AUTHORS
Gabriele Casirati, M.D., Ph.D., Enrico Drago, MD2, Moises Chavez-Navarro, PhD3, Pietro Genovese, PhD4, Danilo Pellin, PhD5
AFFILIATIONS
1 Boston Children's Hospital, Boston, United States, Harvard Medical School, Boston, United States
2 Boston Children's Hospital, Boston, United States, IRCCS Istituto Giannina Gaslini, Genova, Italy
3 Boston Children's Hospital, Boston, United States
4 Boston Children's Hospital, Boston, United States, Harvard Medical School, Boston, United States
5 Boston Children's Hospital, Boston, United States, Harvard Medical School, Boston, United States
ABSTRACT
Epitope editing enables selective immunotherapeutic targeting of difficult-to-treat hematologic malignancies by introducing single amino-acid substitutions that preserve the physiological function of hematopoietic antigens while abrogating recognition by therapeutic antibodies or CAR-T cells. This strategy is particularly relevant for acute myeloid leukemia (AML) targets such as FLT3, CD123, CD33, and KIT, for which expression on healthy hematopoietic cells creates substantial on-target/off-tumor toxicity and limits clinical translation. By engineering these epitopes in hematopoietic stem and progenitor cells (HSPCs), normal hematopoiesis can be rendered resistant to antigen-directed therapy while malignant cells remain susceptible.
Because epitope editing requires precise nucleotide substitutions rather than gene disruption, it is well suited to nuclease-free CRISPR technologies such as base editing, which achieve programmable nucleotide conversion without dsDNA breaks or donor templates. Adenine base editors (ABEs), which couple an evolved TadA deaminase to Cas9n, can install naturally occurring or rationally designed epitope-shielding variants. However, ABE activity window can expose neighboring adenines to undesired bystander editing, potentially altering coding sequence, splice motifs, or cis-regulatory elements. This constraint is especially important for HSC engineering, where unintended sequence changes could impair repopulating capacity or potentially result in malignant transformation.
To overcome this limitation, we developed a human cell-based evolution and profiling platform to engineer TadA variants with enhanced discrimination among adenines within the same protospacer. Central to this platform is BeTag, a programmable live-cell reporter that links single-nucleotide editing outcomes to selectable phenotypes in human cells. Unlike reporters constrained to a fixed protospacer, BeTag can be programmed to model user-defined therapeutic targets, including epitope-editing loci and disease-causing variants. This enables pooled screening of deaminase libraries in the cellular context in which therapeutic editors will be deployed.
We screened TadA mutational libraries against epitope-editing contexts and recovered variants with altered positional and sequence-context preferences. Revalidation revealed substantial functional plasticity within engineered TadAs, with preservation of desired editing with concomitant suppression of bystander activity. At the FLT3 N399G epitope-disrupting site, selected editors increased on-target precision from 32.4% with ABE8e to >82.5% and maintained 87% of parental efficiency. We then tested translational concordance in primary human CD34+ HSPCs. Across 24 candidates, editing precision profiles closely matched those observed in screening cells (R=0.87, p<0.0001), supporting the relevance of this discovery system for hematopoietic engineering.
To define generalizable rules of TadA specificity, we profiled selected candidates in human cells using matched gRNA:target libraries that directly measure editing across >104 surrogate genomic targets. We assembled a deaminase atlas of >300 variants, including publicly available editors and newly discovered deaminases. This atlas revealed that ABE outcomes are governed by the interaction of spacer position, immediate neighboring motifs, and broader protospacer architecture, enabling prediction of editors for complex multi-adenine targets.
We applied this framework to nominate precision ABEs for AML-targeted epitope editing, including FLT3 and CD123, to engineer antigen-modified HSPCs, as well as minimizing potential immunogenicity of bystander codon changes. Bespoke variants enabled near-perfect single-base editing at FLT3, reducing bystander edits by 91%, and redistributed editing activity at CD123, selectively suppressing A8 deamination by 57%.
Together, these findings establish a precision ABE-design framework in which the deaminase is matched to the therapeutic sequence rather than treating a single high-activity editor as universally optimal. By integrating directed evolution, sequence-resolved profiling, predictive modeling, and primary HSPC validation, this work provides a scalable strategy to expand the therapeutic scope and genomic safety of base editing for hematopoietic cell engineering, epitope editing, and gene correction.
DISCLOSURES
Gabriele Casirati, M.D., Ph.D.: No relevant conflicts of interest to declare.