We engineer the IgH locus of B cells using electroporation of RNA-guided nucleases and transduction with recombinant adeno-associated viral vectors (rAAV). The bicistronic cassette encodes an anti-HPV E6 full light chain and the variable domain of the heavy chain. The translated segments are separated by a 2A peptide and terminated with a splice donor. Successful integration into the J–C intron of the IgH locus allows expression and splicing with endogenous constant segments, leading to translation of a full heavy chain ( Figure 1A ). Therefore, EBCs express the anti-HPV E6 antibody as a BCR and can differentiate into plasmablasts and plasma cells, secreting the anti-HPV E6 antibody ( Figures 1B, C ). As a BCR, the antibody facilitates B cell activation and internalization of the bound antigen for processing and presentation on MHC, ultimately leading to antigen-specific T cell activation ( Figure 1D ). When secreted by differentiated EBCs, the antibody further enables antigen-specific T cell activation through the formation of immune complexes ( Figure 1E ). Moreover, since the edited target locus is located upstream of the genomic regions that undergo rearrangement during class-switch recombination, the anti-HPV E6 antibody can undergo isotype switching. Isotype switching enables mucosal tissue protection, systemic clearance, immune complex-mediated T cell activation, and effector functions such as antibody-dependent cellular cytotoxicity (48). Finally, because the cassette is integrated within the IgH locus, it may undergo somatic hypermutation during germinal center (GC) reactions, enabling clonal expansion and potentially leading to affinity maturation.

We engineered B cells to express the anti-E6 antibodies C1P5 or 6F4 (49, 50). CRISPR/Cas9 activity, enabling on-target double-strand breaks for downstream integration of the engineering cassette, was confirmed using TIDE ( Figure 2A ). Since our engineering cassette does not encode the constant domains of the heavy chain, including membrane-anchoring domains, on-target genome integration enables BCR expression via splicing with endogenous heavy chain segments (18, 21, 22). We therefore quantified functional engineering rates by spectral cytometry using E6 peptides containing the target epitope for either C1P5 or 6F4 ( Figures 2B, C ).

We further characterized subsets of the engineered cells by spectral cytometry. No significant phenotypic differences between EBCs and non-engineered B cells were detected. Three main B cell populations were identified. The dominant population exhibited an unswitched, activated B cell phenotype with GC homing potential, comprising up to 63% of the cells. In addition, antibody-secreting cells and class-switched B cells, made up 11% and 3% of the cell populations, respectively ( Supplementary Figure 1 ). Employing i.29 cells, which basally express low levels of the kappa light chain, enabled detection of engineered cells independently of antigen binding. When engineered to express 6F4, the cells could reliably bind the full E6 protein, but not when engineered to express C1P5 ( Supplementary Figures 2A, B ). 6F4 bound the E6 protein at lower concentrations than C1P5 ( Supplementary Figure 2C ). The E6 transcript undergoes alternative splicing to a shorter version called E6*I. This splicing event is thought to support E7 expression (51). Interestingly, the 6F4 binding site—but not the C1P5 binding site—is present in both the full-length E6 and the alternatively spliced E6*I protein product ( Supplementary Figure 2D ).

Therefore, we proceeded with the 6F4 binder and improved the primary B cell engineering protocol by enriching B cells prior to culture, employing standardized culture media and electroporating the cells at a lower concentration ( Supplementary Figures 2E, F ). Taken together, these results confirm that IgH targeting is a robust approach for engineering B cells to target tumor-associated antigens; however, expression level, antibody affinity, binding epitope, and target expression must be considered when selecting a therapeutically relevant antibody for B cell engineering.

B cells are potent antigen-presenting cells (APCs) (52). Interaction between B cells and T cells require HLA compatibility and TCR specificity (53). Previous studies have shown that polyclonal and antigen-specific B cells can activate CD4⁺ and CD8⁺ T cells (42–44, 54). However, mouse models with T cells recognizing E6 via a defined TCR are lacking. To initially demonstrate antigen-induced activation of T cells by mouse EBCs in vitro, we engineered B cells to express the OB-I antibody, a specific binder for ovalbumin (OVA) (55), and employed the OT-I and OT-II systems as cognate CD8⁺ and CD4⁺ T cells, respectively ( Figure 3A ) (56, 57). We achieved approximately 30% engineering efficiency, as detected by flow cytometry ( Figures 3B, C ). OT-I T cells recognize Ova_257–264 presented on MHC class I H-2K^b, and OT-II T cells recognize Ova_323–339 presented on MHC class II I-A^b. Thus, we validated each T cell type by co-culturing with peptide-spiked, non-engineered primary mouse B cells ( Figure 3D ). Next, we co-cultured OB-I EBCs, with the full OVA antigen and either OT-I or OT-II T cells ( Figures 3A, E ). B cells are professional APCs and may therefore present antigens independently of antibody expression, likely via pinocytosis (58). However, antigen presentation to be significantly more efficient for B cells expressing an antibody specific to the antigen (59). Thus, we investigated dose responses to full OVA antigen pre-incubated with EBCs prior to co-culture, as monitored by T cell activation markers and compared to control non-engineered B cells. As detected by flow cytometry and ELISA, significantly higher activation of OT-I and OT-II cells was observed when using OB-I EBCs compared to control non-engineered B cells ( Figures 3F–L ). Indeed, we detected increased frequencies of T cells expressing activation markers CD69 and CD25, along with elevated secretion of IFNg and TNFa in the supernatants of EBC-T cell co-cultures supplemented with OVA antigen. This effect was particularly evident at antigen concentrations of 10 nM–100 nM, indicating that the antibody expressed in EBCs internalizes OVA for processing and presentation on MHC class I and II. At high antigen concentration, pinocytosis may enable B cells to uptake-up antigens for T cell presentation, independent of antibody specificity (44, 59). Consistently, we observed activation of OT-I and OT-II T cells in co-cultures with EPOnly B cells at 1,000 nM OVA. Together, these results demonstrate that at low antigen concentrations, EBCs function as potent antigen-specific-presenting cells.

To generate polyclonal, E6-reactive T cells, we immunized mice with the E6 oncoantigen and extracted total T cells following three immunizations ( Figures 4A, B ). To demonstrate the APC functions of mouse EBCs in vitro, we co-incubated E6-immunized splenic T cells with engineered splenic B cells ( Figures 4C, D ) in the presence of the full E6 oncoantigen. Following internalization, EBCs present peptides derived from the proteolysis of the oncoantigen on MHC class I/II, subsequently activating E6-reactive T cells. As indicated by intracellular flow cytometry on CD4⁺ and CD8⁺ T cells ( Figures 4E, F ; Supplementary Figure 3A ) and confirmed by ELISA ( Figure 4G ) for IFN-γ expression and secretion, EBCs—but not control non-engineered B cells (EPOnly)—activate T cells in the presence of E6 at concentrations as low as 10 nM. Furthermore, EBCs enhanced CD8⁺ cytotoxicity, as indicated by granzyme B and upregulation of the degranulation marker CD107a ( Figures 4H, I ; Supplementary Figure 3B ). We compared T cell activation elicited by mouse EBCs and control EPOnly B cells when cultured at E6 concentrations ranging from 0.1 nM to 100 nM. These experiments indicated that at high E6 concentrations, EPOnly B cells could also activate E6-reactive T cells, possibly via non-specific, low-efficiency pinocytosis. Importantly, EBCs initiated T cell activation at concentrations 10–100 times lower than those required for EPOnly control B cells ( Supplementary Figures 3A–C ).

In general, secreted antibodies may form immune complexes that enable T cell activation. The antibody binds to the soluble antigen, and the resulting immune complexes are internalized by Fc receptor-expressing APCs, enabling cross-presentation (48, 60, 61). Mouse EBCs secrete polyisotypic anti-HPV E6 antibodies, as detected by ELISA ( Figures 5A, B ) and RT-PCR, further confirming molecular evidence of on-target integration ( Figures 5C, D ). To simulate the immune complex process in vitro, we incubated EBC supernatants with the antigen and loaded them onto myeloid-derived dendritic cells. Splenic T cells from immunized mice were subsequently added to the dendritic cell cultures and monitored for activation by ELISA ( Figures 5E, F ). EBC supernatants formed immune complexes that enabled T cell activation ( Figures 5G–I ).

Constructs were designed as previously characterized, specifically for integration into the IgH gene following a nuclease-mediated double-strand break (22, 26). Donor cassettes encode an antibody expressed under the Enhancer Dependent Promoter (EDP, mutated IGH promoter) (22), with a splice donor to the endogenous heavy chain, thus allowing expression only upon on-target integration into the IgH locus. The segment is preceded by a bGH PolyA signal to ablate endogenous VDJ expression. The variable domains of the antibody are preceded by IgK and IgH leaders. The leaders were pre-spliced to avoid intronic coding and splicing interference. The light chain employs the IgK constant and is separated from the heavy chain using a Furin-GSG-P2A sequence, allowing ribosomal skipping and efficient protein separation. The variable region of the heavy chain is followed by the intronic portion of the native IgHJ4 splice donor, enabling splicing with the endogenous heavy chain exons. The sequences were verified to ensure frame conservation throughout the fully expressed RNA, including at the splice junction with the endogenous constant segments. The coding sequences of the constructs were codon-optimized using a custom pipeline to enhance expression in mice cells, ablate cryptic splice sites, remove G-quadruplets, and improve somatic hypermutation hotspots (22). All donor constructs were delivered using single- stranded AAV backbones and packaged into AAV-DJ serotypes (Packgene Biotech). Supplementary Table 1 lists the sequences, origins, and components of an example construct.

For initial studies involving the binders presented in Figure 2 , B cells were engineered as previously described (22). For E6 co-culture experiments, mouse splenic B cells from naïve 5–9-week-old C57BL/6 mice were negatively selected with B cell isolation kit (Miltenyi Biotec) seeded in Immunocult mouse B cell expansion kit (Stem Cell Technologies) and incubated at 37°C and 5% CO₂ for 1 day. Cells were subsequently electroporated using the NEON Transfection System (ThermoFisher Scientific). Each electroporation contained 100 pmol sgRNA and 30.5 pmol Alt-R HiFi CRISPR-Cas9 (IDT) in Buffer R, using settings 1,675 V, 10 ms, three pulses for 0.5e6 cells per 10 uL reaction. AAV-DJ containing custom cassettes (PackGene Biotech) was added to the cells within 5 min post-electroporation at an MOI of 100,000. Following electroporation, cells were incubated for 24 h, then supplemented with 1 mL of Immunocult Mouse B Cell Expansion Kit media and incubated for an additional 24 h before downstream experiments and analyses.

A list of reagents used for cell culture and engineering is provided in Supplementary Tables 2 , 3 .

All mouse experiments complied with ethical regulations under the supervision of the Ministry of Health, Israel. OT-I and OT-II mice were obtained from Charles River Laboratories. Spleen and bone marrow from C56BL/6 mice were obtained from Science in Action Ltd. To generate polyclonal anti-E6 T cells, 6–12-week-old mice (Envigo) were immunized intraperitoneally with recombinant E6 (Abcam) every 2–4 weeks at a dose of 20 ug/200 uL/mouse in 1% Alhydrogel (Invivogen) for a total of three immunizations. Mice were housed at ambient temperatures of 19 ˚C–23 ˚C, 45%–65% humidity, and a 12 h light/12 h dark cycle.

To detect engineering via E6 target epitopes of the respective binders, cells were incubated at 0.5 μM in Cell Staining Buffer (Biolegend) with biotinylated peptides or biotinylated recombinant E6 protein (R&D Systems) after staining for Zombie L/D (Biolegend), washed and stained with streptavidin (Biolegend) and other antibodies for phenotypic characterization. For OB-I engineering, cells were incubated with OVA at 1 mg/mL in Cell Staining Buffer for 10 min at room temperature following Zombie staining, washed, incubated with rabbit polyclonal anti-OVA (1:50 dilution), washed again, and detected using conjugated anti-rabbit (1:50 dilution). In general, L/D staining was performed according to the manufacturer’s instructions (Biolegend), and mouse-specific conjugated antibodies were used at a 1:100 dilution in Cell Staining Buffer (Biolegend) for 10 min at room temperature. Samples were acquired using a Cytek Northern Lights V/B/R spectral cytometer (Cytek Biosciences). For flow cytometry gating of engineered cells, the biological control—commonly EPOnly—was used to set the gate at values<10%. Cells from the EBC samples within that gate were considered engineered. Positive cells in EPOnly controls may result from technical, nonspecific binding, or from naïve polyclonal B cells expressing diverse BCRs and SMVT/SLC5A6, which can bind biotin. Biotinylation of recombinant E6 was performed according to the manufacturer’s instructions (ThermoFisher). N-terminal biotinylated peptides were synthesized by GenScript.

A list of antibodies and reagents used in detection experiments is provided in Supplementary Table 2 . Custom peptide sequences are provided in Supplementary Table 3 .

For unbiased spectral analysis of cell phenotypes, computational analysis was performed using the Spectre R package (73). Samples were initially prepared in FlowJo, and data were exported as raw-value CSV files. The datasets were merged into a single data table, with sample origin (EBCs and EPOnly) annotated, following exclusion of Zombie Yellow⁺ and CD45.1⁻ cells. The FlowSOM algorithm (74) was then applied to the merged dataset to cluster the data into four cell populations. Clustering was based on the expression of the following markers: CD19, CD138, CD27, CD80, CD86, IgD, IgM, and IgG. Cluster annotation was performed manually. Subsequently, data were analyzed using the Uniform Manifold Approximation and Projection (UMAP) algorithm (75) for cellular visualization. The expression distribution of each marker within each cluster was analyzed and exported as csv files for bar plot visualization. Plots were generated using the Spectre::make.colour.plot and ggplot2 functions (76).

For analysis of E6 binding from engineered cells populations, quantification was performed as follows:

Antigen presentation assays of co-cultures were performed at a 1:1 ratio in 96-well plates for 48 h. Cells were seeded in cell culture medium at a 1:1 ratio, with a final total cell concentration of 1e6 cells/mL. For APC experiments using EBCs, B cells 2–3 days post-engineering were pre-incubated with the indicated concentration of E6 (R&D systems) for 20 min at 37 ˚C and 5%CO₂ prior to T cell supplementation. For co-cultures with polyclonal or OT-I/II derived T cells, splenic lymphocytes were negatively enriched for T cells (Miltenyi) prior to seeding onto B cells.

For immune complex experiments, mouse bone marrow cells were lysed to remove red blood cells (Biolegend) and cultured with 20 ngmL GM-CSF (peprotech) and 10 ng/mL IL-4 (Miltenyi) in RPMI 1640 supplemented with 10% FBS (SigmaAldrich), 2 nM L-glutamine (Sartorius), 100 u/mL penicillin, 100 ug/ml streptomycin, and 50 µM β-mercaptoethanol (SigmaAldrich). Suspended cells were carefully removed on day 3, and adherent cells were cultured for an additional 3 days. Immune complexes were formed by incubating E6 (R&D Systems) with antibodies derived from the supernatants of EPOnly or EBCs, collected 2 days post-engineering, at the indicated concentrations for 20 min at 37 ˚C. Immune complexes were then directly loaded onto counted and reseeded myeloid-derived dendritic cells, followed by the addition of T cells as described above.

A list of reagents employed is provided in Supplementary Table 2 .

ELISA for IFNg, TNFa, GrzmB on the supernatant of co-cultures were performed according to the kit manufacturer’s instructions (R&D Systems). For isotype specific ELISA, plates were coated with E6 (R&D Systems) at 2ug/ml in PBS for 24h at 4˚C. Blocking was performed in 5% BSA, for 2h at room temperature. Supernatant dilutions ranged 1:2-1:20 depending on the specific sample. Secondary HRP conjugated anti-mouse IgG or anti-mouse IgM (Jackson Immunoresearch) were utilized at a 1:5000 dilution and development was acquired on a Multiskan FC Microplate photometer (ThermoScientific). ELISA for EBC secreted anti-HPV-E6 antibodies were quantified using a commercially available recombinant version of the antibody (Abcam) serially diluted to produce a standard curve.

For TIDE analysis of on-target double-strand breaks, gDNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen). PCR was performed for 35 cycles using PrimeStar MAX (Takara). Primers sequences are listed in Supplementary Table 3 . Resulting amplicons were purified using AMPure XP beads (Beckman Coulter) at a bead-to-sample 1:1 ratio. Purified amplicons were subjected to Sanger sequencing at Hylabs Ltd. Samples were analyzed using the TIDE algorithm (77).

For reverse translated PCR reactions of cassette integration and expression, total RNA was extracted using the RNeasy Mini Kit (Qiagen) with on-column DNAse treatment. Reverse transcription was performed using RevertAid (ThermoFisher) and oligodT primers. PCR on the resulting complementary DNA was performed for 35 cycles using PrimeStar MAX DNA Polymerase (Takara). Following each PCR, resulting amplicons were analyzed agarose gel electrophoresis using a standard DNA ladder, and imaged with the E-Gel Power Snap Electrophoresis System (ThermoFisher).

Statistical analyses and data visualization were performed on distinct biological samples using GraphPad Prism version 10. Two-tailed unpaired t-tests were performed for comparisons between two groups. For multiple group comparisons, p-values were adjusted for multiplicity using appropriate correction. Following ANOVA, Tukey’s, Šídák’s, or Dunnett’s post hoc tests were applied based on the comparisons indicated in each figure. Each figure legend specifies the statistical test used, the measure of tendency, and the type of error bars. Final visualizations were created using Adobe Illustrator version 29.4.