Turning universal O into rare Bombay type blood

Blood group antigens are critical determinants in transfusion and transplantation compatibility and are distributed beyond red blood cells to various tissues. The ABO system, defined by specific oligosaccharide epitopes synthesized by FUT1/FUT2 and ABO-encoded glycosyltransferases, leads to A, B, AB, or O blood groups depending on terminal sugar additions to the H antigen. The rare Bombay (Oh) phenotype arises from homozygous inactivating mutations in FUT1 and FUT2, resulting in an absence of H antigen on erythrocytes and secretions and the presence of anti-H antibodies. Bombay individuals can only safely receive H-deficient blood, making transfusion challenging due to the phenotype’s low prevalence and geographic skew. The study addresses this clinical need by identifying and characterizing an enzyme, FucOB from Akkermansia muciniphila, that specifically removes α-1,2-linked fucose from H antigens, with the aim of converting group O red cells (which express H) into Bombay-type (H-deficient) cells to enable transfusion in Bombay patients.

Prior approaches have enzymatically converted A or B antigens to O by removing terminal sugars using α-galactosidases, α-N-acetylgalactosaminidases, or endo-β-galactosidases, enabling broader donor compatibility. A metagenomically discovered A-converting pathway (a deacetylase plus a galactosaminidase) functioned effectively at low enzyme concentrations and even in whole blood and organs ex vivo. In contrast, attempts to remove the α-1,2-linked fucose from H antigens to produce Bombay-type RBCs have been limited: fungal α-1,2-fucosidases (e.g., Aspergillus niger) had acidic pH optima and incomplete activity across H chain types; other enzymes affected some H types but not others (e.g., ineffective on Type III), or required non-physiological conditions. Artificial metallopeptides were designed to remove fucose from Type II H antigen, and a membrane α-1,2-fucosidase from Elizabethkingia meningoseptica was reported to act on Type I/II/IV H antigens. Overall, efficient, specific, and practical generation of Bombay-type RBCs remained unmet, motivating discovery of a robust α-1,2-fucosidase with broad H-type specificity and favorable reaction conditions.

Enzyme discovery and production: Genomic context analysis in Akkermansia muciniphila ATCC BAA-835 identified Amuc_1120 (FucOB), a predicted GH95 near an O-glycopeptidase (OgpA). FucOB (residues 24–796; signal peptide removed) and variants were cloned into pET29a with N-terminal His tags and expressed in E. coli BL21(DE3). Purification used Ni2+-affinity, cation exchange, and size-exclusion chromatography.

Substrate specificity assays: Synthetic oligosaccharides representing fucosyl linkages (α-1,2, α-1,3, α-1,4, α-1,6) and blood group epitopes (Type I, II, V H antigens; Type V A and B antigens) were incubated with FucOB (typical enzyme:substrate 1:100,000; 30 min at 37 °C in 20 mM Tris-HCl pH 6.8). Released fucose was quantified by L-fucose dehydrogenase/NADP+ assay. Activities were compared with GH95 homologs BbAfcA (B. bifidum) and BiFuc95A (B. longum CZ0511).

Glycoprotein defucosylation: A TNFα receptor (etanercept) was glycoengineered to present 8–11 α-1,2-fucosylated core 1 O-glycans (Type III H antigen) via recombinant FUT2. FucOB (1 µg) incubations (1 h, 37 °C) were analyzed by reverse-phase LC-MS after IdeS cleavage and reduction/denaturation.

Crystallography: Apo FucOB and catalytic mutant FucOB E541A were crystallized (PEG 3500/3350 conditions). X-ray data were collected at SLS and ALBA beamlines, structures solved by molecular replacement using GH95 BbAfcA (PDB 2EAB), built/refined with PHENIX/CCP4/Coot, and validated (depositions: PDB 7ZNZ, 7ZOO).

Computational analyses: Structural homology searches (DALI), pocket analyses, and domain interfaces were performed. Ligand docking (AutoDock Vina) positioned Type H, A, and B epitopes in the active site guided by BbAfcA–2′FL complexes. Molecular dynamics simulations (AMBER20; ff14SB/GLYCAM06j; TIP3P water; PME electrostatics) of FucOB complexes ran 0.5 µs production following minimization, heating, and equilibration to assess binding stability and interactions.

Mutagenesis: Single alanine substitutions targeted residues predicted to contact H antigen (W378, H383, N385, N387, T443, S444, W453, H613, W655, H693, D699) and catalytic E541. Activities toward Type II and Type V H antigens were measured relative to wild-type.

Red blood cell conversion and testing: O-type RBCs (n=20 donors; 10 O Rh−, 10 O Rh+) were washed and incubated with FucOB at 200, 50, 5.0, 0.5, 0.05, or 0.005 µg mL−1 (30 min, 37 °C, 110 rpm). Controls included no enzyme and inactive FucOB E541A. After washing, agglutination was assessed with (i) DG Gel cards using anti-H antibodies in Bombay serum and (ii) an anti-H lectin (Ulex europaeus) assay. RBC integrity was evaluated by May-Grünwald-Giemsa smears and G6PD activity assays. Flow cytometry (FACS) quantified H antigen expression using anti-H monoclonal antibodies (clones 97-1, 86-M) and FITC-labeled anti-H lectin; RBCs were gated by CD235a.

• FucOB is a highly specific α-1,2-L-fucosidase from A. muciniphila that hydrolyzes α-1,2-linked fucose but shows no detectable activity on α-1,3, α-1,4, or α-1,6 fucosyl linkages. • FucOB efficiently defucosylates H antigens of multiple chain types: Type I, Type II, Type III (shown on a glycoengineered TNFR glycoprotein), and Type V, yielding the afucosylated (Bombay) phenotype. It poorly processes branched A or B antigens. • Compared to GH95 homologs (BbAfcA, BiFuc95A), FucOB is more selective for α-1,2 fucose, whereas the homologs also show activity toward α-1,3 fucose. • Structural biology: FucOB adopts a three-domain architecture with an (α/α)6 helical barrel GH95 catalytic domain and N/C-terminal β-sandwiches. The active site groove is formed by flexible loops (25, 28, 33, 35, 37, 39) and helix 8; catalytic residue E541 functions as the general acid in an inverting mechanism. Key residues (N385, N387, D699) activate water for nucleophilic attack. • Docking and MD indicate stable binding and proper positioning for Type H antigens, while A and B antigens exhibit steric clashes (notably with T443, S444, H383, N385) and do not maintain stable binding. • Alanine scanning identified residues critical for H-antigen processing: W378, H383, N385, N387, H613, W655, H693, D699 are essential; W453 at the +1 galactose subsite significantly contributes; T443 and S444 are largely dispensable. • RBC conversion: O-type RBCs pre-incubated with FucOB at 200, 50, or 5 µg mL−1 showed no agglutination with Bombay serum anti-H and no agglutination with anti-H lectin, indicating successful H removal. Controls (no enzyme or inactive E541A) agglutinated strongly. • Conversion generalized across 20 O-type donors (10 Rh−, 10 Rh+). • Flow cytometry on untreated O RBCs detected H positivity in 64.2% (clone 97-1), 56.7% (clone 86-M), and 95.1% (anti-H lectin); after FucOB treatment (5 µg mL−1), binding by both anti-H antibodies and lectin disappeared, consistent with complete H removal. • RBC integrity was preserved post-treatment: normal biconcave morphology on smears and positive G6PD activity, indicating maintained viability.

The study demonstrates that a single, specific exoglycosidase, FucOB, can remove the terminal α-1,2-linked fucose from H antigens across multiple chain types under near-physiological conditions, effectively converting O-type RBCs into Bombay-type cells. This addresses a longstanding clinical need for safe transfusion options for individuals with the rare Bombay phenotype, who otherwise face severe hemolytic reactions due to anti-H antibodies. Structural and mutational analyses elucidate the molecular basis for FucOB’s selectivity, explaining its lack of activity on A and B antigens due to steric clashes at the +1 subsite and validating key catalytic and binding residues conserved within GH95 enzymes. Compared with prior attempts using enzymes with suboptimal pH/activity profiles or limited H-chain coverage, FucOB operates efficiently at low concentrations (as low as 5 µg mL−1) on washed RBCs and in whole blood conditions, with preserved cell morphology and metabolic integrity. The complete loss of H staining by lectin and monoclonal antibodies following treatment confirms the robustness of conversion. These findings suggest FucOB (and potentially other GH95 members with similar specificity) could be developed as a practical biotechnological tool to generate Bombay-compatible donor units and may extend to other clinical applications where removal of α-1,2 fucose is beneficial.

This work identifies and structurally characterizes FucOB, an α-1,2-L-fucosidase from A. muciniphila, that specifically removes α-1,2-linked fucose from H antigens (Types I, II, III, V) on glycans and red blood cells. Through biochemical assays, crystallography, computational modeling, and alanine scanning, the molecular determinants of substrate recognition and catalysis were defined. Functionally, FucOB converts O-type RBCs to the H-deficient Bombay phenotype, validated by agglutination assays and flow cytometry, while preserving RBC integrity. FucOB thus represents a promising tool to produce Bombay-compatible blood for transfusion. Future studies should evaluate scalability and robustness in blood bank workflows, assess safety and efficacy in preclinical/clinical transfusion models, explore activity on additional H chain types (e.g., Type IV) and clinical matrices, and screen related GH95 enzymes to broaden and optimize enzyme toolkits for blood group engineering.

The conversion was demonstrated ex vivo; no in vivo transfusion studies were performed. Structural complexes with substrate or product could not be crystallized due to crystal packing constraints, and binding models relied on docking and MD. Expression of H antigen varies among donors and antibody reagents, complicating baseline quantification. Activity was shown for H Types I, II, III, and V; Type IV H antigen was not explicitly tested. Only O-type RBCs were converted; extension to A, B, or AB RBCs to generate Bombay-type cells was not assessed. Comprehensive evaluation of potential off-target glycan modifications on RBC surfaces and long-term storage or post-transfusion survival was not included.