Reverse-engineering the anti-MUC1 antibody 139H2 by mass spectrometry-based de novo sequencing

De novo sequencing by bottom-up mass spectrometry

The goal of our study was to obtain the sequence of the full-length 139H2 IgG antibody using a bottom-up proteomics approach. As a starting point, we used 139H2 IgG hybridoma supernatant and purified the antibody using protein G affinity resin. The purified IgG was digested with a panel of four proteases in parallel (trypsin, chymotrypsin, α-lytic protease, and thermolysin) to generate overlapping peptides for the LC-MS/MS analysis, using a hybrid fragmentation scheme with stepped high-energy collision dissociation (sHCD) and electron-transfer high-energy collision dissociation (EThcD) on all peptide precursors. The peptide sequences were predicted from the MS/MS spectra using PEAKS and assembled into the full-length heavy- and light-chain sequences using the in-house–developed software Stitch. This resulted in the identification of a mouse IgG1 antibody with an IGHV1-53 heavy chain paired with an IGKV8-30 light chain (the full sequence is provided in the Supplementary Information). The depth of coverage for the complementarity-determining regions (CDRs) varies from around 10–100, indicating a high sequence accuracy (see Fig S1). Examples of MS/MS spectra supporting the CDRs of both heavy chain and light chain are shown in Fig 2. Comparison with the inferred germline precursors indicates a typical moderate level of somatic hypermutation (3% in the light chain; 10% in the heavy chain), with some notable mutations in the framework regions, also directly flanking CDRH2.

Validation of the experimentally determined 139H2 sequence

The experimentally determined sequences of the 139H2 variable domains were codon-optimized for mammalian expression and subcloned into expression vectors with the mouse IgG1 heavy-chain (with an 8xHis-tag) and the kappa light-chain backbones (see Supplementary Information for the full amino acid sequences). Co-transfection of the two plasmids in HEK293 cells yielded ca. 10 mg from a 1-liter culture after HisTrap purification (see Fig S2A and B). In addition, the fragment antigen-binding (Fab) region was expressed to study the monovalent binding to MUC1. The recombinant 139H2 IgG and Fab were then compared with the hybridoma-derived 139H2 IgG in Western blot and confocal immunofluorescence microscopy.

To investigate the specificity of the recombinant 139H2 antibody for MUC1, we performed immunoblot analysis on lysates of the methotrexate-adapted human colon cancer cell line HT29-MTX, known for its high MUC1 expression, and a MUC1 knockout of the same cell line that was previously described (see Fig 3A) (5). The original hybridoma-derived 139H2 recognizes one predominant band at an estimated molecular weight of 600 kD, corresponding to full-length MUC1, and this band is absent in lysates of the MUC1-knockout cells. The recombinant 139H2 showed the same binding pattern. In confocal immunofluorescence microscopy, original hybridoma-derived 139H2 stains MUC1 at the apical surface in a confluent culture of HT29-MTX, and this signal is reduced to the background in the MUC1-knockout cell line (see Fig 3B). A similar staining is observed with the recombinant 139H2. Western blot and immunofluorescence microscopy using the monovalent Fab fragment also showed specific binding to MUC1 in the WT background but with reduced avidity compared with the full bivalent IgG molecule. These results confirm that the reverse-engineered 139H2 antibody is functional and recognizes the full-length MUC1 glycoprotein at the apical surface of intestinal epithelial cells.

Epitope mapping of 139H2

Using the reverse-engineered 139H2 product, we next characterized binding to the immunodominant epitope APDTRPAPG within the MUC1 VNTR. Binding to the synthetic peptide, including an N-terminal biotin and short peptide linker for immobilization to the SPR substrate (i.e., biotin-GGS-APDTRPAPG), was determined by SPR. Binding of the full IgG was characterized by a high- and low-affinity phase with dissociation constants of 17 × 10⁻⁹ and 43 × 10⁻⁷ M, respectively (Fig S3A and B). We interpret this biphasic binding as an avidity-enhanced bivalent mode (both Fab arms engaged with epitope, high affinity) and a monovalent mode (single Fab arm, low affinity) of binding, respectively. In line with this interpretation, binding to a recombinant monovalent 139H2 Fab yielded a dissociation constant of 45 × 10⁻⁷ M, similar to the low-affinity binding phase of the full IgG.

To better understand the molecular basis of 139H2 binding to the immunodominant epitope within the VNTR, we determined a crystal structure of the Fab fragment in complex with the synthetic APDTRPAPG peptide (without N-terminal biotin or peptide linker). Crystals diffracted to a resolution of 2.5 Å, and a structure was solved using molecular replacement with a ColabFold model of the 139H2 Fab. This also revealed clear density for the peptide epitope in contact with the CDRs of 139H2 (see Table S1 and Fig S4A and B).

The APDTRPAPG peptide binds diagonally across the cleft between the heavy and light chains, making direct contact with all CDRs, except CDRL2 (see Fig 4A and B and Table S2). Contact points between the peptide and the 139H2 Fab include hydrogen bonds with the peptide backbone at six of eight positions. Both the aspartic acid and arginine residues within the epitope make salt bridges with side chains from 139H2. Although D3 interacts with R99 within CDRL1, R5 interacts with E50 and T59 near CDRH2, in addition to a stacking interaction with Y100 in CDRL3. Previous studies on the binding specificity of 139H2 have shown that R5 of the epitope is crucial for 139H2 binding. The interactions of 139H2 with R5 via residues E50/T59 in the heavy chain directly flank CDRH2 but are formally part of the framework regions, and both positions are mutated compared with the inferred germline precursor (see Fig 2). Two additional framework mutations in the heavy chain, that is, Y35 and T97, appear indirectly involved in MUC1 binding by positioning CDRH3 through hydrogen bonds with N106 and the backbone of Y111, respectively (see Fig S5A–C). Finally, the T4 residue of the APDTRPAPG epitope is a known glycosylation site, although 139H2 binding is reported to be unaffected by the presence of a single O-linked GalNAc at this position (14, 25). The crystal structure reported here shows the T4 side chain to be pointing outward from the 139H2 paratope with no indication of potential clashes that would preclude binding of the epitope with glycosylated APDTRPAPG at the T4 position. In line with this previous report and our own structural data, we also found that 139H2 binds equally well to MUC1 reporter constructs with different types of O-linked glycans (see Fig 5A–F). The therapeutic potential of 139H2 therefore lies in targeting the characteristic overexpression of MUC1 in tumor cells, not the aberrant MUC1 glycosylation that is commonly described in MUC1-overexpressing cancers.

Comparison with previously reported structures of monoclonal anti-MUC1 antibodies targeting the VNTR reveals a striking diversity in the modes of binding (Fig 4C; a full overview of reported structures is listed in Table S3) (26 Preprint, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36). Monoclonal antibodies 14A, 16A, and 5E5 all target a different region within the VNTR. Although monoclonal antibodies SM3, SN101, and AR20.5 all bind to the same immunodominant epitope of the VNTR as 139H2, the peptide is either shifted or oriented in the opposite direction relative to the cleft between the heavy and light chains. For SN101 and AR20.5, the peptide runs across this cleft in the opposite direction compared with 139H2. In SM3, the peptide is oriented in a similar direction but shifted by ∼2 residues such that both D3 and R5 are contacting different CDRs. In contrast to 139H2, each of the monoclonals compared above binds stronger to the glycosylated epitope. In the case of AR20.5 and SN101, this specificity can be explained by direct contacts made between the glycan and CDRs of the antibody. However, for SM3 the orientation of the glycosylated T4 residue is more similar to 139H2. In SM3, the GalNAc residue makes an additional hydrogen bond with a tyrosine in CDRL1. A similar interaction is predicted for 139H2, albeit through a different group of the GalNAc residue (see Fig S6).

Purification of 139H2 from the hybridoma culture supernatant

The 139H2 in the hybridoma culture supernatant was a kind gift from John Hilkins from The Netherlands Cancer Institute (NKI). The 139H2 was purified with Protein G Sepharose 4 Fast Flow beads (Merck), washed with PBS, eluted with 0.2 mM glycine buffer, pH 2.5, neutralized with 1 M Tris–HCL, pH 8, and dialyzed against PBS with Pierce Protein Concentrators PES, 30 kD MWCO.

Bottom-up proteomics—in-solution digestion

139H2 was denatured in 2% sodium deoxycholate, 200 mM Tris–HCl, and 10 mM Tris(2-carboxyethyl)phosphine, pH 8.0, at 95°C for 10 min, followed by 30-min incubation at 37°C for reduction. The samples were then alkylated by adding iodoacetic acid to a final concentration of 40 mM and incubated in the dark at RT for 45 min. 3 μg sample was then digested by trypsin (Promega) and elastase (Sigma-Aldrich) in a 1:50 ratio (w/w) in a total volume of 100 μl of 50 mM ammonium bicarbonate at 37°C for 4 h. After digestion, sodium deoxycholate was removed by adding 2 μl of formic acid (FA) and centrifuged at 14,000g for 20 min. After centrifugation, the supernatant containing the peptides was collected for desalting on a 30-μm Oasis HLB 96-well plate (Waters). The Oasis HLB sorbent was activated with 100% acetonitrile and subsequently equilibrated with 10% formic acid in water. Next, peptides were bound to the sorbent, washed twice with 10% formic acid in water, and eluted with 100 μl of 50% acetonitrile/5% formic acid in water (vol/vol).

Bottom-up proteomics—in-gel digestion

The hybridoma 139H2 was loaded on a 4–12% Bis-Tris precast gel (Bio-Rad) in non-reducing conditions and run at 120 V in 3-MOPS buffer (Bio-Rad). Bands were visualized with Imperial Protein Stain (Thermo Fisher Scientific), and the size of the fragments was evaluated by running a protein standard ladder (Bio-Rad). The IgG bands were cut and reduced by 10 mM Tris(2-carboxyethyl)phosphine at 37°C, followed by alkylation in 40 mM IAA at RT in the dark. The gel bands were digested by chymotrypsin and thermolysin at 37°C overnight in 50 mM ammonium bicarbonate buffer. The peptides were extracted with two-step incubation at RT in 50% ACN and 0.01% TFA, and then 100% ACN, respectively.

Bottom-up proteomics—LC-MS/MS

The peptides obtained by in-solution and in-gel digestion were vacuum-dried and reconstituted in 100 μl of 2% FA. The digested peptides were separated by online reversed-phase chromatography on an Agilent 1290 Ultra-high-performance LC (UHPLC) or Dionex UltiMate 3000 (Thermo Fisher Scientific) coupled to a Thermo Fisher Scientific Orbitrap Fusion mass spectrometer. Peptides were separated using a Poroshell 120 EC-C18 2.7-μm analytical column (ZORBAX Chromatographic Packing, Agilent) and a C18 PepMap 100 trap column (5 mm × 300, 5 μm; Thermo Fisher Scientific). Samples were eluted over a 90-min gradient from 0% to 35% acetonitrile at a flow rate of 0.3 μl/min. Peptides were analyzed with a resolution setting of 60,000 in MS1. MS1 scans were obtained with a standard automatic gain control target, a maximum injection time of 50 ms, and a scan range of 350–2,000. The precursors were selected with a 3 m/z window and fragmented by sHCD and EThcD. The sHCD fragmentation included steps of 25%, 35%, and 50% normalized collision energies (NCE). EThcD fragmentation was performed with calibrated charge-dependent electron-transfer dissociation parameters and 27% NCE supplemental activation. For both fragmentation types, MS2 scans were acquired at a 30,000 resolution, a 4 × 10⁻⁵ automatic gain control target, a 250-ms maximum injection time, and a scan range of 120–3,500.

Bottom-up proteomics—peptide sequencing from MS/MS spectra

MS/MS spectra were used to determine de novo peptide sequences using PEAKS Studio X (version 10.6) (39, 40). We used a tolerance of 20 ppm and 0.02 Da for MS1 and MS2, respectively. Carboxymethylation was set as fixed modification of cysteine and variable modification of peptide N-termini and lysine. Oxidation of methionine and tryptophan and pyroglutamic acid modification of N-terminal glutamic acid and glutamine were set as additional variable modifications. The CSV file containing all the de novo–sequenced peptide was exported for further analysis.

Bottom-up proteomics—template-based assembly via Stitch

Stitch (nightly version 1.4.0+802a5ba) was used for the template-based assembly (41). The mouse antibody database from IMGT was used as a template (42). The cutoff score for the de novo–sequenced peptide was set as 90, and the cutoff score for the template matching was set as 10. All the peptides supporting the sequences were examined manually.

Cloning and expression of recombinant 139H2 IgG and Fab

To recombinantly express full-length anti-MUC1 antibodies, the proteomics sequences of both the light and heavy chains were reverse-translated and codon-optimized for expression in human cells using the Thermo Fisher Scientific web tool (https://www.thermofisher.com/order/gene-design/index.html). For the linker and Fc region of the heavy chain, the standard mouse Ig γ-1 (IGHG1) amino acid sequence (UniProt P01868.1) was used. An N-terminal secretion signal peptide derived from the human IgG light chain (MEAPAQLLFLLLLWLPDTTG) was added to the N-termini of both heavy and light chains. BamHI and NotI restriction sites were added to the 5′ and 3′ ends of the coding regions, respectively. Only for the light chain, a double stop codon was introduced at the 3′ site before the NotI restriction site. The coding regions were subcloned using BamHI and NotI restriction–ligation into a pRK5 expression vector with a C-terminal octahistidine tag between the NotI site and a double stop codon 3′ of the insert so that only the heavy chain has a C-terminal AAAHHHHHHHH sequence for nickel-affinity purification (the triple alanine resulting from the NotI site). After the sequence was validated by Sanger sequencing, the HC/LC were mixed in a 1:1 DNA ratio and expressed in HEK293 cells by the ImmunoPrecise Antibodies (Europe) B.V. company. After expression, the culture supernatant of the cells was harvested and purified using a prepacked HisTrap Excel column (Cytiva), following the standard protocols (see Fig S2).

To recombinantly express anti-MUC1 Fab, the coding regions of the HC variable region were subcloned using AgeI and NheI restriction–ligation into a pRK5 expression vector. The subcloned region contains the mouse Ig γ-1 (IGHG1) Fab constant region with a C-terminal octahistidine tag followed by a double stop codon 3′ of the insert so that only the heavy chain has a C-terminal AAAHHHHHHHH sequence for nickel-affinity purification (the triple alanine resulting from the NotI site). After the sequence was validated by Sanger sequencing, the HC/LC were mixed in a 1:1 (m/m) DNA ratio and expressed in HEK293 cells by the ImmunoPrecise Antibodies (Europe) B.V. company. After expression, the culture supernatant was loaded onto a 5-ml HisTrap Excel column (Cytiva) using a peristaltic pump. A column was reconnected to the ÄktaGo system (Cytiva) for column wash (50 mM Tris at pH = 8, 150 mM NaCl) and step elution (50 mM Tris at pH = 8, 150 mM NaCl, 300 mM imidazole). Fractions from the peak corresponding to the Fab were concentrated using Amicon Ultra-15 (Millipore) and further purified by size-exclusion chromatography using Superdex 200 Increase 10/300 GL (Cytiva) in buffer containing 50 mM Tris (pH = 8), 150 mM NaCl.

Mammalian cell lines and culture conditions

The human gastrointestinal epithelial cell lines HT29-MTX (43) and HT29-MTX ∆MUC1 (5) were cultured in 25-cm² flasks in DMEM containing 10% FCS at 37°C in 10% CO².

Western blot

HT29-MTX and HT29-MTX ∆MUC1 lysates were prepared from cells grown to full confluency for 7 d in a six-well plate. Cells were harvested by scraping and lysed with lysis buffer (10% SDS in PBS with 1× Halt Protease Inhibitor Cocktail). The concentration was measured by BCA assay, 5× Laemmli buffer was added, and the sample was boiled for 15 min at 95°C. A mucin–SDS gel was made according to reference 5; 40 μg of protein was added to each well and run in boric acid–Tris buffer (192 mM boric acid, Merck; 1 mM EDTA, Merck; and 0.1% SDS, to pH 7.6 with Tris) at 25 mA for 1.5 h. Proteins were transferred to a polyvinylidene fluoride (PVDF) membrane using wet transfer for 3 h at 90 V/4°C in transfer buffer (25 mM Tris; 192 mM glycine, Merck; and 20% methanol, Merck). Afterward, membranes were blocked with 5% BSA in TSMT (20 mM Tris; 150 mM NaCl, Merck; 1 mM CaCl₂ [Sigma-Aldrich]; 2 mM MgCl₂, Merck; adjusted to pH 7 with HCl; and 0.1% Tween-20 [Sigma-Aldrich]) overnight at 4°C. The next day, membranes were washed with TSMT and incubated with 139H2 WT, synthetic, or FAB antibodies (1:1,000) in TSMT containing 1% BSA for 1 h at RT. Membranes were washed again with TSMT and incubated with α-mouse IgG secondary antibody (A2304; Sigma-Aldrich) diluted 1:8,000 in TSMT with 1% BSA for 1 h at RT, and washed with TSMT followed by TSM. For detection of actin, cell lysates were loaded onto a 10% SDS–PAGE, transferred to PVDF membranes, and incubated with α-actin antibody (1:2,000; bs-0061R; Bioss) and α-rabbit IgG (1: 10,000; A4914; Sigma-Aldrich). Blots were developed with the Clarity Western ECL kit (Bio-Rad) and imaged in a Gel-Doc system (Bio-Rad).

Western blot of MUC1 reporter constructs

Four MUC1 reporter constructs, expressed in engineered HEK293 cells, were a kind gift from Christian Büll of the Copenhagen Center for Glycomics. Each reporter construct in 1×, PBS was boiled in 5× Laemmli buffer. 10 ng/25 ng of each construct was loaded per well on a 10% Bis–acrylamide SDS gel for the 139H2/6× His-tag blots, respectively. Samples were run in 1× Novex Tris–Glycine SDS Running Buffer (Thermo Fisher Scientific) for 1.5 h at 120 V. Proteins were transferred to a 0.2-μm Trans-Blot PVDF membrane (Bio-Rad) and transferred at 1.3 A/25 V for 7 min using the Trans-Blot Turbo system (Bio-Rad). Afterward, membranes were blocked with 5% BSA in TSMT (20 mM Tris; 150 mM NaCl, Merck; 1 mM CaCl₂, Sigma-Aldrich; 2 mM MgCl₂, Merck, adjusted to pH 7 with HCl; and 0.1% Tween-20, Sigma-Aldrich) overnight at 4°C. The next day, membranes were washed with TSMT and incubated with 139H2 WT, synthetic antibody (1:1,000) or HisProbe-HRP Conjugate (15165, 1:5,000; Thermo Fisher Scientific) in TSMT containing 1% BSA for 1 h at RT. The 6× His-tag blots were washed with TMST and TSM, developed with the Clarity Western ECL kit (Bio-Rad), and imaged in a Gel-Doc system (Bio-Rad). The 139H2 membranes were washed again with TSMT and incubated with α-mouse IgG secondary antibody (A2304; Sigma-Aldrich) diluted 1:8,000 in TSMT with 1% BSA for 1 h at RT, washed with TSMT followed by TSM, and developed.

Confocal microscopy

HT29-MTX and HT29-MTX ∆MUC1 cells were grown for 7 d to reach a confluent monolayer on coverslips (8 mm diameter #1.5) in 24-well plates. Cells were washed with Dulbecco’s Phosphate-Buffered Saline (DPBS, D8537) and fixed with 4% paraformaldehyde in PBS (Affymetrix) for 30 min at RT. Fixation was stopped by adding 50 mM NH₄Cl in PBS for 15 min. Cells were washed two times and permeabilized in binding buffer (0.1% saponin [Sigma-Aldrich] and 0.2% BSA [Sigma-Aldrich] in DPBS) for 30 min. Coverslips were incubated with 139H2 WT, synthetic FAB at 1:100 dilution for 1 h, washed 3× with binding buffer, and incubated with Alexa Fluor 488–conjugated α-mouse IgG secondary antibodies (1:200; A11029; Thermo Fisher Scientific) and DAPI at 2 μg/ml (D21490; Invitrogen) for 1 h. Coverslips were washed 3× with DPBS, desalted in Milli-Q, dried and embedded in ProLong Diamond Mounting Solution (Thermo Fisher Scientific), and allowed to harden. Images were collected on a Leica SPE-II confocal microscope with a 63× objective (NA 1.3, HCX PLANAPO oil) and controlled by Leica LAS AF software with default settings to detect DAPI, Alexa Fluor 488, Alexa Fluor 568, and Alexa Fluor 647. Axial series were collected with step sizes of 0.29 μm.

SPR

N-terminally biotinylated synthetic MUC1 peptide with the sequence biotin-GGS-APDTRPAPG was ordered from GenScript. This was dissolved in PBS and printed on a planar streptavidin-coated SPR chip (P-Strep, SSens B.V.) using a continuous flow microfluidics spotter (Wasatch), flowing for 1 h at RT, after which it was washed with SPR buffer (150 mM NaCl, and 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid [Hepes] with 0.005% Tween-20) for 15 min and quenched with biotin solution (10 mM biotin in SPR buffer). SPR experiments were performed using an IBIS-MX96 system (IBIS Technologies) with SPR buffer as the running buffer. A dilution series of 2× steps of the full recombinant 139H2 or Fab were prepared, starting from a 10.0-μM stock for full IgG and a 7.88-μM stock for the Fab, diluting with SPR buffer. 20 dilution steps (including the stock) were used for the full IgG, and 10 dilutions were used for the Fab. SPR experiments were performed as a kinetic titration without regenerating in between association/dissociation cycles, with 30-min association and 10-min dissociation time for the full IgG and 6-min association and 4-min dissociation for the Fab. Binding affinity was determined by fitting data at binding equilibrium to a two-site binding model for the full IgG and a one-site (Langmuir) binding model for the Fab, using Scrubber 2.0 (BioLogic Software) and GraphPad Prism 5 (GraphPad Software, Inc.).

Crystallization and data collection

Sitting-drop vapor diffusion crystallization trials were set up at 20°C by mixing 150 nl of complex with 150 nl of reservoir solution. The complex sample consisted of purified 139H2 Fab and MUC1 epitope peptide (APDTRPAPG; GenScript) in a 1:2.5 M ratio, at a total concentration of 3.8 mg/ml in a buffer of 50 mM trisaminomethane at pH 8.0 and 150 mM NaCl. The diffracting crystals grew in a condition of 0.2 M NaCl, 0.1 M sodium–phosphocitrate, and 20% wt/vol polyethylene glycol (PEG) 8,000 used as reservoir solution. A 3:1 mixture of reservoir solution and glycerol was added as a cryoprotectant to the crystals before plunge-freezing them in liquid nitrogen. Datasets were collected at 100 K at Diamond Light Source beamline I24, equipped with an Eiger 9M detector (Dectris), at a wavelength of 0.6199 Å.

Structure determination and refinement

Collected datasets were integrated using the xia2.multiplex pipeline (44), and the three best datasets were subsequently merged and scaled in AIMLESS to a maximum resolution of 2.5 Å. The resolution limit cutoff was determined based on the mean intensity correlation coefficient of half-datasets, CC_1/2. An initial model of 139H2 Fab was generated using ColabFold (45). The variable region and constant region were placed in subsequent PHASER (46) runs, the short linkers between the two regions were built manually, and the CDRs were adjusted in COOT (47). Clear density for the MUC1 peptide was present in the Fo-Fc map, and the peptide was built manually in COOT. The structure was refined by iterative rounds of manual model building in COOT and refinement in REFMAC5 (48). The final model was assessed using MolProbity (49). All programs were used as implemented in CCP4i2 version 1.1.0 (50).