Comprehensive Characterization of Subtype-Specific V(D)J Recombination Features of B-Cell Acute Lymphoblastic Leukemia

Introduction: B cell acute lymphoblastic leukemia (B-ALL) is a highly diverse disease in both phenotype and prognosis. Its heterogeneity lies in its oncogenesis, where B cell differentiation in bone marrow is disrupted by various subtype-defining genetic lesions. For example, KMT2A-rearranged (KMT2A-r) leukemia typically arrests at the early pro-B stage, whereas leukemia driven by BCL2 and/or MYC translocations (BCL2/MYC) emerges from more differentiated B cells. In parallel to differentiation, somatic recombination of immunoglobulin (Ig) V(D)J gene segments drives assembly of unique B cell receptors (BCR) in distinct stages. Thus, thorough study of V(D)J recombination features of B-ALL can assist in profiling genetic subtypes, disease clonal structures, and blocked B cell stages in leukemia, as well as advancing our understanding of normal BCR assembly.

Methods: We analyzed RNA-Seq data of 2,955 B-ALL samples, along with 24 normal B lineage control samples, to identify their V(D)J repertoires using the TRUST4 algorithm (Nat Methods, 2021). The B-ALL cohort was annotated into 26 subtypes in our previous work (Haematologica, 2023). To identify unique V, D, J and junction combinations of Ig heavy chain (IGH) representing clonally expanded leukemic cells (disease clones), we used a frequency-based linear thresholding model with fine-tuned parameters. To identify ongoing V gene replacement evolved from the same clone (different V; same DJ), we applied strict thresholds of independent assembly probability. To study the connections of KMT2A-r and MEF2D-r driver lesions to V gene usage in leukemic BCR, we performed in vivo B cell differentiation assays in mice.

Results: IGH recombination marks the earliest B cell stages and often occurs alongside B-ALL oncogenesis. We found that 76.4% (n=2257/2955) of B-ALL samples had at least 1 IGH disease clone. The BCL2/MYC subtype showed the highest monoclonality (72.4%), in line with late onset of B cell transformation. The PAX5-altered (PAX5alt) subtype, exhibiting pro/pre-B cell characteristics, had the highest ratio of multiclonal populations (41.4%). In contrast, the pro-B-enriched Low Hypodiploid subtype showed the lowest ratio (47.6%) of BCR disease clones. Subtypes such as TCF3::PBX1, NUTM1, ETV6::RUNX1-like, HLF, and MEF2D-r, all associated with the pre-B stage, showed the most active ongoing V gene replacement (>70%).

Due to its ubiquity, we profiled IGH V gene distributions of B-ALL versus the normal B cells to identify subtype-specific V gene selection. We observed biased proximal V gene (V6-1) use in the PAX5alt and PAX5 P80R subtypes defined by PAX5 alterations. PAX5, a key transcription factor in B cell development, represses WAPL expression in pro-B and pre-B cells, allowing selection from the full range of V genes via extended loop extrusion. PAX5 alterations, as in these 2 subtypes, lead to high WAPL expression, resulting in proximal V gene bias. Additionally, we saw even stronger V6-1 use in KMT2A-r and MEF2D-r subtypes, neither fully characterized. Through in vivo assays, we confirmed that both rearrangements can induce a proximal V gene bias. In human B-ALL and mouse B cells carrying MEF2D fusions, we saw significant WAPL overexpression, explaining V6-1’s enriched use in MEF2D-r B-ALL. However, no WAPL expression changes were seen in B cells carrying KMT2A fusions, indicating an unknown process driving V6-1 use in KMT2A-r B-ALL.

Further, we analyzed IGH D and J gene features across subtypes. For several, namely PAX5 P80R, PAX5alt, KMT2A-r, and MEF2D-r, we found significantly enriched near-proximal D2-2 selection. In contrast, distal most J6 gene use was significantly enriched in PAX5 P80R, PAX5alt, and MEF2D-r subtypes, but not KMT2A-r. Despite the V, D, and J gene biases’ enrichment in the same subtypes, they were not correlated in the same IGH disease clones, suggesting different mechanisms underlie each bias.

Conclusion: In summary, we thoroughly profiled V(D)J recombination features in 26 B-ALL subtypes using the largest B-ALL cohort to date. Our analysis revealed significant V, D, and J use biases in specific subtypes. We validated the involvement of KMT2A-r and MEF2D-r in aberrant V gene selection, furthering our comprehension of (ab)normal BCR development. Our findings provide new research directions to discern the regulatory networks governing V(D)J recombination and how biased selection may contribute to leukemogenesis.