Identification of a conformational epitope on the E antigen implicated in anti-E alloimmunization

TO THE EDITOR:

Antierythrocyte antibodies are generated by alloimmune responses, which are themselves triggered by exposure to nonself red blood cell (RBC) antigens, typically through transfusion or pregnancy. These antibodies form when the host possesses HLA alleles capable of presenting the antigen to T cells¹ together with additional processes involving antigen-presenting cells and B cells; however, not all exposed individuals develop such antibodies. Indeed, only 1.1% to 1.4% of pregnant women and 4.5% to 12.2% of patients who have received multiple transfusions develop alloantibodies,2, 3, 4, 5 suggesting genetic or immunological variability.

Previous studies proposed that HLA genotype influences antibody formation, but most research focuses on allele frequency correlations without identifying specific antigenic determinants.6, 7, 8, 9, 10 The E antigen, a component of the RhCE protein located on the RBC membrane, is the most frequently detected alloantibody specificity in Japan.² To elucidate the mechanisms that drive the development of anti-E antibodies, we aimed to identify B- and T-cell epitope regions on the E antigen involved in immune recognition. Because recent findings indicate that the 3-dimensional structure of an antigen influences immunogenicity,¹¹ we hypothesized that conformational features of the E antigen contribute critically to the induction of anti-E antibodies. Therefore, we combined patient serology with in silico epitope prediction, structural modeling, and docking simulations to identify candidate epitopes potentially responsible for both B-cell and T-cell responses to the E antigen.

Samples from 28 patients with confirmed anti-E antibodies were analyzed. Conformational epitope mapping of serum from 1 patient with a strong antibody response was performed using the PEPperMAP Conformational Epitope Mapping service (PEPperPRINT GmbH, Heidelberg, Germany), which uses overlapping cyclic peptides synthesized on peptide microarrays and fluorescence-based antibody-binding detection. HLA-DRB1 genotyping was performed for all 28 patients with anti-E antibodies using the WAKFlow HLA Typing Kit (Wakunaga, Osaka, Japan). The complete data set of genotyping results is provided in supplemental Table 1. Prediction of major histocompatibility complex class II peptide binding was conducted using NetMHCIIpan 4.0, and prediction of linear B-cell epitopes was performed using BepiPred 3.0.

A 3-dimensional structural model of the E antigen was analyzed using the experimentally determined crystal structure (Protein Data Bank [PDB] ID: 7UZQ) and visualized in PyMOL 3.10. The structure of the anti-E antibody was modeled using AbodyBuilder2,¹² based on a reported single-chain variable fragment (scFv) from phage display.¹³ Antigen-antibody interactions were analyzed via docking simulations using HADDOCK2.4.¹⁴

Serologic epitope mapping (PEPperMAP analysis) identified 4 candidate peptides, only 1 of which (residues 345-357) was located on the extracellular domain and accessible to antibodies (Figure 1A,C). The critical proline at position 226 (specific for the E antigen) was absent in all 4 peptides, including 345 to 357, suggesting that 345 to 357 does not constitute a continuous linear epitope with residue 226 due to the distance between them.

Figure 1.

Structural and sequence-based analysis of RhCE epitopes recognized by anti-E antibodies. (A) Linear schematic of the RhCE protein showing peptides identified by epitope mapping. Mapped regions along the primary amino acid sequence are highlighted in cyan. A 6-residue segment centered on proline at position 226 is highlighted in red. (B) Core peptides predicted to bind HLA-DRB1∗13:02. Residues 165 to 173 (green) and 390 to 398 (gray) are shown, along with mapped regions indicated along the primary amino acid sequence. (C) Topological model of the RhCE protein showing the extracellular location of the mapped peptide (residues 345-357, cyan) and residues 223 to 229. Among these, position 226 is shown in blue, whereas the remaining residues are red. (D) Localization of the core peptide that binds HLA-DRB1∗13:02 (residues 165-173, green) on the topological model of the RhCE protein. This peptide, located on the third extracellular loop, lies adjacent to residue 226 (blue), a known determinant of E antigen specificity. A 6-residue segment centered on position 226 is shown in red.

HLA-DRB1 genotyping revealed enrichment of HLA-DRB1∗13:02 and HLA-DRB1∗09:01 alleles compared with frequencies in the Japanese population. A total of 28 patients with confirmed anti-E antibodies were analyzed. HLA-DRB1∗13:02 showed an odds ratio of 3.01 (95% confidence interval, 1.06-8.53; P = .04) and HLA-DRB1∗09:01 an odds ratio of 2.59 (95% confidence interval, 1.23-5.44; P = .02), suggesting a genetic predisposition to anti-E antibody formation (supplemental Table 1). Although previous studies primarily examined associations between certain HLA alleles, such as HLA-DRB1∗04, ∗15, and ∗07, and antierythrocyte antibodies,6, 7, 8, 9, 10 our findings were distinct and identified enrichment of HLA-DRB1∗13:02 and ∗09:01 in individuals with anti-E antibodies. Moreover, unlike previous studies that focused solely on allele frequency correlations, this study integrated HLA genotyping with antibody formation and epitope-level structural analyses, providing a novel and comprehensive perspective on RBC alloimmunization.

T-cell epitopes were predicted using NetMHCIIpan for HLA-DRB1∗13:02. Two strong-binding peptides were identified: YHMNLRHFY (165-173) and LKIWKAPHV (390-398; Figure 1B,D). These peptides did not include or overlap with residue 226, indicating they cannot explain the specificity of anti-E antibodies by themselves.

Prediction of B-cell linear epitopes with a default threshold¹⁵ yielded no significant hits. Lowering the threshold to 0.15 identified residues 192 to 200, but this region is intracellular and therefore inaccessible to antibodies.

Docking simulations showed that the modeled anti-E scFv interacts with regions corresponding to residues 165 to 173, 223 to 229 (including residue 226), and 345 to 357 (Figure 2). There were 92 antigen and antibody residue pairs within 5 Å, with the closest pair at 1.6 Å, which supports a tight and specific conformational interaction.16, 17, 18, 19, 20

Figure 2.

Docking model of the anti-E antibody and RhCE protein. (A) Three-dimensional model of RhCE protein showing candidate epitope regions obtained from mapping and HLA-binding prediction. The panels show the front (left), side (middle), and top (right) views of the E antigen, with epitope regions 165 to 173 (cyan), 223 to 229 (green), and 345 to 357 (red). (B) Docking model of RhCE (gray) and anti-E scFv (light orange). The panels show the front (left), side (middle), and top (right) views and highlight antibody residues within 5 Å of the epitope regions in magenta, indicating close multisite antigen-antibody interactions.

Structural analysis of the crystal structure of RhCE (PDB ID: 7UZQ) revealed that, despite linear separation, residues 345 to 357 and 165 to 173 are spatially close to residue 226 in the folded protein. The distances from residue 226 to residues 346 and 166 were 16.4 Å and 6.6 Å, respectively, which were within the conformational epitope range.^21,22

Notably, the apparent discrepancy between the HADDOCK2.4 results and the experimental PEPperMAP findings may be explained by methodological and structural factors. HADDOCK defines antigen-antibody interfaces by atomic proximity (typically within 5 Å), which can yield a broader contact map than that detected experimentally, particularly when local conformational flexibility exists around residue 226 and the 223 to 229 loop. Conversely, PEPperMAP uses immobilized cyclic or linear peptides. Even when probed with polyclonal serum, it preferentially detects localized, sequence-contiguous epitopes and may underrepresent discontinuous but spatially proximate regions such as 165 to 173 and 345 to 357.

Additional docking simulations using the e form of the RhCE antigen (with Ala226) were also performed to evaluate binding specificity. Unlike the E form, the e antigen did not exhibit any stable interaction with anti-E scFv, and no consistent atomic contacts were detected within 5 Å. This lack of binding is consistent with the known serologic specificity of anti-E antibodies and can be attributed to the local structural alteration induced by the Pro→Ala substitution at residue 226. The amino acid substitution alters the local conformation, thereby disrupting the spatial complementarity required for recognition of the conformational epitope.

Collectively, these results suggest that although residue 226 determines E antigen specificity, it is likely part of a broader conformational epitope composed of noncontiguous, spatially adjacent segments. Our findings are consistent with the prediction of Howe and Stack,¹¹ who proposed that the B-cell epitope of the E antigen is conformational and that the substitution site at residue 226 is structurally buried. This agreement further supports the conclusion that the specificity of anti-E antibodies arises because of conformational, rather than linear, determinants of the E antigen. These may include the 165 to 173 T-cell peptide and the 345 to 357 antibody-mapped peptide, potentially forming a composite epitope structure that contributes to both T-cell activation and B-cell recognition.

This may help explain the lack of sequence-based epitope overlap with residue 226, and suggests how antibodies could maintain specificity through structural proximity. Importantly, these findings highlight the importance of integrating multiple bioinformatic approaches, including structural modeling, to better understand complex antigenic landscapes, particularly when linear sequence data are insufficient.

However, this study has limitations. The small sample size limits the statistical power and generalizability of the HLA association findings. Moreover, the study population was restricted to Japanese individuals; therefore, the findings may not be generalizable to other populations. Serologic B-cell epitope mapping was performed using serum from a single individual with HLA-DRB1∗13:02 and was not performed in participants carrying HLA-DRB1∗09:01, which is also associated with anti-E alloimmunization. In addition, our study relied on in silico prediction methods for both B- and T-cell epitopes, which have limited accuracy due to inherent uncertainties in computational modeling and peptide-HLA affinity estimation. Therefore, these predictions should be interpreted with caution and ideally validated by further experimental analyses.

We further note that database searches (gnomAD version 4.1.0) revealed no prevalent amino acid substitutions in residues 165 to 173 or 390 to 398, suggesting that these peptides predicted by NetMHCIIpan are unlikely to represent physiologically relevant T-cell epitopes. These observations underscore the need for future experimental validation and cross-population studies. Furthermore, the study focused exclusively on anti-E antibodies; it remains to be determined whether similar structural features are involved in other antierythrocyte antibody responses.

Despite these limitations, this integrative approach identified a structurally contiguous region that is potentially responsible for both T-cell activation and antibody recognition. These findings contribute to a broader understanding of alloimmunization mechanisms and may inform future strategies that predict patient risk and allow the development of targeted therapies in the field of transfusion medicine.

Conflict-of-interest disclosure: The authors declare no competing financial interests.