Immunophenotypic Profiling of Childhood B-Cell Precursor Acute Lymphoblastic Helps Identifying Genetic Subtypes, Including Recently Identified Ones

BackgroundB-cell precursor acute lymphoblastic leukemia (B-ALL) comprises distinct entities characterized by recurring genetic alterations. These classifying alterations are associated with gene expression profiles, reflecting the specific biology of each genetic subgroup. In patients with none of these classifying alterations, referred as "B Other", RNAseq recently led to the identification of a huge number of in frame fusions which are likely to define novel genetic subgroups. Leukemic blasts often show aberrant antigen expression, giving rise to leukemia-associated immunophenotypic profiles (LAIP) distinct from their normal immature B-lineage counterparts. These variations in protein expression are likely to reflect the specific biology of B-ALL. We evaluated how LAIP were associated with the various classifying genetic lesions, including recently identified ones.

Materials and MethodsA cohort of 299 childhood B-ALL was studied prospectively. Genetic typing by cytogenetics, MLPA and RT-PCR identified high hyperdiploidy (Heh, n=77), ETV6-RUNX1 (n=81), TCF3-PBX1 (n=21), TCF3-HLF (n=3), MLL rearrangement (n=25), BCR-ABL/Phie-like fusions (n=10), ERGdel (n=8), TSLPR (n=8), iAMP21 (n=6), Hypodiploidy (n=6). The remaining patients were considered "B-Other" (n=54). Immunophenotyping was performed by 6-8-colors flow cytometry with a panel of 29 antibodies targeting surface proteins routinely used for ALL diagnosis and MRD assessment. The log of Ratio Fluorescence Intensity (RFI) of B-ALL/normal B precursor cells was calculated for each marker. RNAseq was performed in 18 B-Other and 8 BCR-ABL/Phie-like. Fusions were detected using TopHat2. Samples were classified using hierarchical clustering and/or principal component analyses (R software).

ResultsUnsupervised clustering identified 6 groups with shared pattern of LAIP. Four groups were associated with known classifying lesions: ETV6-RUNX1 (of which 87% co-clustered), Heh (82%), TCF3-PBX1 (100%) and MLL (96%). ALL that did not cluster according to their genetic group mostly clustered (23/26; 89%) in two groups, hereafter named "B Other-1" and "B Other-2", which mainly contained ALL lacking any previously known cytogenetic lesion. Interestingly, 3 of 5 ETV6-RUNX1 with IKZF1del clustered in the "B Other 2", suggesting that misclassifications may not be due to poor LAIP specificity but rather reveal distinct biological features in some ALL.

Although comprising too few cases to define specific groups, 8/10 B-ALL with BCR-ABL/Phie-like, 3/3 TCF3-HLF, and 8/8 TSLPR co-clustered in "B Other 2", and 7/8 ERGdel in "B Other 1". No clear clustering was observed for Hypodiploid and iAMP21. B-Other ALL were allocated either to oncogenic groups, e.g. TCF3-PBX1 (n=7) or MLL (n=7), or to the "B Other" groups (n=35). Only one B Other, harboring an IGH-IL3 fusion, co-clustered with BCR-ABL/Phie-like ALL. RNAseq of 18 B-Other identified no additional Phie-like fusion and none of these B-Other co-clustered with a group of 8 controls BCR-ABL/Phie-like. This makes unlikely that some Phie-like cases remained unidentified. However, other fusions involving MEF2D (n=3), PAX5 (n=5) or ZNF384 (n=2) were evidenced in B Other ALL. Cases with MEF2D or ZNF384 clustered both at the transcriptome and protein level. Using LAIP clustering, MEF2D ALL layed close to those with TCF3 fusions. ZNF384 cases clustered in the MLL group, consistent with a lack of CD10 expression combined to myeloid antigen expression. PAX5 ALL were allocated to the "B Other 2" group with no clear clustering and, more surprisingly, no consistent effect on CD19 expression. Secondary aberrations such as deletions in IKFZ1 or CDKN2A/B genes did not significantly affect clustering.

ConclusionUnsupervised clustering based on surface protein expression allowed to classify most childhood ALL genetic groups, including Phie-like, MEF2D or ZNF384, which are challenging to identify due to the diversity of the underlying gene fusions. LAIP can thus rapidly orientate diagnosis, in particular for Phie-like which are candidates to targeted therapies. Interestingly, B-Other that clustered in known genetic groups may present biologically equivalent alterations undetected in our routine workup. Further studies will allow to assess whether cases co-clustering with ETV6-RUNX1 (n=2) or ERGdel (n=6) correspond to the recently described ETV6-RUNX1-like or DUX4 group respectively.