Mitigating the impact of antidrug antibodies against insulin on ELISA assay

(1)

Biomedical Science

Faculty of Health and Society Malmö University

SE-205 06 Malmö Sweden

Master programme in Biomedical Surface Science http://edu.mah.se/en/Program/VABSE

2021-05-26 Master degree thesis, 30 ECTS

Examensarbete, 30 hp

Mitigating the impact of antidrug antibodies against insulin on ELISA assay

Thea Sofie Bøwadt

SUPERVISORS: Prof. Lennart Ljunggren Biomedical Science

(2)

2 AUTHOR: Thea Sofie Bøwadt

ABSTRACT

Diabetes has, in the past three decades, surged immensely. Because of this, new insulin analogues are constantly in the making.

In clinical studies, the presence of antidrug antibodies can prove a challenge when measuring insulin. In order to overcome the interference from antidrug antibody complexes on the total insulin measurement in human serum, several pre-treatment methods on insulin and polyclonal antibodies spiked samples were tried using ELISA analysis.

Several different methods were tried, acid dissociation using a glycine buffer with and without ethanol in different concentrations, high ionic strength dissociation using MgCl2, Polyethylene glycol (PEG) and filtration.

The best results were found when using the acid dissociation technique. Using glycine promising results were achieved, especially when 20 % ethanol was added to the acid mixture. Pre-treatment using PEG, MgCl2 and filtration was unsuccessful with the methods used.

The main goal was reached through the use of glycine with the addition of 20 % ethanol for acid dissociation. The proposed method still leaves significant room for optimisation and needs further verification on real patient samples. However, it is a good step in the direction of a global methodology using ELISA to overcome antidrug antibody

interference for total insulin measurement in human serum.

Keywords: Enzyme-Linked Immunosorbent Assay, ELISA, Antidrug antibodies, ADA, Insulin, Polyethylene glycol, PEG, Acid dissociation, MgCl2.

This master thesis has been defended on June 3, 2021 at the Faculty of Health and Society, Malmö University.

Opponent: Prof. Tommy Eriksson Biomedical Science

Examiner: Prof. Sergey Sheev Biomedical Science

(3)

3

ABBREVIATIONS

Abbreviation Explanation

ELISA Enzyme-Linked Immunosorbent Assay

ADA Antidrug antibody

In Insulin

AB Antibody

Neg AB Negative IgG antibodies

Poly AB Polyclonal anti-insulin antibodies Mono AB Monoclonal anti-insulin antibodies

Cal Calibration Solution

MgCl2 Magnesium dichloride

PEG Polyethylene glycol

IU International Units

(5)

5

INTRODUCTION

Background

Diabetes mellitus (diabetes) is a primeval disease, as there are descriptions of it from ancient Egyptian and Greek writings [1, 2].

The name is derived from Greek and Latin and means “go through” and “sweet”. This refers to the oldest diagnosing symptom, glucose-rich urine or as sweet urine in ancient Egyptian and Greek texts.

Diabetes has, in the past three decades scourged immensely. The International Diabetes Federation estimated that globally 415 million adults had diabetes in 2015, and it is estimated to rise to 642 million by 2040. Of the 415 million adults with diabetes, about 46.5% lived in three countries: China, India, and the USA [3].

The reasons for the escalating scourge of diabetes are numerous: population ageing, economic development, urbanisation, unhealthy eating habits and inactive lifestyles, to name a few. It is associated with a reduction in lifespan of 8-years in the USA and has a negative effect on the quality of life as most patients also have complications [3].

Treatment for diabetes includes changing unhealthy eating habits and sedentary lifestyle, monitoring blood glucose levels regularly though the day and taking insulin to regulate the level. Monitoring blood glucose levels is done on a small point of care device that needs a drop of blood taken from the finger or ear. Insulin needs to be injected subcutaneously in the upper arms, abdomen, buttocks, or upper thigh. There is, therefore, a lot of invasive measuring and treatment through the skin [4]. Because of the massive number of people who need to take insulin, it is still intensively researched how to increase the quality of life and reduce the amount of needed medications.

In someone with diabetes, the β-cells in the pancreas are not functioning correctly. They have been destroyed by an auto-immune reaction or have developed insulin resistance either in the liver, muscles, or fatty tissue. This results in the development of

hyperglycaemia, leading to excessive glucose circulating in the blood [3]. This excessive blood glucose level leads to the oldest known symptom, sweet urine.

There are two types of diabetes: Type 1 and type 2. In type 1 diabetes, the β cells in the pancreas have been destroyed, meaning that the body cannot make insulin. In type 2 diabetes, insulin resistance has developed in the body.

Over 90% of diabetes cases are type 2 diabetes. Though there is increasing evidence that a percentage of people initially diagnosed with type 2 may, in reality, have a slowly

progressing and less severe form of type 1 (estimated at 5–15%) [1, 3].

Hyperglycaemia in people with type 2 diabetes can lead to alterations in sensing of glucose by the β-cell and a reduction in β-cell mass. According to Mathis et al. [1] the destruction of β cells may be involved in the occasional evolution of type 2 into type 1 diabetes.

(6)

6 Insulin

Insulin is a dipeptide consisting of an A chain with 21 amino acids and a B chain with 30 amino acids held together by disulphide bonds. It is secreted from the β-cells in the pancreas. Insulin reduces the liver's glucose output and increases the glucose uptake by muscle and fat tissue around the body [5].

Figure 1 Depiction of the insulin chains, A chain with 21 amino acids and a B chain with 30 amino acids held together by disulphide bonds [5].

Insulin concentration is most often given in International Units, written IU, or sometimes simply just U. This unit is based on bio-efficacy instead of the more normal mass/count- based units of the International System of Units.

The definition of insulin IU is the amount of insulin required to achieve a standard glycaemic effect. This has changed several times since insulin were discovered in 1922.

It was originally defined as the amount of insulin required to cause convulsive

hypoglycaemia in a 2kg rabbit that had fasted, which shows the standard for potency (IU per mg or equivalent). However, it has changed with improvements in the preparation and stabilisation of insulin in solution. Because of this, there have been different conversion factors between the conventional IU and SI units which can be traced back to differences in standards and purity of insulin preparations [6]. In this report, the SI unit that is used is in molar concentration in pmol/L. In order to better compare between experiments as well as the results in the articles, which also use pmol/L.

Antibodies

An antibody, also called immunoglobulin, is a Y-shaped protein. There are five main types of immunoglobulins: IgG, IgA, IgM, IgD and IgE, however for analytical methods like ELISA, IgG is most commonly used.

The basic structure contains two identical heavy chains and light chains linked together by di-sulﬁde bonds. The two antigen-binding sites on the antibody are called paratopes, while the place where the antibodies bind on the antigen is called the epitope. Antibodies are mainly produced by plasma cells [7, 8].

Antibodies bind antigens through weak chemical interactions. The interactions are non- covalent, e.g., hydrogen bonds, electrostatic interactions (ionic bonds and salt bridges), van der Waals forces, and hydrophobic interactions. Hydrogen bondings are considered to be the critical determining factor.

The strength of a single antigen-antibody bond is termed the antibody aﬃnity, which is a summation of the attractive and repulsive forces between the molecules and the number of attachments.

The avidity of an antibody for its antigen is determined by the sum of all individual interactions between individual antigen-binding sites of antibodies and determinants on the antigens. The difference between monoclonal and polyclonal antibodies, is that

(7)

7 monoclonal antibodies are all identical, and the polyclonal binds the same antigen but the paratopes are not identical thereby are able to binding to the antigen in different places [8].

Antidrug antibodies

Antidrug antibodies (ADA) are IgG antibodies the body makes that are targeted towards therapeutic proteins, i.e. a protein-based drug. Although ADA can be IgE, IgM and IgG, ADA mainly consists of IgG [9].

One of the problems with taking insulin analogues or any long-lasting protein-based drugs is that it increases the risk of the body making antibodies against it [10]. Any protein- based drug can cause the body to produce antibodies against it depending on the lifetime in the body and its concentration. This includes insulin since insulin is a protein.

Therefore, the body can interpret these insulin analogues as foreign and trigger the immune system.

ADA can be separated into two distinct classes: neutralising and binding antibodies.

The neutralising ADA’s recognise regions that are critical for the biological activity, therebyneutralising any effect, where the binding ADA binds to the protein but does not hinder the binding to the receptor, thereby letting the protein stay active [9].

In this study, both polyclonal and monoclonal IgG are used, but polyclonal were used primarily.

Being able to measure the total drug concentration even in the presence of ADA is important in clinical trials in order to get a drug approved [9].

Enzyme-Linked Immunosorbent Assay (ELISA)

The scientists Engvall and Perlmann essentially pioneered the ELISA method [11]. They developed the first ELISA method in 1971 by modifying the first-ever made

immunoassay, the radioimmunoassay (RIA), by conjugating the tagged antigen and antibody with enzymes instead of radioactive iodine.

The ELISA method was developed as an alternative to the much more expensive, labour intensive and radioactive RIA methods.

This technique pioneered other ELISA types and is also called Direct ELISA. It is suitable for determining the amount of high molecule-weight antigens and is used for antigen screening.

The surface of a microtiter plate is coated directly with the antibody or antigen. An enzyme tagged antibody or antigen enables the measurement. A sample is incubated for a time. The incubation is followed by washing which removes the unbound antigens or antibodies from the medium. Next, a substrate is added, most often horseradish peroxidase, to the medium to produce a signal through colouration. The signal is then measured to determine the amount of the antigen or antibody [11-13].

Following the invention, several researchers used the method to identify infections caused by influenza, parainfluenza, and mumps viruses in 1978, 1979, and 1981, respectively.

The method quickly found different fields of application and grew, and over time it has become a routine method in research and diagnosis in laboratories around the world [11, 12].

(8)

8 Sandwich ELISA

The technique was developed in 1977 by Kato and his co-workers. The method is called a Sandwich ELISA because the target protein is captured between two antibodies [14].

In a Sandwich ELISA, the wells are coated with a capture antibody and blocked to capture only the target antigen. Samples are added to the wells and incubated. After incubation, the wells are washed, thereby removing anything that is not bound to the antibodies in the well. Following a washing step, tagged antibodies with another specific enzyme to the antigen are added and incubated. After that, rewashed and the enzyme-substrate is added to reveal the enzyme activity by colouration [15].

This method's benefit is that it has an improved sensitivity, reported to be 2–5 times more sensitive than the other ELISA methods [15].

The drawback of this method is that for a sandwich ELISA to work, two antigen-specific antibodies are required that binds to two different epitopes on the antigen, which must be far enough from each other so as not to hinder the binding of one another. It is also possible to perform this technic in the opposite form, two different antibodies that capture a specific antigen [12].

The ELISA Insulin kit used was the Mercodia Insulin ELISA kit which is a solid phase two-site enzyme immunoassay.

The kit uses a direct sandwich technique utilising two monoclonal antibodies directed against different parts of the insulin molecule.

During incubation, the insulin in the sample reacts with the peroxidase-conjugated anti- insulin antibodies and anti-insulin antibodies bound to the microplate wells, making a kind of sandwich.

The washing step removes all the unbound enzyme labelled antibody. The bound conjugate is detected by reacting with 3,3’,5,5’-tetramethylbenzidine (TMB).

The reaction is then stopped by adding a stop solution of 0.5M Sulphuric acid (H2SO4), giving a colourimetric endpoint that is then read spectrophotometrically [16].

Internal and external standards

The Mercodia Insulin ELISA kit comes with multi-point external standards that are used to make a calibration curve. The calibration curve was prepared/used together with the samples every time the ELISA Test procedure was performed. It is impossible to calculate the amount of insulin in the samples if a standard curve is not measured simultaneously.

Slight differences occur every time an analysis is done as a few minutes more or less can change the absorbance measurements.

However, a problem with a calibration curve is that it can be challenging to detect a matrix effect. In this case serum is used that should not give a matrix effect, because the kit is made to measure in both serum and plasma. However, the samples are spiked with both insulin and antibodies as well as pre-treated with different solutions to remove antibodies from the sample. This could potentially cause a matrix effect.

A different method is the use of internal multi-point standard addition. This method has the benefit that it shows if there is a matrix effect, but a drawback of the method is that this must be done for every sample matrix and is labour intensive.

(9)

9 Therefore, this was only done in two samples, human serum that was not spiked with insulin and a sample spiked both with insulin and antibodies.

Research attempts to break the antidrug antibodies complex

Several different approaches have been tried to solve the interference of ADA in the measurement of therapeutic protein-based drugs over the years, but none of these have seemingly been compared to each other.

Casesnoves et al. [17] and von Schenck et al. [18] both tried to mitigate the problem using a combination of acid and PEG. They both used RIA, but Casesnoves et al. [17] used immunoradiometric assay (IRMA), and they used an ethanol-acid mixture of 87.5%

ethanol and 12.5% 2M HCl to dissociate the antigen antibody complex. They focused on which measuring method was best, RIA vs IRMA, rather than the dissociation method used [17].

Jordan G. et al. [19] compared the use of acid and MgCl2 to dissociate ADA complexes.

However, they were looking for a way to test which samples potentially contained the complexes rather than trying to measure the antigen concentration. In their article, the treatment with MgCl2 seem to work, but they are not using insulin as in our case for the investigated antigen, and the samples are consistently diluted 10 times, whereby lowering the dynamic range of the method.

The Aim

The study aims to mitigate the antidrug antibodies' interferences in quantifying insulin concentration in human serum and investigate possible sample pre-treatments to overcome antidrug antibodies (ADA) interference, using ELISA as an analytical tool.

The inspiration for this study came from a problem that originated in a clinical trial at Novo Nordisk, where they were using an ELISA type test to measure a new insulin analogue. However, due to antidrug antibodies' interference, the quantification of the insulin concentration was challenged in subjects with a high amount of antidrug antibodies.

In this study the interferes from antidrug antibodies is simulated using serum spiked with normal human insulin and anti-insulin antibodies.

Problem definition and objective

Is it possible to mitigate the problem of interference from the antidrug antibodies by pre- treating samples through different methods?

Relevance to the biomedical surface science area

The antibodies and proteins are dependent on the surface interactions of the molecules.

ELISA uses these surface interactions to make sensitive and stable analytical complexes.

ELISA is one of the most commonly used, analytical methods worldwide that is strongly dependent on biomedical surface science, through the surface interaction between the antigen epitope and the antibody paratope.

(10)

10 The combination of biomedical surface sciences, the ELISA technique and molecule surface manipulation through different chemical treatments makes the methodology highly relevant in the biomedical surface science area.

Ethical considerations

Since this project does not contain any test subjects or material related to specific persons, and the used biomaterials consist of serum donated from Novo Nordisk and insulin that Novo Nordisk has manufactured, there are no ethical concerns.

Being able to measure the correct amount of insulin even in the presence of antidrug antibodies is potentially important, especially during clinical trials, where the presence and amount of antidrug antibodies play a part in approving the drugs for commercial use. In addition, later potentially occurring antidrug complexes in patients increase the dose needed and can make the adjustment of drug doses in patients more difficult. As such, there are no ethical considerations.

(11)

11

MATERIALS and METHODS

ELISA procedure

Insulin in serum samples was determined with the Mercodia Insulin ELISA kit (10-1113- 01) donated by Mercodia, Uppsala, Sweden,

The procedure was run according to the manufacturer's instructions Mercodia Insulin ELISA - Directions for Use.

A calibration curve was made for each experiment from the premixed standard solutions contained in the Mercodia Insulin ELISA kit.

The calibration curve consisted of 6 points, a blank Cal 0 - 0 mIU/L (0 pmol/L) and the five other calibration mixtures with increasing amount of insulin: Cal 1 – 3.3 mIU/L (19.8 pmol/L), Cal 2 - 11.0 mIU/L (66.0 pmol/L), Cal 3 - 33.6 mIU/L (201.6 pmol/L), Cal 4 - 113 mIU/L (678.0 pmol/L) and Cal 5 - 217 mIU/L (1302.0 pmol/L). These came premade and ready to use. The concentration of the calibration standards was not written in the kit's instruction or anywhere else but instead written on the bottles included. The kit's

instructions also mandated that sample dilutions due to high insulin concentration should be performed using Cal 0 to ensure against interference to the assay.

The kit's measuring range is stated to be between 3-200 mIU/L (18-1200 pmol/L).

The microtiter plates were read at 450 nm using the Biotek Power wave XS microplate reader using the Gen 5, reader software from Biotek Life science Instruments Inc, Winooski, VT, USA.

Insulin

10 mL of certified human insulin standard was provided by Novo Nordisk, Måløv,

Denmark, in a concentration of 99.9 IU/ml, with a molar mass of 5808 g/mol, a potency of 28.8 IU/mg. Human serum was also received as a standard from Novo Nordisk.

The insulin was diluted in human serum using the conversion factor (1mIU/L=6.00 pmol/L) by Knopp et al. [6] into the first stock solution of 59 940 pmol/L, which was then diluted further to achieve three more stock solution, with a dilution factor of 1:250 000, 1:500 000 and 1:1 000 000.

At a later step, these three stock solutions will be diluted one final time, one to one, before use to achieve an experimental concentration within the range of the ELISA kit.

Thereby having an end dilution of 1:500 000, 1:1 000 000 and 1:2 000 000, with a

calculated insulin concentration of ~1200, ~600 and ~300 pmol/L, respectively (calculated without taking the pre-existing insulin in the serum into account).

All insulin stocks and the insulin spiked serum samples were stored at -80°C, and before use, were thawed, vortexed, and left at room temperature for one hour.

The non-spiked serum was measured in every experiment to determine the storage stability.

(12)

12 Standard addition

The standard addition experiment was made by adding 0, 1, 2 and 3 µl portions of 646 pmol/L insulin to individual 30 µl samples of serum and insulin-antibody spiked serum, resulting in an increase of the insulin concentration of 20.8, 40.4 and 58.7 pmol/L, respectively. Thereafter, the resulting insulin amount was measured by ELISA.

Antibodies

Three different types of mouse antibodies were provided by Mercodia AB, Uppsala, Sweden, Monoclonal Mouse anti-insulin (20.0 µmol/L), Polyclonal Mouse anti-insulin (29.3 µmol/L) and polyclonal Negative (non-reacting) Mouse Antibodies (4.0 µmol/L).

The antibodies were diluted in human serum. The molecular weight for mouse IgG

=150 000 g/mol.

Initial experiments were done to determine the amount of antibodies needed to complex (neutralise) the spiked samples.

The antibodies were diluted in serum in a series dilution, and before measurement, all samples were vortexed and incubated for 1 hour.

The series dilution started by diluting the polyclonal antibodies 1:10 to a concentration of

~2 935 290pmol/L. Following, a one-to-one dilution was done a total of 14 times, thereby spanning from a concentration of ~2 935 290 pmol/L to a concentration of ~358 pmol/L.

All samples, serum and stock solutions were stored at -80°C and, upon use, thawed, vortexed and left for one hour before use.

After the initial experiments, it was decided to use a 15 000 pmol/L concentration of antibodies for all experiments containing antibodies.

The final step to making the spiked insulin samples was to mix one to one insulin and antibodies (or serum where antibodies were not used), thereby making sure the same amount of insulin is the same in both spiked insulin-AB and spiked insulin samples. This was done for the three stock insulin dilutions (1:250 000, 1:500 000 and 1:1 000 000) whereby obtaining a calculated insulin concentration of ~300, ~600 and 1 200 pmol/L, respectively to the human serum. Hereafter referred to as spiked insulin samples with or without antibodies.

Serum pH change with glycine

Serum acts as a natural buffer, therefore an experiment to determine the effect of acid addition on pH was needed to ensure adequately low pH in the actual experiments. Sample volumes in ELISA were too small to allow pH measurement, and as such, a titration experiment was performed using serum and glycine 0.1 M, analytical grade, from Fisher Scientific, Leicester, UK, adjusted to pH 2.0 with 1 M HCl.

To 1 ml of pure serum,1 ml portions of 0.1 M glycine pH 2.0 were added and the pH of the solution recorded before and after each addition (8 times).

After arrival at a near minimum pH (pH 2.2) for the glycine solution, Tris 1M pH 8.5, Tris(hydroxymethyl)aminomethane (TRIS), ultra-pure, Saveen & Werner AB, Malmö,

(13)

13 Sweden was added in 100 µl increments (11 times) until arriving back at the initial pH of the serum (pH: ~7.8).

Dissociation experiments

Acid dissociation using glycine and glycine-ethanol

Acid dissociation experiments were performed using glycine 0.1 M, analytical grade, from Fisher Scientific, Leicester, UK, adjusted to pH 2.0 with 1 M HCl.

The different mixtures of glycine-ethanol were made using glycine 0.1 M pH 2, analytical grade, Fisher Scientific, Leicester, United Kingdom, were made with ethanol 99.5% from analytical grade, Solveco, AB, Rosenberg, Sweden, in percentage v/v%.

The 10% mixture was made with 10 mL ethanol 99.5% and added 90 mL of glycine 0.1 M, pH 2.0. The 20% mixture contained 20 mL ethanol 99.5% added 80 mL of glycine.

Neutralisation of the acid was each time performed using Tris 1 M buffer.

Using 0.1M glycine with a pH of 2 and 1M Tris with a pH of 8.5 was inspired by

Hoffmann et al. [20], where they diluted the sample 1:5 with the glycine. Adding ethanol to the acid was inspired by Casesnoves et al. [17].

Samples were diluted 1:2, 1:3, 1:4 1:5, 1:6 and 1:7 in either acid or acid-ethanol, vortexed and incubated for 30 min, followed by centrifugation for 2 min at 10 000 x g. Some were transferred to another Eppendorf tube and neutralised with Tris as below, vortexed to ensure homogeneity and then measured via the Insulin ELISA kit.

Two different approaches were used, either 50 µL of the sample-acid mixture was taken after incubation, and 5 µL base was added to neutralise the sample, or 60 µL of the sample-acid mixture was taken after incubation, and 10 µL base was added to neutralise the sample. However, in practice, the resulting pH in the samples was the same.

High ionic strength dissociation with MgCl2

The aqueous MgCl2 solution was made by dissolving 44.66 g of Magnesium Chloride hexahydrate, analytical grade, Merck, Darmstadt, Germany in 55 mL Millipore H2O, resulting in a concentration of 4 M MgCl2 (used as stock solution).

Spiked insulin samples were diluted 1:2, 1:3, 1:4 and 1:5 with MgCl2 solution. Then it was vortexed and incubated for 30 minutes, followed by centrifugation for 2 minutes at 10 000 x g. Thereafter, 50 µL were transferred to another Eppendorf tube and diluted with 50 µL of Cal 0 (in line with kit instruction), vortexed to ensure that it is homogenous and then measured via the Insulin ELISA kit. [19].

Verification of potential interference of MgCl2 on the ELISA kit was done through a dilution of the MgCl2 then mixed with the spiked insulin samples.

The MgCl2 stock solution was first diluted 1:5 and 1:7.5 in Cal 0, then further adding an equal amount of spiked insulin sample so that the MgCl2 end dilution of MgCl2 was 1:10 and 1:15 at the stage of measurement.

(14)

14 Precipitation and Filtration

Polyethylene Glycol precipitation

Polyethylene Glycol (PEG) was used after the dissociation step in order to remove the antibodies without removing the insulin together with the antibodies since any insulin- antibody complexes would also precipitate. The PEG 40% was prepared by dissolving 20g of PEG6000 from Sigma-Aldrich Merck, Darmstadt, Germany, into 50mL of Millipore H2O.

PEG was used after either acid or MgCl2 had been added, incubated, and centrifuged.

50µL was transferred to another Eppendorf tube, 25µL PEG was added, then vortexed well, centrifuged for 2 minutes at 10 000 x g, then analysed.

Using the PEG in combination with acid was inspired by the articles by Hayford et al. [21]

and von Schenck et al. [18].

Filtration experiment

Spin column Filtration was tried using Centrifugal Filters, Amicon® Ultra – 0.5 mL 50K from Sigma-Aldrich Merck, Darmstadt, Germany.

The procedure was used according to the manufacturer's instructions, and both phases were measured using the ELISA insulin kit.

Potential interference from pre-treatment chemicals on ELISA

The insulin calibration solutions 201.6, 678.0 and 1302.0 pmol/L (Cal 3, 4 and 5) were used to detect potential interference from the pre-treatment with acid, MgCl2 and PEG, respectively. All done as single experiments except for the highest concentration that was repeated.

In the case of the acid, 30 µL of 0.1M glycine pH 2 was added to 10 µL 1M pH 8.5 Tris buffer added to 30 µL of each of the calibration solutions and subsequently measured by the ELISA kit.

For the MgCl2 the 4M stock solution was diluted 1:2, making a concentration of 2M and 30 µL of this was added to 30 µL of each of the calibration solutions and subsequently measured by the ELISA kit.

30 µL PEG stock solution (40 %) was added to 30 µL of each of the calibration solutions and subsequently measured by the ELISA kit.

For each experiment, a control blank was made with Cal 0 in a 1:2 dilution to mimic the different experiments' end concentrations.

For all the experiments performed in this study, the individual dilution factors were used to calculate the insulin concentrations.

(15)

15

RESULTS

Insulin in serum and insulin dilution comparison

The difference in the insulin measurement of the non-spiked serum on the different experimentation days was measured 18 different times, giving an average of 110.2 ± 10.2 pmol/L. This also confirmed the stability of the insulin in the serum used throughout the experiments.

The spiked insulin samples throughout the experiments had three different concentrations of insulin, and they measured at 1084.2±35.1 pmol/L (n=4), 646.1±34.0pmol/L (n=6) and 412.3±19.9pmol/L (n=7).

From/through all the tests, it was evident that the calibration curve from the kit was linear from 10–1000 pmol/L and had a linear regression coefficient better than 0.97.

Antibody experiments

The objective of these experiments was to verify the ability of the insulin and the

antibodies to bind. A reduction of 50 to 80% of insulin measurement by adding antibodies was aimed at. Table 1 below shows the pre-experiment of the antibody concentrations. It shows that there were too little antibodies in Test 1. Test 2 had a much large amount of antibodies in order to ensure that insulin-antibody complexes were forming.

To this end, a serial dilution of the antibodies was used. First, the polyclonal antibodies were diluted in serum using serial dilution, then the antibody samples were mixed with the insulin spiked samples, vortexed well and incubated for one hour. After incubation, the samples were measured via the Mercodia Insulin ELISA kit. Figure 2 on the next page, show the measured insulin concentration was decreased by more than 50%, showing that a concentration of antibodies of 15 000pmol/L is sufficient to decrease the available insulin concentration with the desired amount.

Table 1. Pre-experiment of the antibody concentration needed. Test 1 clearly shows that the difference between sample without AB and with Poly AB and Neg AB was negligible. Test 2 was made to test that a high amount of antibodies completely bound the insulin. Test 3 gave a reduction of insulin that is within the target set for the experiment. n=1.

Test 1

Sample without AB 415 0

Sample + Neg AB 414 75

Sample + Poly AB 420 73

Test 2

Sample + Neg AB 406 100068

Sample + Poly AB 0 366911

Test 3

Sample + Poly AB 1 153 22932 Sample + Poly AB 2 215 11466 Sample + Poly AB 3 289 5733

AB conc.

pmol/L In Conc.

pmol/L

(16)

16 Figure 2. Results of the Insulin-Antibody experiment that was made to determine the amount of antibodies that was needed to bind insulin in the spiked samples within 1 hour. There need to be enough antibodies to impact the measured insulin amount, but not so much that no insulin could be measured.

A 1:2000 dilution of the concentrated polyclonal antibody solution gives a concentration of antibodies of 15000 pmol/L, which fits well within the selected interval for decrease of insulin in serum. n=1

Standard addition

In order to determine if the serum matrix had an influence on the measurements of insulin in the test samples, a standard addition experiment was conducted both with the blank serum as well as in a sample spiked with antibodies and insulin. Results are shown in Figure 3, below.

Figure 3. Results of the Standard addition. The X-axis shows the added amount of insulin in pmol/L, and the Y-axis shows the absorbance that was measured. The starting point for the blue curve is pure serum, to which 1µL containing 20 pmol/L insulin was added repeatedly. In the orange curve, the same was done with insulin-antibody spiked serum. N=1.

(17)

17 Experiment how serum pH change with glycine

The titration experiment was performed in order to determine serum pH changes with acid addition, since serum is a natural buffer. The results are shown in Figure 4 and Table 2 below.

Figure 4. Graph of the Serum and Acid experiment on how pH changes with addition of glycine (pH 2) to serum. The X-axis shows the acid molar concentration, and the Y-axis is the pH. This experiment gives an approximate pH in the real samples (too low volume to allow measurement, 0.5-1mL). n=1

Table 2. Measured pH results as well as the calculated concentration of the glycine and the amount of Serum to glycine.

Results from acid pre-treatment

Results from the acid dissociation of the complex of insulin and the antibodies are shown in Table 3 on the next page. The best results obtained were in the first test conducted, where the sample:acid ratio was 1:7. Here, the recovery percentages for the samples with antibodies present were roughly the same as those without any added antibodies.

However, in the second test, where a higher concentration of insulin used, the recovery was lower for the higher sample:acid ratios.

Serum 0 7.78 N/A 1000

Glysine 0.050 3.90 1:2 2000

Glysine 0.067 2.91 1:3 3000

Glysine 0.075 2.60 1:4 4000

Glysine 0.080 2.45 1:5 5000

Glysine 0.083 2.36 1:6 6000

Glysine 0.086 2.30 1:7 7000

Glysine 0.088 2.26 1:8 8000

Glysine 0.089 2.22 1:9 9000

volume µl Ratio of

Serum:Acid Measured

pH Calculated

Gycine conc.

M

(18)

18 Table 3. Results of the two experiments performed with glycine. The test was made two times with different spiked insulin concentrations. Test 1 had a concentration of 682 pmol/L, and test 2had 1065 pmol/L. (n=1).

Results from acid/ethanol pre-treatment

The final results shown in the Table 4 below are the results from the experiments using both acid and ethanol to dissociate the insulin-antibody complex. This was conducted with polyclonal antibodies as well as with monoclonal antibodies. In order to verify that the ethanol acid addition did not induce any bias, the experiment was also performed without the addition of antibodies. Monoclonal antibodies are, in general, expected to bind with a higher affinity to insulin than polyclonal antibodies. As such, it is not surprising that the recovery percentage is not as high as it is for the polyclonal antibodies. It is noteworthy that recovery between 97 and 105% can be achieved if 20% ethanol is added to the acid. If only 10% ethanol is added, the results are comparable to the acid experiments without the addition of ethanol.

Table 4. Part A, the top part of the table, shows the results when using 10% ethanol (EtOH) in the acid, glycine pH 2.0, to make the acid dissociation of the Poly or Monoclonal antibodies towards insulin. The spiked insulin samples were measured to 1084.2. The average was measured four different times (n=4).

Part B shows 20% ethanol in the acid used to make the acid dissociation of the Poly or Monoclonal antibodies towards insulin. The spiked insulin samples were the same as in Table 1A (n=4).

Sample:Acid no AB poly AB no AB poly AB

1:4 655.2 504.0 96.1 73.9

Average: 682.1 1:5 733.5 532.7 107.5 78.1

SD: 31.3 1:6 630.6 579.6 92.5 85.0

SD % 4.6 1:7 626.5 634.7 91.9 93.1

1:4 964.8 796.5 90.6 74.8

Average: 1064.9 1:5 973.8 828.0 91.4 77.8

SD: 49.0 1:6 972.0 865.6 91.3 81.3

SD % 4.6 1:7 965.1 828.3 90.6 77.8

recovery %

Test 2 Test 1

messured

No Antibodies Polyclonal Monoclonal A

1:5 1061.4 ± 38.4 985.9 ± 44.7 667.4 ± 244.6 98 ± 6 91 ± 2 61 ± 33 1:6 1067.7 ± 46.9 1012.7 ± 44.8 696.4 ± 266.0 99 ± 6 93 ± 4 63 ± 34 1:7 1047.4 ± 53.8 989.9 ± 49.9 764.4 ± 209.9 97 ± 8 91 ± 8 70 ± 24 B

1:5 1116.8 ± 38.3 1048.1 ± 80.3 818.3 ± 228.4 103 ± 6 97 ± 8 75 ± 24 1:6 1134.5 ± 98.2 1089.9 ± 50.1 813.2 ± 209.4 105 ± 9 100 ± 5 74 ± 22 1:7 1127.9 ± 93.8 1137.1 ± 80.5 780.9 ± 271.8 104 ± 9 105 ± 7 71 ± 31

Sample : EtOH/Acid Recovery %

No Antibodies Polyclonal Monoclonal Insulin pmol/L

(19)

19 Potential interference from pre-treatment chemicals on ELISA

To check if the different chemical solutions used for potential dissociation and

precipitation experiments would interfere with the insulin measurements, an experiment was made with three different calibration solutions. The results are shown in Figure 5 below shows that especially PEG and MgCl2 has a significant influence on the results obtained. Results were all calculated using the dilution factor.

Figure 5 Illustration of the interference from the different methods used. Different reagents were tested to see if they interfered with the measurement of insulin. The green column is the stated amount of insulin from the kits Instruction. The blue column shows the control (measured amount of insulin with nothing to interfere). The yellow column shows interference from the glycine and Tris. The orange and purple columns show that PEG and MgCl2 interferes. (n=1)

(20)

20

DISCUSSION

Basic starting decisions and experiments

At the start of the experimental part of the project, everything had to be set up from scratch, and several things were still unknown. Working with insulin in the lab was new, the ELISA kit had not been used before, and no previous projects about separating insulin- antibody complexes had been done prior to this project. Therefore, all of the applied procedures had to be established in the laboratory.

Several key decisions had to be made prior to starting in the laboratory, based on

availability of ingredients and equipment in the lab and the project's overall aim: to be able to mitigate effects from ADA complexes for insulin measurements, all based on the

available literature.

The first decision was to use human serum instead of phosphate-buffered saline (PBS) as the medium for all experiments and dilutions. This decision was taken to allow the method to mimic a real situation with ADA complexes from patients as closely as possible. In principle, we could also have used plasma, but decided against this to avoid possible interference from the addition of anticoagulants (e.g. heparin or EDTA) as well as possible effects from the remaining fibrinogens. As such, human serum was used throughout the project, and as a test of the comparability between the experiments, it was decided to include a serum blank (pure human serum) in each series. These analyses showed that the standard serum supplied by Novo Nordisk contained ~110 pmol/L. Over 18 independent measurements, the value was determined to be 110 ± 10 pmol/L. This lies well within the normal level of insulin in fasting and healthy humans [22].

Insulin and antibody concentrations used throughout the experiments In a second step the insulin and antibody concentrations to be used in the experiments had to be determined.

In order to have the optimal dynamic range of the insulin concentration, it was decided to aim for a concentration in the upper part of the dynamic range at 500-600 pmol/L.

Accordingly, it was decided to use an equivalent concentration of polyclonal antibodies as this was believed to lead to a complete complexation of the insulin within a 1-hour time interval. However, the results showed that the addition of the equivalent antibody concentration did not lead to any measurable lowering of the insulin content in the samples. As such, it was decided to add, after linear dilution of the antibodies in serum, different levels of antibodies to determine how much had to be added in order to find the dilution that would achieve a near 100% complexation of insulin by the antibodies within a 1-hour time interval, as seen in the methods.

The result of this experiment can be seen in Figure 2, on page 16. From this experiment, it was decided to use a 20-fold excess of antibodies resulting in a concentration of ~15 000 pmol/L. This should ensure that the ADA complexes would be achieved within a

reasonable time interval, and most of the available insulin would be bound by the antibodies.

Figure 2, on page 16, clearly shows that there is a logarithmic relationship for the complexing reaction [23]:

𝐼𝑛𝑠𝑢𝑙𝑖𝑛 + 𝐴𝑛𝑡𝑖𝑏𝑜𝑑𝑦 𝐴𝐷𝐴

(21)

21 This is in line with the energy relationship for the reaction:

∆𝐺 = −𝑅𝑇𝐿𝑛𝐾

∆G must be negative for the reaction to proceed towards the complex.

The experiment shows that the time to achieve a chemical equilibrium under the conditions chosen is likely higher than 1 hour, because otherwise the decrease of the insulin concentration in the samples would be more pronounced and/or a smaller concentration of antibodies would be needed.

Standard addition experiment

Standard addition is an internal multi-point standard addition, meaning a known amount of analyte is added to a sample and measured to see if something in that sample system has an impact on the measurement. Thereby showing if there is a matrix effect.

In this case, only insulin was measured and was a test to see if we could find the initial insulin concentration in the serum by looking at the absorbance and plotting it against the added insulin concentration.

To minimises the effect the volume change would have on the concentration, 1, 2 or 3 µL was added to individual 30µL serum samples, changing the volume by a maximum of 10%.

The resulting curve (blue line), as visualised in Figure 3 on page 16, has a regression coefficient better than 0.95.

Tracing the line back to where it intercepts the x-axis (y=0), it can be seen that we have approximately 85 pmol/L of insulin in the serum. This is expected as all the individual measurements gave an average value of 110 ± 10 pmol/L, and given the uncertainty of the standard addition curve, this must be considered as reasonable. In fact, on this particular day, the independently analysed blank serum gave a result of 90 pmol/L, which verifies the results obtained in the standard addition experiment.

As such, it can be deduced from the results obtained that there was no matrix effect when blank serum was used. This was expected since the kit has been conceived for serum and plasma samples.

To determine if the matrix influenced the measurements of insulin in test samples containing both added insulin and antibodies (insulin conc. 650 pmol/L and polyclonal antibody conc. 15000 pmol/L), a standard addition experiment was further conducted in the same manner as above for the pure serum sample. The resulting curve (orange line), as visualised in Figure 3 on page 16, has a regression coefficient marginally below 0.95.

Tracing the line back to where it incepts the x-axis (y=0) shows that there should have been approximately 285 pmol/L of insulin in the serum spiked with insulin and antibody.

This seems to be in line with the results shown in Figure 2, where ~ 42 % of the insulin is free in solution, which is equivalent to ~ 270 pmol/L when starting with 650 pmol/L of spiked insulin to the 15 000 pmol/L antibodies. However, the slope of the curve is significantly lower than in the previous experiment with pure serum. As such, it is clear from the results obtained that in this case, there is a rotational matrix effect when serum spiked with both insulin and antibody was used [24]. This is likely due to the insulin reacting with the antibodies to form ADA complexes and hence shifting the equilibrium for the reaction.

(22)

22 In retrospect, changing the volume even less by using a larger sample volume to more concentrated insulin could have given more accurate results and a better linear regression coefficient. Also, adding more points on both curves, or making duplicates would have made the results more trustworthy.

This is a time-consuming experiment, but it could have been beneficial to do this

experiment with the different pre-treatment chemicals to determine if a matrix effect was involved with the pre-treatment chemicals. Unfortunately, it was not possible due to lack of time and the number of remaining ELISA kits.

Serum pH change with glycine

As the bonding between insulin and antibody depends on several different weak bonding types, mainly hydrogen bonding, it was decided to check the pH of mixtures of serum and different amounts of glycine, the acid chosen for the acid dissociation experiments. Table 2, which can be seen on page 17, shows the pH of the different amounts of a 0.1 M glycine with serum. As serum is well known to act as a buffer[25], it can be seen that it takes a surplus of acid to achieve a pH believed to be low enough to achieve an efficient dissociation of the complexes [20].

Acid affects the pH of blood, serum, or plasma in a very complex manner as at least two buffer systems are in operation [25]. Therefore, a serum-acid experiment was performed as the sample volumes used in all the ELISA experiments were too small to be able to be measured with a pH meter. Making samples of a larger volume for all acid experiments would be a waste of insulin and antibodies, but at the same time, it was critical to know the approximate pH when adding a certain glycine acid to serum ratio. Therefore, this experiment was performed instead.

Antibody-antigen complexes are hard to break, but should be achievable at a certain low pH or if other substances, such as salt, disturbs the complex bindings [20]. Therefore, it was paramount to be able to achieve an approximate pH in the samples to see when pH was low enough to achieve dissociation. At the same time, it is also very important to ensure that the low pH does not denature the insulin. As shown in Figure 4 and Table 2, on page 17, pH did not go below pH 2.3 at any point, so the acidity should not affect the stability of insulin [26, 27].

Acid Dissociation

The most promising method to dissociate ADA complexes was found to be the addition of acid to disrupt the bonds in the complex [9, 18, 20, 21, 28]. One of the most common middle strong organic acids that are commonly used throughout bioanalytical labs and in protein experiments is glycine, and it was decided to use a 0.1 M glycine adjusted to pH 2 as the starting point for the experiments. After measuring the approximate pH in the different ratios of 0.1 M glycine with serum, a series of experiments were performed in an interval from 1:4 to 1:7 ratios of samples with ADA complexes to 0.1 M glycine buffer.

As can be seen from Table 4, it is clear that a high ratio of acid to sample was necessary to achieve a more complete dissociation. In essence, it seems that 1:7 yields the best results with up to >90% recovery depending on the starting insulin concentration. The negative antibodies were used to show that no spontaneous ADA complexes with insulin were

(23)

23 made with high concentrations of non-binding AB, e.g., that only specific anti-insulin antibodies interfered with the insulin measurements.

Potentially better results could have been achieved with a higher ratio of glycine to sample or with a lower pH [26]. However, instead, it was tried to add ethanol to the glycine buffer to see if this would improve the recovery. The reason for this was an inspiration from the article by Gelsema et al. [29], where ethanol had been used together with acid for the dissociation of insulin antibody complexes with good results.

As shown in Table 4 on page 18, already, the addition of 10% ethanol gave significantly better results than the pure glycine experiments, even at a lower sample:acid ratio. With 20% ethanol addition, the results improved to achieve a near quantitative recovery of insulin from the polyclonal ADA complexes, while the monoclonal complex dissociation was improved in comparison to the experiments with 10% ethanol. Further increasing the concentration of ethanol in the acid buffer mixture was attempted (up to 50%), however, the results did not improve (results not shown). Rather, they became more unreliable, which was in line with results from Gelsema et al. [29] that showed the pH of the mixtures to change significantly above 30% ethanol. As such, in our experiments, the sample:0.1 M glycine / 20% ethanol ration of 1:6 seemed to yield the best recovery and the most reliable results.

The near quantitative recovery of insulin from the polyclonal ADA complexes was between 90 to 100%, while the monoclonal ADA complex dissociation was only between 60 and 75%. This could be because the monoclonal antibodies have a higher affinity to the insulin, whereby making it harder to dissociate the complex.

Centrifugal filtration

Centrifugal filtration was also attempted for the separation of insulin and antibodies in the samples. The filtration tests were tried several times, but they yielded inconsistent results.

Most of the time, the insulin was found in both the supernatant and in the filter while in some experiment the insulin was only found in the filter, which was where the antibodies also were located. Therefore after a few tests, the filtration method was discarded.

The filtration kit system was not intended for separation. It was intended to concentrate proteins that were larger than 50 kDa, e.g. antibodies that are 150 kDa. The hope was that these large proteins would be retained in the filter and that the much smaller insulin, with only 5808 Da, would pass through the filter. This turned out not to be the case.

Potential interferences on the ELISA kit

Based on previous literature, PEG had shown promise for the separation of protein complexes [17, 18, 21, 30], but it was discarded after the interference test results, which can be seen in Figure 5 on page 19 due to significant interference from PEG on the ELISA kit.

In Figure 5, it can also be seen that for PEG, the results for the interference were

inconsistent, e.g. for Cal 3, results showed that adding PEG gave approximately the right amount of insulin, while for both Cal 4 and Cal 5, a lower amount of insulin was

measured.

(24)

24 Perhaps if PEG had been diluted further or if the PEG could have been filtered out, it might have changed the amount of interference, but this was not possible due to time constraints and the available amount of ELISA kits.

In the same experiment, MgCl2 showed an even higher interference than PEG, whereby strongly lowering the amount of insulin measured. This is likely due to having too much salt in the mixture, effectively preventing the binding of the antibodies in the ELISA kit to the insulin, yielding a basically blank result at an MgCl2 concentration > 3M. Even with a 1:15 dilution resulting in a 0.3 M MgCl2 concentration, the result for insulin was still diminished >15% (result not shown).

Further it was considered that the use of MgCl2 for ADA dissociation experiments would lead to a significant reduction of the dynamic range of the ELISA kit (factor of >10).

Potential further improvements of the method – the way forward to a standard method

Higher glycine concentrations or use of a different acid.

Additional experiments could be made, trying out more concentrated glycine at pH 2 to dilute samples as little as possible. Following this line, the acid/ethanol balance should also be tested with the more concentrated acid, while testing that this would not lead to ester formation between glycine and ethanol.

The acid experiments showed very promising results, and an alternative would be to use different acids. This is definitely worth exploring in order to lower the amount of added acid and thereby avoid the lowering of the dynamic range of the ELISA kit.

Different acids, such as a stronger acid could be considered including acids commonly used in biochemical labs, while controlling the pH accurately. Alternatively, glacial acetic acid used in the article by Hoffmann et al. [20] or HCl in the article from von Schenck et al. [18]. In both cases, it would be optimal to try and use a high concentration of these acids so as not to dilute the samples more than needed, but still reaching a pH low enough to achieve the dissociation of the ADA complex. At the same time, it is important to make sure that the insulin does not denaturate during the experiments [26, 27]. So having an untreated sample, and a sample without added antibodies as control is paramount in the setup.

Use of other filtration systems.

The concept with filtration systems could still be valid and should be tested further, whereby trying other filters since being able to filter out the antibodies after dissociation could make the measurement of insulin easier as there would not be a time factor involved where antibodies could potentially re-attach to the insulin. Thereby potentially making insulin measurement more efficient through different analytical techniques, such as ELISA, Chemiluminescence immunoassay (CLIA) or AlphaLISA, as well as make it possible to measure in sample flow systems.

Longer incubation time.

Different incubation times for the acid dissociation experiments should also be looked into. In the experiments with acid, PEG and MgCL2, the incubation time was only 30 minutes due to time constraints. However, it would be interesting to see if results change simply by incubating longer with the dissociation reagents, thereby possibly allowing for a

(25)

25 better dissociation of the ADA complex. Lengthening the incubation time might yield better and more stable results since the end goal was to remove the anti-insulin antibody consistently from the samples.

Heat treatment

Heat was tried at the beginning of the project on insulin samples but was very quickly abandoned. The reason was that the heat in the sample could not be controlled precisely enough with the equipment available. Dissociation via heat is probably a more theoretical method than a truly practical method due to the small interval between the denaturation of IgG (62 °C) and insulin (70 °C) [31, 32].

Acid dissociation while heating the sample, has also been considered, but according to Whittingham et al. [26] heating insulin while the pH is low results in unstable insulin and degradation and/or denaturation.

Storage of insulin and the antibodies in the fridge instead of a freezer.

Determining how long the insulin is stable at 4°C in the fridge versus frozen at -80°C could be beneficial in order to lower the time for the experiments and, in general avoiding the many freeze-thaw processes.

Testing on real samples

Finally, the method should be tested on real samples to verify the overall methodology.

The overall goal of the method is to verify real-world applications, and as such, it is mandatory to try it on real patient samples. Even if the method seems to work on the spiked samples used in this study, it cannot be assured that it would also work with the naturally occurring ADA complexes that could have a substantially different binding affinity.

This is especially true if the method should work universally in laboratories across the world where minor differences in the application of the method could lead to different results due to different surface interactions of the insulin-antibody system.

(26)

26

CONCLUSIONS

The biomedical problem dealt with in this thesis originates from a real clinical trial, that had to be changed/terminated due to development of ADA in the participants. The aim for this project was therefore to find a possible way to overcome the interference and to determine the total amount of insulin in samples contained free in the serum and bound in ADA complexes by help of a “simple” ELISA method. The project tried to mimic the problem in the lab using spiked samples, and then to reverse the ADA interferences problem in measuring the insulin concentration after pre-treatment with reactants in different ways.

The main goal was reached with the use of glycine for acid dissociation and the addition of 20% ethanol. This method showed very promising results, although more work needs to be done in order to be able to determine if it would work also on real-world samples, which were not part of this project. From the onset the project was quite ambitious

considering the available time and resources and it became clear that the project was more labour intensive than anticipated due to limited previous experiences with the sample type and the methodology used.

There was a lot of preparatory work that had to be done before the experiment to separate the ADA complexes could be initiated. This also meant that a good deal of time and resources were spent “just to get started”, which however is the nature of a good research project.

In order to obtain a simple, reliable and stable method that can be used globally for removing anti-insulin antibodies consistently from samples and measure the total insulin without interference, a number of items still have to be further checked and optimised.

The most important extension of the work and last step is to verify the method on actual patient samples containing ADA complexes.

Acknowledgements

Making a Master Project during the corona pandemic has shown to be a challenge larger than expected. Of course, every project has ups and downs and unforeseen challenges, but unforeseen challenges seemed to have been waiting at every corner through this pandemic.

Nevertheless, with help and support and hard work, it became a reality.

Thanks to Malmö University for everything throughout the whole master program. Thanks to Novo Nordisk for donating insulin, ELISA Insulin kits and Serum and to Mercodia for donating antibodies for the experiments.

A big thank you to my wonderful supervisor from Malmö University, Prof. Lennart Ljunggren.

Thanks to PhD Jen Sigh and PhD Lene Andersen from Novo Nordisk for your guidance and help.

A special thanks go to my family for putting up with me during this time and the people who helped me proofread. Also, thanks for the heart-warming help from friends who gave me a place to stay when I got stranded in Sweden on the second last day of my lab work when the border between Sweden and Denmark was closed.

I am very grateful for all the help and support received.

(27)

27

REFERENCES

1. Mathis D, Vence L, Benoist C. beta-Cell death during progression to diabetes.

Nature. 2001;414(6865):792-8.

2. Stylianou C, Kelnar C. The introduction of successful treatment of diabetes mellitus with insulin. Journal of the Royal Society of Medicine. 2009;102(7):298-303.

3. Zheng Y, Ley SH, Hu FB. Global aetiology and epidemiology of type 2 diabetes mellitus and its complications. Nature Reviews Endocrinology. 2018;14(2):88-98.

4. Quianzon CC, Cheikh I. History of insulin. Journal of Community Hospital Internal Medicine Perspectives. 2012;2(2):18701.

5. Shen Y, Prinyawiwatkul W, Xu Z. Insulin: a review of analytical methods. The Analyst. 2019;144(14):4139-48.

6. Knopp JL, Holder-Pearson L, Chase JG. Insulin Units and Conversion Factors: A Story of Truth, Boots, and Faster Half-Truths. Journal of Diabetes Science and Technology.

2019;13(3):597-600.

7. Campbell NA, Reece JB. Biology: International Edition. 8 ed. San Francisco: Pearson Benjamin Cummings; 2008. 1393 p.

8. Kumagai I, Tsumoto K. Antigen-Antibody Binding. Encyclopedia of Life Sciences2002.

9. Dingman R, Balu-Iyer SV. Immunogenicity of Protein Pharmaceuticals. Journal of Pharmaceutical Sciences. 2019;108(5):1637-54.

10. Vecchio I, Tornali C, Bragazzi NL, Martini M. The Discovery of Insulin: An Important Milestone in the History of Medicine. Frontiers in Endocrinology. 2018;9.

11. Engvall E, Perlmann P. Enzyme-linked immunosorbent assay (ELISA). Quantitative assay of immunoglobulin G. Immunochemistry. 1971;8(9):871-4.

12. Grange RD, Thompson JP, Lambert DG. Radioimmunoassay, enzyme and non- enzyme-based immunoassays. British Journal of Anaesthesia. 2014;112(2):213-6.

13. Hornbeck P. Enzyme-linked immunosorbent assays. Curr Protoc Immunol.

2001;Chapter 2(1):Unit 2 1.

14. KATO K, HAMAGUCHI Y, OKAWA S, ISHIKAWA E, KOBAYASHI K, KATUNUMA N. Use of Rabbit Antibody IgG Bound onto Plain and Aminoalkylsilyl Glass Surface for the Enzyme-Linked Sandwich Immunoassay. The Journal of Biochemistry. 1977;82(1):261-6.

15. Aydin S. A short history, principles, and types of ELISA, and our laboratory experience with peptide/protein analyses using ELISA. Peptides. 2015;72:4-15.

16. Mercodia. Mercodia Insulin ELISA - Directions for Use. 15.0 ed2019.

17. Casesnoves A, Mauri M, Dominguez JR, Alfayate R, Pico AM. Influence of Anti- Insulin Antibodies on Insulin Immunoassays in the Autoimmune Insulin Syndrome.

1998;35(6):768-74.

18. von Schenck H, Olsson PO, Arnqvist H. Free and total insulin as determined after precipitation with polyethylene glycol: analytical characteristics and effects of sample handling and storage. Clinical Chemistry. 1987;33(1):93-6.

19. Jordan G, Pöhler A, Guilhot F, Zaspel M, Staack RF. High ionic strength dissociation assay (HISDA) for high drug tolerant immunogenicity testing. Bioanalysis. 2020.

20. Hoffmann E, Jordan G, Lauer M, Ringler P, Kusznir EA, Rufer AC, et al. Generation, Characterization, and Quantitative Bioanalysis of Drug/Anti-drug Antibody Immune Complexes to Facilitate Dedicated In Vivo Studies. Pharmaceutical Research. 2019;36(9):129.

21. Hayford JT, Thompson RG. Free and total insulin integrated concentrations in insulin dependent diabetes. Metabolism. 1982;31(4):387-97.

22. Larsen PB, Linneberg A, Hansen T, Friis-Hansen L. Reference intervals for C-peptide and insulin derived from a general adult Danish population. Clinical Biochemistry. 2017;50(7- 8):408-13.

(28)

28 23. Reverberi R, Reverberi L. Factors affecting the antigen-antibody reaction. Blood Transfus. 2007;5(4):227-40.

24. Ellison SLR, Thompson M. Standard additions: myth and reality. The Analyst.

2008;133(8):992.

25. Gilbert DL. Buffering of blood plasma. Yale J Biol Med. 1960;32(5):378-89.

26. Whittingham J, Scott D, Chance K, Wilson A, Finch J, Brange J, et al. Insulin at pH 2:

Structural Analysis of the Conditions Promoting Insulin Fibre Formation. Journal of molecular biology. 2002;318:479-90.

27. M. Hach NN. Personal communication. October 2nd 2020.

28. Zoghbi J, Xu Y, Grabert R, Theobald V, Richards S. A breakthrough novel method to resolve the drug and target interference problem in immunogenicity assays. J Immunol Methods.

2015;426:62-9.

29. Gelsema WJ, DeLigny CL, Remijnse AG, Blijleven HA. pH-Measurements in alcohol- water mixtures, using aqueous standard buffer solutions for calibration. RECUEIL.

1966;85(7):647-60.

30. Sim S-L, He T, Tscheliessnig A, Mueller M, Tan RBH, Jungbauer A. Protein

precipitation by polyethylene glycol: A generalized model based on hydrodynamic radius. Journal of Biotechnology. 2012;157(2):315-9.

31. Vermeer AWP, Norde W. The Thermal Stability of Immunoglobulin: Unfolding and Aggregation of a Multi-Domain Protein. Biophysical Journal. 2000;78(1):394-404.

32. Huus K, Havelund S, Olsen HB, van de Weert M, Frokjaer S. Thermal dissociation and unfolding of insulin. Biochemistry. 2005;44(33):11171-7.

Mitigating the impact of antidrug antibodies against insulin on ELISA assay

Mitigating the impact of antidrug antibodies against insulin on ELISA assay

TABLE OF CONTENTS

ABBREVIATIONS

INTRODUCTION

MATERIALS and METHODS

RESULTS

DISCUSSION

CONCLUSIONS

Acknowledgements

REFERENCES