1. Introduction
B lymphoma is the most frequent type of hematolymphoid tumor. The emergence of adoptive cell therapy (ACT), particularly CD19-specific chimeric antigen receptor T (CAR-T) cells, has revolutionized the patient care paradigm [
1,
2]. These cost-ineffective cellular immunotherapies leverage naturally occurring or engineered immune killers, such as natural killer (NK) and cytotoxic T cells, to recognize and destroy B lymphoma [
3,
4]. Unfortunately, intrinsic challenges remain in their clinical application. Intensive attempts to optimize ACT functional persistence in the tumor microenvironment (TME) have led to a comprehensive understanding of effector immune responses, including their inherent cytotoxicity against target cells, antigen sensitivity, tumor infiltration capacity, hypofunctional transformation, and suppressive circuits [
5]. Converging immune cell biology and genetic/nongenetic engineering tools to improve ACT effector responses against cancers represents an emerging research field [
6].
In addition to the well-known CD19, CD22 is another critical therapeutic antigen for B lymphoma and belongs to the sialic acid (Sia)-binding immunoglobulin-like lectin (Siglec) family [
7,
8]. Both antibody and CAR-T therapies targeting CD22 are currently under intensive clinical investigation [
8,
9]. Among the treatments, TriCAR-acute lymphoblastic leukemia (ALL) T-cells, which are designed to target CD19, CD20, and CD22, bring renewed hope to patients with relapsed or refractory B-cell lymphoma [
10]. Moreover, several researchers have screened high-avidity ligands derived from Sia analogs to target CD22 [
11]. Examples include 9-
N-
m-phenoxybenzamide-
N-acetylneuraminic acid (
MPBNeu5Ac) and 9-
N-biphenyl carboxamide-Neu5Ac (
BPCNeu5Ac), both of which feature substituents at the C9 position with selective affinities for CD22 in the micromolar range [
12]. In addition, paucivalent ligands demonstrate significantly higher avidity (up to 1500-fold) than monovalent ligands do, offering cost-effective alternatives to traditional antibodies [
13]. Despite these advances, therapies using antibodies or paucivalent ligands are still associated with serious on-target/off-disease side effects [
14,
15]. Recently, Wang et al. [
16] cultured NK-92 cells with
MPBNeu5Ac, providing this off-the-shelf therapy with the ability to target CD22 after metabolic incorporation. They reported that the injection of
MPBNeu5Ac-modified NK-92 cells improved the growth control of B lymphoma. In brief, Hong et al. [
17] developed a chemoenzymatic strategy to modify NK-92MI cells with
BPCNeu5Ac, effectively targeting NK-92MI cells to treat B lymphoma. Importantly, these Sia-ligand-armed NK therapies leverage their inherent immune discrimination for non-self invaders while leaving self-cells untouched, making them a viable option despite their limited efficacy in the intravenous adoptive treatment of B lymphoma. By combining biorthogonal chemistry, glycocalyx engineering facilitates a convenient upload of various agencies to customize anticancer responses, such as targeting multiple antigens or improving TME trafficking [[
18], [
19], [
20], [
21], [
22], [
23]]. Given that glycans are structurally diverse molecules involved in many biological processes [[
24], [
25], [
26]], we believe that in-depth insight into sialylated glycan profiles could significantly broaden the applications of this technology to create cellular immunotherapies with the desired properties.
To this end, we characterized the molecular details of the NK-92MI cells that define the functionality-loading efficiency of metabolic glycocalyx engineering (MGE) of sialylated glycans and chemoenzymatic glycocalyx engineering (CeGE) mediated by β-galactoside α2,6 sialyltransferase 1 (ST6Gal1). Subsequently, the functionality-loading capacity for the MGE approach was probed with the newly developed unnatural sugar 1,6-di-
O-propionyl-
N-azidoacetylmannosamine (1,6Pr
2ManNAz), and the accessibility of the ST6Gal1-assisted CeGE was elucidated by an azide-bearing substrate (
Fig. 1(a)). The corresponding glycoproteins were subsequently enriched and identified by mass spectrometry. Proteomic analysis further aligned the promise of ST6Gal1-based engineering of NK-92MI cells with supervisory loading efficiency and coverage for immune synapse components that might facilitate the engaging and killing of B lymphoma by NK-92MI cells. We also assessed the orthogonality of the ST6Gal1-assisted incorporation of
MPBNeu5Ac for CD22 targeting with the
in situ generation of selectin ligands by exogenous α1,3-fucosylation. In a murine xenograft tumor model, increased levels of selectin ligands were essential for the
in vivo control of tumor growth by
MPBNeu5Ac-armed NK-92MI cells. In addition, the incorporation of
MPBNeu5Ac onto CD19-CAR-T cells can similarly enhance their targeting and killing of CD22-positive malignancies. These findings establish the use of live-cell glycocalyx engineering as a versatile strategy to refine cellular immunotherapies and overcome barriers in cancer treatment.
2. Materials and methods
2.1. Reagents
Antibiotin-horseradish peroxidase (HRP) was purchased from Cell Signal Technology (USA). Alkyne-polyethylene glycol (PEG)4-cyanine 5 (Cy5), alkyne-PEG4-biotin, water-soluble dibenzo-cyclooctyne (DBCO)-biotin, and water-soluble DBCO-Cy5 were purchased from Click Chemistry Tools (USA). Cytidine monophosphate-N-azidoacetylneuraminic acid (CMP-Neu5Az) and guanosine-5′-diphosphate-β-L-fucose (GDP-Fuc) were purchased from Shanghai Biochemsyn (China). Celltracker Deep Red and Celltracker 5-chloromethylfluorescein diacetate (CMFDA) were obtained from Thermo Fisher Scientific (USA). The lactate dehydrogenase (LDH) detection kit was from Promega (USA). Aleuria aurantia lectin (AAL)-biotin, Sambucus nigra lectin (SNA)-biotin, and Maackia amurensis lectin II (MAL-II)-biotin were purchased from VectorLabs (USA). Fluorescein isothiocyanate (FITC)-conjugated anti-human/mouse cutaneous lymphocyte antigen (CLA), anti-human CD19, and anti-human CD22 antibodies were obtained from Biolegend (USA). Streptavidin-Alexa Fluor 647 was purchased from Invitrogen (USA). FITC-conjugated anti-human immunoglobulin G (IgG) antibody was purchased from Abcam (USA). CD22-fragment crystallizable (Fc) was obtained from Sinobiological (China). An inferno-γ (IFN-γ) enzyme linked immunosorbent assay (ELISA) kit was purchased from Solarbio (China).
2.2. Cell culture
Daudi, Raji, Jurkat, BJAB, Ramos, HEK293-FT, and NK-92MI cells were obtained from the laboratory of Dr. Xing Chen (Peking University, China). GM12878 was purchased from the BeNa Culture Collection (China). A Raji cell line stably expressing firefly luciferase and tdTomato (Raji-Luc) was established by lentiviral infection. Daudi, Raji, Jurkat, BJAB, GM12878, and Ramos cells were kept in Roswell Park Memorial Institute (RPMI)-1640 media (Gibco, USA) supplemented with 10% fetal bovine serum (FBS; Gibco), 100 U∙mL−1 penicillin G (Gibco), and 100 μg∙mL−1 streptomycin (Gibco). HEK293-FT cells were cultured in high-glucose complete Dulbecco’s modified Eagle’s medium (DMEM) containing 10% FBS, 100 U∙mL−1 penicillin G, and 100 μg∙mL−1 streptomycin. NK-92MI cells were cultured in α-minimal essential medium (αMEM) supplemented with 12.5% FBS, 12.5% horse serum, 0.2 mmol∙L-1 inositol, 0.1 mmol∙L-1 2-mercaptoethanol, 0.02 mmol∙L-1 folic acid, 100 U∙mL−1 penicillin G, and 100 μg∙mL−1 streptomycin. All the cells were cultured in a water-saturated 5% CO2 incubator at 37 °C.
2.3. Glycocalyx engineering method
1,6Pr
2ManNAz, per-
O-acetylated
N-azdioacetylmannosamine (Ac
4ManNAz),
MPBNeu5Ac, and CMP-
MPBNeu5Ac were synthesized as previously reported [[
27], [
28], [
29]]. ST6Gal1 was purified following established protocols [
30]. For CeGE, the cultured cells were washed twice using phosphate-buffered saline (PBS) and resuspended at a concentration of 1 × 10
7 cells∙mL
−1 in labeling buffer (Hanks’ balanced salt solution (HBSS) with 3 mmol∙L
-1 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) and 20 mmol∙L
-1 MgSO
4) supplemented with 40 μg∙mL
−1 enzymes and various concentrations of nucleotide sugar donors. After incubation at 37 °C for 30 min, the cells were washed twice with PBS and used for subsequent detection. For MGE, the cells were cultured with probes (1,6Pr
2ManNAz, Ac
4ManNAz, or
MPBNeu5Ac) at different time points, then washed twice with PBS and used for subsequent detection.
2.4. Proteomics sample preparation
For the CeGE group, biotin-labeled cells were washed three times with PBS and lysed with radioimmunoprecipitation assay (RIPA) lysis buffer containing a protease inhibitor cocktail. For the MGE group, azide-labeled cells were washed three times with PBS and lysed with RIPA lysis buffer containing a protease inhibitor cocktail. The resulting protein lysates were quantified using a BCA kit (Thermo Fisher Scientific). The MGE proteins were then tagged with biotin via click chemistry. Biotinylated proteins were isolated by incubation with avidin-agarose beads (Thermo Fisher Scientific) at 4 °C for 4 h. The beads were subsequently washed three times with PBS and once with PBS containing 2% sodium dodecyl sulphate (SDS), followed by resuspension in 200 μL of 8 mol∙L-1 urea buffer (100 mmol∙L-1 Tris-HCl, pH 8.5). The proteins were reduced with 10 μL of 100 mmol∙L-1 tris(2-carboxyethyl)phosphine (TCEP) for 20 min and alkylated with 5 μL of 400 mmol∙L-1 iodoacetamide (IAA) for 30 min in the dark. The reaction was quenched with 585 μL of 100 mmol∙L-1 Tris-HCl (pH 8.5). After the supernatant was discarded, the proteins on the beads were trypsinized overnight at 37 °C in 200 μL of 2 mol∙L-1 urea, 100 mmol∙L-1 Tris-HCl (pH 8.5), 2 μL of 100 mmol∙L-1 CaCl2, and trypsin (1:25 ratio to total protein). The supernatant was collected, and the beads were washed twice with 100 mmol∙L-1 Tris-HCl (pH 8.5). All the supernatants were combined, and the digestion was stopped with 5% formic acid.
2.5. Mass spectrometry
Peptides were pressure-loaded onto a 250 μm capillary column packed with 2 cm of 10 μm Jupiter C18-A material and 2 cm of a 5 μm Partisphere strong cation exchanger. The buffer solutions used were 5% acetonitrile/0.1% formic acid (Buffer A), 80% acetonitrile/0.1% formic acid (Buffer B), and 500 mmol∙L-1 ammonium acetate/5% acetonitrile/0.1% formic acid (Buffer C). The column was subsequently washed with Buffer A. A 100 μm capillary with a 5 μm pulled tip packed with 15 cm of 4 μm Jupiter C18 material was attached via a union, and the assembly was connected to an Agilent 1100 quaternary high-performance liquid chromatography (HPLC; USA). Peptides were separated using a modified four step gradient: Step 1 (0-100% Buffer B over 90 min), Step 2 (20% Buffer C, 0-100% Buffer B over 100 min), Steps 3-9 (X% Buffer C, 0-100% Buffer B over 110 min, with X = 30%-100%), and the final step (90% Buffer C, 0-100% Buffer B over 100 min). Peptides were electrosprayed into an Orbitrap Elite mass spectrometer (2.4 kV spray voltage; USA) and analyzed using a cycle of one full-scan Fourier-transform (FT) mass spectrum (300-1600 m/z, 240 000 resolution) followed by 20 data-dependent tandem mass spectrometry (MS/MS) spectra (35% collision energy).
2.6. Mass data analysis
MS/MS spectra were extracted from the XCalibur data system format (.RAW) files using RAW_Xtractor [
31] and searched with ProLuCID [
32] against a UniProt human database concatenated with a decoy database [
33]. Searches were performed on a 100 central processing unit (CPU) Beowulf cluster with no enzyme specificity. Static cysteine modification (57.02146) was included. ProLuCID results were filtered using DTASelect 2.0 [
34], and linear discriminant analysis was applied to achieve a protein-level false discovery rate (FDR) < 1% and peptide-level FDR < 1%. Peptides were required to be partially tryptic with < 5 parts per million (ppm) mass deviation. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted using the DAVID tool with default parameters.
2.7. CD19-CAR-T cell preparation
For lentiviral production, HEK293-FT cells were transfected with a plasmid encoding CD19-CAR together with packaging and envelope vectors (pMDLg/pRRE, pMD2.G, and pRSV-Rev) using polyethylenimine (PEI). The culture supernatant was collected at 48 h post-transfection and concentrated by ultracentrifugation before being resuspended in RPMI-1640 media. Peripheral blood mononuclear cells (PBMCs) from healthy donors were isolated by density gradient centrifugation with Ficoll-Paque Plus (Cytiva, Sweden). CD3-positive T-cells were isolated from PBMCs using a CD3 T-cell isolation kit (BioLegend) and activated using T-cell activation beads (Thermo Fisher Scientific). Activated T-cells were cultured at a concentration of 1 × 106 cells∙mL−1 in RPMI-1640 media supplemented with 10% FBS, 100 IU∙mL−1 humanized interleukin-2 (hIL-2), 1% gluta-MAX (Thermo Fisher Scientific), 100 U∙mL−1 penicillin G, and 100 µg∙mL−1 streptomycin. After 24 h, activated T-cells (0.5 × 106 cells∙well−1) were infected with lentivirus in RPMI-1640 medium supplemented with 8 µg∙mL−1 polybrene and 100 IU∙mL−1 hIL-2 under centrifugation (1000g, 1.5 h, 33 °C). The cells were expanded in fresh IL-2-containing medium for 7 d before use or cryopreservation. All individuals provided informed consent to participate in this study, and approval was provided by the Medical Research Ethics Committee of Peking University Health Science Center (approval No. 2024188).
2.8. Quantification of specific lysis of B lymphoma cells via LDH release assay
NK-92MI cells or T-cells and CD19-CAR-T cells were mixed with different types of cancer cells at the indicated effector/target cell (E/T) ratios for 4-8 h in a 96-well plate. Specific cancer cell lysis was detected by LDH secretion in the supernatant. Specific lysis of target cells was performed according to the manufacturer’s instructions.
2.9. Quantification of the formation of cell-cell conjugates via flow cytometry
Lymphoma and engineered immune cells were washed twice with PBS and then incubated in RPMI-1640 with Celltracker Deep Red or Celltracker CMFDA for 30 min at 37 °C. After staining, the cells were mixed at various E/T ratios for 1 h at 37 °C and directly analyzed by flow cytometry using a Attune NxT flow cytometer (Thermo Fisher Scientific). The double-positive signals indicate the formation of cell-cell conjugates. Quantification was performed by calculating the percentage of double-positive (CMFDA+/Deep Red+) cells relative to the total number of Deep Red+ target cells.
2.10. Western blot
Azide-labeled proteins were precipitated with methanol/chloroform and resuspended in 50 μL of PBS containing 1% SDS per 200 μg of protein. Fifty microliter suspensions were incubated with 50 μmol∙L-1 CuSO4-2-(4-((bis((1-(tert-butyl)-1H-1,2,3-triazol-4-yl)methyl)amino)methyl)-1H-1,2,3-triazol-1-yl)acetic acid (BTTAA) premixed complex (1:2 mole ratio), 100 μmol∙L-1 alkyne-PEG4-biotin, and 2.5 mmol∙L-1 fresh sodium ascorbate at 37 °C for 2 h to perform the click reaction. The samples were resolved by 10% SDS-polyacrylamide gel electrophoresis (PAGE) and imaged by ChemiDoc (Bio-Rad, USA). Biotin was detected using an antibody (1:1000; Cell Signaling Technology), and actin (1:5000; Abcam) was used as a control. Anti-rabbit IgG-HRP (1:5000; ABclonal, China) was used for secondary detection.
2.11. In-gel fluorescence assay
For protein samples, azide-labeled proteins were precipitated with methanol/chloroform and resuspended in 40 μL of PBS containing 1% SDS per 200 μg of protein. Fifty microliter suspensions were incubated with 50 μmol∙L-1 CuSO4-BTTAA premixed complex (1:2 mole ratio), 100 μmol∙L-1 alkyne-PEG4-Cy5, and 2.5 mmol∙L-1 fresh sodium ascorbate at 37 °C for 2 h to perform the click reaction. The samples were resolved by 10% SDS-PAGE and imaged with ChemiDoc. The gels were then stained with Coomassie brilliant blue (CBB).
2.12. Tumor models
Male NOD.Cg-Prkdcscid IL2rgtm1Wjl/SzJ (NSG) mice (8-10 weeks old) were inoculated with 1 × 106 Raji-Luc cells through tail vein injection (Day 0). On Day 2, the mice were randomly divided into four groups and treated with HBSS, NK-92MI, MPBNeu5Ac-modified NK-92MI, or MPBNeu5Ac/sialyl Lewis X (sLeX) dual-labeled NK-92MI cells through tail vein injection (7 × 106 NK-92MI cells per mouse). On Day 5 after tumor challenge, the mice were injected with 200 μL of D-luciferin (15 mg∙mL−1) through intraperitoneal (i.p.) injection. After 10 min, the bioluminescence signal was measured using an IVIS Lumina Series III (PerkinElmer, USA). The total number of photons indicating the tumor was quantified with IVIS software. To follow the progression of B lymphoma, NSG mice received three additional doses of NK-92MI cells on Days 6, 10, and 14. Luciferin-assisted IVIS monitoring of tumor growth was performed on Days 5, 8, 11, and 13. NSG mice were purchased from Biocytongen Pharmaceuticals (China). All animal study procedures were approved by the Institutional Animal Care and Use Committee of Peking University Health Science Center and performed in accordance with ethical standards (No. DLASBD0224).
3. Results and discussion
3.1. The glycan repertoire defines the engineering efficiency of CeGE vs MGE
In addition to the lipid anchor-assisted incorporation of glycoconjugates, the outermost glycocalyx of live cells provides viable scaffolds for loading functional agents to influence their cellular properties. To date, two major glycan-centric strategies for the functional engineering of live cells, MGE and CeGE, have been intensively evaluated in this field. MGE strategies, which rely on intracellular glycosylation machinery to incorporate small functionalities onto the glycocalyx, have a limited tolerance for substrate modifications. As previously documented, MGE of sialylated glycans using unnatural sugar building blocks (e.g., azide-modified analogs of Sia or
N-acetylmannosamine) has robust efficiency. In contrast, CeGE approaches exhibit excellent biocompatibility and broad substrate scope, and their engineering efficiency is largely determined by the membrane content of tunable glycan moieties. Accordingly, we first compared the efficiency of MGE-based sialoside engineering of NK-92MI cells with that of ST6Gal1-assisted CeGE engineering (
Fig. 1(a)). The analog of Sia (e.g., the azide derivative Neu5Az) had insufficient plasma membrane penetration, we thus assessed the labeling of NK-92MI cells by 1,6Pr
2ManNAz, a recently developed probe. The dose-dependent incorporation of azides onto NK-92MI cells was probed by conjugation with DBCO-biotin, followed by streptavidin-Alexa Fluor 647 staining and flow cytometry quantification (
Fig. 1(b)). As expected, the ST6Gal1-based addition of Neu5Az onto NK-92MI cells was concentration dependent (
Fig. 1(c)). The antibiotin western blot results revealed that the intensity of ST6Gal1 labeling was significantly greater than that resulting from metabolic labeling using 1,6Pr
2ManNAz or the previous well-established Ac
4ManNAz under saturated conditions (
Figs. 1(d) and (e)).
3.2. Different glycoprotein networks control downstream cellular signaling beyond functionalization
Next, we conducted a comparative glycoproteomic analysis to identify the molecules underlying the differential labeling efficiency, given that the MGE and CeGE approaches involve distinct mechanisms. Following engineering and biotin conjugation, the modified glycoproteins were enriched with avidin-agarose beads for trypsin digestion and subsequent mass spectrometry analysis (
Fig. 2(a)). In the ST6Gal1-CeGE group, we confidently identified a total of 463 membrane-associated glycoproteins, which was 2.5-fold greater than that documented in the MGE group (with 184 targets). Compared with that of the MGE approach, the ST6Gal1-assisted CeGE approach covered a significantly greater diversity of glycoproteins (i.e., 344 unique targets), with 119 proteins shared by both types of glycocalyx engineering (
Fig. 2(b)). The shared subgroup of proteins accounted for 81.8% of the total protein intensity in the MGE group versus 46.4% in the CeGE group (
Fig. 2(c)), suggesting that the CeGE group exhibited a distinctive glycoproteome.
Notably, the top 40 glycoproteins ranked by normalized abundance included key regulators associated with immune cytotoxicity and adhesion: CD44, CD45, CD47, lymhotactin (XCL1), dipeptidyl peptidase 4 (DPP4), and cadherin 17 (CDH17) (
Fig. 2(d), Figs. S1(a) and (b) in Appendix A) [[
35], [
36], [
37], [
38], [
39], [
40]]. GO and KEGG functional enrichment analyses revealed that the shared proteins were involved in NK cell-mediated cytotoxicity and cell adhesion pathways (
Fig. 2(e), Fig. S1(c) in Appendix A). Interestingly, the ST6Gal1-CeGE method uniquely modified critical proteins involved in immune synapse formation and NK activation, including CD2, 3-phosphoinositide-dependent protein kinase 1 (PDPK1), Nck-associated protein 1-like (NCKAP1L), signaling lymphocyte activation molecule (SLAM) family receptors (e.g., SLAMF1), and Src family kinases (e.g., LCK) [[
41], [
42], [
43], [
44], [
45]]. These findings suggest that glycoengineering may attach functionalities to glycoproteins critical for NK cell effector functions. After ST6Gal1-CeGE treatment, the selective enrichment of agents (i.e., CD22-target glycan ligands) to immune synapse components may spatially facilitate the formation of cytotoxic machinery at the effector-target interface. Together, the broader glycoprotein coverage and preferential labeling of cytotoxic effectors by ST6Gal1-CeGE make it the preferred method for improving NK cell therapies with CD22-targeting capability.
3.3. Chemoenzymatic incorporation of MPBNeu5Ac promotes the targeting of B lymphoma cells by NK-92MI cells
Inspired by these mechanistic insights, we explored the feasibility of modifying NK-92MI cells with
MPBNeu5Ac to target CD22 through ST6Gal1-assisted CeGE (
Fig. 3(a)). Compared with those of
BPCNeu5Ac,
MPBNeu5Ac has comparable avidity for CD22 but minimal potential off-target side effects because of its negligible affinity for other Siglecs [
12]. NK-92MI cells were subsequently treated with ST6Gal1 and CMP-
MPBNeu5Ac. The successful chemoenzymatic incorporation of
MPBNeu5Ac onto NK-92MI cells significantly strengthened their binding to CD22 (
Fig. 3(b)). Interestingly, compared with that of cells treated with
MPBNeu5Ac-MGE, NK-92MI cells treated with ST6Gal1 for the incorporation of
MPBNeu5Ac demonstrated an approximately 5-fold increase in the avidity of CD22 under saturated conditions (
Fig. 3(c)). In addition, neither the MGE- nor the CeGE-mediated installation of
MPBNeu5Ac for CD22 targeting affected the viability or proliferation of NK-92MI cells, suggesting their biocompatibility (Figs. S2(a) and (b) in Appendix A). To assess the ability of
MPBNeu5Ac-modified NK-92MI cells to target cancer cells, we cocultured CMFDA-labeled NK-92MI cells with Deep Red-labeled lymphoma cells (i.e., Ramos, Daudi, Raji, and BJAB cells). The formation of effector-target “cell-cell” conjugates was quantified by flow cytometry (
Fig. 3(d), Figs. S2(c) and (d) in Appendix A). The results indicated that
MPBNeu5Ac-NK-92MI cells improved the specific targeting of CD22-positive B lymphoma cells but did not affect binding to CD22-negative Jurkat cells (
Fig. 3(e)). Consistent with the findings of Hong et al. [
17] using
BPCNeu5Ac-modified NK-92MI cells,
MPBNeu5Ac-modified NK-92MI cells exhibited enhanced cytotoxicity against CD22-positive lymphomas without damaging normal cells (i.e., GM12878 B lymphocytes and PBMCs), as measured by an LDH release assay and flow cytometry (Figs. 3(f)-(i), Figs. S2(e)-(h) in Appendix A). To assess ligand dependence, we prepared corresponding high-affinity ligands on
MPBNeu5Ac-armed NK-92MI cells by α2,6-linking
MPBNeu5Ac onto
N-acetyllactosamine (LacNAc) as previously reported [
46]. As expected, the better CD22 ligand-mediated lysis of B lymphoma cells was specifically blocked by the addition of
MPBNeu5Ac-LacNAc but not free
MPBNeu5Ac (Fig. S3(a) in Appendix A). In addition to CD22-negative target cells (i.e., Jurkat cells), we further evaluated CD22 dependence by pretreating target B lymphoma cells (i.e., BJAB cells) with CD22-specific blocking antibodies. We observed that the presence of antibodies significantly inhibited the specific lysis of B lymphoma cells by
MPBNeu5Ac-NK-92MI cells, suggesting that CD22 antigens on B lymphoma cells can be targeted by ligands (Fig. S3(b) in Appendix A). These data demonstrated that ST6Gal1-assisted CeGE effectively incorporated
MPBNeu5Ac onto NK-92MI cells to generate CD22-specific high-affinity ligands, thereby enhancing their interaction with and specific lysis of CD22-positive B lymphoma cells. Moreover, improved
in vitro B lymphoma eradication was achieved using primary NK cells after ST6Gal1-mediated incorporation of
MPBNeu5Ac to generate CD22-high-affinity ligands (Fig. S3(c) in Appendix A), supporting the feasibility of engineering allogenic NK cells for B lymphoma treatment.
3.4. Orthogonal CeGE modifications for the eradication of B lymphoma in a murine model
Owing to the greater selectivity of
MPBNeu5Ac for CD22 compared with that of
BPCNeu5Ac, we evaluated the
in vivo antitumor efficacy of
MPBNeu5Ac-NK-92MI cells. Effective tumor clearance by NK-92MI cells relies on both their ability to target CD22 and their ability to infiltrate tumors. Adhesion factors, such as E-selectin on vascular endothelial cells, bind to their ligand sLe
X (Neu5Acα2,3Galβ1,4(Fucα1,3)GlcNAc) and facilitate the migration of leukocytes to bone marrow and inflamed cancer tissues [
47]. Scientists previously demonstrated that the increased expression of sLe
X by human fucosyltransferase VI (hFUT6) is essential for the
in vivo anti-B lymphoma potency of
BPCNeu5Ac-armed NK-92MI cells [
16]. Therefore, we treated
MPBNeu5Ac-modified NK-92MI cells with hFUT6 and its substrate GDP-Fuc. The exogenously added α1,3-Fuc was stained with AAL, and the strong staining of hFUT6-treated cells relative to that of control cells confirmed the success of α1,3-fucosylation (
Fig. 4(a)). Notably, the application of hFUT6, which added α1,3-Fuc to NK-92MI cells, resulted in a significant increase in the level of sLe
X, as clearly detected by anti-sLe
X antibody staining, which revealed an approximately 20-fold increase in sLe
X staining of α1,3-fucosylated NK-92MI cells compared with that of untreated cells (
Fig. 4(b)). Moreover, the CD22-Fc binding efficacy of α1,3-fucosylated NK-92MI cells was comparable to that of
MPBNeu5Ac-NK-92MI cells, highlighting the orthogonality of CD22 ligands generated by ST6Gal1 and sLe
X moieties generated by hFUT6 (
Fig. 4(c)).
Next, we evaluated the therapeutic efficacy of
MPBNeu5Ac-NK-92MI cells using a murine model of Raji B lymphoma. On Day 0, Raji-Luc cells, a B lymphoma cell line that constantly expresses firefly luciferase and the tdTomato we developed, were intravenously (i.v.) xenografted into NSG mice. Subsequently, the mice received four rounds of NK-92MI cell treatments (i.v.) on Days 2, 6, 10, and 14 after tumor inoculation (
Fig. 4(d)). Bioluminescence imaging was performed on Days 5, 8, 11, and 13. The images clearly revealed that compared with the HBSS control group, the group treated with α1,3-fucosylated
MPBNeu5Ac-NK-92MI cells had better tumor growth control after Day 11 (
Figs. 4(e) and
(f)). However, no significant differences were observed between the groups treated with NK-92MI or
MPBNeu5Ac-NK-92MI cells and the HBSS control group. At the terminal stage on Day 17, peripheral blood mononuclear cell analysis confirmed a lower tumor burden only in mice treated with α1,3-fucosylated
MPBNeu5Ac-modified NK-92MI cells (
Fig. 4(g)), which further aligned with the survival statistics, indicating a therapeutic effect in this group (
Fig. 4(h)). Together with the results of a previous study on
BPCNeu5Ac-NK-92MI therapeutics, these results highlight the potential of dual orthogonal glycoengineering—targeting both CD22 and selectins—to increase the antitumor efficacy of NK-92MI cells, which is indispensable for improving the survival of tumor-inoculated mice.
3.5. MPBNeu5Ac modification targeted CD19-CAR-T cells to kill B-cells in lymphoma
Inspired by the improved targeting of CD22-positive B lymphoma by
MPBNeu5Ac-armed NK cells, we explored whether
MPBNeu5Ac could improve CD19-CAR-T therapy. Previous success with multiple-antigen targeting strategies suggests that this approach can effectively address antigen escape and reduce the on-target/off-tumor side effects associated with CD19-CAR-T treatment [
9,
48]. We hypothesized that the addition of
MPBNeu5Ac could offer a versatile solution. We constructed CD19-CAR-T cells and validated their cytotoxic activity against B-cell lymphomas (Fig. S4 in Appendix A). We then quantified the CD19 antigens present on CD22-positive B lymphoma cell lines, including BJAB, Ramos, and Daudi, using Jurkat cells as CD19-negative controls (Fig. S4(c)). Dose-dependent incorporation of Neu5Az through ST6Gal1-assisted engineering confirmed a tunable glycan subset on CD19-CAR-T cells (
Fig. 5(a)). The ST6Gal1-mediated addition of
MPBNeu5Ac enhanced the binding of CD19-CAR-T cells, despite the loading efficiency not being as pronounced as that observed in NK-92MI cells (
Figs. 3(b) and
Figs. 5(b)). This difference may be due to the significantly higher sialylation levels on CAR-T cells relative to NK cells, which results in fewer available scaffolds for Sia modification. This speculation was supported by staining with SNA (specific for α2,6-linked Sia) and MAL-II (specific for α2,3-linked Sia), which revealed high levels of both α2,3- and α2,6-linked Sia residues on CAR-T cells (Fig. S5(a) in Appendix A). Moreover, we observed donor-dependent variability in labeling efficacy, suggesting differences in glycoengineering outcomes among donors (Fig. S5(b) in Appendix A). Given the high levels of surface sialylation on CAR-T cells, we compared the
MPBNeu5Ac engineering efficiency between ST6Gal1-CeGE and
MPBNeu5Ac-MGE for CAR-T cell modification. CD22-Fc staining revealed marginally superior modification efficiency with MGE versus CeGE (
Fig. 5(c)).
We subsequently assessed the formation of cell
-cell conjugates between CAR-T cells and cancer cells by coculturing CMFDA-labeled CAR-T cells with Deep Red-labeled lymphoma cells.
MPBNeu5Ac modification significantly increased the binding of CD19-CAR-T cells to CD22-positive Ramos cells but did not improve the binding of CD22-negative Jurkat cells (
Fig. 5(d), Fig. S5(c) in Appendix A). Expanded testing across different donors and tumor lines confirmed increased targeting of CD22-expressing malignancies, despite variation in the efficacy among CD22-positive tumors (
Figs. 5(e) and
(f), Fig. S5(d) in Appendix A). Moreover,
MPBNeu5Ac-CD19-CAR-T cells exhibited enhanced cytotoxicity against CD22/CD19-positive lymphoma cells (
Figs. 5(g)-(i)). No significant difference in cytotoxic efficacy was found between CeGE- and MGE-engineered CAR-T cells. This finding may be attributed to the efficient formation of immune synapses associated with CeGE engineering. Collectively, these results demonstrate that
MPBNeu5Ac ligands on engineered CD19-CAR-T cells improve their ability to recognize and kill multiple types of CD22-positive B lymphoma cells; however, the potential of using CD22-ligand-armed CD19-CAR-T cells to mitigate antigen escape needs further
in vivo exploration.
4. Conclusions
It is well established that glycosylation in the secretory pathway defines the molecular repertoire of the surface glycocalyx on live cells by linking carbohydrate moieties to proteins and lipids. Recently, the discovery of RNA N-glycosylation has complicated these cell-surface molecular networks, in addition to inspiring concerns about DNA glycosylation [
49,
50]. Owing to their remarkable structure-to-information coding divergence, recognized as the third alphabet of life, abnormal glycans play vital (patho)physiological roles in diseases such as cancers and neurodegenerative diseases [
26]. Characterizing glycan fingerprints with high-throughput omics has thus empowered the discovery of markers on glycoconjugates and live cells to differentiate health and disease [
51,
52]. These outermost scaffolds present ideal frames to load functionalities, and live-cell glycocalyx engineering has emerged as a highly compatible and flexible means for loading agents to endow cellular immunotherapies with desired functions, such as via MGE and CeGE. In the current study, we compared the labeling efficiency and molecular characteristics of 1,6Pr
2ManNAz-assisted MGE and ST6Gal1-assisted CeGE in NK-92MI cells. ST6Gal1-CeGE exhibited comparable or even superior ligand-loading efficiency (about 2.5-fold), with many immune synapse components (e.g., LCK and SLAMF1) modified, which may improve the spatial recognition of target cells. The superiority of ST6Gal1-assisted CeGE compared with the MGE approach for functionalization of immune cells (i.e., NK-92MI cells, primary NK cells, and CD19-CAR-T cells) to treat B lymphoma was elucidated and expanded. CeGE-assisted CD22-specific
MPBNeu5Ac-NK-92MI cells exhibited superior targeting and cytotoxicity against B lymphoma
in vitro, and fucosylated
MPBNeu5Ac-NK-92MI cells showed significant antitumor efficacy
in vivo. In addition to addressing the potential off-target effects of CD19-CAR-T-cell therapy in clinical settings, MGE- and ST6Gal1-mediated CeGE glycocalyx engineering were applied to incorporate
MPBNeu5Ac onto CD19-CAR-T cells. The
in situ generation of CD22-specific high-affinity ligands on CD19-CAR-T cells facilitated their lysis of CD19/CD22 B lymphoma cells in a cell coculture assay. Compared with cost-effective MGE, which relies on intracellular glycosylation machinery to incorporate ligands, culture-free CeGE engineering is time-saving and easy to transfer in many settings (e.g., engineering of live cell-free cellular derivatives). Notably, the elimination of multiple target cells by
MPBNeu5Ac-NK-92MI cells significantly increased when B lymphoma and effector cells were cocultured. However, no substantial
in vivo tumor eradication was observed unless
MPBNeu5Ac-modified effector NK-92MI cells were also modified by hFUT6 to add α1,3-Fuc, which acts as an E-selectin ligand to guide them to tumor sites. For clinical applications, our proof-of-concept studies still require in-depth assessment for such synergistic treatment (e.g.,
in vivo evaluations). Additionally,
MPBNeu5Ac exhibits limited avidity for CD22, and more substantial benefits may be achieved by other high-avidity structures [
11]. This study is expected to stimulate interest in the preparation of novel glycoligand-armed immunotherapies via glycocalyx engineering, thus revealing their potential to improve B lymphoma treatment. This could be helpful in meeting the growing demand for the nongenetic engineering of immune cells that exhibit high compatibility. The glycoproteomic profile provides fundamental information for immune cell engineering, and CeGE offers ideal tools for functionalizing immune cells for clinical applications.
CRediT authorship contribution statement
Yuxin Li: Writing - review & editing, Writing - original draft, Formal analysis, Data curation, Conceptualization. Tao Gao: Writing - original draft, Methodology, Formal analysis, Data curation. Zhaoxin Han: Data curation. Valeria M. Stepanova: Resources. Han Wang: Data curation. Hongmin Chen: Data curation. Alexey Stepanov: Supervision, Resources. Senlian Hong: Writing - review & editing, Writing - original draft, Supervision, Funding acquisition, Conceptualization.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
We gratefully acknowledge the financial support from the National Natural Science Foundation of China (32150027 and 22177002) and the National Key Research and Development Pro-gram of China (2022YFC3400803 and 2023YFC2308003). Valeria M. Stepanova and Alexey Stepanov were partially supported by the Russian Science Foundation (5-74-30002).
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.eng.2025.08.037.
PDF
(4902KB)
