Hydrogen deuterium exchange mass spectrometry identifies the dominant paratope in CD20 antigen binding to the NCD1.2 monoclonal antibody

Age-related diseases in humans include enhanced susceptibility to virus infections, arthritis, metabolic diseases, cognitive dysfunction, and cancer development. These pathologies are controlled, in part, by the genetic background, exposure to environmental factors, age-related epigenetic, proteomic, and genetic changes in the cell, and by the integrity of the host immune response [1]. The paucity of physiological models that could drive an understanding of these complex multi-factorial pathways in vivo make it quite challenging to improve the treatment of disease. ‘One World Health’ proposes the unification of medical and veterinary sciences to develop cross-species research into spontaneous and infectious disease pathogenesis [2,3]. Acquiring a deeper understanding of the in vivo function and regulation of naturally occurring spontaneous disease, for example, the changes in immunity in age-related diseases like cancer, will be important to improve animal health and welfare.

Immunotherapeutics have emerged as a compelling treatment choice for some human cancers such as those that express CD20, CTLA-4, or the immune checkpoint axis PD1/PDL1 [4–6]. Such agents attempt to exploit, mimic, or stimulate the natural immune defenses. Additional monoclonal antibodies that are used to build on our physiological knowledge of cell signalling, target receptors including VEGF, EGFR, ERBB2, CD40, CD33, and CD52 [7]. However, unexpected physiological problems can emerge in the use of antibody therapeutics. The antibody Yervoy was reported to induce a cytokine storm in healthy humans [8]. In addition, resistance to immunotherapies can emerge [9]. This highlights the difficulty in predicting monoclonal antibody responses in human patients. One major limitation in developing monoclonal antibody therapeutics is that the host immune system in ageing and/or ill human populations have very few ‘physiological’ preclinical models that can be used to predict antibody efficacy. Current robust rodent models such as xenografts in immune-deficient mice do not successfully predict clinical efficacy in complex immune-competent diseases. Improved physiological and age-correlated disease models would be of high value to stimulate more innovative immunotherapeutics of benefit to a larger population of patients.

Breeds of the domestic dog Canis lupus suffer from many spontaneous and age-related diseases such as cancer, arthritis, and viralopathies. This offers an opportunity to develop the spontaneous canine disease as a preclinical model that better represents human disease states [10]. Spontaneous tumours in dogs share important clinical, pathological, immunologic, molecular, diagnostic, and therapeutic characteristics with corresponding human disease and may be treated with similar anti-cancer modalities as in humans. Spontaneous tumours in dogs can provide an improved and more relevant model for developing innovative cancer therapeutic concepts that are closer to man than rodent models [11,12]. For example, EGFR is emerging as a target for the development of imaging modalities that can be experimentally examined in spontaneous canine cancers [13,14]. In addition, the genetic features of spontaneous canine glioma and its similarities to paediatric and adult human glioma [15] which provides a foundation for using canine as a physiological model for the human counterpart. Evidence for how canine cancer models can drive experimental therapeutics for potential use in human cancers was shown by data highlighting that doxorubicin can be delivered to spontaneous canine brain tumuors using synthetic minicells [16]. Such innovative approaches not only aim to improve the health of this animal, but it also provides a roadmap for developing novel preclinical models for testing new modalities for use in human populations.

One of the first paradigms for the success of monoclonal antibody therapeutics in human cancer patients has been formed by an antibody that targets the CD20 receptor [17]. CD20 is a four-pass transmembrane receptor expressed in B-cells and lymphoma [17,18]. In humans, CD20 is targeted successfully with the human–mouse chimeric monoclonal antibody, Rituximab [19,20]. Additional evidence indicates that Rituximab can be used in the treatment of other diseases including arthritis [21], immune-related indications [22–24], and in transplantation [25]. There are additional antibodies used that target CD20 in humans including the ‘biosimilars’ [26] Ibritumomab and Tositumomab [27,28]. Although the Rituximab and chemotherapy regime (R-CHOP) has had a positive impact on patient management, there are views towards ever-improving anti-CD20 modalities mainly because up to one-third of patients suffer a degree of mild to severe toxicity using Rituximab [29]. Reducing Rituximab toxicity using novel combined treatments or improving immunological responses with this monoclonal antibody is a key aim of translational research.

Canine lymphomas have been shown to have a similar molecular signature to the human equivalent [30,31]. This provides an opportunity to develop a canine ‘Rituximab mimetic’ and use it to develop innovative approaches to reduce antibody toxicity using combined agents. There is currently no clinically approved anti-CD20 monoclonal antibody used for the treatment of canine lymphoma and Rituximab does not bind to the canine CD20 epitope [32]. There have been anti-CD20 mouse monoclonal antibodies developed that can target canine CD20 [33] but clinical trials have not been forthcoming. We had previously reported the development of a novel anti-CD20 murine monoclonal antibody derived from a murine hybridoma that targets canine CD20 protein [32]. The monoclonal can bind CD20 in immunohistochemical platforms and a bioactive scFV can be cloned suggesting the possibility for its development as a recombinant antibody. However, the hybridoma from which the antibody was derived produces two light chain mRNA [32], questioning its utility as a recombinant biologic. In this report, we determine whether both light chains assemble into the authentic full-length IgG produced by this anti-CD20 hybridoma; whether recombinant mouse–dog chimeric full-length IgG versions of the antibody are bioactive; and define the dominant CDR contact to drive future structure–function studies. Together, these approaches will guide this antibody's future utility as a recombinant canine-ized monoclonal antibody for developing novel combined therapeutics for the treatment of canine and human diseases.

The hybridoma NCD1.2 [32] was grown in DMEM with 10% foetal bovine serum. The NCD1.2 monoclonal antibody was purified using Protein A. RNA was extracted using Trizol from the NCD1.2 hybridoma cells and reverse transcription (Qiagen) was used to obtain cDNA. HEK293 (human embryonic kidney cells, GEO accession no. GSE119929) were obtained from ATCC and grown in DMEM plus 10% foetal bovine serum. CLBL1 [34] (dog B-cell lymphoma, GEO accession no. GSE81110) grown in IMDM plus 20% foetal bovine serum and SUDHL4 (human B-cell lymphoma, GEO accession no. GSE132365) grown in RPMI and 10% foetal bovine serum. Synthetic peptides were from Mimotopes (Australia). 5(6)-FAM was from Anaspec (AS-81002).

Flow cytometry (FCM) expression of CD20 was evaluated using FACSVerse (BD-Biosciences, Belgium). ELISA was performed using biotinylated CD20 peptide (as described [32]). Briefly, 96 well Costar plates (white, high protein binding) were coated with 100 μl of streptavidin (1 μg/ml in H₂O) overnight at 37°C. The wells were blocked with PBS containing tween-20 (0.1%) and 3% BSA (PBSTB), and the CD20 peptide was incubated at 1 μg/ml in PBSTB for 1 h to capture the peptide. The recombinant NCD1.2 monoclonal antibody or the NCD1.2 mouse IgG was titrated into the wells. Detection was performed using anti-canine IgG coupled to peroxidase for detecting the recombinant antibody containing the canine Fc (A6792, Sigma) or anti-mouse IgG (DAKO, P260) coupled to peroxidase for detecting the mouse IgG NCD1.2. The reaction was developed using ECL and read using a PerkinElmer Fluoroskan.

Rapigest SF powder (1 mg, Waters) was suspended in 50 mM Ammonium Bicarbonate to give a 0.2% (w/v) solution. It was used for dilution of the anti-CD20 canine monoclonal antibody (NCD1.2) in a final ratio of 1 : 1 (v/v, Rapigest : NCD1.2) to give 0.1% Rapigest solution and 1 mg/ml for the final concentration of NCD1.2. The sample was heated at 90°C for 5 min and then cooled to room temperature and shortly centrifuged. The reduction in the di-sulfide bonds of the antibody was performed using DTT (final concentration 5 mM) by incubating at 60°C for 30 min. After cooling, the cysteines were alkylated using iodoacetamide (final concentration 15 mM) in the dark for 30 min. DTT was added again in the same concentration to reduce over-alkylation and the sample was incubated at room temperature for 15 min. Proteins were enzymatically digested overnight at 37°C using three enzymes in order to acquire the optimal peptide coverage for mapping; (1) trypsin (sequencing grade modified trypsin, Promega) in a 1 : 100 ratio (w/w) of trypsin : protein; (2) trypsin with LysC (sequencing grade, Roche) in a final ratio of 1 : 10 (w/w) LysC : protein; and (3) Pepsin (Promega) in a final ratio of 1 : 100 (w/w) pepsin : protein. The next day, the sample was centrifuged and formic acid (FA) was alternatively added to acidify the sample solution to pH 3–4 with subsequent incubation at 37°C for 45 min when Rapigest was hydrolyzed to prevent suppression of peptide ionization. The peptides were purified using Micro SpinColumns C-18 (Harvard Apparatus, U.S.A.) prior to mass spectrometry analysis.

Peptides from in-solution digestion was injected using a robotic system based on the HTS-XT platform (CTC Analytics, Zwingen, Switzerland) were trapped and desalted on-line on a peptide microtrap (Michrom Bioresources, Auburn, CA) for 3 min at flow rate 20 µl/min and eluted onto an analytical column (Jupiter C18, 1.0 × 50 mm, 5 µm, 300 Å, Phenomenex, Torrance, CA) and separated using a linear gradient elution of 10% B in 2 min, followed by 31 min isocratic elution at 40% B. Solvents were: A — 0.1% FA in water, B — 80% ACN/0.08% FA. The trap cartridge and the analytical column were kept at 1°C and flow rates were directed by nano- and loading-pumps from LC-system (UltiMate 3000 RSLCnano, Thermo Scientific, Dionex, Massachusetts, U.S.A.). The mass spectrometric analysis was carried out using an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific, Massachusetts, U.S.A.) with ESI ionization on-line connected with the instrument was operated in a data-dependent mode for peptide mapping (LC–MS/MS). Each MS scan was followed by MS/MS scans of the top three most intensive ions from both CID and HCD fragmentation spectra. Tandem mass spectra were searched using SequestHT against the cRap protein database (ftp://ftp.thegpm.org/fasta/cRAP) containing the sequence of the anti-CD20 antibody with the following search settings: mass tolerance for precursor ions of 10 ppm, mass tolerance for fragment ions of 0.6 Da, no enzyme specificity, two maximum missed cleavage sites and no-fixed or variable modifications were applied. The false discovery rate (FDR) at peptide identification level was set to 1%. Sequence coverage was analyzed with Proteome Discoverer software version 1.4 (Thermo Fisher Scientific, Massachusetts, U.S.A.).

Samples were separated on Eksigent Ekspert nanoLC 400 system (SCIEX, Dublin, California) liquid chromatograph. The output from the separation column was coupled to a nano-electrospray source TripleTOF 5600+ (SCIEX, Toronto, Canada). Nitrogen was used as a drying and nebulizing gas; temperature and flow of drying gas were set to 150°C and 12 psi. the capillary voltage was 2.65 kV. Peptides were trapped on a C18 PepMap100 pre-column (Thermo Fisher Scientific, U.S.A.) with a particle size 5 µm, 100 Å, 300 µm ×  5 mm. A 10 min washing with a solvent consisting of 2% ACN and 0.05% TFA (v/v) was followed by peptide elution into an analytical column. The analytical separation was performed on a PicoTip emitter 75 μm × 210 mm (New Objective, Woburn, U.S.A.) self-packed with ProntoSIL 120-3-C18 AQ sorbent (3 μm particles, Bischoff, Leonberg, Germany). For chromatography, the mobile phase A was composed of 0.1% FA in water (v/v), and the mobile phase B was composed of 0.1% FA in ACN (v/v). The analytical column was equilibrated first for 30 min in the buffer containing 5% mobile phase B. Peptide separation was performed using a gradient elution of peptides from 5% mobile phase B and linearly increased up to 40% mobile phase B over 120 min. A TripleTOF 5600+ mass spectrometer was operated in information-dependent mode (IDA). The TOF-MS spectrum (250 ms accumulation time) was measured in each cycle followed by collision-induced dissociation (CID) fragmentation of 15 of the most intense precursor ions and their MS/MS spectra measurement. Minimum precursor ion intensity was 100 cps, accumulation time was 75 ms and cycle time of IDA method was 1425 ms. Shotgun data were searched using ProteinPilot™ version 5.0.1 (SCIEX, Canada) against the custom-made reference protein database with the sequence of the anti-CD20 antibody. ProteinPilot™ 5.0.1 settings were as follows: digestion criteria were set with the agent as trypsin and alkylation with iodoacetamide. Samples were searched both with and without additional ID focuses of variants: Empirical Swiss-Prot, SNP-derived, Zhang Misincorporations, and amino acid substitutions. A FDR was produced with an error rate of 0.05 (5%).

NCD1.2 IgG (1 μM final concentration) was diluted 10-fold with phosphate buffer in H₂O (75 mM Na₂HPO_4, 25 mM NaH₂PO_4, 50 mM KCl, pH 7.4) and subsequently 2-fold diluted with quench buffer (0.5 M TCEP-HCl, 4 M urea, 1 M glycine and 3 μl pepsin with concentration 1 mg/ml) that acidified the samples to pH 2.3. These samples were used for the preparation of the undeuterated control sample and for the peptide mapping. Deuterated samples used were either (i) the anti-CD20 antibody free or (ii) in complex with CD20 peptide (the biotin-SGSG-DPANPSEKNSLSIQYCGSIR-amide) incubated with 1 mM DTT overnight at 37°C before the interaction. There is one cysteine residue in the CD20 peptide used in this report (biotin-SGSG-DPANPSEKNSLSIQYCGSIR) so DTT was included to ensure the peptide epitope was not perturbed via the formation of intermolecular di-sulfide bridges). The antibody : peptide complex was mixed at a molar ratio of 1 : 10 and then preincubated in H₂O phosphate buffer for 60 min at room temperature. Exchange of deuterium was initiated by 10-fold dilution with phosphate buffer prepared in D₂O (pD 7.0) carried out at room temperature. All chemicals used for making this phosphate buffer (Na₂HPO_4, NaH₂PO₄, and KCl) were directly added to the D₂O. Deuteration was quenched at 10 s, 1 min, 10 min, and 24 h (fully deuterated control sample) by the addition quench buffer (0.5 M TCEP-HCl, 4 M urea, 1 M glycine and 3 μl pepsin with concentration 1 mg/ml) at room temperature, followed with 3 min incubation to allow pepsin to partially proteolyze the antibody, and then the sample was snap frozen in liquid nitrogen. All samples were prepared in triplicates. Blank injections were performed between each injection to confirm the absence of peptide carryover from previous runs. Peptic peptides were identified in separate LC–MS/MS analysis. Replicates and fully deuterated samples were prepared according to the recommendations introduced recently for HDX methodology [35].

Each sample was thawed and injected onto an immobilized nepenthesin-1 column (AffiPro s.r.o. CZ), 15 μl bed volume, flow rate 20 μl/min, 2% acetonitrile/0.05% TFA. Peptides were trapped and desalted on-line on a peptide microtrap (Michrom Bioresources, Auburn, CA) for 2 min at flow rate 20 μl/min. Next, the peptides were eluted onto an analytical column (Jupiter C18, 1.0 × 50 mm, 5 μm, 300 Å, PheRomenex, CA) and separated using a 2 min linear gradient elution of 10–40% B buffer in A buffer, followed by 31 min isocratic elution at 40% B buffer. Solvents used were: A buffer — 0.1% FA in water, B buffer — 80% acetonitrile/0.08% FA. The immobilized nepenthesin-1 column, trap cartridge and the analytical column were kept at 1°C using a refrigerated system (Science Instruments and Software, s.r.o., CZ). The mass spectrometric analysis was carried out using an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific, Massachusetts, U.S.A.) with ESI ionization on-line connected with a robotic system based on the HTS-XT platform (CTC Analytics, Zwingen, Switzerland). The instrument was operated in a data-dependent mode for peptide mapping (HPLC–MS/MS). Each MS scan was followed by MS/MS scans of the top three most intensive ions from both CID and HCD fragmentation spectra. Analysis of deuterated samples was done in HPLC–MS mode with ion detection in the orbital ion trap.

Tandem mass spectra were searched using SequestHT against the cRap protein database (ftp://ftp.thegpm.org/fasta/cRAP) containing the sequence of the anti-CD20 antibody with the following search settings: mass tolerance for precursor ions of 10 ppm, mass tolerance for fragment ions of 0.6 Da, no enzyme specificity, two maximum missed cleavage sites and no-fixed or variable modifications were applied. The FDR at peptide identification level was set to 1%. Sequence coverage was analyzed with Proteome Discoverer software version 1.4 (Thermo Fisher Scientific, Massachusetts, U.S.A.) and graphically visualized with MS Tools application [36], 95% of anti-CD20 with the light chain-7 antibody sequence was covered with 111 identified unique peptides. Analysis of deuterated samples was done in HPLC–MS mode with ion detection in the orbital ion trap. The MS raw files together with the list of peptides (peptide pool) identified with high confidence characterized by requested parameters (amino acid sequence of each peptide, its retention time, XCorr, and ion charge) were processed using HDExaminer version 2.2 (Sierra Analytics, Modesto, CA). The software used to analyze protein and peptide behaviour, create the uptake plots that measured peptide deuteration over time with calculated confidence level (high, medium confidence are accepted, low confidence is rejected). The average standard deviation (repeatability) was 0.067 (Supplementary Table S1). The results from the peptide pool were displayed as graph showing the evolution of deuteration at individual parts of the protein at the same time and different protein states plotted using the GraphPad Prism version 5.03 for Windows (GraphPad Software, San Diego, CA, U.S.A.). Additional web-based application named ‘MS Tools’ [36] was used for the creation of a graphical representation of H/D protection plots. The results from HDX analysis of the standard protein, myoglobin, are summarized in Supplementary Table S2. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [37] partner repository with the dataset identifier PXD018510; Username: [email protected], Password: eOgipxVu (mapping the antibody peptide fragments) and PXD018550; Username: [email protected], Password: IiYmwPfe (HDX-MS data). This data includes files for HDX, IDA-TripleTOF 5600+, and DDA analyses-Orbitrap Elite and is listed in Supplementary Table S3.

We had previously used degenerate PCR primers derived from the murine variable framework regions to clone and create a M13 scFV-phage library derived from NCD1.2 murine hybridoma cell line [32]. The scFV-phage library derived from this hybridoma was used to biopan against the canine CD20 peptide. Two different scFV light chain amino acid sequences were acquired that bind to the CD20 epitope suggesting that the variable domains can function when fused as a single chain. However, the utility of this recombinant NCD1.2 monoclonal antibody as a potential therapeutic for canine lymphoma was unclear because two different light chains were produced by the hybridoma cell. As such, we did not know whether the authentic high-affinity IgG was composed of a mixture of two light chains and/or what the proportion of each light chain might be represented in the authentic IgG. If the authentic full-length IgG was composed of two light chains, this would complicate its utility as a recombinant therapeutic monoclonal antibody in the future. In this study, we aimed to (i) define the dominant IgG light chain polypeptides produced in this NCD1.2 hybridoma cell line using mass spectrometry; (ii) use gene synthesis to validate the bioactivity of any recombinant chimeric versions of the monoclonal antibody; and (iii) then map the CDR regions using hydrogen deuterium exchange mass spectrometry implicated in antibody binding to its CD20 epitope to define its paratope.

Mass spectrometry can be used to de novo sequence monoclonal antibodies but this is not trivial [38]; for example, Rituximab subunits were analyzed recently using the state-of-the-art top and middle down mass spectrometry [39]. To use mass spectrometry to identify the primary amino acid sequence of the authentic NCD1.2 monoclonal antibody, we first created a naïve reference database of both light chains. The reference protein library was derived from the original mRNA from the NCD1.2 hybridoma cell line. A common C-terminal reverse PCR primer was used to capture the common framework-4 region of the light chains (Figure 1A; (i)) and N-terminal forward PCR primers were used that capture the N-terminal framework-1 region of the light chain-3 or light chain-7 (Figure 1A; (ii) and (iii)). This amplified mRNA was sequenced to create a reference polypeptide database (Figure 1B; the underlined sequences represent the site encoded by the PCR primers). Prior to analysis using mass spectrometry, we confirmed that the authentic NCD1.2 IgG purified from the hybridoma cell can bind to canine CD20 expressed on canine lymphoma cells (Figure 1C). As reported previously [30], the NCD1.2 IgG does not bind to human CD20 (Figure 1C). This bioactive IgG was subjected to proteolysis using a range of enzymes and then the proteolytic peptides were analyzed using mass spectrometry to define the actual contribution of light chain-3 and/or light chain-7 polypeptides to the authentic IgG makeup (Figure 2).

The homology of light chain-3 and light chain-7 is depicted in Figure 3A. The IgG was subjected to pepsinization, processed using LC–MS/MS, and spectral data were processed using Proteome Discoverer 1.4. This workflow led to nearly complete protein sequence coverage of light chain-7 (Figure 3B). The use of trypsin and trypsin with LysC gave less than 40% coverage (data not shown). We did not detect any unique peptide coverage from sequences in light chain-3 (Figure 3C,D). The sequences that diverge between light chain-3 and light chain-7 reside in FR1, CDR1, FR2, and the sequence QMSN in CDR2 of light chain-7 vs DTSK in the CDR2 of light chain-3. The remaining protein sequence is identical in FR3 and CDR3 of both light chains (Figure 3A). The only missing sequence after mass spectrometric analysis in light chain-7 was Y86 in FR3 (Figure 3B–D). These data suggest that NCD1.2 IgG is composed predominately of, within our limits of detection, light chain-7.

To highlight the spectral data of light chain-7, the MS/MS spectrum of a peptide from the CDR1 of light chain-7 is shown; the HRNGFTYL, a double-charged peptide (z = 2) was acquired on both mass spectrometers, an Orbitrap Elite (Figure 4A) and a TripleTOF 5600+ (Figure 4B). The MS/MS spectrum of a peptide sequence from the heavy chain; the FMSWVRQPPGKSLEWLGL, a double-charged peptide with a (M + H)⁺ 2131.1158 Da and a retention time 20.09 min, was acquired on an Orbitrap Elite (Figure 5A), and the WVRQPPGKS, a double-charged peptide (z = 2) with a (M + H)⁺ 1053.5737 Da and a retention time = 52.3 min was acquired on a TripleTOF5600+ (Figure 5B). The resulting peaks in the MS/MS spectrum (Figures 4 and 5) correspond to a series of b and y ions (introduced in the table, right up in the spectrum) that resulted from CID fragmentation. Data of these peaks clearly covered the amino acid sequence of the peptide.

Although we observed no peptic coverage of light chain-3, this negative result does not allow us to conclude there is no light chain-3 peptide present in the IgG. As such we employed a different mass spectrometer to search for peptide fragments of light chain-3. The NCD1.2 IgG (1 μg/μl concentration) was trypsinized or pepsinized, alkylated and reduced, followed by processing using a 120 min LC gradient composed from 0.1% FA in water (v/v)/0.1% FA in acetonitrile (v/v) and analyzed on a Triple TOF 5600+ mass spectrometer. The instrument settings included fragmentation of top 15 precursor ions and measurement of their TOF-MS/MS spectra. Minimum precursor ion intensity was 100 cps, accumulation time per precursor was 75 ms and cycle time of IDA method was 1425 ms. Using this methodology, we were able to map tryptic peptides (Figure 6A,C,D) or peptic peptides (Figure 6B,E,F,G) to the variable domains, with differing degrees of confidence.

The entire heavy chain amino acid sequence was identified with a high degree of confidence with ProteinPilot using pepsin, except for the QM sequence in FR3 (Figure 6G, in red). The entire light chain-7 sequence was also identified with a high confidence using pepsin, except for the LRI sequence in FR3 (Figure 6F). However, by comparison, the light chain-3 FR1, CDR1 FR2, and CDR2 were not identified with a high degree of confidence (Figure 6E). Using trypsin (Figure 6C,D), the mapping was not as complete as that with pepsin (Figure 6E,F), although there were more high confidence peptides mapped onto light chain-7 than that for light chain-3. Together, these data suggest that light chain-7 is the predominating light chain in the NCD1.2 IgG.

Next, gene synthesis of the entire IgG was used as an orthogonal method to determine which light chain was bioactive when assembled as an IgG. Chimeric recombinant plasmids encoding of the light chain-3, light chain-7, and heavy chain IgG were synthesized to see if we could reconstitute activity when the heavy and light chain plasmids are expressed by co-transfection from plasmid vectors and in a non-hybridoma cell type (HEK393). The sequence of the light chain-7 and the heavy chain plasmid is shown in Figure 7A–F.

The IL2 leader sequence is from primate; the murine variable light chain was fused to the canine λ constant chain to create a murine–canine chimeric light chain; and the variable heavy chain was fused to the canine-Fc-receptor-positive canine ‘B-chain’ [40] to create a murine–canine chimera that would in principle be active in complement-mediated and Fcγ receptors (FcγRs)-binding mediated cell death. This feature of the Fc is based on the mouse–human chimeric antibody that targets human CD20 used in the human R-CHOP therapies (Rituximab) which induces indirect cell death via complement-mediated cell death and immunological attack from FcR-expressing innate effectors [41]. However, complement-dependent cytotoxicity (CDC) was originally thought to be the main mechanism of anti-tumour activity of Rituximab, which binds to C1q. Rituximab causes many diverse biological responses through its Fc domain, resulting in different immune response depending on the binding partner [42,43]. Cell surface transmembrane dimeric Fcγ receptors are expressed on multiple immune cells. Natural killer cells (NK) stimulated through FcγRIIIA, attack antibody opsonized targets, releasing cytolytic compounds such as granzyme B and perforin in a process called antibody-dependent cell-mediated cytotoxicity (ADCC). Additionally, binding of FcγR on phagocytic cells like macrophages, neutrophils, and monocytes leads to phagocytosis and destruction of opsonized targets, a process known as antibody-dependent cell-mediated phagocytosis (ADCP). Another mechanism stimulated through the Fc portion of Rituximab were described as ‘hypercrosslinking’ and include direct killing related to extracellular calcium influx and generation of reactive oxygen species. For these reasons, the canine Fc B-chain was used to capture some of the immunological benefits thought to be involved in the mouse–human chimeric antibody Rituximab effects.

The plasmid structure of light chain-7 and heavy chain chimeras are shown in Figure 7G,H. The heavy and light chains were co-transfected into HEK293 cells for 1 week, and the supernatant was measured for activity in a CD20 ELISA with anti-canine Fc HRP-conjugated secondary antibody. A titration of the light chain-7 and heavy chain reconstituted system produced detectable activity in binding to the CD20 peptide (Figure 7I). The inclusion of DTT in the ELISA ensured reduction in the single cysteine in the CD20 peptide (the biotin-SGSG-DPANPSEKNSLSIQYCGSIR) prevented inter-molecular cross-linking of the peptide (Figure 7I); there was no difference in binding to the CD20 peptide without DTT. The light chain-3 + heavy chain reconstituted assay did not result in detectable activity under these same conditions (data not shown). These data together suggest that light chain-7 forms the predominant light chain in the IgG preparation and that this chain is also the bioactive component of the hybridoma derived IgG. A parallel titration of the original NCD1.2 mouse monoclonal antibody and the recombinant canine–mouse chimera reveal a similar efficiency of binding to the CD20 antigen (Supplementary Figure S1). We next mapped the CDR contacts in the IgG using hydrogen deuterium exchange mass spectrometry, using the light chain-7 sequence and heavy chain sequence as reference databases.

Given that the light chain-7 dominates in the IgG preparation (Figure 3) and that light chain-7 can reconstitute a biochemically active antibody when co-expressed with the heavy chain as a recombinant protein in HEK293 cells (Figure 7), we used this sequence as a reference database to map the CDR contacts on the antibody when it interacts with peptide epitope. NCD1.2 IgG (1 μM final concentration) was diluted 10-fold with phosphate buffer (pH 7.4) in H₂O and subsequently 2-fold diluted with quench buffer then used in measurements as a non-deuterated control for the peptic peptide mapping. The sample was processed by acidification and proteolytic digestion with pepsin and nepenthesin-1 as described in the experimental procedures. The mock-quenched samples were measured using LC–MS/MS to determine the pepsin coverage under conditions used for hydrogen deuterium exchange mass spectrometry (Figure 8). Blue boxes highlight the antibody-derived peptides detected as a function of the amino acid sequence of the variable domains containing light chain-7 (sequence coverage was 91% with 109 unique peptides). The primary structure of light and heavy chains is also depicted together with FR (framework) and CDR (Complementarity-determining regions) regions. The light chain-7 polypeptide sequence was well represented with the digestion of both enzymes pepsin and nepenthesin-1 using the reference database (Figure 8), further affirming that the single chain-7 peptide sequence dominates the IgG preparation.

The CDR regions from the NCD1.2 antibody interacting with its’ epitope peptide was next mapped using hydrogen deuterium exchange mass spectrometry. Samples of either the ligand-free NCD1.2 antibody or the NCD1.2 antibody bound to its CD20 peptide epitope (biotin-SGSG-DPANPSEKNSLSIQYCGSIR-amide) were pre-incubated in molar ratio 1 : 10 (antibody : peptide) for 60 min in H₂O phosphate buffer at room temperature. At these concentrations used, fluorescence polarization using a 5(6)-FAM labelled CD20 peptide as a tracer was used to calculate that the antibody occupancy was 98.77% (Supplementary Figure S2). Deuterium exchange was initiated by a 10-fold dilution with phosphate buffer in D₂O (pD 7.0) carried out at room temperature. Deuteration was quenched at 10 s, 1 min, and 10 min by adding the acid quench buffer at room temperature, incubating on ice for 3 min, then immediately snap-freezing in liquid nitrogen. After mass spectrometry, the data were processed using HDExaminer software to evaluate changes in deuteration of the peptides as a function of time in the absence (purple) or presence (blue) of the CD20 peptide (Figure 9). The data are plotted as % deuteration of the peptide as a function of the numbering of the amino acids 1–221 where the peptides were subjected to evaluation with HDExaminer software. The diagram highlights the positions of the CDR and FR domains from the light and heavy chain as a single polypeptide sequence of 221 amino acids for simplicity. Peptide deuteration is plotted directly above the CDR and FR domains for ease of comparison. The asterisks highlight the peptides whose deuteration was suppressed by the CD20 peptide at either of the times points. Some peptides were highly deuterated (presumably solvent exposed such as both CDR2 domains) and others less deuterated (presumably less solvent exposed such as the CDR1 domain from the light chain; Figure 9A–C). In addition, baseline deuteration generally increases in all domains over the time course from 10 to 600 s (Figure 9A–C). The results over this time course were captured in triplicate, analyzed using HDExaminer, and are summarized in Supplementary Tables S1 and S4.

One major feature in IgG deuteration rates after CD20 peptide binding was notable. At the earliest time point of deuteration (10 s), only the CDR3 (V_H) domain exhibited suppressed hydrogen deuterium exchange, suggesting that this is the dominant paratope (Figure 9A). There was no detectable binding at CDR3 (V_L) at the early or late time points (Figure 9A–C). At the 60 s time point there was additional deuterium suppression at CDR2 (V_L) (Figure 9B). By the 600 s time point, CDR3 (V_H), CDR2 (V_H), CDR2 (V_L), and CDR1 (V_L) domains exhibited decreased hydrogen deuterium exchange (Figure 9C). CDR3 (V_L), FR2 (V_L), FR3 (V_L), FR2 (V_H), and FR3 (V_H) did not exhibit suppressed hydrogen deuterium exchange, suggesting that they do not comprise peptide binding sites. In fact, most FR regions exhibited elevated hydrogen deuterium exchange after peptide binding (Figure 9A–C).

The binding of the antibody to the CD20 peptide was quantified in more detail (Figure 10A,B). Early in the exchange reaction, the major contact point for the CD20 epitope peptide can be inferred to be CDR3 (V_H; D value = 0.574, peptide YYCVRASTGTSF). In the later time points, deuterium suppression is observed at both CDR2 domains; the D value for CDR2 (V_H) increases from −0.533 and 0.137, whilst the D value for CDR2 (V_L) increases from −0.232 to 0.150 (Figure 10B). CDR3 (V_L) interactions are not detected (Figure 10). The major sites exhibiting deuterium suppression late in the reaction are: (i) CDR2 (V_H) with a D value of 0.891; (ii) CDR2 (V_L) with a D values of being 0.608; and (iii) CDR3 (V_H) with a D value of 0.852. A summary of the deuteration mapping (Figure 11A) and the homology modelled variable domain (Figure 11B) highlights the regions whose deuteration is suppressed by peptide (in dark blue, CDR3 (V_H); in light blue, CDR2 (V_H), CDR2 (V_L), and CDR1 (V_L)) and those whose deuteration is increased (in red; CDR3,V_L).

We can propose an interpretation of these data. First, CDR3 (V_H) is the likely dominant paratope. In addition, the elevated deuteration in the Framework regions at later time points, as well as suppressed deuteration at the late time points in CDR2 regions, is suggestive of allosteric effects inducing conformational changes including aggregation as well as local unfolding over the time course. For example, a prior study has shown that deuteration exchange-protected segments in the CDR domains of aggregated Bevacizumab likely represent intermolecular protein–protein interactions in ligand-induced aggregates [44].

To evaluate the quality and reproducibility of the methodology, we also prepared two alternative sets of the samples in triplicates: (i) one set at a molar ratio of 1 : 1 (antibody : peptide) incubated for 60 min at room temperature before the time course of deuteration and (ii) another set at a molar ratio 1 : 10 (antibody : peptide) without the 60 min pre-incubation of the peptide and antibody (Supplementary Figure S3). These experimental variations gave rise to data similar to that observed in Figure 9 (Supplementary Figure S3). The fact that deuterium uptake over the entire variable region of the IgG was nearly identical when there was no pre-incubation with antigen, or after a 60 min preincubation with antigen, also suggests that binding was established rapidly after mixing of the monoclonal antibody and CD20 peptide. Rapid equilibration is also consistent with what would be expected given the rapid ‘on’ rates associated with most antibody/antigen interactions limited by diffusion [45].

Canine and human B-cell lymphoma treatment undergo a chemotherapeutic regime termed CHOP; cyclophosphamide, hydroxydaunorubicin, oncovin, and prednisone. The addition of the anti-human CD20 antibody Rituximab increases disease-free survival rates in human patients [46]. This R-CHOP therapy is not available for canine lymphoma patients who only can access CHOP because Rituximab does not bind the canine CD20 epitope. In this report, we have characterized an anti-CD20 antibody named NCD1.2 that can bind canine CD20 [32]. This offers an opportunity to develop recombinant IgG suitable for use in veterinary medicine and also to develop more complex clinical trials to inform novel human treatments for lymphoma. However, an interesting feature of the NCD1.2 hybridoma is that it produces two variable light chain genes that potentially complicates the adaptation of this antibody as a therapeutic. Because hybridomas are not used as often as phage-antibody libraries, it is difficult to say how widespread this phenomenon might be. However, one study has shown that an anti-transferrin receptor antibody producing hybridoma cell produces three heavy-chains and three kappa (κ)-chain transcripts of which two light chains and one heavy chain were functional [47]. As such, it might not be unusual for a hybridoma to produce a complex IgG protein.

Because a mixed IgG chain would complicate or even preclude the development of a synthetic antibody based on NCD1.2 domain structure, the first aim of this report was to use mass spectrometry to determine the extent to which both light chains assemble into the full-length IgG. De novo protein sequencing of a protein typically can be performed using Edman degradation and/or mass spectrometry. Edman sequencing does not easily allow full sequence coverage and has more recently been used to sequence small regions of interest [48]. Mass spectrometry has emerged as a more robust tool, often in combination with Edman sequencing, with which to sequence proteins for which there is no reference dataset. A recent report highlighted the difficulty in acquiring full sequence coverage of an antibody using the Edman method [49]; this new method used protease XIII (Aspergillus saitoi) whilst electrospray voltage was applied to the digestion mixture through the emitter and Tandem MS data were captured as a function of proteolytic cleavage time.

More recently, two mass spectrometric approaches have been used to identify the sequence of a protein; bottom-up or top-down mass spectrometry. There are advantages and disadvantages to these collective methods. De novo antibody sequencing applying both a middle-down and top-down strategy used multiple ion activation stages to fragment and sequence the monoclonal antibody. Innovations in the bottom up method, have for example, applied the Candida albicans secreted aspartic protease Sap9 protease to enhance sequence coverage of peptides by producing larger peptides than trypsin, in the 3–5 kDa mass range [50]. The method described in our current study using NCD1.2 is described as ‘bottom-up’ approach that produces a peptide library that can be mapped onto a reference translated DNA sequence. Using this bottom-up mass spectrometric approach, combined with in-solution digestion of the NCD1.2 monoclonal antibody, we found that proteolysis either with trypsin or with trypsin and LysC were not successful in producing a high coverage of peptide fragments across the protein (data not shown). In contrast, the addition of Rapigest [44] to the NCD1.2 antibody, acidic conditions, and pepsin provided a 98% match of the amino acid sequence to the reference DNA sequence in one mass spectrometric run with using two fragmentation methods (CID and HCD). Using this bottom up method, we concluded that only one of the two light chains (Figure 1; Figure 3) predominates in the authentic hybridoma-derived full-length NCD1.2 IgG.

We then used hydrogen deuterium exchange mass spectrometry to identify the dominant paratope implicated in antigen binding. Hydrogen deuterium exchange mass spectrometry has been used previously to understand how a monoclonal antibody interacts with its antigen; for example, a monoclonal antibody that targets the Vaccinia virus L1 protein uses heavy chain CDR1 and CDR2 [51]. Hydrogen deuterium exchange mass spectrometry was used to identify oxidation states in CDR regions of a monoclonal antibody and infer action mechanism [52]. Using hydrogen deuterium exchange mass spectrometry on the anti-CD20 monoclonal antibody, there was one major feature observed (Figure 11A); suppression of CDR3 (V_H) deuteration in the presence of the CD20 antigen early in the exchange reaction (Figures 9A and 10A,B). In contrast, there was little interaction observed between the antigen and the light chain-7 CDR2(V_L) or CDR3 (V_L) (Figure 9A and 10A,B). These data suggest that CDR3 (V_H) forms the dominant docking site in the antigen-IgG. Future mutagenesis of the CDR3 (V_H) domain, the Fc domain, as well as ‘canine-izing’ the variable framework domains, might optimize the immunochemical properties of the recombinant NCD1.2 antibody. In addition, as complement mutants are emerging that might augment the human anti-CD20 monoclonal antibody [53], canine cancers might form a physiological preclinical testbed for translating such novel combined synthetic biologics for use in human lymphoma.

The authors declare that there are no competing interests associated with the manuscript.

This work was supported in part by the BBSRC UK (BB/J00751X/1; BBSRC IAA, TEC3706; BBSRC IAA PIII-024, BB/R012385/1 (pathfinder)), by the European Regional Development Fund — Project ENOCH (No. CZ.02.1.01/0.0/0.0/16_019/0000868), by the Ministry of Health Development of Research Organization, MH CZ — DRO (MMCI, 00209805) and by The International Centre for Cancer Vaccine Science, a project is carried out within the International Research Agendas programme of the Foundation for Polish Science co-financed by the European Union under the European Regional Development Fund. MN was supported by Czech Science Foundation project number 18-23773Y. P.M. was supported by grant AZV NV18-03-00339.

Open access for this article was enabled by the participation of University of Edinburgh in an all-inclusive Read & Publish pilot with Portland Press and the Biochemical Society under a transformative agreement with JISC.

Lenka Hernychova: Investigation, Methodology, Writing — original draft, Writing — review and editing. Lukas Uhrik: Data curation, Investigation. Petr Muller: Resources, Validation. Umesh Kalathiya: Formal analysis, Visualization. Malgorzata M. Lisowska: Resources, Methodology. Mikolaj Kocikowski: Resources, Visualization. Maciej Parys: Resources, Data curation. Jakub Faktor: Data curation, Writing — original draft. Marta Nekulova: Validation, Visualization. Chris Nortcliffe: Data curation. Pavlina Zatloukalova: Data curation, Visualization. Barbara Ruetgen: Supervision, Validation, Visualization, Project administration. Robin Fahraeus: Supervision, Project administration. Kathryn L. Ball: Supervision, Funding acquisition, Project administration. David J. Argyle: Supervision, Funding acquisition, Project administration, Writing — review and editing. Borivoj Vojtesek: Conceptualization, Supervision, Funding acquisition, Writing — original draft, Writing — review and editing. Ted Hupp: Conceptualization, Supervision, Funding acquisition, Validation, Visualization, Writing — original draft.

ACN

acetonitrile

ADCC

antibody-dependent cell-mediated cytotoxicity

ADCP

antibody-dependent cell-mediated phagocytosis

CD

cluster of differentiation

CDC

complement-dependent cytotoxicity

CDR

complementarity-determining regions

CID

collision-induced dissociation

CTLA-4

cytotoxic T-lymphocyte antigen 4

ERBB2

erb-b2 receptor tyrosine kinase 2

FA

formic acid

FAM

6-carboxyfluorescein (6-FAM)

Fc

fragment crystallizable region

FcR

Fc receptor

FDR

false discovery rate

FR

framework region

HCD

higher-energy collisional dissociation

HDX

hydrogen deuterium exchange

IDA

information-dependent mode

IgG

immunoglobulin G

LC–MS/MS

liquid chromatography–tandem mass spectrometry

MS

mass spectrometry

NK

natural killer

PD1

programmed cell death

PDL1

programmed death-ligand 1

scFV

single-chain variable fragment

TCEP

tris(2-carboxyethyl)phosphine hydrochloride

TFA

trifluoroacetic acid

VEGF

vascular endothelial growth factor

VEGFR

vascular endothelial growth factor receptor