Structural Basis of a High-Affinity Antibody Binding to Glycoprotein with Glycocalyx Decay

Abstract

Cancer-associated defects in mucin-type O-glycosylation expose truncated glycans on cell-surface glycoproteins and create tumor-associated glycopeptide epitopes. However, the structural principles by which antibodies recognize specific MUC1 glycopeptide sequences remain incompletely defined. Here, we report the molecular basis of MUC1 recognition by 16A, a high-affinity antibody isolated against hypoglycosylated MUC1. Co-crystal structures of 16A Fab bound to site-selectively glycosylated MUC1 peptides revealed that 16A recognizes an STAPPAHG-centered epitope within the MUC1 variable number tandem repeat region. Surface plasmon resonance analysis showed that GalNAc modification at the threonine site in this sequence increased Fab binding affinity by 30.6-fold relative to the nonglycosylated peptide. Structural and mutational analyses identified a hydrogen bond between the threonine-linked GalNAc and Trp34 in heavy-chain CDR1 as a key contributor to this affinity enhancement. Cell-based binding assays further showed that COSMC deficiency enhanced 16A recognition of MUC1-expressing cells, while molecular dynamics simulations suggested that Ser-linked Core 1 extension can create local steric incompatibility near CDRH1 Arg32. Together, these results define a peptide-register-driven recognition mode in which the MUC1 peptide backbone determines specificity, whereas a site-compatible GalNAc acts as an affinity-enhancing element at the edge of the paratope. This structural framework may guide the design of MUC1 glycopeptide-based vaccines and antibody therapeutics.

Introduction

Glycocalyx decay has been reported in cancer cells, as represented by the Tn antigen caused by deficiency of COSMC, the chaperone protein for the core-1 elongation enzyme during O-glycan synthesis. , Truncated glycans are involved in cancer cell signaling and immune suppression. Among tumor-associated glycoproteins bearing such truncated O-glycan epitopes, mucin 1 (MUC1) has emerged as one of the most prominent and clinically relevant targets for cancer immunotherapy. MUC1 is a type I transmembrane protein with a heavily glycosylated extracellular domain composed of 20–120 tandem repeats of a 20-amino-acid sequence, GSTAPPAHGVTSAPDTRPAP, with five potential O-glycosylation sites at Ser and Thr residues. , Its O-glycosylation is initiated by UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferases and further elongated by core-1 and branched core-2 enzymes. In most human epithelial cancers, MUC1 is overexpressed, loses its apical polarity, and displays aberrantly truncated O-glycans, including Tn antigen (GalNAc-O-Ser/Thr) and sialyl-Tn antigen, as a result of defective COSMC-dependent core-1 O-glycan extension.

Although aberrantly glycosylated MUC1 is an attractive target for anticancer immunotherapy, , vaccines based on unglycosylated MUC1 peptides have shown limited success, likely because they do not fully mimic the conformational and antigenic features of tumor-associated glycosylated MUC1. , Accordingly, MUC1 glycopeptide-based vaccine strategies have been explored over the past decades, and conformational analyses of MUC1-derived glycopeptides have provided key insights into antigen presentation and antibody recognition. In parallel, various antibodies have been reported to recognize the glycan moiety, the peptide backbone, or composite glycan–peptide epitopes of aberrantly glycosylated MUC1. − In particular, antibody recognition of MUC1-derived Tn glycopeptides is strongly influenced by both the peptide context and the site of glycosylation. Recent studies have further shown that structure-guided antigen engineering, dendritic-cell targeting with GalNAc glycoclusters, and backbone modification can improve the stability, antigenicity, and immunogenicity of MUC1 glycopeptide vaccine candidates. ,, However, the structural principles governing MUC1 glycopeptide immunogenicity and antibody recognition remain incompletely understood. To date, only a few crystal structures of MUC1-specific antibodies in complex with MUC1-derived peptide or Tn-glycopeptide epitopes have been reported, including SM3, AR20.5, SN-101, and 5E5. SM3, AR20.5, and SN-101 recognize epitopes containing the immunodominant PDT­(GalNAc) motif, whereas 5E5 targets a distinct Tn-MUC1 epitope in the GSTAP region. Thus, structural studies of antibodies recognizing distinct MUC1 glycopeptide epitopes are needed to better understand how site-specific O-glycosylation shapes antibody recognition.

We previously isolated two murine antibodies, 14A and 16A, from mice immunized with MUC1-expressing mutant cells lacking mucin-type core-1 β1–3 galactosyltransferase activity, using the synthetic glycopeptide RPAPGS­(Tn)­TAPPAHG as a screening antigen. Although that study established 14A and 16A as binders of hypoglycosylated MUC1-related glycopeptides, their minimal epitopes within the tandem repeat, the effects of site-specific GalNAc modification at Ser versus Thr, and the structural basis of their recognition remained unclear. Here, we define the molecular recognition principles of 14A and 16A by integrating glycopeptide array mapping, SPR analysis with site-selectively glycosylated ligands, X-ray crystallography of Fab/(glyco)­peptide complexes, mutagenesis, and molecular dynamics simulations. Our results show that both antibodies recognize a STAPPAHG-centered epitope within the MUC1 tandem repeat, distinct from the canonical PDTR motif. We further demonstrate that Thr-linked GalNAc, but not equivalently Ser-linked GalNAc, enhances binding through direct and context-dependent interactions. These findings provide an atomic-level framework for understanding how site-specific O-glycosylation modulates antibody recognition of tumor-associated MUC1.

Results

COSMC Deficiency Enhances 16A Recognition of Cell-Surface Hypoglycosylated MUC1

We first used Western blot analysis to compare the recognition of MUC1 produced in COSMC-deficient and parental HEK293T cells by 16A and 14A. Both antibodies detected MUC1 from HEK293T-COSMC^–/– cells, but not MUC1 from parental HEK293T cells (Figure A), indicating that 16A and 14A preferentially recognize hypoglycosylated MUC1 under Western blot conditions. We next examined antibody binding to MUC1 in its native cell-surface context by flow cytometry. COSMC deficiency markedly increased 16A staining of MUC1-expressing cells, whereas 14A staining was not substantially enhanced (Figure B). Thus, although both antibodies preferentially recognized COSMC-deficient MUC1 by Western blotting, only 16A clearly distinguished cell-surface MUC1 expressed in HEK293T-COSMC^–/– cells from that expressed in parental HEK293T cells. These data suggest that 14A is more compatible with COSMC-dependent O-glycan extension on cell-surface MUC1, whereas 16A preferentially recognizes exposed, hypoglycosylated MUC1 tandem-repeat glycoepitopes.

1.

Differential recognition of hypoglycosylated MUC1 by 14A and 16A. (A) Western blot analysis of MUC1 in protein lysates from MUC1-expressing parental HEK293T cells and HEK293T-COSMC^–/– cells using 14A and 16A antibodies. (B) Flow cytometric analysis of 14A and 16A binding to MUC1-expressing parental HEK293T cells and HEK293T-COSMC^–/– cells. Cells were stained with 14A or 16A mAbs (mouse IgG1, 0.2 μg/mL), followed by an APC-conjugated antimouse IgG secondary antibody.

Epitope Mapping of 14A and 16A mAbs

Because 14A and 16A were originally selected using the glycopeptide RPAPGS­(Tn)­TAPPAHG, the precise location of their epitope within the MUC1 tandem repeat remained unresolved. To define their binding specificity more systematically, we profiled both antibodies on a glycopeptide microarray comprising 73 MUC1-derived glycopeptides spanning different tandem-repeat regions and glycosylation states (Table S1). Vicia villosa agglutinin (VVA), a Tn-binding lectin, was used as a Tn-recognition control. In contrast to the broad VVA reactivity toward Tn-bearing peptides, 14A and 16A showed restricted and highly similar binding profiles, indicating that these antibodies do not recognize the Tn moiety alone but instead require a defined MUC1 peptide context (Figure ). Strong binding was observed for glycopeptides 15, 16, 32, and 35, all of which contain an intact GSTAPPAHG region. Among these, glycopeptide 35 showed the highest binding signal and contained a Tn modification at the threonine within the GSTAPPAHG motif, suggesting that Tn glycosylation at this position enhances antibody recognition. In contrast, glycopeptides lacking the complete GSTAPPAHG region showed substantially reduced binding. For example, glycopeptide 33, which contains a GST­(Tn)­AP motif in an upstream STn-glycosylated context similar to that of glycopeptide 35 but lacks the downstream PPAHG extension, showed much weaker binding. Together, these data identify the GSTAPPAHG region as the principal epitope recognized by 14A and 16A, consistent with the original selection glycopeptide RPAPGS­(Tn)­TAPPAHG.

2.

Glycopeptide microarray analysis of 14A and 16A epitope specificity. 14A and 16A mAbs were profiled on a microarray containing 73 MUC1-derived glycopeptides spanning different tandem-repeat regions and glycosylation states. Vicia villosa agglutinin (VVA), a Tn-binding lectin, was included as a control for Tn recognition. Glycopeptide sequences and glycosylation patterns are listed in Table S1.

Affinity and Specificity of Fab Fragments of 14A and 16A

Having more precisely defined the epitope, we next examined how site-specific GalNAc installation within this region affects antibody binding. In our previous study, the initial characterization of 14A and 16A was based on a limited surface plasmon resonance (SPR) analysis in which intact IgG molecules were tested against a single screening glycopeptide pair, RPAPGS­(Tn)­TAPPAHG and its nonglycosylated counterpart RPAPGSTAPPAHG. That analysis therefore did not resolve the individual contributions of glycosylation at the Ser and Thr sites. In the present study, we extended this analysis by systematically comparing ligands that differ only in glycosylation state. Because RPAPGSTAPPAHG contains two potential O-glycosylation sites, Ser and Thr, we synthesized four ligands: the nonglycosylated peptide, designated Peptide; the Ser-glycosylated peptide, designated Glyco-S; the Thr-glycosylated peptide, designated Glyco-T; and the doubly glycosylated peptide, designated Glyco-ST (Figure A). SPR was then used to quantify the binding kinetics and affinities of 14A and 16A Fab fragments toward these ligands (Figure B).

3.

Affinity and specificity of Fab fragments of 14A and 16A mAbs. (A) Chemical structure of synthesized (glyco)­peptides used for SPR analysis and crystallization. (B) SPR sensorgram overlays of (glyco)­peptide (concentrations of 1, 2.5, 5, 10, 25, 50, 100, 250, 500, 1000, 2500, and 5000 nM) binding to 16A and 14A Fab. Black lines indicate observed data points; red lines indicate fitted data.

Specifically, 14A bound to the nonglycosylated peptide with a KD of 114.6 nM, whereas 16A bound with a KD of 220.0 nM (Table ). Addition of a single GalNAc residue at the Thr site dramatically enhanced antigen binding by both 14A and 16A, decreasing the KD by 11.6-fold to 9.9 nM and by 30.6-fold to 7.2 nM, respectively, relative to the nonglycosylated peptide. Kinetic analysis showed that this affinity enhancement was driven by both faster association and slower dissociation, with the reduced dissociation rate making the larger contribution. For 14A, Thr-linked GalNAc increased ka by approximately 1.9-fold and decreased kd by approximately 6.0-fold. For 16A, Thr-linked GalNAc increased ka by approximately 2.2-fold and decreased kd by approximately 14.1-fold, indicating that stabilization of the antibody–glycopeptide complex was the dominant kinetic basis for the improved affinity.

1. SPR Kinetic and Equilibrium Binding Parameters for 14A and 16A Fab Binding to MUC1 (Glyco)­peptides.

(glyco) peptide	ka (M–1 s–1)	kd (s–1)	KD (nM)	R max (RU)
16A WT
Peptide	153494.6	0.033787	220.0	86.6
Glyco-S	150642.8	0.033116	219.8	105.2
Glyco-T	333366.1	0.002392	7.2	114.9
Glyco-ST	289028.4	0.001937	6.7	124.8
16A W34F
Peptide	86160.0	0.295400	3429.0	59.6
Glyco-S	86210.0	0.336800	3907.0	85.1
Glyco-T	73610.0	0.089950	1222.0	112.4
Glyco-ST	79700.0	0.090090	1130.0	127.6
16A W34A
Peptide	–	–	–	–
Glyco-S	–	–	–	–
Glyco-T	–	–	–	–
Glyco-ST	–	–	–	–
16A R32G
Peptide	162951.3	0.053169	326.3	79.4
Glyco-S	174472.5	0.029398	168.5	94.4
Glyco-T	359105.5	0.008231	22.9	108.1
Glyco-ST	533726.7	0.001825	3.4	104.8
14A WT
Peptide	268707.5	0.030806	114.6	68.0
Glyco-S	640975.0	0.047107	73.5	94.0
Glyco-T	521101.4	0.005157	9.9	99.3
Glyco-ST	928676.0	0.001298	1.4	98.0
14A G32R
Peptide	191198.1	0.023278	121.7	74.2
Glyco-S	179521.0	0.022909	127.6	86.6
Glyco-T	313691.6	0.001590	5.1	91.9
Glyco-ST	302506.7	0.001371	4.5	106.1

In contrast, addition of a single GalNAc residue at the Ser site only slightly enhanced 14A binding by 1.6-fold, with a KD of 73.5 nM, whereas it had almost no effect on 16A binding, with a KD of 219.8 nM. For 14A, the modest affinity improvement caused by Ser-linked GalNAc was mainly attributable to a faster association rate, as ka increased by approximately 2.4-fold; however, this favorable effect was partly offset by an approximately 1.5-fold increase in kd, indicating faster dissociation. Consistently, Ser-linked GalNAc had little effect on either ka or kd for 16A, explaining the lack of a measurable affinity gain.

The doubly Tn-modified glycopeptide bearing GalNAc residues at both the Ser and Thr sites bound 14A with the highest affinity, with a KD of 1.4 nM. This represented a 52.5-fold enhancement relative to Glyco-S and a 7.1-fold enhancement relative to Glyco-T, indicating that Thr-linked GalNAc is the major contributor to 14A binding, whereas Ser-linked GalNAc further enhances affinity in the doubly glycosylated ligand. Kinetic analysis showed that the improved binding of Glyco-ST was mainly driven by a slower dissociation rate, together with an increased association rate. Compared with Glyco-S, Glyco-ST showed a 1.4-fold increase in ka and a 36.3-fold decrease in kd. Compared with Glyco-T, Glyco-ST showed a 1.8-fold increase in ka and a 4.0-fold decrease in kd. In contrast, 16A bound Glyco-ST with a KD of 6.7 nM, nearly identical to that for Glyco-T, further confirming that GalNAc addition at Ser has little effect on 16A binding.

Crystal Structures of 16A Fab in Complex with Antigen (Glyco)­peptides

To define the structural basis by which site-specific GalNAc glycosylation enhances antibody binding, we screened cocrystallization conditions for 16A Fab with MUC1 (glyco)­peptides. The 16A/Peptide, 16A/Glyco-T, and 16A/Glyco-ST complexes crystallized in space groups P12₁1, P2₁2₁2₁, and P2₁2₁2₁, respectively, and the structures were solved and refined at resolutions of 2.10, 1.56, and 2.20 Å, respectively (Table S2). Each asymmetric unit contained one 16A Fab–(glyco)­peptide complex. The overall Fab backbone conformations were nearly identical among the three structures, as illustrated by the representative 16A/Glyco-ST complex (Figure A). Fab residues were numbered according to the Kabat system, and peptide residues were numbered sequentially from R1 to G13 based on the RPAPGSTAPPAHG sequence.

4.

Structural characterization of the 16A Fab/(glyco)­peptide complexes. (A) Overall structure of the 16A Fab/Glyco-ST complex. Glyco-ST is shown with carbon atoms colored green, and the light and heavy chains of 16A are colored light blue and light pink, respectively. The light-chain and heavy-chain CDRs involved in antigen binding are highlighted in blue and hot pink, respectively. (B) Superposition of the peptide backbones from the 16A/Peptide, 16A/Glyco-T, and 16A/Glyco-ST complexes. (C) Surface representation of the 16A Fab/Glyco-ST complex. (D) Detailed interactions between 16A Fab and Glyco-ST. Hydrogen bonds are shown as dashed lines.

In the 16A/Peptide structure, clear 2F _o–F _c electron density was observed for the TAPPAHG segment, whereas the six N-terminal residues, RPAPGS, were disordered, suggesting that they do not form stable contacts with 16A Fab. In the 16A/Glyco-T structure, unambiguous 2F _o–F _c electron density was observed for the ST­(GalNAc)­APPAHG segment and the Thr-linked GalNAc moiety. In the 16A/Glyco-ST structure, clear electron density was observed for the PGS­(GalNAc)­T­(GalNAc)­APPAHG region, including both GalNAc moieties. Superposition of the peptide backbones from the 16A/Peptide, 16A/Glyco-T, and 16A/Glyco-ST complexes showed that the bound peptide regions adopted nearly identical conformations and binding registers (Figure B). The (glyco)­peptide binds in a surface groove composed of CDRs L1, L2, L3, H1, H2, and H3 of 16A (Figure C). The MUC1 peptide epitope forms only a few polar contacts with the 16A paratope, including two light-chain-mediated hydrogen bonds, CDRL1^N36–A11 and CDRL2^R52–H12, and three heavy-chain-mediated hydrogen bonds, CDRH1 ^R32–A8, CDRH2^E51–H12, and CDRH3^E102–P9 (Figure D). In both the 16A/Glyco-T and 16A/Glyco-ST structures, an additional hydrogen bond was observed between CDRH1^R32 and S6. Although the Thr-linked GalNAc points away from the central peptide-binding groove toward the solvent, its endocyclic oxygen O5 forms a hydrogen bond with the Nε1–H group of CDRH1^W34 (Figure D). Notably, in the 16A/Glyco-ST structure, the Ser-linked GalNAc does not appear to directly participate in antibody binding.

We further measured the glycosidic torsion angles φ and ψ of the GalNAc residues in the 16A-bound glycopeptide complexes (Table S3). In the 16A/Glyco-T complex, the GalNAc–Thr linkage adopted φ/ψ values of 74.0°/117.1°, closely matching the typical Thr-linked eclipsed conformation with ψ around 120°. In the 16A/Glyco-ST complex, the Ser6-linked GalNAc showed φ/ψ values of 29.6°/161.0°, consistent with the alternate conformation characteristic of GalNAc–Ser linkages, whereas the Thr7-linked GalNAc adopted φ/ψ values of −90.6°/–116.5°, indicating an alternative bound-state rotamer. Thus, the characteristic Ser- and Thr-dependent glycosidic preferences are evident in the 16A-bound structures, although the precise GalNAc orientation can be modulated by adjacent glycosylation and the local antibody-binding environment.

Crystal Structures of 14A Fab in Complex with Antigen (Glyco)­peptides

To investigate the structural basis of the site-specific glycosylation preference of 14A and to compare its recognition mode with that of 16A, we cocrystallized 14A Fab with three MUC1 glycopeptides: Glyco-S, Glyco-T, and Glyco-ST. The 14A/Glyco-S, 14A/Glyco-T, and 14A/Glyco-ST complexes crystallized in space groups P12₁1, P2₁2₁2₁, and P2₁2₁2₁, and the structures were refined to 2.73, 3.20, and 3.50 Å, respectively (Table S2). Each asymmetric unit contained three Fab–glycopeptide complexes arranged in a head-to-tail manner (Figure A). The overall Fab backbone conformations were nearly identical among the three 14A complexes and were also highly similar to that of 16A Fab, with an RMSD of 0.21 Å between the 14A/Glyco-ST and 16A/Glyco-ST complexes. Clear 2F _o–F _c electron density was observed for the PGSTAPPAHG region in the 14A/Glyco-S structure and for the GSTAPPAHG region in the 14A/Glyco-T and 14A/Glyco-ST structures. The carbohydrate moieties of all three glycopeptide antigens also showed well-defined electron density in the 2F _o–F _c maps. Superposition of the bound glycopeptides showed that their peptide backbones adopted nearly identical conformations, which were also highly similar to those observed in the corresponding 16A complexes (Figure B).

5.

Structural characterization of the 14A Fab/(glyco)­peptide complexes. (A) Overall structure of the 14A Fab/Glyco-ST complex. Glyco-ST is shown with carbon atoms colored green, and the light and heavy chains of 14A are colored light blue and light pink, respectively. The light-chain and heavy-chain CDRs involved in antigen binding are highlighted in blue and hot pink, respectively. (B) Superposition of the peptide backbones from the 14A/Glyco-S, 14A/Glyco-T, and 14A/Glyco-ST complexes. (C) Surface representation of the 14A Fab/Glyco-ST complex. (D) Detailed interactions between 14A Fab and Glyco-ST. Hydrogen bonds are shown as dashed lines.

The glycopeptide antigen is accommodated in a surface groove formed by CDRs L1, L2, L3, H1, H2, and H3 of 14A (Figure C). Close inspection of the ligand-binding site showed that the peptide portion of the glycopeptide forms only a limited number of polar contacts with the antibody-combining site. The light chain contributes two hydrogen bonds, CDRL1^N36–A11 and CDRL2^R52–H12, whereas the heavy chain contributes three hydrogen bonds, CDRH1^G32–A8, CDRH2^E51–H12, and CDRH3^E102–P9 (Figure D). In the 14A/Glyco-S structure, the Ser-linked GalNAc does not appear to directly participate in binding, as it forms no specific polar contacts with the antibody paratope. In contrast, in both the 14A/Glyco-T and 14A/Glyco-ST structures, the endocyclic oxygen O5 of the Thr-linked GalNAc forms an intermolecular hydrogen bond with the Nε1–H group of CDRH1^W34 (Figure D). These observations provide a structural basis for the enhanced binding of 14A to Glyco-T and Glyco-ST.

We further measured the glycosidic torsion angles φ and ψ of the GalNAc residues in the 14A-bound glycopeptide complexes (Table S3). In the 14A/Glyco-S complex, the GalNAc–Ser linkage adopted φ/ψ values of 52.9°/161.3°, consistent with the alternate conformation characteristic of GalNAc–Ser linkages. The GalNAc–Thr linkage in 14A/Glyco-T showed φ/ψ values of −2.3°/–168.2°, whereas the Ser6- and Thr7-linked GalNAc residues in 14A/Glyco-ST adopted φ/ψ values of −60.2°/–77.9° and −122.7°/–70.0°, respectively. These data suggest that the 14A-bound GalNAc linkages can adopt distinct local rotamers, with their precise orientations modulated by the antibody-binding environment and adjacent glycosylation.

Mutation Analysis

Structural analyses of 14A and 16A in complex with MUC1 (glyco)­peptides identified several antibody–antigen contacts that may contribute to enhanced binding to Thr-glycosylated ligands. To further assess the roles of these residues in site-specific glycopeptide recognition, we generated a series of Fab mutants and measured their binding to the four MUC1 (glyco)­peptides by SPR (Table ).

Both structures of 14A and 16A in complex with Glyco-T suggested that the higher affinity of 14A and 16A toward the Thr-glycosylated peptide was mediated by a hydrogen bond between the endocyclic oxygen O5 of GalNAc and the Nε1–H group of the CDRH1 Trp34 indole ring. We chose 16A as the model system to study the role of CDRH1 Trp34. First, we mutated Trp34 to alanine to eliminate the hydrogen bond interaction. The W34A mutant completely lost binding to all four tested (glyco)­peptides, including the nonglycosylated peptide. This result indicates that Trp34 is not only involved in glycan-specific recognition but also plays a critical structural role in maintaining the integrity of the binding groove and stabilizing the local paratope conformation required for peptide accommodation. We next examined the W34F mutant, in which the hydrogen-bonding capacity of the indole Nε1–H group is removed, whereas a bulky aromatic side chain is retained. Although the W34F mutation reduced binding affinity toward all tested ligands, it markedly weakened the preference for Glyco-T over the nonglycosylated peptide. The KD_Peptide/KD_Glyco‑T ratio decreased from 30.6-fold in wild-type 16A to 2.8-fold in W34F. This loss of glycosylation-dependent affinity enhancement was mainly associated with a reduced difference in the dissociation rate constant (kd) between the nonglycosylated peptide and Glyco-T.

Sequence alignment (Figure S1) and structural comparison showed that heavy-chain residue 32 is the only residue within the antigen-binding site that differs between 14A and 16A, with Gly32 in 14A and Arg32 in 16A. This difference suggested that residue 32 may contribute to the distinct responses of 14A and 16A to Ser-linked GalNAc. We therefore examined the role of residue 32 by generating the reciprocal mutants 16A R32G and 14A G32R. As shown in Table , mutation of CDRH1 Arg32 to glycine in 16A increased the relative preference for Glyco-S, with a KD_Peptide/KD_Glyco‑S ratio of approximately 1.9-fold. In contrast, the 14A G32R mutant showed nearly identical KD values for the nonglycosylated peptide and Glyco-S, 121.7 and 127.6 nM, respectively, indicating that introduction of Arg32 abolished the modest Glyco-S preference observed for wild-type 14A. Together, these results demonstrate that heavy-chain residue 32 modulates recognition of Ser-linked GalNAc and contributes to the different glycosylation preferences of 14A and 16A.

Simulations Revealing Dynamic Features of MUC1 (Glyco)­peptide Recognition by 14A and 16A

Molecular dynamics (MD) simulation systems were constructed using PGSTAPPAHG, a 10-residue truncated model of the 13-residue crystallographic ligand RPAPGSTAPPAHG, and its glycosylated counterparts, including Ser6-GalNAc, Thr7-GalNAc, Ser6/Thr7-GalNAc, Ser6-Core 1, and Thr7-Core 1 glycopeptides. Across these simulated complexes, residues Thr7–Gly13 displayed relatively low conformational fluctuations, whereas residues Pro4–Ser6 near the peptide N-terminus were more mobile (Figure A). This fluctuation pattern is consistent with the crystallographic observation that the N-terminal residues are less well resolved. Hydrogen-bond analysis showed that, in most simulated complexes, the peptide backbone formed recurrent hydrogen bonds with both 14A and 16A (Table S4). The dominant interactions were observed between peptide residue Pro9 and heavy-chain Glu102, and between peptide Ala8 and heavy-chain residue 32, corresponding to Gly32 in 14A and Arg32 in 16A. These interactions showed occupancies of approximately 54–90% and 70–89%, respectively, across different simulation systems. Additional recurrent hydrogen bonds involved peptide residue Ala11 with light-chain Asn36 and His12 with light-chain Arg52, with occupancies of approximately 49–71% and 18–40%, respectively. These persistent peptide-anchoring contacts support a peptide-register-driven mode of antigen recognition.

6.

Molecular dynamics analysis of MUC1 (glyco)­peptide recognition by 14A and 16A. (A) Root-mean-square fluctuations (RMSFs) of peptide residues in the simulated 14A and 16A complexes with MUC1 peptides bearing different glycosylation patterns. (B) Structural definition of the Arg32­(Cz)–Tyr33­(Cz) distance used to monitor local conformational changes around CDRH1 Arg32 in 16A. (C) Distribution of the Arg32­(Cz)–Tyr33­(Cz) distance in different 16A/(glyco)­peptide systems. (D) Representative conformations of the 16A/MUC1-peptide complex corresponding to the two major distance populations identified in panel C.

Further MD analysis provided a dynamic explanation for the observation that 14A exhibited higher association rate constants (ka) toward MUC1 (glyco)­peptides than 16A. We inferred that the bulky and flexible side chain of Arg32 in 16A may impose greater conformational and steric constraints during antigen binding, whereas Gly32 in 14A lacks such a side chain. To quantitatively evaluate this effect, we used the distance between Arg32­(Cz) and Tyr33­(Cz) in 16A as a structural parameter (Figure B) and analyzed its distribution in different 16A/(glyco)­peptide MD systems (Figure C). In apo 16A, the Arg32­(Cz)–Tyr33­(Cz) distance showed a broad distribution from approximately 3 to 13 Å, indicating substantial local flexibility around Arg32. In the 16A/MUC1-peptide complex, however, this distance was concentrated in two major populations centered at approximately 5 and 10 Å (Figure C). These two distance populations correspond to two representative bound conformations: in one state, the peptide lies over Arg32, whereas in the other, the peptide inserts between Arg32 and Tyr33 (Figure D). Thus, productive binding to 16A requires Arg32 to adopt a compatible conformation. By contrast, 14A contains Gly32 at the corresponding position and is therefore less restricted by this local steric requirement.

The MD simulations also revealed distinct effects of Ser6 and Thr7 glycosylation on MUC1 glycopeptide binding by 14A and 16A. In both 14A and 16A complexes, the Thr7-linked GalNAc formed relatively stable hydrogen bonds with the antibody, mainly involving CDRH1 Trp34 and heavy-chain Asp55, with combined occupancies of approximately 36–65% (Table S4). These interactions are consistent with the markedly reduced dissociation rate constants (kd) observed for Thr-glycosylated ligands in the SPR experiments for both 14A and 16A. By contrast, the Ser6-linked GalNAc formed less stable hydrogen bonds, including interactions with heavy-chain Asp29 and Tyr101, with combined occupancies of only 13–30% (Table S4). These less persistent contacts likely explain why Ser6 glycosylation provides only a modest binding advantage for 14A. In contrast, Ser6 glycosylation has little or even an unfavorable effect on 16A binding. As shown in Figure C, the Arg32­(Cz)–Tyr33­(Cz) distance distribution in the 16A/MUC1-peptide_Ser6-GalNAc system retained two major populations, but the short-distance population shifted slightly toward larger values relative to the nonglycosylated 16A/MUC1-peptide system. This suggests that the Ser6-linked GalNAc introduces local steric crowding near Arg32, as reflected by partial expansion of the Arg32–Tyr33 region. In 16A, the persistent Arg32–Ala8 hydrogen bond, with an occupancy greater than 85% (Table S4), may restrict peptide repositioning within the binding groove. Thus, the Ser6-linked glycan cannot be efficiently accommodated, and the associated steric cost likely offsets the weak favorable hydrogen bonds formed by the Ser6-linked GalNAc with the antibody.

The steric constraint caused by Ser6-linked glycosylation in 16A became more pronounced when the glycan was extended to the Core 1 structure. In the 16A/MUC1-peptide_Ser6-Core 1 system, the additional galactose required more space near the entrance of the binding groove. Accordingly, the Arg32­(Cz)–Tyr33­(Cz) distance distribution shifted mainly toward values around 10 Å (Figure C), indicating that the elongated Ser6-linked glycan favors expanded conformations of the Arg32–Tyr33 region. Because the Arg32–Ala8 interaction constrains the peptide position, the peptide has limited freedom to move away and relieve this local crowding. Therefore, extension of the Ser6-linked glycan to Core 1 further increases steric incompatibility around Arg32. In contrast, 14A contains Gly32 instead of Arg32 and therefore lacks this bulky, conformationally restrictive side chain near the Ser6 glycosylation site. As a result, 14A is less affected by the steric constraints imposed by Ser6-linked glycan extension. This difference provides a structural rationale consistent with the flow cytometry observation that COSMC deficiency significantly increased 16A binding but had little effect on 14A binding.

Overall, the MD simulations support a model in which MUC1 recognition by 14A and 16A is primarily peptide-register-driven, while glycosylation functions as a site-dependent affinity modulator. The TAPPAHG segment is stably anchored by recurrent peptide-mediated contacts with conserved paratope residues. Thr7-linked GalNAc strengthens binding by forming favorable contacts with the antibody, whereas Ser6-linked GalNAc forms less persistent contacts. In 16A, the bulky Arg32 residue limits accommodation of Ser6-linked glycans, and this steric constraint is amplified upon Core 1 extension. Therefore, glycosylation enhances binding only when its position and orientation are compatible with the paratope, but can weaken binding when glycan extension creates local steric incompatibility.

Cell-Based Competition Analysis of 16A Recognition of Cell-Surface MUC1

To assess the functional relevance of the structurally defined STAPPAHG-centered glycopeptide epitope in a cellular context, we performed a cell-based competition assay using MUC1-positive B16-OVA-COSMC^–/– cells. Binding of 16A to the cells was competed with three synthetic ligands: the nonglycosylated peptide, the Ser-GalNAc glycopeptide Glyco-S, and the Thr-GalNAc glycopeptide Glyco-T. All three ligands inhibited 16A binding in a concentration-dependent manner (Figure S2), consistent with the contribution of the STAPPAHG-centered peptide epitope to cell-surface MUC1 recognition. However, the potency of inhibition differed markedly among the ligands. The Glyco-T showed the strongest inhibition, with an IC50 of 28.5 nM, whereas the Glyco-S and the nonglycosylated peptide showed weaker inhibition, with IC50 values of 0.6 and 2.1 μM, respectively. Thus, these cell-based competition data indicate that the structurally defined STAPPAHG-centered glycopeptide epitope is functionally relevant to 16A binding in a cellular context, and support the conclusion that GalNAc modification at the Thr site enhances the binding affinity of this epitope for 16A.

Discussion

Antibodies that distinguish tumor-associated glycoforms from their normal counterparts can employ diverse recognition strategies. , The first structurally characterized tumor-specific anti-Tn glycopeptide antibody, mAb 237, recognizes the GalNAc (Tn) moiety of the epitope through a deep binding pocket, while accommodating the peptide moiety in a comparatively shallow surface groove. Among MUC1-targeting antibodies, SM3 and AR20.5 recognize their epitopes primarily through the peptide backbone, with glycosylation contributing mainly by modulating glycopeptide conformation rather than through extensive direct contacts with the paratope. , SN-101 represents a more conformationally complex recognition mode, in which the GalNAc moiety contributes to binding together with a noncontiguous peptide epitope. 5E5 recognizes a GSTAP-region Tn-MUC1 epitope and displays a lectin-like binding mode in which the GalNAc moiety serves as the dominant recognition determinant. ,

The 14A and 16A antibodies described here define a distinct mode of MUC1 glycopeptide recognition. Both antibodies recognize an STAPPAHG-centered epitope through a CDR-defined peptide-binding groove, indicating that the peptide sequence establishes the primary binding register. The Thr-linked GalNAc positioned adjacent to this core peptide epitope engages the edge of the paratope through direct interactions with CDRH1 Trp34. The crystal structures reveal a hydrogen bond between the GalNAc endocyclic oxygen and the indole NH of Trp34, providing a structural basis for the pronounced affinity enhancement observed with Thr-glycosylated ligands. Molecular dynamics simulations further support this model by showing that glycosylation contributes favorable conformational and dynamical effects. Thus, 14A and 16A use a recognition principle in which the peptide defines specificity, whereas a site-proximal GalNAc functions as a positional affinity enhancer. Notably, although 14A and 16A were obtained by immunization with core-1-deficient MUC1-expressing cells and screening with the Ser-Tn-containing glycopeptide RPAPGS­(Tn)­TAPPAHG, antibody binding to this screening antigen could be driven mainly by the peptide portion rather than by the Ser-linked glycan. This explains why 14A and 16A were selected using RPAPGS­(Tn)­TAPPAHG but ultimately showed stronger affinity enhancement by Thr-linked GalNAc, which is better positioned to interact with the antibody paratope. Cell-based competition further extended this model to a cellular setting, showing that STAPPAHG-containing ligands compete with cell-surface MUC1 for 16A binding. The stronger competition by the Thr-GalNAc glycopeptide further supports the conclusion that GalNAc modification at the Thr site enhances 16A recognition of this epitope. This finding suggests that future glycopeptide-based antibody discovery may benefit from site-selectively glycosylated peptide panels to assess how site-specific glycosylation modulates affinity within a peptide-driven recognition mode.

Although the STAPPAHG-centered epitope recognized by 14A/16A partially overlaps with the GSTAP-region epitope recognized by 5E5 within the MUC1 tandem-repeat region, the underlying recognition modes are substantially different. , In 5E5, the Thr-linked GalNAc is deeply engaged by the antibody through multiple hydrogen bonds and CH−π interactions and constitutes the major epitope, whereas the peptide contributes only limited contacts, mainly involving the glycosylated Thr residue and the downstream Pro residue. By contrast, 14A and 16A recognize the STAPPAHG region through an extended peptide-binding groove, with GalNAc enhancing binding only when installed at a compatible position near the edge of the paratope. Therefore, whereas 5E5 behaves as a lectin-like antibody with GalNAc-dominant recognition, 14A and 16A use a composite but peptide-register-driven mechanism. This comparison highlights that antibodies targeting related Tn-MUC1 regions can adopt fundamentally different paratope architectures and glycan/peptide recognition strategies.

The comparison between 14A and 16A further shows how minimal sequence variation can tune glycoform selectivity. Although the two antibodies recognize the same core peptide region and adopt highly similar overall binding architectures, 16A discriminates more effectively at the cell-surface level between hypoglycosylated tumor-associated MUC1 and the more fully glycosylated form expressed by normal cells. Our structural and mutational analyses identify residue 32 in CDRH1 as a key contributor to this difference. In 16A, Arg32 is positioned near the entrance of the binding groove, where it is well placed to restrict accommodation of neighboring glycosylation states; in 14A, the corresponding Gly32 creates a less constraining local environment. This difference provides a plausible structural basis for the stronger glycoform selectivity of 16A. Additional framework mutations present in 16A but absent from 14A may further modulate Fab geometry or flexibility, as reported for other antibody systems, although this possibility remains to be tested directly.

These findings provide structural guidance for Tn-based vaccine development and anti-Tn antibody design. Previous NMR studies suggested that the STAPPAHG-containing region of MUC1 adopts an organized solution conformation shaped jointly by peptide sequence and glycosylation. , The present structures further identify this region as an antibody-accessible tumor-associated glycoepitope. In this recognition mode, the peptide backbone provides the main MUC1-specific binding determinant, whereas the adjacent Thr-linked GalNAc enhances affinity through direct interaction with the antibody paratope. These structures suggest that effective Tn-MUC1 immunogens should preserve the MUC1 peptide epitope while optimizing the position and orientation of GalNAc to enhance antibody binding. This may explain why modified Tn antigens can enhance immune responses against native tumor-associated MUC1, as suitable glycan modification may improve epitope presentation and provide additional affinity-enhancing contacts without compromising MUC1 peptide specificity. More broadly, defining GalNAc-modified peptide epitopes at site-specific resolution may also provide a conceptual basis for understanding antibody recognition of other truncated O-glycan neoepitopes on self-derived proteins, including aberrantly glycosylated IgA1 and Tn-exposing blood cell antigens. ,

The crystal structures have been deposited in the Protein Data Bank under accession codes 7 V3Q (16A), 7 V4W (16A/Peptide), 7 V64 (16A/Glyco-T), 7 V7K (16A/Glyco-ST), 7VAZ (14A/Glyco-S), 7 V8Q (14A/Glyco-T), and 7VAC (14A/Glyco-ST).

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.6c00598.

• Additional materials and methods, MUC1 glycopeptide sequences and glycosylation patterns used for microarray analysis, crystallographic data collection and refinement statistics, glycosidic torsion-angle measurements, molecular dynamics hydrogen-bond occupancy analyses, antibody VH/VL sequence alignment, and cell-based competition analysis of 16A binding to cell-surface MUC1 (PDF)

∇.

Global Health Drug Discovery Institute, Haidian, Beijing 100192, China

The authors declare no competing financial interest.