2D-LC-UV/MS FOR THE ANALYSIS OF MABS AND ADCS
MSc Chemistry
Analytical Sciences
Literature thesis
By
Sanne Pot
2588708 - 11200014
November 2019
12 ECTS
Examiner:
Dr. Bob Pirok
Supervisor/Examiner:
Dr. Rob Haselberg
ABSTRACT
An emerging class in the biopharmaceutical industry are mAbs and ADCs. Their success is attributed to their high specificity for their targeting antigen. MAbs and ADCs are highly complex glycoproteins, exhibiting a wide range of microheterogeneities due to post-translational modifications (PTMs). These PTMs can occur during expression, purification and storage. Besides PTMs, the protein can also undergo aggregation, truncation and degradation. Since these modifications may significantly impact the pharmacological properties of the protein, regulation agencies demand extensive analysis in order to ensure the product quality and consistency. Many analytical techniques have been proposed for the characterization of mAbs and ADCs. However, characterization of therapeutic proteins involves many off-line processes – such as fractionation, reduction, alkylation, digestion – and multiple separation techniques to ensure the product quality and consistency. This makes the characterization very time consuming and labor intensive. Besides, sample loss, contamination and degradation may arise. To overcome these issues, the use of 2D-LC is rapidly increasing for the characterization of therapeutic proteins. 2D-LC combines two different separation mechanisms which can tremendously improve the separation power in terms of peak capacity and selectivity. In addition, it allows hyphenation of separation techniques using non-volatile buffers to MS by using a second separation mode.
This literature thesis describes the competence of 2D-LC for the characterization of mAbs and ADCs. Based on this literature study, both heart-cutting and comprehensive 2D-LC-UV/MS seem to be very promising for the analysis of mAbs and ADCs. Heart-cutting 2D-LC seems to be very suitable for monitoring bioreactor samples and small drug species in ADC samples. A great advantage of 2D-LC over 1D-LC is the direct coupling that can be achieved between chromatographic techniques using non-volatile buffers - such as, HIC, SEC and CEX – and MS. This allows direct MS-identification of size and/or charge variants and different DARs. Besides, online reduction and digestion steps can be incorporated, which is very powerful since its fast, can be automated and allows no offline processes. Online comprehensive 2D-LC-UV/MS is very powerful for the separation of peptides since a high resolving power can be achieved by combining two orthogonal separation modes. Once the characterization of mAbs and ADCs is completed by MS, 2D-UV contour plots can be used for fast comparison of different batches and biosimilars. Overall, 2D-LC-UV/MS is a very powerful technique for the characterization of mAbs and ADCs since it allows automatization, a high resolving power can be achieved, it involves less offline processes and allows coupling of non-volatile chromatographic techniques to MS.
ABBREVATIONS
ACN - Acetonitrile
ADC - Antibody drug conjugate
AEX - Anion exchange chromatography ASM - Active solvent modulator
CE - Capillary electrophoresis
CEX - Cation exchange chromatography DDT - Dithiothreitol
Da - Dalton
DAR - Drug-to-antibody ratio EMA - European Medicines Agency ESI - Electrospray ionization
FA - Formic acid
FDA - Food and Drug Administration
HIC - Hydrophobic interaction chromatography HILIC - Hydrophilic interaction liquid chromatography HMW - High molecular weight
IEC - Ion-exchange chromatography
IgG - Immunoglobulin G
IdeS - Immunoglobulin-degrading enzyme of Steptococcus pyogenes Kav - Partition coefficient
LC - Liquid chromatography LMW - Low molecular weight mAb - Monoclonal antibody
MS - Mass spectrometry
PI - Isoelectric point
ProA - Protein A affinity chromatography PTM - Post-translational modification
RPLC - Reversed phase liquid chromatography RSD - Relative standard deviation
SCX - Strong cation exchange SEC - Size-exclusion chromatography SPAM - Stationary phase assisted modulation TCEP - Tris (2-carboxyethyl) phospine TFA - Trifluoro acetic acid
TOF - Time-of-flight
UHPLC - Ultra-high-performance liquid chromatography UV - Ultraviolet
TABLE OF CONTENT
Abstract...III Abbrevations...IV
1. Introduction...1
2. MABS and ADCs...3
3. 2D-LC...4
3.1. 2D-LC techniques...5
3.2. Modulations...7
3.2.2. Passive modulation...7
3.2.3. Active modulations...7
3.3. Analysis of mAbs and ADCs by 2D-LC...8
4. 2D-LC-UV/MS for the analysis of mAbs...12
4.1. (Multiple) heart-cutting-2D-LC...12
4.2. Comprehensive 2D-LC...17
5. 2D-LC-UV/MS for the analysis of ADCs...21
5.1. (Multiple) Heart-cutting 2D-LC...21
5.2. Comprehensive 2D-LC...25
6. Discussion...29
6.1. 2D-LC-UV/MS for the analyis of mAbs...29
6.1.1. (Multiple) heart-cutting 2D-LC...29
6.1.2. Comprehensive 2D-LC...29
6.2. 2D-LC-UV/MS for the analyis of ADCs...30
6.2.1. (Multiple) heart-cutting 2D-LC...30 6.2.2. Comprehensive 2D-LC...31 7. Conclusion...32 8. Future perspectives...33 References...34 Appendix...I Appendix I: 2D-LC applications mAbs...I Appendix II: 2D-LC applications ADCs...V
1. INTRODUCTION
The fastest growing class in the biopharmaceutical industry are monoclonal antibodies (mAbs).2 The remarkable success of mAbs and related products is attributed to their high specificity to their targeting antigen.2–4 The field of application is mainly oncology and autoimmune disease treatment (such as Crohn disease and rheumatoid arthritis).5,6 Since mAbs that are currently on the market are soon evolving out of patent, the market will be reshaped by biosimilars. Biosimilars are a subclass of mAbs, which are similar to the original product regarding quality, safety and efficiency.2,3,7 The success of mAbs has also led to the development of a new generation of therapeutic proteins, such as antibody drug conjugates (ADCs).3 ADCs are composed by a specific mAb that is bound to a small cytotoxic drug molecule (payload) via a linker.8 ADCs look very promising as biotherapeutics in the field of oncology9, due to their high specificity and the ability to selectively deliver a potent drug to their targeting antigen.10–12
MAbs and related products are highly complex glycoproteins, exhibiting a wide range of microheterogeneities due to post-translational modifications (PTMs). These PTMs can occur during expression, purification and storage.2 PTMs - including glycosylation, glycation, deamidation and oxidation - can affect the size, charge and structure of the protein.3 The protein can also undergo aggregation, truncation and degradation.3,13 Consequently, the original biopharmaceutical product consists of many proteoforms instead of one unique molecule.14 Since these modifications may significantly impact the pharmacological properties of the protein2, regulation agencies demand extensive analysis in order to ensure the product quality and consistency. Compared to small molecule-based medicines (150-600 Da)10, detailed characterization of mAbs represents a challenge due to their large size (150 kDa), complexity and the various aspects which define their structure. For ADCs, the level of complexity is further increasing due to the linker-payload conjugation that is attached to the mAb.12
The need for analytical techniques to perform detailed characterization has increased, due to the expansion of therapeutic proteins in the biopharmaceutical industry. The last few years, many analytical techniques have been implied for the characterization of therapeutic proteins. Liquid chromatography (LC) is a very popular separation technique, because of the different separation modes that can be used within this technique. Due to developments such as ultrahigh performance pressure liquid chromatography (UHPLC) in combination with small particle sized columns – the peak capacity in LC has been tremendously increased. Other innovations include improvement of column features, such as the pore size, porous particles and chemical inertness. These technical improvements have led to faster characterization of therapeutic proteins by LC.14,29 Another popular separation technique for the analysis of therapeutic proteins is capillary electrophoresis (CE). This technique is very powerful, due to its high efficiency, resolution and fast separation.12 Mass spectrometry (MS) has also become an important detection technique for the characterization of therapeutic proteins. Due to the introduction of electrospray ionization (ESI) it became possible to couple LC and CE to MS, which allows mass identification on the separated compounds.15,16
Yet, boundaries are being pushed since there is a high demand for increasing the overall performance of analytical methodologies in terms of throughput. To date, characterization of therapeutic proteins involves many off-line processes – such as fractionation, reduction, alkylation, digestion – and multiple separation techniques in order to ensure the product quality and consistency.17 Besides,
incompatibility issues between separation and detection techniques is an often-encountered problem. For example, well established methods for the characterization of therapeutic proteins, such as ion-exchange chromatography and size exclusion chromatography, using non-volatile buffers hampers direct coupling with MS. Therefore, off-line fractionation with subsequent buffer exchange and concentration steps must be performed prior to MS. This is very time-consuming, increases the risk of unwanted modifications and requires large amount of samples.4 2D-LC might be a suitable technique to overcome the described issues.17 In the last decade, the use of 2D-LC for the characterization of therapeutic proteins is rapidly increasing due to the introduction of commercially 2D-LC systems from several manufacturers. Therefore, 2D-LC is no longer considered a specialist tool.18 2D-LC combines two different separation mechanisms which can tremendously improve the separation power in terms of peak capacity and selectivity.1 In addition, it allows hyphenation of separation techniques using non-volatile buffers to MS by using a second separation mode.1,17
The aim of this literature study, is to discuss the performance and suitability of 2D-LC for the characterization of therapeutic proteins, such as mAbs and ADCs. Additionally, advantages over 1D-LC will be discussed. In this literature study both heart-cutting and comprehensive 2D-LC for the analysis of mAbs and ADCs will be discussed.
2. MABS AND ADCS
MAbs that are currently on the market are humanized immunoglobulin G (IgG), which structure consists of a Y-shaped molecule. MAbs are produced in mammalian cell cultures using host cell, such as Chinese hamster ovary (CHO) and myeloma cells.19 MAbs are composed by four polypeptide chains, two identical heavy chains (~50 kDa) and two identical light chains (~25 kDa). This gives a molecular weight of approximately 150 kDa.14,20 These polypeptide chains are maintained by noncovalent interactions and disulfide bridges. MAbs can be divided into three different IgG subclasses, depending on the amino acid sequence of the heavy chains (IgG1, IgG2 or IgG4). IgG3 are not used for therapeutic treatment, since they have a reduced half-life time. Figure 1A shows the structure of an IgG1 mAb. The light chain is divided by one variable region (VL) and one constant region (CL). The heavy chain can be divided into one variable region (VH) and three constant regions (CH1, CH2 and CH3). The antigen binding side is comprised by the variable regions (VL and VH) of the heavy and light chain and known as the complementary determining region (CDR). For IgG1 and IgG4, the polypeptide chains are held together by 16 disulfide bonds, while for IgG2 the polypeptide chains are held together by 18 disulfide bonds. The heavy and light chain are linked by one disulfide bond and the heavy chains are linked by two (IgG1 and IgG4) or four (IgG2) and are in the flexible hinge domain (Hi). The other 12 cysteine bridges are intramolecular and divide the light and heavy chains in variable and/or constant regions.20 In addition, most mAbs consists of two N-glycosylation sites in the CH2 domain of the heavy chain where N-glycans can bind.3,20 These N-glycans represent an average of 2-3% of the total antibody molecular weight.3 Glycans are oligosaccharides that can consist of a variety of sugars such as galactose, mannose, fucose and sialic acid.
Figure 1B shows the general structure of an ADC. As is shown, ADCs can be divided into three important structural components: the mAb, a payload (small drug molecule of ~1 kDa) and a linker that covalently binds the payload to the mAb.8 The mAb is important for the recognition of the targeting antigen, while the conjugated payload plays an important role in the elimination of their targeting antigen.21 MAbs used in ADCs are mainly IgG1 isotypes. Payloads in ADCs are mainly microtubule inhibitors and DNA-damaging drugs to eliminate their targeting antigen.22 To date, four ADCs - brentuximab vedotin, T-DM1, inotuzumab ozogamicin and gemtuzumab ozogamicin - are approved for therapeutic treatments.12
Figure 1: General structure of an IgG1 mAb (A) and IgG1 ADC (B). Glycan sugar groups are denotated as green: mannose, yellow: galactose, blue: N-acetylglucosamine and red: fucose. Reproduced from Ref. 20 and modified from ref. 23.
An important feature of an ADC is the synthetic linker, that covalently binds the mAb to a payload.21 The synthetic linker is important, since it has an impact on the physicochemical properties of the ADC. Ideally, the linker is stable and does not release the payload before the targeting antigen is reached.22 This can be achieved by using cleavable or non-cleavable linkers. Cleavable linkers are designed to release the active payload when it reaches their targeting antigen by a change in environment. For example, acid-sensitive linkers (e.g. hydrazone) are stable at neutral pH and therefore can transport the drug through the blood circulation, but release the drugs at an acidic pH.21 Non-cleavable linkers involve linkers that are released upon degradation of the mAb. Consequently, the linker is still being attached to the active payload.22 The conjugation of the synthetic linker is also particularly important, since it determines the amount of drug that is loaded onto the mAb, namely drug-antibody ratio (DAR). A small number of drugs attached to the mAb lead to a decrease in efficiency, while too many drugs attached to the mAb lead to unstable physicochemical properties.22 Currently, the linkers are conjugated to the mAb via amino acid residues on lysines or via thiols on reduced hinge cysteines.12,21 Conjugation leads to a biopharmaceutical product consisting of a heterogeneous population of ADCs with various numbers of drugs ranging from 0 to 8 drugs per antibody. Antibodies with a DAR of zero are considered impurities, whereas DAR species with odd numbers are considered degradation products.24 Figure 2 shows the potential isomers of DAR species of a cysteine conjugated ADCs. Whereas for cysteine conjugation only eight interchain cysteine residues are available, for lysine conjugation 80 to 100 lysines are available. Consequently, lysine-conjugated ADCs consist of a more heterogeneous mixture. Yet, the overall level of drugs per antibody is the same as for cysteine conjugation (0 to 8 drugs per antibody).25 Since ADCs consist of various DAR species, the average DAR is often calculated (Equation 1) since it gives an indication of the amount of payload that is delivered to a tumor cell. The ‘optimal’ average DAR in an ADC mixture is 3.5 and 4, respectively.21,22,26,27
DAR
avg=
∑
0 8nA
DARn∑
0 8A
DARn (Equation 1) 24Where n = number of loaded drugs, ADARn = peak area of DAR n and n = DAR number
Figure 2: The positional isomers for different DARs of a cysteine-conjugated ADC. DAR zero is considered an impurity. Reproduced from ref. 28.
3. 2D-LC
2D-LC experiments were already performed in the late 1970’s and early 1980’s. Since then, 2D-LC has been widely explored for a variety of applications in a variety of fields including proteomics.1 The last decade, the use of 2D-LC is rapidly increasing due to the introduction of commercially 2D-LC systems from several manufacturers.18 2D-LC can be defined as a technique in which analytes are separated by two completely independent separation mechanisms to obtain an orthogonal separation, which enhances the selectivity and peak capacity.1,17,29
As can be observed in Figure 3, a conventional 2D-LC system consists of a first dimension pump, column, detector (optional) and a second dimension pump, sampler (i.e. modulator), column and detector.18 The sample mixture is separated in the first-dimension column (1D column) using an isocratic or gradient elution followed by a separation in the second-dimension column (2D column). 2D-LC for the analysis of mAbs and ADCs is mainly done in combination with MS and UV-detection. The transfer of the eluate of the first dimension (1D effluent) to the second dimension is achieved by the sampler (i.e. modulator). To enhance the selectivity and peak capacity, the first dimension and second dimension of a 2D-LC system should have two completely different separation mechanisms.17,29 Thus ideally, there is no correlation in the retention time obtained in the first dimension and second dimension.18,30,31
Figure 3: Schematic representation of a conventional 2D-LC system. Reproduced from ref. 18. 3.1. 2D-LC TECHNIQUES
2D-LC can be divided into two techniques, namely cutting and comprehensive 2D-LC. In heart-cutting 2D-LC, only one fraction (LC-LC, Fig 4A) or multiple fractions (mLC-LC, Fig 4B) of 1D effluent are collected, stored and sampled onto the second dimension.17,30,18 This means that only parts of the 1D
effluent are selected for the second dimension separation, which is useful when only a few compounds in the sample mixture are of interest.17 In comprehensive 2D-LC, a continuous stream of 1D effluent is transferred to the second dimension, which can be done selectively or in full comprehensive mode. In selective comprehensive 2D-LC (sLCxLC, Fig 4C) there is a continuous stream of 1D effluent, but only a selective part is subjected to the 2D column. Full comprehensive 2D-LC (LCxLC, Fig 5D) involves the analysis of the whole sample in the first and second dimension (Fig. 4D).17,30,18 Comprehensive 2D-LC is frequently used for the separation of complex samples due to the high peak capacity that can be achieved.29
Figure 4: Schematic representation of the four different modes that can be used in 2D-LC. Reproduced from ref 29.
In 2D-LC, the 1D effluent can be transferred from the first dimension to the second dimension in online or offline mode. In offline mode, the eluate from the first dimension is stored and fractions of interest are injected into the 2D column at a certain time point.18 In online mode, the eluate from the first dimension is immediately loaded onto the 2D column after collection. As summarized in Table 1, both online and offline 2D-LC have pros and cons. For example, online configuration can be very complex due to the modulation time that has to be equal to the 2D separation, yet it has the advantage that it can be automated. While offline configuration cannot be automated, it is easy to operate since no time constraints are needed and conventional HPLC systems can be used.4 Another 2D-LC mode is stop-and-go 2D-LC, where the 1D separation is stopped after the 1D effluent is collected and the 2D separation is carried out.18,30,33
Table 1: Advantages and disadvantages of off-line and on-line 2D-LC.30, 33
Offline 2D-LC Online 2D-LC
Advantages
Conventional HPLC system can be used
No time constraints in 2D-LC
High peak capacity
Fast
Automation possible
High reproducibility
Time-consuming
Automatization not possible
Low reproducibility
Sample loss
Lower peak capacity
Complex set-up
Modulation time must be equal to 2D separation
Solvent compatibility issues
3.2. MODULATIONS
The transfer of the 1D effluent from the first to the second dimension is achieved by an interface, also known as a modulator. The 1D effluent can be transferred by an active or passive modulator. Both modulators will be discussed in this paragraph.
3.2.2. PASSIVE MODULATION
A conventional passive modulator is shown in Figure 5A. As is shown, fractions from the 1D effluent are transferred and sampled into the 2D column using an 8- or 10- port valve equipped with two storage loops. The loops are used for sampling of the 1D effluent and loading of the 1D effluent onto the 2D column. In the case of mLC-LC, loops can be replaced by multiple sampling loops (also named ‘’decks’’) to allow storage of multiple fractions.29 This modulation interface is known as passive modulation, since the volume of the 1D effluent and/or concentration of the analytes is not adjusted prior to injection into the 2D column.31
Passive modulation cannot overcome the incompatibility issues that can occur when two different separation mechanism are combined.18 This can influence the overall performance of a 2D-LC separation.34 For instance, when the first and second dimension have mobile phases consisting of different viscosities, this can lead to flow instabilities and an effect which is known as viscous fingering.29 Viscous fingering is a high-viscosity injection plug that moves through the medium of a 2D column due to the lower viscosity of the 2D eluent in comparison to the 1D effluent. Viscous fingering leads to peak deformation and/or peak splitting. In addition, different solvent strengths between the 1D effluent and 2D eluent also have a tremendous effect on the retention mechanism. For example, if the solvent strength of the 1D effluent is higher than the 2D eluent this can lead to unretained peaks, which subsequently lead to smaller retained peaks at the expected location. This phenomenon is also called breakthrough. Another problem arises when salts present in the 1D effluent are adsorbed onto the stationary phase of the 2D column. As a consequence, the salts are concentrated on the column which can complicate detection when MS is used.29,34
3.2.3. ACTIVE MODULATIONS
To overcome the limitations of passive modulation, various researchers have been looking into active modulation. Active modulation is an interface where ‘’action’’ takes place29, such as reducing the volume of the 1D effluent or concentrating analytes.29,31 However, this section will only discuss the most applied active modulators which are the active solvent modulator (ASM) and stationary-phase active-assisted modulation (SPAM).
Active solvent modulation (ASM)
Recently, Stoll et al. introduced ASM to overcome ‘’solvent mismatch’’ between the first and second dimension.34 ASM is based on the existing valve technology in 2D-LC (Fig.5B), which makes it a
relatively simple approach. The general principle of ASM is decreasing the solvent strength of the 1D effluent by dilution with 2D eluent using an ASM capillary prior 2D analysis.29,34 The 2D eluent coming from the 2D pump is divided into two flows to allow transfer of the 1D effluent in Deck A by the 2D eluent (port 8) and flow through the ASM capillary (port 9). Both flows will be transferred to the 2D column via port 5 to allow dilution of the 1D solvent prior analysis.35
Stationary phase-assisted modulation (SPAM)
Another commonly used active modulator is SPAM (Fig. 5C). This modulator uses trapping columns – which are commonly guard columns – instead of storage loops.29,31 This approach was first introduced by Vonk et al.31 SPAM traps analytes present in the 1D effluent to concentrate the analytes and remove the 1D eluent. As a result, higher sensitivity can be obtained and less injection volumes are needed.29 Figure 5C shows the set-up of SPAM for a 2D-LC system.31 As is shown in the figure, SPAM consist of a mixer (optional), two trap column and a valve. First, the 1D effluent is mixed with weak solvent using a mixer to decrease the solvent strength of the 1D effluent and enhance the retention of analytes on the trap column.29 Next, the 1D effluent goes through the trap column. The trap column ensures retention of the analytes that pass through the trap column, while 1D eluent is unretained and goes to waste. Next, the valve is switched and 2D eluent elutes the trapped analytes to the 2D column. When a sample consists of analytes with different chemical properties (hydrophobic and hydrophilic species) it can be challenging to retain all analytes on the trap column, because not all analytes will interact with the stationary phase of the trap column. Consequently, SPAM can lower the recovery of the analytes when a sample containing different chemical properties is separated. It is also very important that both trap columns have the same stationary phase to prevent reduced robustness of the separation.29
Figure 5: Schematic representation of a passive modulator (A), ASM (B) and SPAM (C). Reproduced from ref. 29,31,35. 3.3. ANALYSIS OF MABS AND ADCS BY 2D-LC
The analysis of mAbs and ADCs is performed at various levels, including protein, peptide, glycan and amino acid level.2 This is necessary in order to ensure the product quality and consistence without unwanted modifications.19 The analysis of protein and peptide level by LC-UV/MS is achieved by the separation of intact, digested and/or reduced protein (Fig. 6).2 Digestion and reduction of proteins result in protein fragments, which allow localization of PTMs. Digestion with pepsin or the immunoglobulin-degrading enzyme of Streptococcus pyogenes (IdeS) leads to generation of large fragments (25 – 100 kDa). Reduction of the mAb is done to allow identification and localization of PTMs on the light and heavy chains. Reduction of the disulfide bonds can be accomplished by dithiothreitol (DTT) or tris(2-carboxyethyl) phosphine (TCEP). In order to characterize proteins at peptide levels, proteins are commonly treated with trypsin.2,36
This paragraph will briefly discuss the most commonly used separation modes for the characterization of mAbs and ADCs. This includes protein affinity chromatography (ProA), ion-exchange chromatography (IEX), size-exclusion chromatography (SEC), reversed phase liquid chromatography (RPLC), hydrophobic interaction chromatography (HIC) and hydrophilic interaction liquid chromatography (HILIC) . Detection of the separated compounds with these techniques is mainly done by UV-detection and ESI-MS.
PROTEIN AFFINITY CHROMATOGRAPHY (PROA)
ProA is commonly used as first chromatographic step for the purification of mAbs from their cell culture.37,38 ProA consists of a basic column support with immobilized native or recombinant protein ligands obtained from staphylococcus aureus or Escherichia coli. These ProA ligands have a specific binding affinity towards the Fc region of the mAb, and, therefore, the mAb reversible binds to the ProA ligands. Interaction between the ProA ligands and the Fc region of the mAb is achieved under basic conditions (pH ~7) while elution of the mAb is achieved by lowering the pH in the column to acidic conditions.39 After purification of the intact protein, other chromatographic techniques can be used in order to ensure the quality of the mAb.37
ION-EXCHANGE CHROMATOGRAPHY (IEX)
The separation mechanism in IEX is based on electrostatic interaction between the stationary phase of the column and charged protein variants. IEX can be divided into two modes; anion exchange chromatography (AEX) and cation exchange chromatography (CEX). In AEX the stationary phase is positively charged, while in CEX the stationary phase is negatively charged. For the analysis of mAbs, CEX is mainly used since the isoelectric points (the pH at which the protein has no net charge) are between pH 7 and 9.40 In order to obtain electrostatic interaction between the therapeutic protein Figure 6: Characterization of an IgG1 mAb and cysteine-conjugated ADC at different levels (protein, , glycan, peptide and amino acid level). Reproduced and modified from ref. 2,9,12,26.
and CEX, the pH should be below the isoelectric point of the mAb to obtain an overall positive charge. CEX can be carried out by applying a salt or pH gradient. With a salt gradient, the concentration of salt is increased at constant pH. With a pH gradient, the pH is increased (CEX) while the salt concentration is constant.12 In CEX, peaks eluting before the main peak are referred to as acidic variants, while peaks after the main peak are referred to as basic variants. IEX uses native running conditions which allows the protein in its native state. Due to the use of non-volatile buffers, direct coupling with MS is hampered.
CEX is a very suitable technique for the separation of charge variants of intact proteins.12 Charge variants are a consequence of several amino acid modifications, including deamidation of asparagine, glycation of lysine residues, isomerization of aspartate, C-terminal lysine truncation, oxidation and certain glycosylation variants.41,42 Additionally, CEX can be used for the separation of different DAR species of lysine-conjugated ADCs.25
SIZE-EXCLUSION CHROMATOGRAPHY (SEC)
The separation mechanism in SEC is based on the hydrodynamic radius. Depending on the size and pore size of the stationary phase in SEC the compounds will have different diffusion inside these pores. Compounds with a small size will spend more time in these pores than compounds with a large size, which will result in different retention times.12 In SEC-UV, peaks eluting before the main peak (predominantly the monomer) are referred to as high molecular weight (HMW) compounds, whereas peaks eluting after the main peak are considered low molecular weight (LMW) compounds.43 SEC can be used in non-denaturing (aqueous SEC) and denaturing conditions (organic SEC) using isocratic conditions.12 Just as for IEX, SEC uses non-volatile buffers which hampers direct hyphenation with MS. In literature it is shown that the use of volatile buffers in SEC leads to a decrease in overall SEC performance.43 SEC is commonly used for the separation of intact proteins to access aggregation and fragments that occur as a consequence of incomplete formation of disulfide bridges.2,12,41
HYDROPHOBIC INTERACTION CHROMATOGRAPHY (HIC)
HIC separates protein variants based on their hydrophobic surface (amino acid residues or other hydrophobic functional groups) using a weak hydrophobic stationary phase (e.g. butyl). The separation mechanism in HIC is based on the salting-out principle. First, proteins are retained on the stationary phase by applying a mobile phase with a high salt concentration. Next, proteins are eluted by decreasing the salt concentration with a gradient using an aqueous mobile phase of a non-denaturing salt at physiological pH (pH 6-7). Proteins with a relative low hydrophobicity will elute first, while proteins with a relative high hydrophobicity elute last. For ADC species, the retention increases with increasing DAR, due to the increase in hydrophobic cytotoxic drug.24,28 A main advantage of HIC is the use of native running conditions (physiological pH, room temperature, weak stationary phase) which allows activity measurements after the separation in order to study structure-function relationships. The main limitation of HIC is the incompatibility with MS, due to the high salt concentration used in this separation mode.2
REVERSED-PHASE LIQUID CHROMATOGRAPHY (RPLC)
RPLC consists of a hydrophobic stationary phase (C4, C8 or C18 chemistries).2 Similarly to HIC, the separation is based on the hydrophobicity of protein variants.2,12 In comparison to HIC, RPLC has the major advantage of being MS compatible, due to the volatile buffers that are used in this approach.12
In combination with its good robustness and high resolution, it’s a very suitable approach for routine analysis. RP uses denaturing conditions, due to the organic modifiers used in the mobile phases and the high temperature (60-90 ⁰C) at which the separation is carried out.2 Due to the denaturing conditions, the non-covalent bindings in mAbs are disrupted, which allows identification of positional isoforms.24 RPLC is the most widely used separation technique in LC for the characterization of therapeutic proteins on both protein and peptide level.2
HYDROPHOBILIC INTERACTION CHROMATOGRAPHY (HILIC)
HILIC consists of a hydrophilic stationary phase, which allows the separation of polar neutral groups on proteins. These polar neutral groups are retained by applying a water gradient, resulting in a water-enriched solvent layer adsorbed onto the surface of the stationary phase. By using a mobile phase consisting of acetonitrile, hydrophilic compounds can be separated by differential partitioning between the acetonitrile rich mobile phase and water-layer. HILIC is frequently used for the analysis of amino acids, (released) N-glycans and glycopeptides.2,44 The retention increases with an increasing number of sugar units and size. The analysis of intact proteins by HILIC is limited due to the poor solubility of proteins in the mobile phase and adsorption issues.45 However, due to the introduction of wide-pore neutral amide-bonded stationary phases, HILIC-MS might be a promising application for the analysis of intact glycosylated proteins.45
Figure 7 shows a summary of the discussed chromatographic techniques (ProA, CEX, SEC, HIC, RPLC and HILIC) for the characterization of a therapeutic mAb (i.e. Trastuzumab) at protein level.
Acidic variants Basic variants
HMW region LMW region HILIC
ProA
Figure 7: Analysis of intact trastuzumab using different chromatographic techniques (ProA, RPLC, CEX, SEC and HIC) and partial digested trastuzumab by IdeS using HILIC. Reproduced from ref. 2,37,44.
4. 2D-LC-UV/MS FOR THE ANALYSIS OF MABS
Characterization of mAbs can be very complex and time-consuming due to the offline processes, separation and detection techniques that are involved. Therefore, there is a need for improving analytical technologies in terms of throughput, automatization and resolving power. 2D-LC is an attractive technique, since it can be automated, and it can tremendously increase the separation power. Besides, it allows coupling between non-volatile chromatographic techniques and MS. Depending on the analytical question, heart-cutting or comprehensive 2D-LC can be used. This paragraph will discuss heart-cutting and comprehensive 2D-LC applications separately. A detailed overview of the mobile phases, columns and detection settings used for each application can be found in Appendix I and II.
4.1. (MULTIPLE) HEART-CUTTING-2D-LC
Protein modifications can already arise during the cell culture process by slight changes in culture conditions. For instance, a change in pH, temperature and/or dissolved oxygen, within the bioreactor system can lead to protein modifications.37,46 Therefore, it is important to ensure the product quality as early as the cell culture process.37 Amand et al. (2014) proposed an automated heart-cut ProA-CEX-UV method for near real-time monitoring of mAbs during production.46 As shown in Figure 8, samples from the bioreactor were transferred to 2D-LC using an automated reactor sampling system. In 1D ProA, the mAb was purified from the host cells, whereas the 2D CEX column was used for the separation of mAb charge variants. In ProA, elution of the mAb was achieved by lowering the pH in the column by acetic acid. ProA and CEX were highly compatible, because the acidic 1D effluent led to a pH below the isoelectric point of the protein leading to an overall positive charge. This method allowed the analysis of charge variants within two hours. However, the limit of detection (LOD) must be determined for each specific mAb. Since this method is not coupled to MS, for each mAb a reference charge profile should be available.46
Figure 8: Flow diagram of the near real-time monitoring of charge variants (Left) Separation of the charge variants of the mAb (Right). Reproduced from ref. 46.
Another method for the analysis of bioreactor samples was proposed by Williams et al. (2017).47 In this study they implemented an automated multiple-heart-cutting ProA-SEC-UV method for the analysis of soluble protein aggregates of bioreactor samples. By implementing an in-line fraction collection device, they were able to collect up to twelve fractions from the 1D separation in 100 μL
Samples from the bioreactor
Automated reactor sampling system
loops instead of one. The applicability of the method was demonstrated by the analysis of a thermally aggregated mAb. Elution of the mAb from the 1D ProA column was achieved by using a linear acetic acid gradient, rather than using a step gradient.As is illustrated in Figure 9A, the gradient led to a pre-peak and main pre-peak which were fractionated and sequentially loaded onto the 2D SEC column. The SEC chromatogram for both peaks can be observed in Figure 9B. The main peak was assigned to a mAb monomer, several fragments and a buffer peak. The pre-peak, which has a rather complex profile, was assigned to aggregates, a monomer and small fragments. The pre-peak contained mainly impurities and was therefore considered undesirable, while the main peak could be used for further purification and additional analysis. This 2D-LC method might be suitable for mAb clone selection by initial removal of impurities (pre peak) and isolation of the desired product (main peak) to minimize aggregation in the final product.47
In order to put the 2D-LC performances for the analysis of bioreactor samples in perspective, Sandra et al. (2017) developed an entire workflow using both heart-cutting and comprehensive 2D-LC.37 Heart-cut ProA-SEC-UV was used for the analysis of size variants and determination of the mAb titer of trastuzumab and its CHO clones supernatants. As can be observed from the chromatogram (Fig. 10A), the CHO clones differ in mAb titer. The mAb titer in clone 3, 8, 9 and 10 were 196, 243, 572 and 809 μg/mL, respectively. Approximately 80% of the retained ProA peak was collected and subsequently loaded onto the SEC column for further separation of LMW- and HMW compounds (Fig 10B). For clone 3, 8, 9 and 10, HMW compounds were present for 2.7%, 0%, 1% and 4%, respectively. The calculated mAb titer using the peak areas in SEC were comparable to the mAb titer obtained with ProA (202, 233, 578 and 819 μg/mL, respectively, for, clone 3, 8, 9 and 10). The first and second dimension showed excellent repeatability (RSD <4%, respectively) for both retention time and peak area.
Heart-cut ProA-CEX-UV allowed insight into various modifications which lead to mAb charge variants. The CEX chromatogram for tocilizumab and its CHO clones supernatants is shown in Figure 11. As can be observed from the figure, clone B and C show slightly different charge state profiles compared to the originator and other clones, suggesting unwanted modifications. For clone B more basic variants were observed, while for clone C the whole charge state profile shifted towards the acidic side of the chromatogram. An excellent repeatability of the retention time and peak areas were obtained (RSD <3%, respectively). Heart-cut ProA-RPLC-MS was used for the conformation of the molecular weight as well as for identification of the major modifications, such as glycosylation, of tocilizumab and its CHO clones supernatants. The fraction from the ProA separation was desalted by RP followed by detection with ESI-MS. N-glycans (G0F, G1F and G2F) with cyclic and lysine truncation on the N- and C-termini of the heavy chains were identified. Also, for clone C, a 68 Da lower MW was observed compared to its originator. To assign the modification that led to a mass difference of 68 Da between Clone C and the originator, peptide mapping was carried out by RPxRP-MS. Prior to analysis by 2D-LC-MS, samples were purified by off-line ProA, reduced and digested. To obtain a high orthogonality, a high mobile phase pH was used in the first dimension, while for the second dimension a low pH was used. The mass difference between clone C and its originator was assigned to a mutated variant of peptide NQFSLR (NQI/LSLR). Upon comparison of the 2D-UV contour plots and the MS data of both clone C and B to the originator, differences were observed. The difference in charge state between clone C and the originator was due to a higher abundant lysine truncated C-terminal peptide, whereas for clone B the more basic charge profile was due to a higher abundant lysine containing C-terminal peptide (SLSLPGK).37 However, the necessary offline steps before RPxRPMS analysis
-Figure 9: Purification and separation of the mAb variants in the first dimension (ProA) resulting in a pre-peak and main peak (A) and further separation of the pre-peak and main peak in the second dimension (SEC) (B). The analysis time in the first dimension was 6.5 minutes and the analysis time in the second dimension for one fraction was 10 minutes. Reproduced from ref. 47.
purification, fractionation, reduction (T=1 h) and digestion (T=16 h) – significantly increase the analysis time of the total workflow.
Figure 10: Heart-cutting ProA-SEC of supernatant of trastuzumab producing CHO clones. ProA chromatogram at UV 280 nm of supernatant of trastuzumab producing CHO clones (A) and SEC chromatogram at UV 280 nm of supernatant of trastuzumab producing CHO clones (B). Reproduced from ref. 37.
Figure 11: Heart-cutting ProA-CEX of supernatants of tocilizumab and its CHO clones supernatants. CEX chromatogram at UV 280 nm of supernatant of tocilizumab producing CHO clones. Reproduced from ref. 37.
The previously described applications used CEX and SEC in combination with UV, which does not allow characterization of the separated species. To enable direct characterization of the separated species, the separation modes should be directly coupled to MS. However, online coupling of these separation modes in a 1D-LC set-up and MS is hampered due to the use of non-volatile buffers. Therefore, Alvarez et al. (2011) implemented a 2D-LC set-up to allow direct coupling between IEX and SEC to MS.4 As can be observed in figure 12A, the first dimension consisted of a CEX or SEC column while the second dimension consisted of multiple RP trap cartridges (T1 to T6). Multiple fractions from the 1D effluent were trapped in the RP cartridges and sequentially desalted prior detection with ESI-MS. Indeed, the RP cartridges in this set-up could be interpreted as SPAM rather than a second separation mode. In order to achieve minimal sample handling, an on-line disulfide reduction was also implemented by flushing the RP cartridges with 15 mM DTT.
SEC-RP-MS was used for the intact analysis of mAb3. Upon MS identification, the main peak was assigned to the intact mAb and LMW compounds. For the HMW peak, duplicate injections were needed to pre-concentrate the fraction prior MS analysis. The HMW peak was assigned to an antibody fragment with a loss of 93 amino acids from the N-terminal end of the LCs. It was assumed that due to a noncovalent complex between the antibody fragment and free LC species in the sample, the antibody fragment was eluting in the HMW region. CEX-RP-MS was used for the intact analysis of mAb1 and reduced analysis of mAb2. This method allowed separation of the different charge variants (mAb1, Fig. 12B). The peaks were assigned to the mAb consisting of neutral glycans (main peak), glycans exhibiting two to three sialic acids (first acidic peak) and both neutral glycans and glycans
exhibiting one sialic acid (second acidic peak). For the base peak, masses could not be immediately assigned. This might be due to the presence of LMW modifications or it might be in the tailing part of the main peak. The analysis of reduced mAb2 allowed localization of large PTMs on the heavy and light chains.4 While the 2D-LC method was able to overcome the incompatibility issues between SEC and CEX, low abundant and/or poorly resolved peaks were observed and only large modifications were identified. In addition, pre-concentration on the RP traps was needed using duplicate injections, which is not time-efficient and consumes more sample. Also, multiple RP cartridges were used for trapping and desalting which makes the method set-up rather complex.
The potential of implementing online reduction in multiple-heart-cutting CEX-RP-MS was also demonstrated by Bathke et al. (2018).48 Rather than using multiple RP cartridges in the second dimension, fractions were sequentially loaded onto one RP column for trapping and desalting. Online reduction was performed by flushing the column with 0.02M DDT. Next, the reduced fractions were further separated by an ACN gradient prior ESI-MS analysis. The on-column reduction allowed reduction of all the intra- and interchain disulfide bridges of an antibody, which was comprised by two different heavy chains (77 and 51 kDa) and two different light chains (22 kDa and 2x 24 kDa). The suitability of the 2D-LC-MS method for the characterization of mAbs was further demonstrated by the characterization of low abundant acidic and base peaks of various mAbs. This method allowed identification of isomerization of aspartic acid to succinimide in the heavy chain (M-17 Da, basic variant) and glycation (M+162 Da, acidic variant). This method was fully automated and enabled rapid characterization of unknown peaks (analysis time was 3h). Although this setup allowed separation and identification of the major charge variants, it was not able to detect charge differences caused by LMW modifications, such as deamidation (M+1 Da).48 Therefore, this method was further developed by Gstöttner et al. (2018), who implemented an automated multiple-heart-cutting 4D-LC/MS method (Fig. 13).42 In the first dimension (CEX), mAb charge variants were separated using a salt gradient. Peaks of interest were fractionated and stored in one of the decks. Directly after cutting, the first fraction was loaded onto the second dimension (RP) where the fraction was trapped and reduced by 0.2M DTT. Next, the reduced mAb chains were transferred to the third dimension (trypsin column) and digested, using a digestion reagent for 1 minute. The generated peptides were transferred to the fourth dimension (RP-C18) where they were trapped, desalted and subsequently separated by an ACN gradient prior to analysis by ESI-MS. This process was repeated for every fraction. In a single run it was possible to fractionate nine peaks.
Figure 12: Schematic set-up of the online 2D-LC configuration (A) and the separation of the charge variants of mAb1 using WCX (B). Reproduced from ref. 4.
The applicability of the 4D-LC/MS method was demonstrated by comparing the method to a traditional offline workflow, reported by Haberger et al. (2014).49 The aim was to detect and localize all the previously reported modifications on the same stressed and non-stressed mAb. Compared to the traditional offline process, the characterization of the peaks observed in the WCX chromatogram was much faster (9 versus 52 hours). In addition, the entire workflow for an unknown peak could be performed in less than 1.5 hours. Due to the online digestion step, identification of charge differences caused by LMW modifications, such as deamidation, were possible. In addition, this method was developed on a commercially available system and extended with standard modules. As described by the authors, the 4D-LC/MS method allows implementation of commercial quality control separation methods without any need for further adaption, which is a great time saving in terms of method transfer. However, some further improvements regarding trapping must be made in order to ensure retention of small hydrophilic peptides. In the current study, not all peptides were retained on the fourth dimension (breakthrough) and therefore not detected during peptide mapping.42
Figure 13: Schematic representation of the automated 4D-LC/MS set-up that allowed separation, reduction, digestion and detection. Comprehensive 2D-LC. First the mAbs charge variants are separated by CEX (first dimension) and subsequently fractionated and stored in Deck A or B. Fractions are sequentially loaded onto the RP column (second dimension) where the samples are desalted and reduced. Next, upon switching valve 1, the fraction is transferred to the digest column (third dimension). Afterwards, valve 2 is switched to allow trapping, desalting and separation of the generated peptides prior ESI-TOF-MS. Reproduced from ref. 42.
4.2. COMPREHENSIVE 2D-LC
Stoll et al. (2015) developed an online selective comprehensive 2D-LC-MS method for the analysis of intact and (partially) digested mAbs.41 To obtain a high resolving power and allow hyphenation with ESI-MS, CEX was used in the first dimension and RP was used in the second dimension. In a single analysis, 11 fractions of 1D effluent in the chromatogram were collected in a continuous way, stored and reinjected into the RP column. While two completely different separation mechanisms were used in the first and second dimension, for the intact analysis of rituximab, the resolving power of the RP separation was insufficient. This might be due to the large size of the mAb and/or due to the high separation speed and short column in the second dimension. For the analysis of digested rituximab, the second dimension became more relevant which is due to the smaller fragments (~25 kDa) that were analyzed.41 The same set-up was used by Sorenson et al. (2015) for the comparison of digested cetuximab, trastuzumab and infliximab to their biosimilars (namely, cetuximab-B, trastuzmab-B and
Valve 1
infliximab-B).50 The CEXxRP-MS allowed the same 2D-LC-MS conditions (except for the 1D CEX gradient) for all mAbs, which is a great advantage regarding throughput. In CEX, mAb variants containing sialic acids had a shorter retention time while mAb variants containing lysines had a longer retention time. Upon comparing the 2D-UV contour plots of the originators to the biosimilars, differences in peak profiles and intensities were observed (Fig. 14). Upon MS identification, the differences in the 2D-UV plots were assigned to differences in glycosylation patterns (cetuximab/cetuximab-B), mass differences caused by one amino acid (change of a lysine residue to arginine B)) and the presence of deamidation (trastuzumab/trastuzumab-B, infliximab/infliximab-B).50
Compared to 1D-LC, comprehensive 2D-LC allows a higher separation power. This makes comprehensive 2D-LC very suitable for peptide mapping since a higher sequence coverage can be achieved. Vanhounacker et al. (2015) proposed comprehensive 2D-LC as a novel tool for peptide mapping, since it can tremendously increase the resolving power by using two orthogonal methods.51 Three methodologies, CEXxRP-MS, RPxRP-MS and HILICxRP-MS were developed for the analysis of different batches of trastuzumab (stressed and non-stressed) was analyzed. CEXxRP and RPxRP were compatible, because of the similar mobile phase compositions used in the first and second dimension. The coupling between HILIC and RPLC was more challenging, due to the high acetonitrile content (~90%) of the HILIC mobile phases, causing breakthrough in the RP separation. Therefore, several adjustments were made – such as reduction of the flow rate and insertion of a flow splitter before the modulator in order to decrease the amount of acetonitrile in the 1D effluent. Improvements were made, but breakthrough could not be eliminated. As can be observed in Figure 15, RPLCxRPLC gave the highest resolving power in terms of orthogonality (highest surface coverage), while HILICxRPLC gave the lowest resolving power. Also, in the HILICxRPLC separation, a faint horizontal band is observed in the dead time of the 2D separation, which is due to rapid changes in the mobile phase composition. The high resolving power in RPxRP is mainly due to differences in mobile phase pH between the first and second dimension (high versus low pH) and the zwitterionic nature of the peptides.
SCXxRPLC-MS was used for the comparison of non-stressed and oxidatively stressed trastuzumab. This method allowed determination of oxidation on peptide T41 (T41ox). The oxidation product (T41ox), which was not fully separated from the original product (T41) in the first dimension, was fully resolved in the second dimension, which points out the feasibility of comprehensive 2D-LC. RPxRP-MS was used for the analysis of the originator to its biosimilar (Fig. 16). Upon peptide mapping, differences in product-related impurities (deamidation and C-terminal lysines) were observed between the originator and biosimilar. The deamidation product of peptide T3 was slightly higher in Figure 14: Comparison of 2D-LC-UV separation after IdeS digestion and DTT reduction of cetuximab and cetuximab-B. Reproduced from ref. 50.
the biosimilar (13 versus 8%, respectively) and the C-terminal lysine (T62+K) was only removed for 87% in the biosimilar, while in the originator the lysine was almost eliminated. It was shown that CEXxRPLC and RPLCxRPLC were both needed to assess different modifications. Once the originator is completely characterized by MS, 2D-UV contour plots can be used for fast comparison with biosimilars. One limitation of this method was the high dilution factor due to the second dimension. This made it difficult to detect low levels of impurities.51
Figure 15: Comparison of 2D-LC plots of SXCxRPLC-UV, RPLCxRPLC-UV and HILICxRPLC-UV peptide maps of trastuzumab. Reproduced from ref. 51.
Figure 16: Comparison of trastuzumab to its biosimilar using RPxRP-MS. Upon peptide mapping, deamidation products of peptide T3 were identified in both the originator (8%) and biosimilar (13%). In addition, the C-terminal lysine (T62+K) was almost eliminated from the originator and only removed for 87% from the biosimilar. Reproduced from ref. 51.
As is described by Vanhounacker et al. (2015), coupling between HILIC and RPLC can be challenging due to the high acetonitrile content used in HILIC.51 However, Stoll et al. (2018) showed that ASM can overcome the challenges that are observed between HILICxRPLC.52 Figure 17 shows the influence of ASM on the 2D separation. As can be observed in Figure 17B, breakthrough is observed at the dead time of the 2D column when ASM is not used, regardless of the injection volume. Because a large amount of protein breaks through in the dead volume of the 2D column, less detection sensitivity is observed for the Fc/2 and Fd/subunits (Fig 17C-G). The highest detection sensitivity is observed when ASM is used between the two dimensions, since it eliminates breakthrough. This clearly shows that ASM can decrease the ACN content of the 1D effluent fractions prior to injection into the 2D column, allowing sensitive detection of separated protein variants. For two N-glycosylated mAbs, HILICxRPLC was able to selectivity separate glycoforms on both the Fc/2 and Fd subunits. In less than 2 hours HILICxRP-MS was able to separate and characterize various glycoforms based on their glycan size (HILIC) and hydrophobicity (RPLC).52
Figure 17: (A) 1D HILIC separation of the sub-units (Fc/2, LC and Fd) of an IdeS-digested and reduced aglycosylated mAb (atezolizumab) (A);
(B-C) Overlay of the 2D-UV chromatograms of the Fc/2, Fd sub-units from the 14-19 min of the 1D; (D-G) and 2D-UV contour plots using
different injection volumes and ASM. Reproduced from ref. 52.
While hearcutting is mainly used to allow coupling between non-volatile chromatographic techniques to MS, Ehkirch et al. (2018) used comprehensive 2D-LC.43 In this study they used a 4-dimensional tool (SECxSEC-IMxMS) for the characterization of mAb size variants. The applicability of this method was demonstrated by the analysis of a variety of mAbs using stressed and non-stressed conditions. First, the use of non-volatile (phosphate) and volatile buffers (ammonium acetate) in SEC was investigated. During this experiment it was shown that the use of volatile buffers leads to a poor separation (peak broadening, longer elution times) and under estimation of HMW- and LMW- compounds. Therefore, the SEC column in the first dimension consisted of a non-volatile salt separation to allow proper quantitation of HMW and LMW compounds. The SEC cartridge in the second dimension was used as a fast (1 minute) on-line desalting step, thereby making the method compatible with online IMxMS analysis. The applicability of the method was demonstrated by the analysis of three mAbs (adalimumab, pembrolizumab and bevacizumab). For adalimumab, 0.8% and 0.4% HMW- and LMW-compounds in 1D SEC were separated (Fig. 18A). A stressed sample led to an increase of HMW- and LMW- compounds (3.2% and 6.7%, respectively) (Fig. 18B). Native-MS allowed identification of the different peaks (Figure 18C). As expected, the main peak (Peak II) was assigned to the monomer (148,249 Da). Peak I was assigned to a dimeric aggregate (296 704 Da, respectively) and peak III was identified as Fc-Fab fragment (100,876 Da) and peak IV to a Fab fragment (44,248 Da). Upon IM analysis, the conformational characterization of the different size variants was determined through collisional cross section (TWCCSN2) (Figure 18D). The monomer had a TWCCSN2 value of 79.6 nm2, which agrees with literature 43.
Figure 18: SECxSEC-IMxMS of adalimumab. (A) Comparison between SEC using non-volatile (phosphate buffer) and volatile mobile phase (ammonium acetate); (B) 1D SEC chromatogram of stressed and non-stressed adalimumab; (C) Deconvoluted mass spectra of the 1D-SEC
peaks obtained from native MS; (D) Ion mobility arrival time distributions (ATDs) and charge states of the dimer (39+), monomer (26+), Fc/Fab (22+) and Fab (13+) and their TWCCS
N2 obtained from native IM-MS. Reproduced from Ref. 43.
The feasibility of the SECxSEC-IMxMS method was further shown by comparison to SEC-UV using forced degraded pembrolizumab. In the SEC-UV of stressed pembrolizumab (Fig. 19), two peaks were observed before the main peak, indicating HMW compounds. However, with native MS, these peaks were assigned to two oxidized monomeric species (149,092 Da and 149,076 Da, respectively) instead of multimers. IM-MS measured similar CCS values for the 26+ charge state of the main peak (peak III)
and peak II, indicating similar conformations, while for peak I a significantly different conformation was observed. This could be explained by the fact that peak I was the most oxidized monomer. Upon comparison to SEC-MALS/RI and cryo-electron microscopy, similar values were observed, which shows that SECxSEC-IMxMS agrees to the current reference methods.
The proposed SECxSEC-IMXMS methodology is a very powerful tool for structural and conformation characterization as well as quantitation of mAbs size variants under native conditions. 43
Figure 19: 1D SEC profile of pembrolizumab.
5. 2D-LC-UV/MS FOR THE ANALYSIS OF ADCS
Compared to mAbs, ADCs are increasing the structural complexity due to the linker-payload conjugation that is attached to the mAb.12 The heterogeneity is also significantly increased, since both free drug species and different DARs can be present in the biopharmaceutical product. Therefore, it is very important to determine the glycoprofile, unconjugated mAbs, free drug species, average DAR and potential DAR isomers. This requires many offline processes, separation techniques as well as detection techniques which make characterization very time-consuming. By implementing 2D-LC, the amount of offline processes can be tremendously decreased. Also, the resolving power can be significantly increased, allowing more in-depth characterization.12 Depending on the analytical question, heart-cutting or comprehensive 2D-LC can be used. This paragraph will discuss heart-cutting and comprehensive 2D-LC applications separately. A detailed overview of the mobile phases, columns and detection settings used for each application can be found in Appendix II.
5.1. (MULTIPLE) HEART-CUTTING 2D-LC
The analysis of unconjugated small molecule drugs29, linker-drugs and/or impurities present in the drug formulation is very important.53 This is important since free drugs and related impurities can decrease the safety and efficiency of the biopharmaceutical product.53,54,55 Therefore, Goyon et al. (2018) developed an online multiple-heart-cutting SEC-RP-UV method for the analysis of small drug impurities.56 The applicability of the method was tested by the analysis of lysine-conjugated ADCs from various purification stages. In SEC, both isocratic and gradient conditions were used. Isocratic conditions were used for the separation of HMW compounds, whereas a gradient was used in order to obtain separation between LMW compounds (<1.5 kDa). Peaks of interest in the LMW region were fractionated and loaded onto a RP-C18 column for further separation of the impurities. The repeatability of the SEC-RPLC was below an RSD of 4% and the recovery was 107%.56 Li et al. (2015) also implemented an SEC-RPLC-UV method for the analysis of unconjugated small drugs and impurities.53 The applicability of the method was demonstrated by the analysis of a mixture containing ADC, free drugs, linker-drugs and a N-acetylcysteine (NAC) adduct. NAC is commonly added to drug formulations during the process for removal of reactive linker drugs. Due to incomplete removal by purification steps, NAC can still be in the final product and, therefore, must be monitored. As is shown in Figure 20A, 1D SEC separated unconjugated small molecule impurities from the ADC and antibody dimers. The peak containing the unconjugated small molecule impurities was fractionated and further separated byRP-C18 into the free drug, linker drug and NAC adduct (Fig. 20B). The method showed a good precision (RSD <2%), low limit to quantitation (5 ppm) and very good recovery (95 – 105%).53 In 2016, Li et al. used their SEC-RP-UV method to investigate the recovery issues that are commonly observed when RP is validated for the analysis of drug impurities. Validation is mainly done by spike recovery experiments of ADC samples with spiked free drugs and drug-linkers. During their study, the recovery of the free drug sample was monitored over time using 2D-LC. A decrease in recovery was observed and was caused by the reaction between the free linker drug and residual reactive groups on ADC, leading to lower free drugs. This was also confirmed by HIC-UV, where an increase in DAR was observed over time.55
Figure 20: The analysis of unconjugated small molecules in ADC samples. (a) 1D SEC chromatogram in which the small unconjugated small
molecules are separated from the ADC and HMW compounds and (b) the 2D RPLC chromatogram of the separated unconjugated small
molecules. Reproduced from ref. 55.
Another application that can be used for the analysis of free drug species in ADC samples was proposed by Birdsall et al. (2015), where they used SPE-RPLC-UV.54 They demonstrated the suitability of this method by the analysis of an antibody-fluorophore conjugate (AFC), mimicking physicochemical characteristics of a cysteine-conjugated ADC (Brentuximab Vedotin). The AFC sample was spiked with non-toxic molecules - mal-linker DSEA and NAC-linker-DSEA - possessing the same structural features as drug-linker components. In the first dimension, the sample mixture was extracted under acidic conditions using an SPE mixed mode column (SAX and C18). The AFC was efficiently removed from the column, due to electrostatic repulsion with the SPE stationary phase, whereas the free hydrophobic drug species were adsorbed onto the column. The hydrophobic drug species were eluted by a high concentration of ACN followed by fractionation. To prevent breakthrough, at-column dilution was incorporated prior to separation by RP-C18. Within a 10-minute gradient the NAC-linker-DSEA and mal-linker-DSEA were separated based on their hydrophobic characteristics. The method precision for both linkers was below an RSD of 3% and the concentration of the free-drug component (using MS-based calibration) were 7.19 ng/mL and 3.82 ng/mL, respectively, for, the NAC-linker-DSEA and mal-linker-DSEA.54
HIC (for cysteine-conjugated ADCs) and CEX (for lysine-conjugated ADCs) are very suitable techniques for the determination of average DAR, since these techniques separate based on drug loading. In order to ensure an accurate average DAR, MS is needed to confirm the drug load variants in the separated peaks. However, due to the non-volatile buffers used in HIC and CEX, coupling with MS for direct characterization is hampered. Therefore, Birdsall et al. (2015) proposed an heart-cutting HIC-RP-MS method for the analysis of a cysteine-conjugated ADC.28 The applicability of this method was shown by the identification of low-, medium, and high loaded cysteine conjugated ADC batches. As is shown in Figure 21, it is expected that the retention in HIC increases with increasing number of DAR. Peak A to H were fractionated and loaded onto the 2D RP-C4 column for desalting and further separation using denaturing conditions. Subsequently, heavy and light chains of the ADC were
Figure 21: 1D HIC chromatograms of the
low-, medium- and high cysteine conjugated ADC. Reproduced from ref. 28.
analyzed by MS. As can be observed in the deconvoluted mass spectrum of peak H of the high loaded ADC (Fig. 22), two masses were found that were assigned to the light-chain with one drug (23,580 Da) and the heavy chain (G0F/G1F glycoforms) with three drugs (51,630 Da). This means that peak H consists of ADC species with DAR 8. MS allowed identification of all the peaks observed in the HIC-UV chromatograms and determination of the average DAR values of the different batches. The average DAR value for the low-, medium-, and high- loaded ADC batches were 2.83, 4.44 and 5.97, respectively. 28
Figure 22: The analysis of the cysteine-conjugated ADC using HICxRPLC-MS. (A) 1D HIC chromatogram of the ADC, (B) 2D RPLC
chromatogram of the ADC and (C) the deconvoluted mass spectrum of the peaks separated in the 2D RPLC. Reproduced from ref. 28. Instead of using only RP in the second dimension to allow coupling between HIC and MS, Gilroy et al. (2017) implemented both HIC-SEC and HIC-RP under reduced and non-reduced conditions, for characterization of an cysteine-conjugated ADC.27 HIC-SEC-MS was used for the separation and identification of different DAR species. The peaks observed in the HIC chromatogram (Fig. 23A) were fractionated and loaded onto the SEC column for desalting prior to detection with ESI-MS. The conditions used in this method kept the ADC in native state, which allows direct DAR conformation of each HIC peak using intact masses. However, challenges may arise when low abundant peaks are present in the chromatogram, since native conditions lead to a decrease in ionization efficiency of the mobile phase. The peak with the shortest retention time in the chromatogram was assigned to an unconjugated ADC (DAR 0) and the peak with the longest retention time was assigned to the highest possible conjugated ADC (DAR 8). In addition, ADCs with DAR 4 were observed in two peaks, indicating the presence of potential isomers. HIC-RP-MS (non-reduced) was used for further identification of the two HIC peaks (DAR4A and DAR4B) that were assigned to DAR4. RP was used under denaturing conditions, allowing the analysis of subunits. The DAR4A peak was assigned to an ADC with conjugation at the light and heavy chain in the Fab region (Fig. 23B), while DAR4B was assigned to an ADC with four drugs conjugated in the hinge of the Fc region (Fig, 23C). HIC-RP-MS (reduced) was used for further characterization of the unresolved regions between DAR2 and DAR4A (Fig. 23A). In this set-up, the RP column was used for both reduction and separation. It was assumed that the unresolved region consisted of intact DAR4b species, but differed in retention time due to a different glycan.27
While these applications 27,28 are able to analyze the HIC profile, they had some limitations. One of the limitations was that these applications were used in single heart-cut mode, meaning that multiple injections were needed to fully characterize the HIC chromatograms. This leads to more sample consumption and an increase in analysis time.
