MSc Chemistry
Analytical Sciences
Literature Thesis
Quantification of monoclonal antibodies in serum
and plasma for use in therapeutic drug monitoring
by
Vivian Westland
12333913
20 November 2020
12 EC
September 2020 – November 2020
Supervisor:
Examiners:
M. Vos – van der Meer
Dr. R. Haselberg
Prof. dr. G.W. Somsen
Amsterdam UMC,
pharmacy laboratory
2
Abstract
Therapeutic monoclonal antibodies are known for the high specificity and affinity towards the target antigen in a mixture of molecules, which makes them very attractive for therapy of diseases. Until now, 90 therapeutic monoclonal antibodies have been approved by the European Union and/or the United States for a variety of diseases, with a large number currently under review. With the fast-growing number of monoclonal antibodies, therapeutic drug monitoring is gaining interest for more personalised dosing regimens leading to more efficient treatment. This review focusses on analysis methods for the quantification of monoclonal antibodies in plasma or serum samples for use in therapeutic drug monitoring. Ideally, a quantification method would be universally applicable to all monoclonal antibodies. Enzyme-linked immunosorbent assay (ELISA) was traditionally used for the quantification of monoclonal antibodies in plasma and serum due to the high sensitivity and ease of use. However, it is not universally applicable due to the specific antibodies or ligands that are required for every monoclonal antibody. Alternatively, liquid chromatography mass spectrometry (LC-MS) approaches can be used for monoclonal antibody quantification. Three basic approaches can be distinguished; bottom-up, middle-down and top-down. Nowadays, middle-down and top-down approaches, which quantify a large subunit or the whole protein, are lacking in terms of sensitivity. Bottom-up quantification, in which a unique signature peptide is used for quantification, is most promising for use as universal method for quantification of monoclonal antibodies. Using nano-surface and molecular oriented limited (nSMOL) proteolysis combined with liquid chromatography tandem mass spectrometry (LC-MS/MS), sufficient sensitivity can be obtained for therapeutic drug monitoring of monoclonal antibodies. Furthermore, no specific antibodies or ligands are required, making it appropriate for use in pharmaceutical laboratories as general method for monoclonal antibody quantification in plasma and serum. When no signature peptide is available, the bottom-up approach cannot be used. Nowadays, ELISA is the most attractive alternative, as top-down and middle-top-down quantification lack sensitivity for therapeutic drug monitoring of monoclonal antibodies. However, research on top-down and middle-down quantification is gaining interest nowadays and improvements might be obtained in the future, which may lead to practical implication of these approaches in pharmaceutical laboratories.
3
Abbreviations
ADA Anti-drug antibody AS Ammonium Sulphate
b-TNF-α Biotinylated tumour necrosis factor-alpha
cDNA Copy DNA
CDR Complementarity Determining Region
CH1 – CH4 Constant regions, heavy chain
CL Constant region, light chain
DTT Dithiothreitol E. coli Escherichia coli
ELISA Enzyme-linked immunosorbent assay
EMA European medicines agency ESI Electrospray ionisation FA Formic acid
Fab Antigen-binding fragment Fc Fragment crystallisable FDA Food and drug administration FLD Fluorescence detector
FT-ICR Fourier-transform ion cyclotron resonance
HACA Human anti-chimeric antibody HAHA Human anti-human antibody HAMA Human anti-mouse antibody Hc Heavy chain
hIgG1 Human immunoglobulin G 1
HRMS High resolution mass spectrometers HT-RPLC High temperature reversed phase
liquid chromatography IAA Iodoacetamide
IFA Immunofluorescence assays IFX Infliximab
Ig Immunoglobulin IS Internal standard LBA Ligand-based assay LC Liquid chromatography Lc Light chain
LC-MS Liquid chromatography mass spectrometry
LC-MS/MS
Liquid chromatography tandem mass spectrometry
LLOQ Lower limit of quantification m/z Mass to charge
mAbs Monoclonal Antibodies MeOH Methanol
MEW Mass extraction window MRM Multiple reaction monitoring
MS Mass spectrometry
nSMOL Nano-surface and molecular oriented limited
P Proline
PCI-IS Post-column infused internal standard
PCR Polymerase Chain Reaction PD Pharmacodynamics
PK Pharmacokinetics
Q Quadrupole
QC Quality control QqQ Triple quadrupole
QTOF Quadrupole time-of-flight
R Arginine
RIA Radioimmunoassay RP Reversed phase RPLC Reversed phase liquid
chromatography
scFv Single-chain fragment variable SIL-IS Stable isotope-labelled internal
standard
SRM Single reaction monitoring TCEP Tris-(2-carboxyethyl)phosphine TDM Therapeutic drug monitoring TFA Trifluoroacetic acid
TNF-α Tumour necrosis factor-alpha TOF Time of Flight
VEGF Vascular endothelial growth factor VH Variable region, heavy chain
VL Variable region, light chain
Table of contents
1. Introduction ... 5
2. Basics of monoclonal antibodies ... 6
2.1. Antigens and antibodies ... 6
2.2. Types of monoclonal antibodies ... 7
2.3. Classes of monoclonal antibodies ... 9
3. Therapeutic drug monitoring ... 10
3.1. TDM of monoclonal antibodies ... 10
4. Quantification of monoclonal antibodies ... 12
4.1. Enzyme-linked immunosorbent assay ... 12
4.2. LC-MS analysis ... 14
4.2.1. Bottom-up quantification ... 14
4.2.2. Top-down quantification ... 28
4.2.3. Middle-down quantification ... 36
5. Conclusion & Discussion ... 42
Acknowledgements ... 44
References ... 45
5
1. Introduction
Monoclonal antibodies (mAbs) are defined to be a homogeneous population of antibodies specific to an antigen and function to neutralise pathogens, such as bacteria and viruses [1,2]. MAbs can bind highly specific to a target antigen in a mixture of molecules [1–3]. Due to the high specificity and affinity of mAbs, they are very attractive for therapy of diseases [4]. Nowadays, 90 therapeutic mAbs have been approved by the European Union and/or the United States for a variety of diseases (see appendix 1 for an overview) [5]. This number increases fast, with a large number of mAbs currently under review [5]. Most of the therapeutic mAbs have been first approved for cancer (antitumor mAbs) or immunological disorders (anti-inflammatory mAbs) (see figure 1). After first approval of a mAb for a specific disease, sometimes the mAb is also applied in the therapy of other classes of diseases.
With the increasing number of mAbs, therapeutic drug monitoring (TDM) of mAbs is gaining interest for more personalised dosing regimens. In TDM, the concentration of the mAb in plasma or serum samples is determined, and dosing is adjusted based on pharmacokinetics (PK) and pharmacodynamics (PD) information to obtain optimal treatment efficiency. Up till now, TDM has mainly been used for small drugs rather than for mAbs due to the more complicated PK and PD of mAbs [6–8]. Nevertheless, TDM can be helpful to highlight accelerated drug clearance in mAb therapy leading to lower serum/plasma concentration, which consequently results in loss in treatment efficiency [8–11]. In order to perform TDM of mAbs in plasma and serum samples, robust, precise, sensitive and accurate analytical methods are required [12].
This literature study focusses on the most frequently used analytical methods for the quantification of therapeutic mAbs in plasma and serum for the use in TDM in pharmaceutical hospital laboratories. The basic structure, types and classes of mAbs will be described in chapter 2. Next, more details on TDM of mAbs will be given in chapter 3, including the main challenges. Chapter 4 discusses the most often used analysis methods for the quantification of mAbs in plasma and serum. This includes the description of enzyme-linked immunosorbent assay (ELISA) methods (section 4.1) and liquid chromatography mass spectrometry (LC-MS) approaches (section 4.2). Finally, chapter 5 will describe the conclusion and discussion; a comparison between the described methods will be given, and the most appropriate method for the quantification of mAbs in plasma and serum for use in TDM in pharmaceutical laboratories will be selected. Ideally, the quantification method would be universally applicable such that it can be used for all types of mAbs.
Figure 1 Classification of approved mAbs by the European Union and/or the United States. Classification based on the disease the mAb was first approved for. Data used to prepare the graph is provided in appendix 1 [5,13].
6
2. Basics of monoclonal antibodies
A basic knowledge of antibodies is required to understand the principles described in the later sections. Therefore, this section will give a short introduction on the structure of antibodies, the interaction with antigens, and the types and classes of mAbs.
2.1. Antigens and antibodies
An antigen can be defined as a molecule that binds to a lymphocyte receptor and may induce an immune response. It is often a foreign to the body, for example, due to an infection. However, it is also possible for the body to have its own proteins that act as antigens. In order to elicit an immune response, the antigen should be recognised by the biological system as nonself [2,14,15]. Antigens can bind highly specific to an antibody, which can induce a response to eliminate the antigen. Epitopes are the part of the antigens that can bind highly specific to the antigen binding site of an antibody [14,15].
Several classes of antibodies, also called immunoglobulins (Ig), exist in the human body (see section 2.3). However, the most abundant human antibody is immunoglobulin G (IgG). The basic structure of the IgG antibody is shown in figure 2. The IgG antibody consists of four protein chains held together with disulphide bonds. The part where the disulphide bonds are located is called the hinge region; a flexible area which causes the characteristic Y-shape of the antibody. Each Y-shaped IgG antibody consists of two identical heavy chains (approximately 50 kDa) and two identical light chains (approximately 25 kDa). The structure and amino acid sequence of the two light and two heavy chains are identical. Each chain in an antibody is made of folded regions, called domains. The light chains consist of two domains, VL and CL (figure 2B). The VL region is the
variable region of the light chain; it varies the most from one antibody to another and is involved in the specific antigen binding. CL is the constant domain of the light chain; the amino acid sequence is relatively constant
from one antibody to another. Similar to the light chains, the heavy chains also consist of multiple domains. Depending on the class of the antibody, the heavy chains contain three or four constant regions (CH1 – CH4) and
one variable region (VH) [2,15–18]. Between the CH1 and CH2 domains, the hinge region is located. Due to the
different domains, an antibody can be divided into a constant and a variable part (figure 2A). Inside the variable region of the antibody, the hypervariable regions (also known as Complementarity Determining Regions (CDRs)) are located. This part shows the most variation between antibodies and consequently is the region that forms the antigen binding sites. Due to the high variation of this part of the antibody, it is highly specific for the target antigen [11,14,15].
The antibody can be divided further into the antigen-binding fragment (Fab) and the fragment crystallisable (Fc) region (see figure 2B). The Fab region is the part above the hinge region containing the variable regions and is, therefore, crucial for the actual antigen binding [16]. Antigens can only bind to the antigen binding sites of the antibody when the structure is complementary. Interaction with high affinity is obtained when there is a perfect match between the antigen epitope and the antigen binding site of the antibody. Due to van der Waal’s forces, hydrogen bonds, electrostatic forces and hydrophobic forces, the antigen will be held tightly in place [2,15]. Other antigens may have a lower affinity to the specific antibody or no interaction at all. Binding of antigens to an antibody other than the target antigen is revered to as cross-reactivity. Cross-reactivity may be observed when epitopes are very similar to the epitope of the target antigen [16].
After binding of the antigen to the antibody, it is necessary to get a response that will result in the removal of the antigen and death of the pathogen. To achieve this, the Fc region, which is located below the hinge region, is critical. The Fc region can bind to cell surface receptors such as Fc𝛾 receptors. The process of the binding of an antibody to an effector cell is called the effector function. Different classes of antibodies have diverse effector functions, which generate a different response after binding to the antigen [3,11,14,19].
7
Figure 2 Structure of an IgG (most common) antibody. Reproduced from ref. [18] (A) and ref. [15] (B)
2.2. Types of monoclonal antibodies
Based on the quantity of human protein sequences in a mAb, several types of monoclonal antibodies can be distinguished. To date, four types of mAbs are being used in diagnostics; murine, chimeric, humanized and human mAbs. A short description of every type of mAb will be given in this section.
Murine monoclonal antibody
Murine mAbs are entirely made of protein sequences originating from mice. To produce murine mAbs (see figure 3), a mouse is immunised with the antigen of interest to stimulate the antibody production. The antibody-forming cells are then isolated and fused with cultivated tumour cells to form a type of cells called hybridomas. Next, the hybridomas are screened for antibody production. Antibody-producing hybridomas are selected and cloned, and finally, the produced monoclonal antibodies are isolated [11,16].
Murine mAbs have several disadvantages which makes them less attractive to use in therapeutics. One of the main problems with murine mAbs is immunogenicity. Due to immunogenicity, the introduction of a murine mAb into a human’s body can result in a human anti-mouse antibody (HAMA) response. The body recognises the murine mAb as foreign to the body; consequently, due to an immune response, murine mAbs are cleared out from the bloodstream quickly [1,4,16]. Furthermore, murine mAbs can cause an allergic reaction due to its foreign nature. Another drawback of murine mAbs is the relatively short half-life, resulting in a lower efficiency of the mAb [1,4,14,16,20]. Finally, due to the constant murine Fc part, effector functions may be weak in humans [21]. Due to the several disadvantages, the therapeutic use of murine mAbs is limited; only a small number of the approved mAbs is murine (see figure 4A).
Chimeric monoclonal antibody
Utilising recombinant DNA technology, mAbs with greater amounts of human sequences can be developed, reducing the immunogenicity against the antibody. A chimeric antibody is composed of two origins: murine and human. Murine mAbs consist for 100% of murine protein sequences, while chimeric antibodies contain for approximately 33% of murine sequences with the remaining being human. In chimeric mAbs, the constant region of the antibody is made of human sequences, while the variable region with the antigenic specificity remains murine [16,17].
To produce chimeric mAbs (see figure 3), the variable region genes that are specific for the target antigen are isolated from a murine hybridoma (as in murine mAb production) and then amplified using polymerase chain reaction (PCR). The result is a copy DNA (cDNA) of the variable murine region which can be inserted into a plasmid. In the same way, a cDNA of the human constant region can be obtained. After insertion of the human
8 and murine genes in separate plasmids, they can be combined in a host cell (bacteria) using co-transfection, resulting in the production of chimeric antibodies [16].
Due to the constant human part of chimeric mAbs, the immunogenicity is less compared to murine mAbs. Consequently, chimeric mAbs show an increased half-life compared to murine mAbs [1]. Nevertheless, the murine part of the chimeric mAb can still cause a human anti-chimeric antibody (HACA) response, resulting in fast clearance of the antibody from the bloodstream. As a consequence of the HACA response, the duration effect of the chimeric mAb will be shortened [4,16,22]. As shown in figure 4A, a larger number of mAbs is chimeric compared to murine; however, most of the approved mAbs are humanized and human.
Humanized monoclonal antibody
To reduce the murine part of the antibody even more, and therefore decrease the immunogenicity, humanized mAbs were developed. In humanized mAbs, only the hypervariable regions (CDRs) of the antibody are made of murine sequences while the remainder is human. As a result, humanized mAbs typically contain for only 5% of murine sequences [16,17]. The most common method to produce humanized mAbs is CDR grafting (see figure 3). First, the CDR sequences are produced in a mouse, after which the CDR sequence is transplanted into the human framework by CDR grafting [16,23].
Even though humanized mAbs contain only a small amount of murine proteins, it is still possible to observe a human anti-human antibody (HAHA) response against the humanized mAb. However, due to the low amount of murine proteins, humanized mAbs are less likely to cause an immune response compared to murine and chimeric mAbs [4,16,22–24]. Furthermore, humanized mAbs have a longer half-life than murine and chimeric mAbs [23]. Nowadays, 50% of the approved mAbs is humanized (see figure 4A).
Human monoclonal antibody
Human mAbs consist entirely of human sequences. Fully human mAbs can be produced by transgenic mice or by phage display libraries (see figure 3) [1,11,16]. In transgenic mice, the mouse is genetically engineered by replacing murine Ig genes with human Ig genes. Immunisation of a transgenic mouse results in a human antibody response. Fully human antibodies can then be obtained by following the same procedure as with the traditional murine mAbs (antibody-forming cells and myeloma cells combined to form hybridomas, screening, cloning and isolation) [25].
Phage display libraries is an in vitro selection procedure and utilises an antibody phage display library to select the antibody fragment (Fab or single-chain fragment variable (scFv)) with the highest specificity to the target antigen in a phage virus particle. Using affinity enrichment, antibodies that bind to the antigen can be selected and non-binding antibodies can be washed away. Antibody production can then be amplified by infection in
Escherichia coli (E. coli). Selection rounds are repeated until the desired specificity is obtained. This procedure
is used to obtain the scFv or Fab region of the antibody with the highest specificity to the target antigen. The fully human mAbs can be obtained by co-transfection, which combines the Fab or scFv region with the constant region genes of the antibody [25].
The use of human (and humanized) mAbs results in a reduced, but still possible, change on an immune response against the monoclonal antibody [22,24]. Human mAbs can be a good alternative when a patient shows an immune response against a murine, chimeric or humanized mAb [22]. As shown in figure 4A, approximately 35% of the approved mAbs is fully human. With more than 85% of the approved mAbs being humanized and human, these mAbs are dominant in the field of therapeutic mAbs nowadays.
9
Figure 3 Schematic overview for the production of murine (mouse), chimeric, humanized and human mAbs. Reproduced from ref. [23]
2.3. Classes of monoclonal antibodies
In humans, five different classes of antibodies exist; IgA, IgD, IgE, IgG and IgM. The antibody class is determined by the heavy chain type. IgA antibodies contain an 𝛼-chain, IgD a 𝛿-chain, IgE an 𝜀-chain, IgG a 𝛾-chain and IgM has a 𝜇-chain. IgG is the most abundant class of antibodies in human blood (approximately 80%) [2,4,14]. Consequently, to date, all approved therapeutic mAbs are of the IgG class. Human IgG has four subclasses (isotypes), named in order of decreasing abundance: IgG1, IgG2, IgG3 and IgG4. The subclasses are distinguished by the size of the hinge region and the number and position of the interchain disulphide bonds between the heavy chains [2,4,14,19]. The hinge length (number of amino acids) of IgG1, IgG2, IgG3 and IgG4 are 15, 12, 62 and 12, respectively [4]. IgG1 contains two interchain disulphide bonds, IgG2 and IgG4 four, and IgG3 eleven [14]. The subclass of the antibody strongly influences the effector function of the antibody. Out of the approved mAbs, most are IgG1 isotype, with a lower number being IgG2, IgG4 or hybrid IgG2/4 (see figure 4B). To date, no IgG3 mAbs are approved, possible reasons are the relatively short half-life compared to the other subclasses (2-6 days for IgG3 and 23 days for IgG1, IgG2 and IgG4) and the long hinge region which complicates the bioprocessing [26,27].
Figure 4 Classification of approved mAbs by the European Union and/or the United States. A: classification based on the type of mAb. B: classification based on the subclass (isotype) of mAb. Data used to prepare the graphs is provided in appendix 1 [5,13].
10
3. Therapeutic drug monitoring
In order to have safe and effective drug regimens, often TDM is used to adjust drug dosage of an individual by measuring drug concentrations in biological fluids. Usually, the measured concentration should be within a therapeutic range in which the desired effect of the drug is observed [28]. Concentrations below the therapeutic range may result in a loss of efficiency of the drug, while a too high concentration may lead to toxicity. After measurement of drug concentrations, clinical interpretation of the results is essential. Knowledge of pharmaceutics, PK and PD is required for good clinical interpretation [12,29,30].
PK can be described as the study of what the body does to the drug, which includes drug absorption, distribution, metabolism and excretion. In clinical PK, the PK principles are used for the safe and effective therapeutic management of drugs in individual patients. On the other hand, PD described what the drug does to the body and focusses on the relationship between drug concentration at the site of action and the resulting effect. Variability between individuals for both PK and PD are observed, which result in variations in drug effects. Therefore, to have optimal benefit of TDM for individuals, it is important to take into account both PK and PD in the clinical interpretation. Combination of all information can be used to obtain the appropriate dosage for the optimal response with minimal toxicity for an individual [29–31].
TDM is routinely applied for small drugs such as antibiotics, antiepileptic drugs, immunosuppressants, antidepressants and antipsychotics [12].
3.1. TDM of monoclonal antibodies
High inter-individual variability is observed in the response of mAbs; for many mAbs, it is reported that patients fail to respond to the mAb treatment or lose response over a longer period of treatment [6–8]. The main cause for treatment failure is an inadequate serum/plasma concentration of the mAb [6,8,32]. The low serum/plasma concentration is a consequence of increased mAb clearance, mainly caused by the formation of anti-drug antibodies (ADAs). ADAs against a therapeutic mAb are formed due to immunogenicity, which consequently leads to the HAMA, HACA or HAHA response (see section 2.2). When ADAs are formed against the therapeutic mAb, an increased clearance and/or neutralisation of the biological effect is observed, which subsequently leads to reduced serum/plasma concentration and often in less therapeutic effect [3,6,8]. As mentioned before, humanized and human antibodies develop less immunogenicity compared to chimeric and murine antibodies due to the minimisation of murine sequences. However, even with fully human mAbs, ADAs are formed resulting in an immune response. This is because many elements are identified that enhance the probability of ADA formation such as formulation characteristics, mAb structure, dose regimen, administration route and treatment duration [3,8,11]. For example, a low dose of a mAb often results in a greater immune response compared to a high dose of the same mAb. Furthermore, longer treatment durations also result in an increased chance of an immune response against the therapeutic mAb [8,11]. When ADAs are formed, the addition of immunomodulators during treatment might result in suppression of the ADA formation, which consequently results in increased drug levels [6,8].
Besides the formation of ADAs, factors like high body weight, gender (mAbs usually have a longer half-life and lower clearance in women), low serum albumin concentrations, and high serum C-reactive protein levels are also related to increased mAb clearance [6,8,11]. This increased clearance results in lower serum/plasma concentrations of the mAb, which leads to decreased treatment efficiency [6,8,32]. TDM of mAbs in plasma/serum samples can be helpful to highlight accelerated drug clearance which results in loss of response and the therapeutic effect. However, PK and PD of mAbs are much more complex compared to small molecules making interpretation more difficult [6–8]. Furthermore, the presence of the highly similar endogenous IgGs complicates the determination of the mAb serum/plasma concentrations. Nevertheless,
11 taking into account patient-specific factors that influence the mAb clearance, TDM can be used for more personalised dosing regimens [6]. Nowadays, interest in TDM of mAbs is upcoming, especially because personalised dosing can result in more efficient treatment and reduced treatment costs [7,10].
Therapeutic ranges of mAbs
For small drugs, for which TDM is regularly applied, therapeutic ranges are clearly defined [33]. However, for many mAbs therapeutic ranges have not (yet) been established [32]. Until now, most mAbs are administered in body-size-based dosing schedules, rather than personalised dosing based on mAb serum/plasma concentration levels [34]. Preclinical toxicity studies showed that therapeutic mAbs typically have a much wider therapeutic range compared to small drugs [34,35]. However, therapeutic ranges have only been defined for a small number of mAbs. For example, for infliximab a therapeutic range of 3-7 µg/mL [36] and for adalimumab 5-8 µg/mL [37] was reported. Even though therapeutic ranges have not been defined for many mAbs, El Amrani et al. [9] reported that a lower limit of quantification (LLOQ) of 1 µg/mL is sufficient for TDM application of most mAbs. Therefore, an LLOQ of 1 µg/mL will be used as guideline in terms of the required sensitivity for the analytical techniques described in the following sections.
12
4. Quantification of monoclonal antibodies
To perform TDM for mAbs, robust, precise, sensitive and accurate analytical techniques are required. This section describes the most important analysis methods which are used for the quantification of therapeutic mAbs in plasma and serum. The principle, advantages and disadvantages of ELISA are described in the next section, followed by liquid chromatography mass spectrometry (LC-MS) approaches in section 4.2.
4.1. Enzyme-linked immunosorbent assay
ELISA, which is a type of ligand-based assay (LBA), has been the standard technique for the quantification of mAbs in serum and plasma for the past decades [38,39]. Other LBAs have been used for the quantification of mAbs, including immunofluorescence assays (IFA) using flow cytometry and radioimmunoassay’s (RIA) [38]. IFA has been used for the quantification of alemtuzumab [40] and rituximab [41], and RIA for infliximab [42] quantification, for example. However, the focus in this review will be on ELISA as this is by far the most popular LBA technique for the quantification of mAbs in serum and plasma.
With ELISA, multiple forms of mAbs can be detected depending on the used reagent, including free mAb, antigen-bound mAb, total mAb and ADA-mAb immunocomplex concentrations. Free mAb is assumed to be the driver for biological activities, and therefore, often the most desirable analyte [3,43]. The free mAb concentration includes unbound and partially bound mAbs. Unbound mAbs have two free antigen binding sites. Partially bound mAbs are bound to an antigen at only one of the antigen binding sites, leaving one of the binding sites free [43].
The majority of the ELISA methods measure the free mAb (free and partially bound) concentration by using the ELISA formats shown in figure 5 [38,44]. In the first step of the ELISA, a 96-well plate is coated with a capture reagent which is specific to the mAb of interest. This can be an anti-idiotype antibody (figure 5A) or an antigen specific for the mAb (figure 5B). Next, the plasma or serum sample is added, which will result in binding of the therapeutic mAb to the anti-idiotype antibody or antigen. With several washing steps, non-binding compounds are washed away. After that, a secondary enzyme-conjugated antibody, often an anti-human IgG antibody, is added that will bind to the therapeutic mAb (see figure 5). After washing away the excess of the secondary antibody, a substrate is added, which is converted into a coloured product by the enzyme [6,38,44]. The colour is proportional to the mAb concentration and can be analysed with a spectrophotometric plate reader. MAb concentrations can then be determined by comparing the result with a standard curve [6,38]. This type of ELISA measures the free mAb concentration, as at least one free antigen binding site is required for binding to the anti-idiotype antibody or antigen.
Both ELISA formats shown in figure 5 are often defined as sandwich ELISA. This might be confusing because sandwich ELISA is often used for the quantification of antigens rather than mAbs. Nevertheless, both principles, utilising an anti-idiotype antibody and the antigen, are referred to as sandwich ELISA for commercially available ELISA kits for mAb quantification. For example, a kit available by Matriks Biotek [45] for the quantification of bevacizumab is described to be based on the sandwich ELISA principle. In this case, the 96-well plate is coated with human vascular endothelial growth factor (VEGF), which is the target antigen of bevacizumab [46]. Therefore, the principle from figure 5B is followed. Most of the commercially available kits measure free mAb concentrations by using the sandwich principle with either the target antigen or an anti-idiotype antibody [45,47].
13
Figure 5 Schematic overview of the most used sandwich ELISA formats for quantification of therapeutic mAbs. A: an anti-idiotype antibody is used to capture the mAb. B: the target antigen is used to capture the mAb. Reproduced from ref. [44]
Advantages and disadvantages of ELISA
ELISA has several advantages which cause that it has been the method of choice for quantification of mAbs in plasma and serum for the past decades. First, ELISA has a high sensitivity and is relatively straightforward once the method development is completed, and therefore, easy to use. Furthermore, instrumental costs for ELISA are low, and minimal sample pre-treatment is required. Finally, analysis is relatively quick compared to the alternative LC-MS analysis allowing high throughput analysis [9,38,48,49].
However, ELISA also has multiple disadvantages. Firstly, an ELISA kit is specific for one particular mAb. Therefore, for every mAb, a different kit with specific antibodies and antigens is required making it expensive and not universally applicable [9,50]. Furthermore, method development is time-consuming and costly (up to 5 months) due to the production and selection processes for the specific antibodies required for every therapeutic mAb [9,39,48,50,51]. ELISA also lacks standardisation because no internal standard is used. Consequently, quantitative results may vary between different ELISA kits for the same mAb, making comparison between different assays difficult [6,9,38]. Furthermore, ELISA typically has relatively narrow dynamic ranges resulting in the requirement for dilution, leading to an additional source of potential errors [9,50,52]. Finally, a major problem in ELISA is cross-reactivity. The serum/plasma samples contain the therapeutic mAb but also the highly similar endogenous IgGs. These endogenous IgGs may interfere because ELISA is unable to differentiate between endogenous IgGs and therapeutic mAbs which have a minor difference in the amino acid sequence. This leads to cross-reactivity and inaccurate quantification results [9,38,39,53].
Conclusion ELISA
Ligand-based assays, with most commonly ELISA, have been the method of choice for the past decades for the quantification of mAbs due to the ease of use, high sensitivity, low instrumental cost and quick analysis. However, ELISA methods have several limitations such as time-consuming and costly method development, no standardisation, narrow dynamic range and cross-reactivity. Furthermore, ELISA cannot be used as universal method as for every therapeutic mAb a different kit is required with specific antigens or antibodies. Due to these disadvantages, alternatives to ELISA which overcome these limitations are upcoming. Nowadays, an increasing number of applications are becoming available for the quantification of mAbs using LC-MS.
14
4.2. LC-MS analysis
LC-MS is an upcoming technique for the quantification of therapeutic mAbs in plasma or serum [9,48]. It has several advantages compared to the traditionally used ELISA. First, LC-MS method development times are much shorter (typically a few weeks) [54–56]. Furthermore, LC-MS is easier to standardise and is more specific compared to ELISA due to limited cross-reactivity [50,51,56]. In addition, LC-MS methods typically have a wider dynamic range and multiplexing capabilities [54,56]. Finally, LC-MS analysis does not require specific antibodies or ligands. However, these might be necessary for sample preparation in LC-MS analysis [54–56]. In LC-MS analysis of proteins, three different approaches can be differentiated; bottom-up, middle-down and top-down. All approaches can be used in therapeutic mAb quantification and will be described in the following sections.
4.2.1. Bottom-up quantification
In the bottom-up approach, liquid chromatography tandem mass spectrometry (LC-MS/MS) is used for the quantification of a therapeutic mAb based on a signature peptide. Different from ELISA is that not the whole protein is measured but only a small peptide which is characteristic for the full antibody. A workflow of the steps followed in the method development for the quantification of mAbs using the bottom-up approach is shown in figure 6. Each step in the workflow will be described in detail in the following sections. An overview of the available applications for the quantification of mAbs in plasma and serum using the bottom-up approach is shown in table 3.
Figure 6 Workflow for the quantification of mAbs using the bottom-up approach. First, the signature peptide is selected, followed by selection of the sample purification strategy. The following step includes the determination of the denaturation conditions and trypsin digestion, after which the internal standard is selected. The final step includes LC-MS/MS method development and validation.
Signature peptide selection
The first step in bottom-up quantification of a mAb is the selection of a signature peptide. It is a crucial step, as when no signature peptide can be selected, bottom-up quantification is impossible. A signature peptide is a unique peptide in the therapeutic mAb that cannot be found in human sequences. Consequently, the signature peptide can be used for quantification, as there is no interference with the endogenous IgGs [9]. For chimeric mAbs, tryptic peptides from the entire variable region can be chosen as these are made of murine sequences, and therefore not found in human sequences. However, for human and humanized mAbs, the choice is often limited to the CDRs. To select an appropriate signature peptide, the amino acid sequence of the therapeutic mAb is required, which can be found in the Immunogenetics Information system [57] or the Drugbank [58]. After the amino acid sequence is obtained, often a trypsin digestion is performed and monitored by a time of flight (TOF) MS or Orbitrap MS [36,59–61]. However, when these instruments are not available, it is also possible to use an online tool such as Protein Prospector [62] to predict the peptides formed by trypsin digestion [9]. Next, Blast software [63] can be used to screen the generated peptides. Blast compares the peptide sequences against the biological matrix of the analyte (human serum/plasma) from an appropriate database such as Swissprot [9,51,59,61]. Any peptide scoring below 100% is not found in the human serum/plasma and is, therefore, a potential signature peptide [9]. Consequently, a list of potential signature peptides can be made and screened on length and stability. The signature peptide should preferably have a length between 6 and 20 amino acids. Furthermore, peptides containing methionine or cysteine should be excluded from the selection due to possible oxidation reactions leading to mass shifts. In addition, peptides
15 containing asparagine followed by glycine or serine should also be avoided due to deamidation reactions which also result in a mass shift [9,51,59]. After removal of unstable peptides from the list of potential signature peptides, a trypsin digestion can be performed with a standard solution. By running the sample on the LC-MS/MS, the peptide with the highest intensity can be selected as signature peptide [9,51].
Sample purification
After selection of the signature peptide, the next step is to determine the required sample purification of the plasma and serum samples. Sample purification is necessary to eliminate interfering proteins and to reduce sample complexity. Compared to the therapeutic mAb, the endogenous IgGs have highly similar structures which complicates the analysis [9,36]. An overview of the performance of the most used sample purification strategies for therapeutic mAbs was described by El Amrani et al. [9] and is shown in table 1.
Table 1 Overview of the performance of different sample purification strategies. Reproduced from ref. [9]
Plasma Components Concentration
in plasma [g/L] Targeted Purification Protein A Protein G AS* Precipitation MeOH** Pellet digestion Albumin (60 kDa) 45 ++ ++ ++ ++ + IgG (150 kDa) 10 ++ - - - - Fibrinogen (340 kDa) 2.5 ++ ++ ++ - - Transferrin (80 kDa) 2.5 ++ ++ ++ ++ + IgA (320 kDa) 2 ++ + ++ - - Alpha-1 Anti-Trypsin (54 kDa) 1.5 ++ ++ ++ ++ + Phospholipids (<1 kDa) 1 ++ ++ ++ ++ ++ IgM (900 kDa) 1 ++ + ++ - - IgD (180 kDa) 0.02 ++ ++ ++ - - IgE (200 kDa) 0.0002 ++ ++ ++ - -
*AS = Ammonium Sulphate, **MeOH = Methanol, ++ = efficiently eliminated, + = moderately eliminated, - = not eliminated
Target-specific sample purification
Target-specific sample purification (also called immuno-affinity capture) extracts only a specific mAb of interest from the plasma or serum matrix by utilising an anti-idiotypic antibody or ligand [9]. Target-specific sample purification was, for example, used for the quantification of infliximab in human serum by LC-MS/MS (see figure 7) [36]. Here, tumour necrosis factor-alpha (TNF-α), which is the target antigen of infliximab, is bound to a 96-well plate by a biotin-streptavidin interaction. Active infliximab, with at least one free epitope, will bind highly specific with the biotinylated TNF-α (b-TNF-α). In several washing steps, all non-binding compounds are washed away after which infliximab can be eluted [9,36]. As a result, very clean samples are obtained because all interfering compounds, including the highly similar endogenous IgG, are washed away from the sample (see table 1). Due to the clean extracts, low detection limits can be obtained. However, targeted assays are time-consuming and require different ligands or antibodies for every therapeutic mAb, which makes it an expensive sample preparation technique [9]. Nevertheless, target-specific sample purification has been used in several applications, for example, for the quantification of infliximab [36] and cetuximab [60] in serum.
16
Figure 7 Example of target-specific sample purification. A 96-well plate is prepared for sample purification by binding b-TNF-α to streptavidin. Infliximab (IFX) can bind highly specific to the b-TNF-α while other compounds can be washed away from the solution. Clean IFX extracts are obtained after elution. Reproduced from ref. [36]
Protein A or G beads
Protein A or G beads can also be used for sample purification (see figure 8). In contrast to target-specific sample purification, protein A/G magnetic beads trap the entire IgG fraction, including the therapeutic mAb and endogenous IgG. Protein A/G are bacterial cell wall proteins which can bind with IgG via the Fc region [9]. After binding of the IgG (therapeutic and endogenous) to the protein magnetic beads, other compounds can be washed away in a washing step and finally, trapped antibodies can be eluted from the magnetic beads. Protein G beads are a bit more efficient in the elimination of IgA and IgM from the sample compared to protein A beads (see table 1). The trapping of both endogenous IgG and the therapeutic mAb leads to more complex samples compared to the target-specific sample purification.
No studies about the influence on the analysis and the error caused by the competitive binding between the endogenous IgGs and therapeutic mAb to the magnetic beads have been found. However, recoveries of 100±20% are obtained using protein G/A purification [39,50,64]. Consequently, the competitive binding of the therapeutic mAb and endogenous IgG does not influence the quantification of mAbs. Nevertheless, careful evaluation of the extraction recovery is required during validation to make sure recoveries are reproducible. The use of an internal standard can help to correct for variations in the extraction recovery.
Even though endogenous IgGs are not removed from the sample, high levels of other interfering proteins, such as albumin, will be removed (see table 1) [9,50]. As a result, using protein A/G beads, relatively clean extracts are obtained containing the IgG-based antibodies [9]. Protein A or G beads have been successfully applied in multiple applications for the quantification of mAbs, including infliximab [59], bevacizumab [39,50,64], pembrolizumab and nivolumab [50].
Figure 8 Schematic overview of protein G purification of three therapeutic mAbs (bevacizumab, pembrolizumab and nivolumab) and internal standard (tocilizumab) from human plasma. IgG-based antibodies will bind to the protein G magnetic beads, while other compounds can be washed away. Elution results in extracts containing IgG-based antibodies (therapeutic and endogenous). The procedure for protein A beads is identical. Reproduced from ref. [50]
17
IgG protein precipitation
Ammonium sulphate (AS) or methanol (MeOH) can be used for protein (pellet) precipitation (see figure 9). By adding AS or MeOH, proteins with low solubility, which are usually large proteins, will precipitate. Highly soluble, smaller proteins will remain dissolved. As shown in table 1, all antibodies (IgG, IgA, IgM, IgD and IgE) have a relatively large molecular weight and will therefore be in the precipitate (pellets). Smaller proteins, such as albumin will remain in the supernatant. Reconstruction of the precipitate results in samples containing all antibody types while small proteins are removed. The main advantages of protein precipitation are low costs and speed. However, samples are less clean compared to targeted purification and protein A/G beads (see table 1), which leads to worse LLOQs [9]. Nevertheless, protein precipitation has been used for the quantification of several mAbs, such as infliximab [61] and trastuzumab [65].
Figure 9 Procedure of protein (pellet) precipitation as sample purification. Large proteins such as antibodies will precipitate, and smaller proteins such as albumin will remain in the supernatant. Reconstruction of the precipitate results in a sample from which small proteins are removed while the larger antibodies are retained. Reproduced from ref. [9]
Denaturation and trypsin digestion
After using one of the sample purification procedures for sample clean-up, often antibodies are denatured (unfolded). The tertiary structure of antibodies is maintained by hydrophobic, ionic, hydrogen and disulphide bonds. Denaturation of the antibodies by disruption of these interactions will result in faster and more efficient digestion by trypsin because cleavage sites can be accessed more easily compared to folded proteins [9].
For the denaturation, dithiothreitol (DTT) is mostly used to reduce disulphide bonds due to the neutral pH, which makes it compatible with trypsin digestion. Usually, denaturation procedures are conducted under heating conditions (around 60 oC) [9]. Sometimes, iodoacetamide (IAA) is added after DTT reduction of the
disulphide bonds for alkylation of the free thiol groups. Alkylation with IAA prevents reformation of the disulphide bonds. It is often performed for a short duration at ambient temperature and in the dark to avoid side reactions [56,66].
After denaturation, peptides are formed by adding trypsin. Trypsin cleaves the peptide bonds in the antibody after arginine and lysine. These two basic amino acids are easily ionised during electrospray ionisation (ESI), making trypsin very useful in combination with mass spectrometry. The efficiency of digestion with trypsin depends on several factors such as protein ratio, temperature, time and protein accessibility and should, therefore, be carefully evaluated [9]. Often experiments are done to optimise trypsin digestion by monitoring peak areas of specific peptides. Factors that are often optimised include incubation time, temperature, pH and ionic strength [36,39,50,59,61,65].
For example, Chiu et al. [39] optimised the trypsin digestion time and trypsin/protein ratio for the quantification of bevacizumab (see figure 10). For the trypsin digestion time, a clear increase in the intensity
18 of the signature (surrogate) peptide was observed with longer digestion times. A digestion time of 12 hours was selected to be optimal. For the trypsin/protein ratio, the ratio was calculated using the average amount of IgG in adults. The highest signal intensity was obtained with a trypsin/protein ratio of 1:50 [39].
Figure 10 Optimisation of the trypsin digestion procedure. a: effect of trypsin digestion time, b: effect of trypsin/protein ratio. Reproduced from ref. [39]
Nano-surface and molecular oriented limited proteolysis
Iwamoto et al. [67] developed an alternative, selective sample preparation technique which increases the sensitivity of the quantification of mAbs by LC-MS/MS. The signature peptide, which is used for the quantification of mAbs, often originates from the CDRs in the Fab region of the antibody as this is the region that varies the most between different antibodies. Nano-surface and molecular oriented limited (nSMOL) proteolysis, focusses on the trypsin digestion of only the Fab region (which contains the signature peptide) of IgG antibodies by limited protease accessibility to the mAbs. A schematic overview of the principle of nSMOL principle is shown in figure 11. In figure 12, the basic workflow for the nSMOL sample preparation is shown. First, IgG-based antibodies (endogenous and therapeutic) are immobilised by adding a protein A/G resin slurry with pore sizes of 100 nm to the serum/plasma sample. The Fc region of the IgG antibodies will bind to the protein A/G resin, and consequently, the Fab region will be oriented towards the reaction solution side. Non-binding compounds can be washed away by several washing steps. Next, trypsin immobilised nanoparticles are added to the solution. The diameter of the nanoparticle is larger (200 nm) than the pore diameter (100 nm); consequently, the accessibility of the trypsin nanoparticle to the therapeutic mAb and other IgG antibodies is limited. As a result, trypsin digestion will be limited to the variable region, and therefore only peptides from this variable region (VL and VH, including CDRs) are obtained. The digested peptides can be easily
collected using centrifugation or magnetic precipitation. Finally, the sample containing the signature peptide from the CDRs can be analysed by a standard triple quadrupole (QqQ) LC-MS/MS using multiple reaction monitoring (MRM) [53,67]. nSMOL pre-treatment improves the sensitivity, precision, accuracy and reproducibility of the quantification of mAbs by LC-MS/MS analysis [53]. Furthermore, no denaturation is required and due to reduced sample complexity (sample contains significant fewer peptides compared to protein A/G beads purification and protein precipitation) analysis times can be shortened, and the quantification limits can be improved. Also, no capture antibodies or ligands are required like with target-specific sample purification and LBAs [53,68]. Validation for the quantification of several mAbs using nSMOL and LC-MS/MS analysis has been approved according to the food and drug administration (FDA) guidelines for analysis of drug concentrations in biological samples [53,69]. Published validated methods using nSMOL and LC-MS/MS analysis are nowadays available for trastuzumab [68], bevacizumab [55], cetuximab [70], rituximab [70,71], nivolumab [54], brentuximab vedotin [70], and infliximab [52]. Quantification of multiple antibodies in the same sample is also possible using nSMOL and LC-MS/MS analysis [70,72]. It is expected that validated applications for more mAbs using nSMOL will be available soon as the nSMOL kit is commercially available [73].
19
Figure 11 Schematic overview of the principle of nSMOL proteolysis. IgG based antibodies (endogenous and therapeutic mAbs) are immobilised by binding to protein A/G resin with a porous surface. Trypsin immobilised nanoparticles are added, which have a larger diameter than the pore size. As a result, proteolysis is limited to the variable region of the antibodies. LC-MS/MS (QqQ) can then be used to analyse the peptides. The signature peptides from the CDRs can be used for quantification. Reproduced from ref. [53]
Figure 12 Workflow of nSMOL sample purification for the quantification of mAbs by analysing signature peptides using LC-MS/MS. Reproduced from ref. [73]
Iwamoto et al. [55] compared the use of nSMOL sample purification with protein (pellet) precipitation (see figure 13). A significantly improved LLOQ was obtained using nSMOL compared to pellet digestion for the quantification of bevacizumab in plasma. This clearly shows the increase in sensitivity obtained with nSMOL sample purification. In contrast to the pellet digestion, the LLOQ obtained using nSMOL is sufficient for TDM application (at least 1 µg/mL).
20
Figure 13 Comparison between nSMOL sample purification and pellet digestion for the quantification of bevacizumab in plasma. A significantly improved LLOQ is obtained with nSMOL compared to pellet digestion. Reproduced from ref.[55]
Internal standard
A challenge in the quantification of therapeutic mAbs is the use of an internal standard (IS). Internal standards are used to correct for sample losses during sample preparation and LC-MS/MS analysis. Several types of internal standards are described in the literature to correct for losses in the quantification of mAbs. El Amrani
et al. [9] published an overview of the performance of several internal standards to correct for losses in sample
preparation and LC-MS/MS analysis, see table 2.
Table 2 Overview of the performance of various internal standards for correction during sample preparation and LC-MS/MS analysis. Reproduced from ref. [9]
Internal standard Sample preparation LC-MS/MS analysis
Sample purification
Digestion Clean-up and Enrichment
Injection Ionisation Fragmentation
SIL Protein ++ ++ ++ ++ ++ ++ SIL Peptide - - ++ ++ ++ ++ Flanking SIL (Extended) Peptide - - ++ ++ ++ ++ Analogue Protein ++ + + ++ + +
SIL = Stable isotopically labelled, ++ = Optimum correction, + = Moderate correction, - = No correction
SIL-IS protein
The ideal method to correct for sample losses is the use of a stable isotope-labelled internal standard (SIL-IS). SIL-IS are routinely used for the quantification of small compounds by LC-MS/MS. For the quantification of a mAb, the mAb of interest is isotopically labelled to obtain the SIL-IS protein. Such SIL-IS proteins are typically produced by a cell-based approach in which a labelled precursor is metabolically conversed into the protein. The labelled precursor is an amino acid with one or more 13C, 15N or 2H isotopes, which is incorporated into
the mAb [74]. The SIL-IS protein and therapeutic mAb will have different masses, and therefore can be differentiated from each other by MS. Due to the matching amino acid sequence and conformational folding, the SIL-IS protein can be added before the sample preparation. Consequently, correction for sample losses during sample preparation and analysis is optimal (see table 2) [9,50,75]. El Amrani et al. [36] used the commercially available SIL-IS of infliximab for the quantification of infliximab in human serum. It was shown
21 that the SIL-IS corrects for losses during sample purification, digestion and LC-MS/MS analysis. However, a major disadvantage of SIL-IS proteins is that they are usually extremely expensive. Furthermore, the SIL-IS protein is not universal, and therefore a specific IS is required for every mAb. This makes the use of a SIL-IS protein even more costly, especially as not all SIL-SIL-IS mAbs are commercially available [9,50,56,75]. Therefore, several other strategies are described in the literature for correction of sample losses in the quantification of therapeutic mAbs.
SIL-IS peptide
SIL-IS peptides are most often used as an alternative to a full SIL-IS protein. A SIL-IS peptide is the isotopically labelled signature peptide. SIL-IS peptides are readily available and much cheaper compared to the full SIL-IS proteins. However, SIL-IS peptides are spiked after digestion, and therefore cannot correct for losses in the sample purification and trypsin digestion (see table 2). Consequently, it is assumed that sample purification and digestion have a 100% efficiency, but this is often not true. For example, trypsin digestion is rarely complete, and consequently, a significant bias may be observed [9,56,75]. Therefore, when using a SIL-IS peptide, the sample purification and digestion steps must be optimised to limit the variability [56]. SIL-IS peptides have been used for the quantification of infliximab [59], trastuzumab [65] and cetuximab [76].
SIL-IS extended peptide
To correct for the losses in digestion efficiency, SIL-IS with an extended peptide was introduced. The extended peptide consists of the isotopically labelled signature peptide which is extended with 3-6 flanking residues from both the N- and C-termini. The extended SIL-IS peptide is added prior to digestion and therefore, could correct for variations in the digestion step [75]. However, research on whether the use of an extended SIL-IS peptide results in improved corrections compared to the normal SIL-IS peptide is limited.
Faria et al. [56] and El Amrani et al. [9] described that the extended SIL-IS peptide could not completely replicate the digestion process that is experienced by the analyte mAb. Consequently, the ability to correct for differences in digestion efficiencies is expected to be low with the extended SIL-IS peptide [9,56]. El Amrani et al. [9] described that the regular SIL-IS peptide outperformed the extended SIL-IS peptide. The use of the extended SIL-IS peptide resulted in additional variability during the digestion that was not correlated to the mAb digestion, resulting in worse internal standard corrections [9].
Nouri-Nigjeh et al. [75] compared the use of a SIL-IS protein, SIL-IS peptide and extended SIL-IS peptide. For a chimeric anti-hepatitis C virus mAb, it was shown that the use of a SIL-IS protein resulted in high accuracies, with biases all within ±15% of the nominal concentration. When using the SIL-IS peptide and SIL-IS extended peptide, increased biases were observed compared to the SIL-IS protein. In comparison to the regular SIL-IS peptide, the use of an extended SIL-IS peptide resulted in an improved accuracy; however, a negative bias remained, which could lead to differences in TDM interpretation. Improved corrections of the extended SIL-IS peptide could be linked to the correction of incomplete digestion [75]. However, the extended SIL-IS peptides cannot completely replicate the behaviour of the large mAb, leading to the negative bias [75].
From these conclusions, it can be derived that the SIL-IS extended peptide is not the ideal internal standard for mAb quantification. Corrections are not significantly improved compared to the regular SIL-IS peptides. When trypsin digestion is maximised, the use of the regular SIL-IS peptides is preferred, as less variability is introduced [9].
22
Analogue protein
Analogue proteins can also be used as internal standard; a protein which is a structural analogue to the mAb of interest. The analogue protein is unlabelled and should be chosen based on similarities in physiochemical properties such as size, hydrophobicity and isoelectric point with the mAb of interest [56]. Analogue proteins can be added before the sample preparation and could, therefore, in theory, correct for both sample preparation and analysis. However, the peptides generated from analogue proteins are not identical to the signature peptides of the therapeutic mAb. Consequently, corrections for sample preparation may be suboptimal. Furthermore, LC-MS/MS corrections depend on the elution similarities between the signature peptide and analogue protein peptides; these may vary, which will influence the correction capabilities [9]. For the quantification of infliximab in serum, Willrich et al. [61] utilises an IgG horse antibody as analogue protein internal standard due to the similar tertiary structure and low cost compared to full SIL-IS proteins. The horse IgG could be added before the sample preparation and therefore, corrects for variations in sample purification and trypsin digestion. To verify retention times, SIL-IS peptides were added, which will have the same retention time as the infliximab signature peptides. Combination of an analogue protein and a SIL-IS peptide results in good correction for both the sample preparation (analogue protein) and the analysis (SIL-IS peptide) with all validation results meeting the FDA requirements for bioanalytical method validation [61,69].
LC-MS/MS analysis
For separation of the peptides typically a simple reversed phase (RP) C18 column is used. As a result, separation of the peptides is relatively straightforward, and no complicated separation mechanisms are required. After separation of the peptides in the LC, detection is usually performed with a standard QqQ MS. Single reaction monitoring (SRM) or MRM are used for detection of the signature peptide. The specific ion-pair transition of precursor and product ion of the signature peptide is utilised for quantification of the mAb [53,77]. A QqQ MS is nowadays often available in pharmaceutical laboratories, making it possible to quantify mAbs by the bottom-up procedure without the need to buy expensive instruments such as TOF or Orbitrap mass spectrometers [9,77].
An overview of a small selection of the bottom-up quantification methods available in the literature for different mAbs is shown in table 3. The selection of the methods was based on the possible applicability on multiple mAbs. These methods could potentially be used as general methods for LC-MS/MS quantification of most of the IgG mAbs in plasma or serum.
Chiu et al. [50] claims to have developed an accurate and precise method which could be used as a general method for the quantification of mAbs by LC-MS/MS. It was one of the few methods found in the literature that quantified multiple mAbs with the same method. Chiu et al. [50] used protein G purification combined with two internal standards; a post-column infused internal standard (PCI-IS) was used to correct for potential errors during LC-MS/MS analysis such as matrix effects, and an analogue protein was utilised to correct for errors during sample preparation. The method successfully quantified bevacizumab, nivolumab and pembrolizumab with the accuracy and precision meeting the requirements. However, validation of matrix effects resulted in relatively high deviations for all three mAbs; 75-86% for nivolumab, 77-82% for pembrolizumab and 108-121% for bevacizumab [50]. FDA and European medicines agency (EMA) guidelines for bioanalytical method validation set a maximum bias of 15% for matrix effects, and therefore this requirement is not met [69,78]. The authors did not discuss this, but matrix effect corrections by the internal standard might be insufficient. Consequently, it is difficult to determine whether the developed method could work as a general method for mAb analysis while meeting the guidelines for analysis of drug concentrations in biological samples set by the EMA/FDA.
23 The method developed for the quantification of dinutuximab in human plasma by LC-MS/MS of El Amrani et
al. [79] stands out due to the high precision and low biases obtained over the whole concentration range. A
simple protein precipitation with ammonium sulphate led to good results with all validation results being in agreement with the EMA/FDA guidelines [69,78,79]. Consequently, the method might also be applicable for the quantification of other mAbs. The obtained LLOQ of 1 µg/mL will probably be sufficient for TDM of most mAbs [9].
For the several nSMOL applications that have been published, all use the same basis for the analysis method (column, mobile phase, etc.), with minor changes in gradient, flow rate and MS settings for specific mAbs [52,54,55,68,70–72]. An example of a chromatogram obtained using the nSMOL sample purification and the LC-MS/MS method is shown in figure 14. A 6.5 min gradient was used with solvent A being 0.1% aqueous formic acid (FA) and solvent B acetonitrile with 0.1% FA [72]. The nSMOL LC-MS/MS method could potentially be used as universal method for the quantification of mAbs. nSMOL can be used as sample preparation for all IgG mAbs, and only minor alterations in the LC-MS/MS method might be required for the quantification of a specific mAb.
Figure 14 Example of obtained chromatogram using nSMOL sample purification and analysis by LC-MS/MS (QqQ). A human serum sample contained infliximab, adalimumab, ustekinumab, golimumab, eculizumab, etanercept and abatacept. Reproduced from ref.
[72]
Methods using nSMOL have relatively low quantification limits compared to other methods, probably due to the reduced sample complexity (especially for bevacizumab when compared to other methods, see table 3). The nSMOL method has also been shown to work in multiplex analysis (see figure 14), as multiple mAbs could be quantified during the same analysis while meeting all requirements set by the FDA for bioanalytical method validation [70,72]. However, a disadvantage of the published nSMOL applications is that for none of the methods, a method comparison with an immunoassay was performed. Consequently, it is difficult to say whether results are in agreement with the conventionally used LBAs. Iwamoto et al. [80] confirmed this limitation, and therefore, performed a method comparison between the nSMOL LC-MS/MS method and an ELISA kit for the quantification of bevacizumab using 245 clinical plasma samples (see figure 15). Concentration data was found to be correlated (Pearson correlation coefficient = 0.8765), but a difference of approximately 20% was observed between ELISA and nSMOL LC-MS/MS results. Therefore, careful evaluation of concentration results is required when changing of analytical technique [80].
24
Figure 15 Comparison between nSMOL LC-MS/MS method and ELISA assay for the quantification of bevacizumab by Pearson's correlation analysis. 245 clinical plasma samples were included in the method comparison. Reproduced from ref. [80]
Discussion bottom-up quantification
An overview of the methods using the bottom-up approach for the quantification of mAbs is shown in table 3. Ideally, the bottom-up method could be applied to multiple mAbs, such that it can be used as a universal method for the quantification of therapeutic mAbs in pharmaceutical laboratories. Out of the described methods, the methods using the nSMOL sample preparation is the most promising for the application as universal method for the quantification of IgG therapeutic mAbs in plasma and serum.
Compared to protein A/G beads and protein precipitation, nSMOL results in lower sample complexity due to the limited accessibility of trypsin, resulting in higher sensitivities. Furthermore, in contrast to target-specific sample purification, no specific antibodies or ligands are required for every therapeutic mAb. This makes it a simple sample preparation technique which is universally applicable to all IgG mAbs. Important in nSMOL LC-MS/MS analysis is that the signature peptides originate from the variable regions (CDRs). Sometimes, peptides from other regions of the antibody could be detected as potential signature peptide by the described procedure. However, with nSMOL, only the variable regions are digested, and therefore signature peptides should originate from that region.
For corrections of losses during sample preparation and LC-MS/MS analysis, the golden standard is the use of a SIL-IS protein. However, these are expensive and not always commercially available. Alternatives, such as IS (extended) peptides, analogue proteins and the use of two internal standards have been described. SIL-IS peptides are mostly used due to the similar structure with the signature peptide leading to optimal corrections during LC-MS/MS analysis. However, SIL-IS peptides cannot correct for losses during sample preparation, and therefore, these steps should be optimised to limit variability. Methods which use nSMOL as sample purification use another kind of internal standard; a P14R peptide. P14R consists of 14 proline (P) and
an arginine (R) residue. The internal standard is added after capturing of the antibodies, but before the selective digestion (see figure 12). The use of the P14R peptide internal standard is cheap and straightforward
without the need for a SIL-IS. However, research on the P14R peptide internal standard is limited; whether
internal standard corrections for the LC-MS/MS analysis are as good as with a SIL-IS is not investigated. As described before, to have optimal corrections for losses during LC-MS/MS analysis, a SIL-IS peptide is often used due to the similar structure between the IS and mAb signature peptide. The P14R peptide is not a
structural analogue to the signature peptide of the mAb, and therefore, it remains unclear whether corrections capabilities are comparable with SIL-IS peptides. Nevertheless, Shimadzu claims that the internal standard is
25 capable of corrections in sensitivity and retention time [73]. Furthermore, all published methods are validated according to the FDA guidelines for analysis of drug concentrations in biological samples [69]. All validation requirements were met, which shows that sufficient correction for the quantification of mAbs is obtained using the P14R peptide internal standard.
For analysis, typically a simple reversed phase separation on a C18 column is used, combined with a standard QqQ MS for the detection and quantification of the signature peptide. All methods utilising nSMOL used the same LC-MS/MS method for the detection and quantification of mAbs. Minor changes in the gradient, flow or MS settings might be required for optimal quantification of a specific mAb. Nevertheless, using nSMOL combined with a relatively simple LC-MS/MS method could be universally applicable to IgG mAbs for the quantification in serum and plasma samples, as nSMOL sample preparation is specific for all IgG mAbs.
Conclusion bottom-up quantification
Concluding, to generate a universal bottom-up LC-MS/MS quantification procedure which can be used for most of the IgG mAbs, the use of nSMOL sample preparation is the most attractive. With nSMOL, relatively clean samples with low complexity are obtained, and a commercial kit for serum and plasma samples is available by Shimadzu [73]. When using the nSMOL kit, it is recommended to use the P14R peptide internal
standard. The published validated methods meet all the requirements set by the FDA, and therefore the correction with the P14R peptide internal standard is sufficient for the quantification of mAbs. After sample
preparation using the nSMOL kit, the LC-MS/MS analysis is relatively straightforward with a simple reversed phase separation on a C18 column and detection using MRM by a QqQ MS. The same LC-MS/MS method, with minor alterations in gradient, flow or MS settings, has been used for multiple mAbs using nSMOL as sample purification. This makes it possible to use the nSMOL sample purification combined with the LC-MS/MS analysis as a universal method for IgG mAbs. When a new mAb must be quantified, it is therefore not necessary to start the method development from scratch. In principle, only the signature peptides and MRM transitions of the new mAb needs to be determined. Consequently, the nSMOL kit can be used for sample preparation and the existing LC-MS/MS method can be used for quantification. Minor alterations in gradients, flow rates or MS settings might be required for optimal separation and detection of a specific mAb. Finally, a quick validation of the new method for the new mAb would suffice. Typically, obtained sensitivities in LC-MS/MS analysis combined with nSMOL sample purification are sufficient for TDM application of mAbs (at least 1 µg/mL). When an ELISA method is replaced by LC-MS/MS, a method comparison is required to determine whether concentration results are in agreement; a correction might be necessary.
