Rapid Identification of BCR/ABL1-like Acute Lymphoblastic Leukemia (ALL) Cases By Quantitative Real Time-PCR (Q-RT-PCR). Generation and Validation of a Predictive Statistical Modell

Introduction: Among B-lineage ALL, BCR/ABL1-like cases have clearly emerged as a subgroup with peculiar features. By gene expression profiling (GEP), they have a transcriptional signature similar to that of BCR/ABL1+ cases. Moreover, these cases often have IKZF1 deletions and CRLF2 deregulation, and harbor several lesions involving tyrosine kinases (TKs) that may be patient specific. The incidence varies with age (from 10% of B-lineage ALL in children, to 27% in adolescents/young adults, with a decrease in adults) and they are usually characterized by an inferior outcome, though risk-directed therapy based on minimal residual disease levels has improved prognosis in childhood. Even if GEP promptly recognizes these patients, this technique is expensive, not readily available and not applicable to all centers.

Aims: To generate a simple and reliable tool, based on a Q-RT-PCR approach, aimed at the rapid identification ofBCR/ABL1-like cases, and to build a statistical predictive model. Finally, to correlate the results obtained with the clinico-biologic features and outcome.

Methods and patients: Through a meta-analysis of previously published GEP data focused on BCR/ABL1-like cases, and on our in house results, 9 commonly overexpressed genes were selected and evaluated by Q-RT-PCR; CRLF2was also included, because of its association with this signature (total=10 genes). Q-RT-PCR was performed using the ABI PRISM 7300 Sequence Detection System and FastStart Universal SYBR Green Master (Rox). Overall, 149 B-lineage ALL cases, negative for known major rearrangements, i.e. BCR/ABL1, ETV6/RUNX1, E2A/PBX1 and ALL1 (B-NEG ALL), were tested. Median age was 28.8 years (range 1-78), 59 were females and 90 males. To generate a predictive model for BCR/ABL1-like identification, one round of cross-validation was performed; the full dataset was divided into two panels (discovery and screening). The discovery panel included 52 cases previously evaluated for GEP analysis and comprised 26 BCR/ABL1-like and 26 B-NEG ALL samples, while the screening panel (n=97) was based on 81 cases including 16 B-NEG ALL samples harboring JAK2, IL7R or CRLF2 mutations, tested as internal control, given their association with a BCR/ABL1-like signature.

Results: The statistical model was built on the discovery panel and was validated, in terms of sensitivity and specificity, on the screening panel. Q-RT-PCR values of the 10 genes were grouped in the discovery panel according to principal component analysis to reduce multicollinearity and a logistic regression model was used to examine the association among the first three principal components (accounting for more than 80% of the total variance) andBCR/ABL1-like cases. Finally, a score was computed and validated in the screening panel. The model had a sensitivity and specificity of 93.8% and 93.8%, respectively, with a positive predictive value of 75%. Within the screening panel, 10/81 (12.3%) and 6/16 JAK2, IL7R or CRLF2 mutated cases (37.5%) were predicted as BCR/ABL1-like (total=16.5%). GEP was performed on 14/16 BCR/ABL1-like samples and confirmed the similarity withBCR/ABL1+ ALL cases by unsupervised and supervised analysis. Molecular screening, comprising the screening ofJAK1/2, CRLF2, IL7R (JAK/STAT pathway) mutations and IKZF1 deletions, confirmed a significant enrichment of the JAK/STAT pathway (48% vs 15%, p=0.001) and of IKZF1 deletions (82 vs 46%, p=0.006) in BCR-ABL1-like cases as opposed to the other B-NEG ALL cases. Finally, BCR-ABL1-like patients showed a significantly higher white blood cells (WBC) count (46x10⁹/l vs 12.8x10⁹/l, p=0.003) and poorer prognosis compared to the other B-NEG ALL patients, in terms of complete remission achievement (70.3 vs 87%, p=0.07), overall survival (41.1% vs 59.1%, p=0.06) and disease-free survival (DFS) (31.1 % vs 46%, p=0.008).

Conclusions: We describe a Q-RT-PCR approach and a statistical model capable of rapidly identifying BCR-ABL1-like cases. We confirm a significant association with JAK/STAT mutations, IKZF1 deletions and poorer outcome. The prompt recognition of BCR-ABL1-like ALL at presentation can: i) identify cases in which additional lesion/s involving a TK - that might sustain the BCR/ABL1-like signature - should be further investigated; ii) drive therapeutic decisions, since these patients might benefit by the addition of targeted therapies, particularly TK inhibitors.