functional group on the matrix. Different functional groups are commercially available such as carboxylic acid and primary amine groups, the sensors used during this project all have a carboxylic acid functional group [ CITATION SPR162 \l 1033 ][ CITATION Sse16 \l 1033 ]. Another type of matrix that can be used is streptavidin. The binding between streptavidin and biotinylated ligand is a strong non-covalent binding. After the binding of the biotinylated ligand the sensor is ready for use. Biotinylation is the labelling of macromolecules containing primary amino groups. NHS-activated biotin is used to form an amide bond with the primary amine group of the protein and the biotin [ CITATION The12 \l 1033 ].

Figure 5: Schematic overview of the biotinylation of a protein. The biotin forms an amide bond with the primary amine of the protein [ CITATION The12 \l 1033 ].

(12)

3.4 Bio-Layer Interferometry

Bio-Layer Interferometry (BLI) is a high-throughput technique to determine macromolecular interactions in real-time. BLI is based on the principle of optical interferometry of white light. If two light waves are exactly in phase the amplitude of the wave increases, this is called constructive interference. (Figure 6a) If the two waves are completely out of phase the amplitude will be zero, this is called destructive interference. BLI is compared to SPR and SIC to test if it has higher throughput and flexibility than the other techniques. Also for BLI other sensor surfaces are available which have different matrices. (Figure 6b)[ CITATION For12 \l 1033 ][ CITATION For16 \l 1033 ].

BLI is based on this basic principle of interference. White light goes through the sensor tip and is reflected by the sensor surface. The wave of the reflected light undergoes interference with the wave that is reflected by the intern reference layer. The optical thickness of the sensor surface is measured; this thickness can change if proteins, e.g. IgGs, bind to the sensor surface. The binding of macromolecules causes a shift in interference pattern and thereby causes a wavelength shift (∆λ). The wavelength shift is measured in real-time. Changes in the medium do not affect the interference pattern; only bound molecules cause a wavelength shift. This provides the ability to monitor association, dissociation and binding specificity with high precision [ CITATION For12 \l 1033 ][ CITATION For16 \l 1033 ].

During this project a FortéBio Octet Qk384 is used which has place for 16 sensors that can be immobilized with different proteins or the same proteins. The glass sensors have an optical surface at the tip of the sensor. This optical surface is coated with a biocompatible matrix that can interact with molecules in the sample. In this project sensors coated with carboxylic acid are used, these sensors have to be activated with EDC/s-NHS before immobilization of a protein. Usually 8 tips are not immobilized with ligand to be used as reference; the other 8 tips are used for interaction measurements. The sensor tips are dipped in the samples and protein self-interaction is measured [ CITATION For16 \l 1033 ].

Figure 6: The principle of interference. When two waves are in phase the amplitude of the wave increases, this is called constructive interference (A). When two waves are out of phase the amplitude will be zero, this is called destructive interference (B)[ CITATION For12 \l 1033 ].

Figure 7: BLI; A, no analyte biding, stable interference pattern. B, Binding of ligand to the sensor surface, interference pattern shifts. C, Overlay of interference patterns, wavelength shift (∆λ) caused by binding of analyte to the ligand [ CITATION For16 \l 1033 ].

(13)

3.4.1 Immobilization of BLI sensors

For the experiments with the Octet system two types of sensors were used. The streptavidin (SAX) sensors form a strong non-covalent bond with biotinylated proteins. Then protein self-interaction was measured. The immobilization of the proteins to the AR2G sensors is a bit more complicated.

The sensor surface contains carboxylic acid groups. These carboxylates (-COOH) groups can react with NHS or NHS in the presence of a carbodiimide such as EDC. This will result in a semi-stable NHS or Sulfo-NHS ester, which can react with primary amines of the protein. The activation with Sulfo-NHS causes a decrease in water solubility therefore Sulfo-NHS can be a better option. Sulfo-NHS also provides a more negatively charged surface which causes the positively charged protein to move towards the sensor surface before it forms a stable amide bond [ CITATION Sci09 \l 1033 ].

After immobilization the Sulfo-NHS-groups that did not react are quenched with ethanolamine. This irreversible immobilization is highly stable in a pH range of 0.5 to 11 and salt concentration up to 5M. In figure 9 a schematic overview is given of the immobilization of proteins to the AR2G sensors [ CITATION For11 \l 1033 ].

Figure 8: Reaction scheme for the activation of the carboxylic acid group on the sensor surface with EDC/Sulfo-NHS and the formation of a stable amide bond with the primary amine of the protein[ CITATION Sci09 \l 1033 ].

Figure 9: Schematic overview of the immobilization of antibodies to the AR2G sensor. The Carboxylic acid groups are activated with sulfo-NHS. Then the primary amine of a protein forms a amide bond with the carboxylate group. The sensor is quenched with ethanolamine to block the rest of the activated groups. Then interaction can be measured by dipping the sensor in the analyte [ CITATION For11 \l 1033 ].

(14)

3.5 Self-Interaction Chromatography

The most used method to determine protein self-interaction is Interaction Chromatography (SIC). Self-Interaction Chromatography is a form of affinity chromatography that is based on the interaction of similar proteins. The degree of self-interaction is measured as function of the retention volume. The retention volume will increase when self-interaction is present, see figure 10. The protein of which the self-interaction needs to be determined is immobilized to a chromatographic resin. Immobilization of the protein to the stationary phase can cause structural and/or conformational change of the immobilized protein[ CITATION Aha05 \l 1033 ], this is also possible during immobilization during SPR or BLI experiments. The amount of proteins immobilized to the resin is called the surface coverage; the optimum surface coverage is 15-20% [ CITATION Luc15 \l 1033 ]. When the surface coverage is too low, weak interactions will not be detected sufficiently. If the surface coverage is high, it is possible that proteins in the sample interact multiple times with the immobilized protein on the resin. The immobilized proteins are spatially closer to each other, which will cause the results to appear if there is more protein-self interaction than there actually is [ CITATION Luc15 \l 1033 ].

In this project NHS-activated Sepharose is used as a resin in SIC experiments. The NHS-group provides a binding site for free amine groups of proteins which can form an amide coupling. After protein immobilization the remaining active sites are blocked by coupling of ethanolamine, see figure 11 [ CITATION Luc15 \l 1033 ].

When protein self-interaction takes place between the immobilized protein and the same protein in the sample solution the retention volume increases. It is also possible that the proteins are repulsed by each other, than a decrease in retention volume is possible. See figure 10. The degree of self-interaction depends for example on buffer pH and ionic strength, so the formulation buffer has a large influence on the degree of protein self-interaction [ CITATION Luc15 \l 1033 ].

Figure 10: Principle of Self-Interaction Chromatography. When self-interaction takes place the retention volume of the protein is longer than when no self-interaction takes place and the protein is repulsed by the immobilized protein [ CITATION Luc15 \l 1033 ][ CITATION Uhl14 \l 1033 ].

Figure 11: Schematic overview of protein immobilization to NHS-activated Sepharose. Left, NHS-group dissociates. Center, protein forms amide bond with the carboxylate group. Right, free binding sites are blocked by ethanolamine[ CITATION Luc15 \l 1033 ].

(15)

4 Material and Method

4.1 Chemicals

The IgG was provided by Synthon Biopharmaceuticals BV (Nijmegen, The Netherlands). L-Histidine (CAS: 71-00-1), Sodium Chloride (CAS: 7647-14-5), Sodium Dihydrogen Phosphate Monohydrate (CAS: 10049-21-5), Disodium Succinate (CAS: 150-90-3), Sodium Citrate Dihydrate (CAS: 6132-04-3) and Acetic Acid (CAS: 7365-45-9) were purchased at Merck KGaA (Darmstadt, Germany). Ethanolamine (CAS: 141-43-5), (3-Dimethylaminopropyl-N’-Ethylcarbodiimide hydrochloride (EDC) (CAS: 25952-53-), N-Hydroxysulfosuccinimide sodium salt (sulfo-NHS) (CAS: 106627-54-7), Glycine (CAS: 56-40-6), MES monohydrate (CAS: 145224-94-8), Sodium Hydroxide (CAS: 1310-73-2), Hydrochloric Acid (CAS:7647-01-0), Phosphoric Acid (CAS: 7664-38-2), Tween 80 (CAS: 9005-65-6), Tween 20 (CAS: 9005-64-5), HEPES (CAS: 7365-45-9), Glycerol (CAS: 56-81-5), Bovine Serum Albumin (CAS: 9048-46-8), Casein Sodium salt from bovine milk (CAS: 9005-46-3), Phosphate Buffered Saline, RBS™ neutral and N-Hydroxysuccinimidyl-Sepharose® 4 Fast Flow were purchased at Sigma-Aldrich (Saint Louis, Missouri, United States). NSB-Reducer (Carboxymethyl dextran sodium salt) (CAS: 39422-83-8) was purchased at GE Healthcare, Little Chalfont, United Kingdom.

(16)

4.2 Surface Plasmon Resonance

IBIS MX96 (IBIS Technologies, Enschede, The Netherlands) is used for the SPR measurements. The SensEye sensors were purchased at IBIS Technologies, Enschede, The Netherlands. The SPR sensor prism HLC200M was purchased at Xantec bioanalytics GmbH, Düsseldorf, Germany. The ligand was immobilized using a Continuous Flow Microspotter (Wasatch Microfluidics, Salt Lake City, Utah, United States).

4.2.1 Ligand Immobilization

Before macromolecular interactions can be measured using the IBIS the ligand was immobilized using the CFM. The CFM was turned on and the following solutions were prepared and placed at the appropriate place: 0.5% Tween 20, 20% RBS™ neutral, 10mM Sodium Acetate pH 4.5/0.05% Tween 80 and daily fresh MilliQ water. A clean blotting slide and cleaning slide were placed at the appropriate place. A “System Setup” is executed and the instructions of the program were followed such as installation of the correct printhead and placing an empty well plate into bay 2. When necessary the printhead was cleaned for 5 minutes in an ultrasonic bath in 5% RBS™ neutral followed by 5 minutes in MilliQ water. After finishing the “system setup” the flow through all channels were checked by visually inspecting the buffer volume in the 48 wells. If the wells were not filled properly the “system setup” was repeated, if the wells were filled properly the CFM was ready for use.

To print a sensor the program “print a slide” was used. The print program that was used during this project is called “5minsampleloc1”. If necessary e.g. the print coordinates or printing time could be adjusted. When the adjustments of the print program were finished the option “run print program” was chosen. Then the appropriate print program was selected. The 96-well plate was filled with the ligand. The IgG was diluted to 4ug/ml in 50mM sodium acetate pH 4.5. There had to be at least 150µl ligand solution in each well. For the correct placement of the ROI’s some wells were filled with BSA or a higher concentration IgG. The sensor was placed at room temperature at least one hour before use and was placed in the CFM upon printing. The printing program was started. Immediately after printing the sensor was deactivated in the IBIS MX96. The dummy slider in the IBIS was replaced with the sensor that had to be deactivated. A vial with 500µl 1M ethanolamine was placed at vial location 1. The “Quenching” script in the IBIS MX96 software was started. When this script was finished the sensor was ready for use. The “system rinse” of the CFM was performed after printing. When the CFM was not used for more than three days the full system shutdown was performed.

(17)

4.2.2 SPR analysis

The SPR analysis with the IBIS MX96 was started by selecting a user in the “IBIS iSPR” software. Then a measurement was selected which was made in the “IBIS SUIT” software. If necessary the “Replace sensor” option was selected in the general tab and the sensor was replaced. The temperature of the sample deck and the flowcell was set on 25°C (default). The buffer was replaced and a “system prime” was performed. The system buffer had to be the same as the sample buffer to reduce bulk shift.

Preparation of a sequence in Suit

The Analysis tab was opened and the sample list and method settings were filled in. It is possible to import the sample list form e.g. Excel. The Standard Run script was selected and the following times were selected; baseline 1 minute, association 5 minutes, dissociation 4 minutes, wash step 1. For the regeneration a baseline of 0.5 minute, association of 0.5 minute and a dissociation of 0.5 minute were selected.

Samples were prepared and at least 150µl sample was present in the 96 well-plate. The wells were covered by aluminium foil. The ROI’s were set to the correct position of the printed spot by clicking on the camera image and then dragging the ROI’s to the correct position. It was made sure that buffer and regeneration fluid (Glycine pH 1.5) were present in the correct vials and covered with aluminium foil.

The “Angle offset” was started in the general tab and was adjusted with the knob in the instrument to place the SPR dip around zero. Then the “prep” was started and a baseline of 1 minute was recorded. At this point it was checked if the buffer, samples and reagents were in place and if the waste container was empty. When desired, “proceed to calibration” was selected and two vials were filled, one vial with at least 700µl H2O, this vial was placed in vial position 11. And one vial was filled with at least 250µl 20% Glycerol in system buffer and placed at vial position 12, both vials were covered with aluminium foil. When the baseline during the prep was stable the measurement was started in the “Analysis tab”. Tested samples contained the same IgG in the system buffers as immobilized to the sensor to determine the degree of self-interaction. To test other buffer conditions the system buffer had to be replaced with the same buffer as the sample to prevent bulk shift. The data was analyzed with “SprintX” software.

(18)

4.3 Bio-Layer Interferometry

For the BLI experiment a FortéBio Octet Qk384 and matching sensor tips were used (Pall Corporation, Port Washington, New York, United States). The immobilization program used for the AR2G sensors consisted of the following steps. First a baseline was acquired in water during 1 minute. This was followed by a 5 minute activation step in a 1:1 (v:v) mixture of EDC/NHS (use mixture within 1 hour of preparation). The third step was the loading of the ligand to the sensor. During 30 second the sensors were dipped in a solution of 3µg/ml IgG in a sodium acetate solution pH 4.5. The reference sensors were dipped in a sodium acetate solution pH 4.5 without IgG. Then the sensors were deactivated by dipping them in 1M Ethanolamine for 5 minutes. And at last the sensors were washed 2 times in PBS for 2.5 minutes. Now the sensors were ready for use.

For the determination of the degree of self-interaction the same IgG as immobilized was diluted in different buffer solutions e.g. 5mM phosphate pH6, 5mM Histidine pH6. After the binding the sensors were regenerated in glycine pH 1.5 and neutralized in PBS before re-use. The binding rate was calculated by the software by selecting the “initial slope” at 10 seconds.

4.3.1 Biotinylation

For the biotinylation the EZ-Link® Micro Sulfo-NHS-LC-Biotinylation Kit was used (Thermo Scientific, Waltham, Massachusetts, United States). The Streptavidin (SAX) sensors were purchased at Pall Corporation.

The Sulfo-NHS-LC-Biotin vial was removed from the freezer to let it come to room temperature before opening. The IgG solution was diluted to 1ml of 5mg/ml IgG in PBS. Immediately before use 10mM biotin solution was prepared by dissolving 1mg Sulfo-NHS-LC-Biotin in 180µl MilliQ water. 67µl biotin solution was added to 1ml protein solution (Molar ratio protein: biotin 1:20). This was incubated for 30-60 minutes at room temperature. The biotinylation was stopped by filtration of the solution over a PD miditrap G-25 column (GE Healthcare, Little Chalfont, United Kingdom). The storage solution of the column was discarded as waste and the bottom cap was removed. The column was washed three times with PBS pH 7.4, the flow-through was discarded. 1 ml of biotin/protein solution was added to the column and it was left to enter the packed bed completely, the flow-through was discarded. A test tube was placed under the column to collect the sample and then the sample was eluted with 1.5ml PBS pH 7.4. The biotinylated protein was then ready for coupling with the streptavidin sensor.

The SAX sensortips were equilibrated in PBS during 120 seconds. Then the sensor tips were loaded with 20µg/ml biotinylated IgG during 20 seconds, which gave a binding rate of ±2nm. After the loading a baseline was measured by dipping the sensor in PBS during 60 seconds. For the determination of self-interaction different concentrations of IgG were tested in various buffers e.g. phosphate, histidine and acetate. The regeneration step was done by glycine pH 1.5.

(19)

4.4 Self-Interaction Chromatography

For the Self-Interaction Chromatography measurements the Äkta Explorer was used. The NHS-Activated Sepharose 4 Fast Flow resin (GE Healthcare, Little Chalfont, United Kingdom) was used to immobilize IgG. The resin was washed 5 times with 1mM hydrochloric acid. 3 millilitres of washed particles were incubated in 10ml 1mM hydrochloric acid for swelling at 4°C. A coupling solution was prepared with 0.1M sodium phosphate and 0.5M sodium chloride. The protein solution was made by diluting the IgG to 5mg/ml in coupling solution. The HCl solution was discarded and was replaced with the protein solution. The suspension of resin and protein solution was incubated at 4°C while shaken gently for 12 hours. After incubation the suspension was washed 5 times with coupling solution. The coupling solution was discarded and the resin was suspended in 1M ethanolamine in coupling solution. This was also incubated for 12 hours while shaken gently at 4°C. The suspension was stored in coupling solution.

For the packing of the column an Omnifit column (Sigma-Aldrich, Saint Louis, Missouri, United States) was used. A column of 6.5 cm was packed with 2ml suspension under a flow of 0.75ml/min. Then the flow rate was reduced to ensure further settling of the particle bed (Pressure was held below 100kPa). A column performance test was performed by injecting 2M NaCl solution. If the plate number per meter is >5000 the column should provide enough separation.

Self-interaction was measured by equilibrating the column with the buffer of interest. After that an IgG solution of 1mg/ml was injected and a peak was detected at 280nm. The idea was that the retention volume would differ in each buffer; large retention volumes are expected in buffers with a lot of self-interaction and lower retention volumes will correspond to less self-self-interaction.

(20)

5 Results

In this chapter the results of the project will be given. A start was made by setting up the SPR method; this was followed by setting up the BLI method. During the optimisation process of these methods some SIC measurements were performed. For the development of a high-throughput measurement for the determination of protein self-interaction different approaches have been tried, the results are displayed below.

5.1 Surface Plasmon Resonance

Surface Plasmon Resonance (SPR) measurements were carried out with the IBIS MX96 instrument and the ligand was immobilized on the sensor with the Continuous Flow Microspotter (CFM). The goal was to set-up a high-throughput method to measure protein self-interaction with the least amount of IgG possible and determine the amount of self-interaction in different buffer solutions.

5.1.1 Concentration Optimization

A start was made by defining the concentration of IgG that is needed to measure the degree of protein self-interaction. There are two different IgG concentrations that will be discussed. The first is the concentration of IgG that is immobilized to the surface sensor, after immobilization the protein will stay present on the sensor surface due to a covalent amide bond. The second IgG concentration is the concentration of IgG in the sample solution. The IgGs in the sample flow freely past the sensor and self-interaction can take place. After regeneration, all of the bound IgG is ideally flushed from the surface and they will be discarded through the waste. When the IgG concentrations are set to a certain value, the buffers can be tested for their influence on protein self-interaction and the results can be compared to the other techniques.

To define the concentration of IgG that should be immobilized to measure protein self-interaction a concentration range of IgG was immobilized by the CFM to an Easy2Spot G-type sensor. The camera image of the sensor is shown in figure 12. The concentration range of the immobilized IgG is 2000-0.98 µg/ml.

Three different buffers were analyzed, respectively 5mM histidine pH 6, 5mM histidine pH 6 with 150mM NaCl and 5mM phosphate pH 6. Histidine and phosphate were chosen because it was found in previous research [ CITATION Luc15 \l 1033 ] that this IgG has the tendency to show more self-interaction in phosphate and the least self-interaction in histidine buffer. The IgG concentration range in these buffers was 150-1.17 µg/ml.

No interactions of IgG to the same IgG at the surface were observed, this could be caused by a too low concentration of the IgG in the buffer. Another possible cause was thought to be that the immobilized IgGs are damaged by the regeneration solution, glycine-HCl pH 1.5.

Figure 12: Camera view of the sensor #084. Yellow and red squares are the regions of interest (ROI). The green squares are the reference ROI’s.

(21)

To see if the immobilized proteins are not damaged by the regeneration with glycine, it was tested if protein A shows interaction with the immobilized ligand. Protein A binds very strongly to IgG, this is also visible in the sensorgram (See figure 13).

Also a higher concentration of IgG in phosphate buffer was tested. A concentration range of 9.4-0.6mg/ml IgG in phosphate buffer is tested. The regeneration of the sensor is done with HBS to prevent damage to the immobilized ligand. In figure 14 it is shown that the sensor is not fully regenerated due to a slight increase in base-line at the start and end of the analysis. All samples show response at all concentrations of immobilized ligand. The red lines in figure 14 indicate the binding curve of the selected spot. Due to the high concentration IgG in the analyte, bulk shift was observed. Bulk shift occurs when the refractive index differs a lot from the system buffer which causes a sudden shift in SPR-angle. The green lines in figure 2 indicate binding to the reference spot and bulk shift. Ideally there is no binding observed to the reference spot, but binding is observed which could indicate two things. Namely the “region of interest” (ROI) is not placed partially on the spot with immobilized ligand. Or there is non-specific binding to the sensor surface. Spots were visible on the camera image starting at a concentration of 7.8 µg/ml and higher; for these spots it is easier to place the ‘region of interest’ (ROI) on the correct position, so 7.8 µg/ml immobilized ligand is chosen for further experiments. At the sample concentration of 0.6mg/ml IgG and higher sufficient response was measured so in the next experiment the other buffers were tested with an IgG concentration of 0-1mg/ml. The lower the amount of IgG required the better, due to the limited availability of the IgG.

Next, the phosphate buffer was compared with the histidine buffer, which should show the least amount of protein self-interaction[ CITATION Luc15 \l 1033 ]. Furthermore, 150mM NaCl was added to the histidine buffer to determine the effects of salt addition on the protein self-interaction. The addition of salt causes a reduction in the amount of binding due to a decrease in electrostatic interactions. This principle is also visible in figure 20.

Large bulk shift was observed in each sample, as shown in figure 2. This could be caused by the change in refractive index between the system buffer (HBS) and the sample buffer (phosphate/histidine). To overcome this bulk shift the system buffer was adjusted to be the same as the sample buffer.

Figure 14: Concentration range 0.6-9.5 mg/ml IgG. Spot [9] 7.8 µg/ml immobilized IgG (Red). Reference spot [57] (Green). Figure 13: Binding of 200 µg/ml Protein A. Sensor #084,

(22)

5.1.2 Bulk Shift

Bulk shift is a large change in SPR-shift due to a change in refractive index rather than protein binding to the surface. This can be caused by a change in buffer type, buffer concentration, of additives like sodium chloride. In case of large bulk shift, for example due to buffer mismatch, it is very difficult to measure an accurate specific binding.

A new sensor was made (#085, Easy2Spot-G) with three different concentrations of immobilized ligand, 1, 2 & 4 µg/ml IgG. A lower concentration of ligand than 7.8µg/ml was chosen because the response in earlier experiments was high. It was possible to place the ROI’s correctly on the spots because a high concentration BSA at was placed on the spots next to the spots with immobilized IgG. The system buffer was chosen to be exactly the same as the sample buffer, in this case 5mM phosphate pH 6, and Tween 80 was added to the system buffer to prevent the formation of air bubbles in the tubing of the instrument. The samples contained 0 to 250 µg/ml IgG. The three spots with different ligand concentrations showed self-interaction with 62.5, 125 and 250 µg/ml IgG in the sample. In figure 15 the binding curve of 62.5 and 125 µg/ml is shown. The sample with 125 µg/ml IgG shows bulk shift or non-specific binding. It is visible in figure 15 that the reference spot (green) shows similar binding as the spots with immobilized ligand (red).

Figure 15: Spot [26] 2mg/ml IgG immobilized. Reference spot [62]. Sample: 62.5 µg/ml (green) and 125 µg/ml IgG (red).

(23)

Samples with different concentrations were analyzed in histidine buffer pH 6 and in phosphate buffer pH 6. In each measurement the system buffer was changed to be exactly the same as the analyte buffer. It was looks like that phosphate shows more interaction than histidine, which was expected (figure 16). But it could also be possible that the buffer also influences the amount of non-specific binding and that is what is measured during this experiment because non-specific binding was also demonstrated in figure 15. In figure 16 the black line indicates the response of 250 µg/ml IgG in phosphate buffer and the blue line indicates 250 µg/ml IgG in histidine buffer.

The high-throughput of the SPR measurements is compromised if the system buffer has to be changed for every condition that will be tested. The results shown in figure 18 and figure 16 only are indicative of the effect of the buffer to the amount of self-interaction. The results as shown are corrected by subtracting the signal of the reference spot. Signal on the reference spot as seen in figure 15, indicates non-specific binding. The measures that were taken to reduce non-specific binding are described in the next paragraph.

Figure 16: Amount of self-interaction measured in different buffers.

Figure 17: Black: 2x 250 µg/ml IgG in Phosphate pH 6. Blue: 2x 250 µg/ml IgG in Histidine pH 6.

(24)

5.1.3 Non-Specific Binding

Protein self-interaction was measured as described before, but now with 50mM phosphate as system buffer. Two phosphate concentrations in the sample solution were tested (5 and 50 mM) and the binding signals were corrected by subtracting a blank measurement of the same buffer, see figure 18. This was performed to correct for small changes in refractive index between the two buffer molarities. More IgG interaction was observed in a 5mM phosphate buffer compared to in the 50mM phosphate buffer. The reference spots showed also the same amount of binding. This indicates non-specific binding to the sensor surface.

Addition of sodium chloride is often used to reduce non specific binding. In figure 19 is shown that the amount of binding decreases with increasing concentrations of NaCl. This was also demonstrated by Gagnon[ CITATION Gag09 \l 1033 ], see figure 20, that when the amount of NaCl is increased to ±200mM the amount of electrostatic (ionic) and hydrophobic interactions decreases. The amount of electrostatic interactions will decrease further when more NaCl is added (up to 1M), but the amount of hydrophobic interactions will increase again when more NaCl is added after a minimum amount of hydrophobic interactions was reached at ±200mM NaCl.

Figure 19: Influence of salt concentration on the amount of self-interaction. When the salt concentration is increased up to 150mM the binding is decreased.

Figure 18: Influence of buffer concentration on the amount of self-interaction. When the concentration phosphate increases the amount of binding decreases.

Figure 20: Effects of salt (NaCl) on electrostatic interaction and hydrophobic interaction [ CITATION Gag09 \l 1033 ]. When the NaCl concentration is increased up to 200mM both hydrophobic and ionic interactions decrease. When the concentration salt is further increased the hydrophobic interaction will also increase.

(25)

The addition of salt to the buffer has two disadvantages, although it prevents non-specific binding. The first disadvantage is that the salt concentration is a factor of the buffer composition that has to be screened during buffer development. Secondly, no specific signal is left when ±150mM of salt was added. But because the protein used is very stable and fully developed it is not known if the specific binding was reduced by the salt or that there was no specific binding at all because the good stability of the protein. Thus in future experiments the addition of salt needs to be tested with proteins that are known to show much self-interaction. If only non-specific binding is measured in these experiments and is reduced by the salt, specific binding will remain.

The results mentioned in chapter 5.1.1 and 5.1.2 were questioned because the binding to the reference spots were not taken in to account. In the ideal situation no non-specific binding is observed to the reference spots. Non-specific binding can be caused by different molecular forces such as electrostatic interaction, hydrophobic interactions and hydrogen bonds[ CITATION tma15 \l 1033 ]. A broad range of additives such as 0.1% BSA, 0.1% Tween 80 and 1mg/ml carboxymethyl dextran were tried to prevent non-specific binding. BSA is added because according to literature [ CITATION tma15 \l 1033 ] BSA has hydrophobic and hydrophilic parts which can surround the protein to prevent electrostatic interactions. Tween 80 could disrupt hydrophobic and ionic interactions between the sensor and the analyte [ CITATION tma15 \l 1033 ]. Carboxymethyl dextran has a similar structure as the dextran matrix on the sensor surface and it could cause a reduction in non-specific binding due to a competition effect [ CITATION GEH161 \l 1033 ].

None of these additives were suitable for use as a blocking agent of non-specific binding. In figure 21 the binding curve of 125 µg/ml IgG in 50mM phosphate buffer is show with 0.1% BSA as additive (Green: binding to reference spot. Black: binding to spot with immobilized ligand). It is shown that the binding to the reference spot is similar to the binding to the spot with immobilized ligand. The binding curve of 125 µg/ml IgG with other additives like Tween 80 and carboxmethyl dextran is shown in figure 22. The binding of the IgG to the sensor with additives is the same on reference spots as on the spots with immobilized ligand. In figure 22 the blue line indicates the sample with Tween 80 as additive, the red line indicates the binding of IgG with carboxymethyl dextran as additive in the sample and the green line shows the binding curve of the sample with Tween 80 and carboxymethyl dextran as additive. The black lines show the binding of the corresponding samples to the spot with immobilized IgG.

Also a planar sensor surface was used for the analysis of protein self-interaction. The previous used sensors, the Easy2Spot-G, have a 3D structure which could be a possible cause for non-specific binding because the IgG could bind to the matrix instead of the specific binding with the immobilized IgG. Therefore, a planar 2D, sensor is used. It was demonstrated that there was also non-specific binding to these planar sensors (figure 23). The amount of binding to the reference spot (green line) and the binding to the spot with immobilized ligand (black line) is comparable.

It was also demonstrated with this planar sensor that when high concentrations of NaCl are added (up to 150mM) no more binding is observed. Only high bulk shift is measured due to the high concentration of salt (figure 24). Non-specific binding to the sensor surfaces needs to be removed or at least decreased if the SPR technique is going to be used to determine the amount of self-interaction of an antibody.

(26)

5.1.4 Xantec polycarboxylate sensor

The sensor purchased at Xantec bioanalytics, HLC200M, had a linear polycarboxylate matrix with reduced charge density. This sensor had, according to Xantec, an extremely low level of non-specific interaction [ CITATION SPR163 \l 1033 ]. And out of personal communication with Xantec it was found out that especially electrostatic interaction was reduced because the polycarboxylate sensor has a less charged sensor surface. Activation of the sensor with EDC/sulfo-NHS is required; this was done by the “SensEye-COOH-Spotting” method in the IBIS software. First the sensor is primed with 50mM MES pH 5.4 and PBS. The activation was performed with 0.4M EDC and 0.1M sulfo-NHS in MES buffer pH 5.4. 250µl of both solutions were pipetted into a 0.5ml tube directly before activation. The deactivation was done with 1M ethanolamine.

The ligand immobilized onto this sensor was the IgG used in previous experiment and different Antibody Drug Conjugate (ADC) with various Drug Antibody Ratio (DAR) species were immobilized, namely DAR 0, DAR 2 and DAR 4. Drug Antibody Ratio is the amount of linker drug bound to the antibody. So DAR 2 has two linker drug molecules attached to the antibody and DAR 4 has four linker drug molecules attached to the antibody. These DAR species were chosen because the DAR 4 species shows much self-interaction, thus it should also be detectable in SPR measurements.

Different concentrations IgG in the sample (0.78-25 µg/ml) were measured to determine dose dependency. With a concentration of 0.78 µg/ml the response is relatively high (±700 RU, figure 25), thus it is possible to lower the sample concentration. It is also visible that the binding to the reference sensor has a big variance. With this experiment it could not be determined if the binding of DAR4 to the spot with immobilized DAR4 is specific or not. Therefore NaCl could be added in steps to reduce non-specific binding. The concentration that can be added may go up to 1M NaCl because if the binding is caused by hydrophobic interactions the amount of binding should increase when the concentration exceeds ±200mM (figure 20). Maybe it is also possible to make a distinction between specific and non-specific binding.

Figure 22: Reference spot [64], 125µg/ml IgG in 50mM Phosphate pH6. Additives: Blue line Tween 80, Red line CM-Dextran and Green line Tween 80 + CM-Dextran. The black lines represent the binding of the sample to a spot with immobilized ligand.

Figure 21: Reference spot [64], 125µg/ml IgG + 0.1% BSA in 50mM Phosphate pH6. The black line represents the binding of the sample to a spot with immobilized ligand.

Figure 24: The black lines indicate the increasing bulk shift caused by NaCl, when no protein is present in the sample. The red line indicates the binding of IgG when NaCl is present. The only response observed is due to the bulk shift, not by interaction of the protein.

Figure 23: Binding of 25 µg/ml IgG in 50mM phosphate pH 6 (binding to 4µg/ml immobilized ligand, black line. Binding to reference, green line. Sensor: Easy2spot P-Type

(27)

Figure 25: Non-specific binding of 0.78µg/ml DAR4 to reference spots (green lines) and spot with 4µg/ml immobilized DAR4 (black line).

(28)

5.1.5 Change in Print-Setup

In the previous setup the ligand was immobilized in the CFM and then interaction was measured in the IBIS. Because the high-throughput of the assay is compromised by the need to change the system buffer with each sample buffer type, it was tested whether the self-interaction step could take place in the CFM and to measure an increase in Local Ligand Response (RLL) in the IBIS afterwards. Sensor #085 is used as 1st_print, the 1st_{print means that the ligand was immobilized in the CFM, the RLL of the spots are measured as a} blank. Then the sensor was replaced in the CFM again and a 2nd_{print with IgG was performed. The 2}nd_print means that the IgG was in the buffer solution of interest e.g. 5mM phosphate pH 6 and 5mM histidine pH 6. The solution flows through the micro channels and the protein can interact with the immobilized ligand. Then the RLL is measured again in the IBIS. An increase of RLL would indicate self-interaction of the protein. It was shown during this experiment that phosphate showed higher affinity than histidine but reproducibility was not tested.

To test reproducibility of this set-up a cover coupling was performed with an IgG concentration of 4 µg/ml. In a cover coupling the ligand is homogeneously divided over the sensor (#086). Then a print with the CFM was made with 250 µg/ml IgG in the buffer of interest. The RLL was measured after the printing to determine the amount of self-interaction at the different spots, and a decrease in RLL after regeneration was measured to check whether protein had bound. The change in RLL after regeneration could not be determined because the RLL remained the same after regeneration. The increase in RLL after the 2nd_print could be determined but the variance is very high. A prep plot is obtained during the stabilization of the IBIS, the RLL of each ROI is determined at the end of the stabilization. This was also done after the 2nd_print in the CFM and the results are shown in Figure 26. In Figure 26 it is clear that the variance of the RLL is high, the relative standard deviation is 47.55%. No further experiments were done with this setup as it would not provide a good alternative for buffer screening.

Figure 26: Determination of RLL after 2nd_{print in CFM. A cover}

coupling of the sensor was performed. Then the binding by self-interaction is done in the CFM. The RLL is measured in the IBIS after a prep run of 60 seconds. When the protein self-interacts it is expected that the RLL increases.

(29)

5.2 Bio-Layer Interferometry

Bio-Layer Interferometry (BLI) is another technique that can be used to determine protein self-interaction with a high-throughput. The measurements were carried out on a FortéBio Octet Qk384. The goal was to develop and optimize a high-throughput method to measure the amount of protein self-interaction with the least amount of material.

5.2.1 Concentration Optimization

Similar to the start with the SPR measurements, a start was made by defining the IgG concentration that needs to be immobilized and the IgG concentration that needs to be in the sample to get sufficient response. Eight AR2G sensors were used for the immobilization of a concentration range of IgG (0-200µg/ml). The measured samples contained 0-500µg/ml IgG in 5mM histidine buffer pH 6 or 50 mM phosphate buffer pH 6. Only small differences in binding response of the samples were measured, so possibly that the sensors were overloaded. The immobilization slope started at different speeds but eventually all sensors reached equilibrium (figure 27). The concentration IgG in the samples was lowered and the amount of immobilized IgG to the new sensor tips was also lowered.

Eight new sensors were immobilized with 20µg/ml IgG during 120 seconds and also reference sensors were made where no IgG was immobilized (figure 28). The interaction was measured in samples with 0-150µg/ml IgG in 5 mM histidine buffer pH6 and 5 mM phosphate buffer pH 6. The concentration range showed a linear relationship with the binding rate, but no difference between histidine and phosphate buffer was observed (figure 29). The binding rate is determinate by the initial slope of the association curve. The reproducibility of the 8 sensors with immobilized ligand was good with an average relative standard deviation of 3%. On the downside, the reference sensors showed also similar binding as the other sensors with immobilized ligand. This is probably caused by non-specific binding.

(30)

Figure 28: Immobilization curve BLI. Baseline 60 sec with H2O. Activation 300 seconds with

EDC/sulfo-NHS. Loading 30 seconds with 20µg/ml IgG in acetate pH 4.5. Quenching 300 seconds with ethanolamine. 2 times 150 seconds wash step with PBS.

(31)

5.2.2 Non-Specific Binding

It was tried to overcome non-specific binding by adding 0.1% BSA. At the same time the concentration of IgG in solution were lowered (0-20µg/ml). With a lower concentration of IgG in the samples and with the addition of 0.1% BSA there was still binding observed to the reference tips. The binding rate was even higher when BSA was added to the samples (figure 30).

Figure 29: BLI measurement of binding rate vs. IgG concentration. 0-150 µg/ml in 5mM Histidine buffer pH 6 and in Phosphate buffer pH 6. Immobilization 20µg/ml during 120 seconds.

Figure 30: Binding rate measured on the reference tips with and without BSA. When 0.1% BSA is added more binding is observed, thus non-specific binding is not reduced.

(32)

New sensors were immobilized with 3µg/ml IgG to prevent overloading of the ligand and the immobilization was stopped at 30 seconds to get a binding rate of 0.5nm immobilized IgG, similar to what is shown in literature [ CITATION Sun \l 1033 ]. 0.1% Casein was tested as additive with these new sensors. Casein is, like BSA, a protein with hydrophobic and hydrophilic sites that can surround the protein to prevent electrostatic interactions. The amount of binding to the specific sensor and the reference sensor did decrease with the addition of 0.1% casein. In figure 31 the binding rate of 15 and 50 µg/ml in 5mM phosphate pH 6 is displayed. The binding to the reference sensors and the specific sensors remains the same, but because a stable and fully developed protein was used it could not be said if specific self-interaction was present or not. Thus the addition of casein has to be tested with proteins that show much self-interaction like the DAR 4 specie.

Figure 31: The effect of the additive Casein on non-specific binding. The total amount of binding is reduced, but the reference tips show the same amount of binding as the sensor tips with immobilized ligand.

(33)

5.2.3 Sensor Setup

Different sensor setups were tried to determine if the amount of non-specific binding could have something to do with the setup of the sensors. It was found in literature [ CITATION Rei14 \l 1033 ] that quenching with ethylenediamine could prevent non-specific binding because the sensor is less negatively charged than after a quenching with ethanolamine. Also the old setup was tried where the quenching was performed with ethanolamine for comparison. Furthermore, the completeness of the quenching with ethanolamine was tested by binding with BSA or lysine after quenching. If the quenching was incomplete the BSA or lysine would bind to the free NHS-groups. A blank sensor was made to test if the antibody would stick to the sensor regardless of the activation or deactivation step. All sensors that were tested during this experiment are listed in table 1, an overview is provided with the steps of the immobilization for these sensors.

Table1: Overview of sensor immobilization procedure

Immobilization procedure Activatio n Immobilizatio n Deactivation Chec k Was h Blank PBS PBS PBS PBS PBS

BSA EDC/NHS BSA Ethanolamine PBS PBS

Check BSA EDC/NHS PBS Ethanolamine BSA PBS Check lysine EDC/NHS PBS Ethanolamine

Lysin

e PBS

Ethanolamine IgG EDC/NHS IgG Ethanolamine PBS PBS Ethanolamine Blank EDC/NHS PBS Ethanolamine PBS PBS Ethylenediamine IgG EDC/NHS IgG

Ethylenediamin

e PBS PBS

Ethylenediamine Blank EDC/NHS PBS

Ethylenediamin

e PBS PBS

Immobilization

During the immobilization the blank showed no binding as expected (figure 32, sensor: A1& A2). Immobilization of BSA to the sensor caused a wavelength shift of ±3.5nm (figure 32, sensor: B1 & B2). The immobilization of 3µg/ml IgG to the sensor caused a wavelength shift of ±0.8nm (figure 32, sensor: E1, F1, G1 & H1). After the blocking with ethanolamine, binding of BSA was observed to the Check

BSA sensor (figure 32, sensor: C1 & C2).

(34)

Self-Interaction analysis

The sensors with the immobilized IgG showed a linear correlation between the binding rate and the concentrations (0-150 µg/ml) IgG in 5mM phosphate pH 6. There was no difference observed between the sensors quenched with ethanolamine and ethylenediamine. There was also no difference in binding rate between the sensors with immobilized IgG and the blank sensors (figure 33). There was also binding observed by the IgG in the sample to the sensor with the immobilized BSA. And there was binding observed to the sensors that are dipped in lysine or BSA after quenching as a check to see if the sensor was entirely quenched. This all indicates that there is non-specific binding to the sensors no matter what the setup of the sensor immobilization is. As a follow up to prevent non-specific binding other types of sensors are used.

.

Figure 33: Binding of 150 µg/ml IgG in 5mM Phosphate pH 6 to sensors with different quenching solutions. The sensors with immobilized IgG and “blank” reference sensor were deactivated with ethanolamine and ethylenediamine. These quenching solutions do not affect the amount of binding.

Evaluation of Techniwues for the Assessment of Protein Self-Interaction

Evaluation of Techniques for the

Assessment of Protein Self-Interaction

Final Report version 1.0

Evaluation of Techniques for the

Assessment of Protein Self-Interaction

Final Report version 1.0

1.1 Abstract

1.2 Samenvatting

2 Introduction

Table of Contents

3 Background

3.1 Monoclonal Antibodies

3.2 Protein Stability

3.3 Surface Plasmon Resonance

4.1 Chemicals

4.2 Surface Plasmon Resonance

4.3 Bio-Layer Interferometry

4.4 Self-Interaction Chromatography

5 Results

5.1 Surface Plasmon Resonance