The Strong Prognostic Effect of Concurrent Deletions of IKZF1 and PAX5, CDKN2A, CDKN2B or PAR1 in the Absence of ERG Deletions (IKZF1^plus) in Pediatric Acute Lymphoblastic Leukemia Strongly Depends on Minimal Residual Disease Burden after Induction Treatment

Technical advances in the field of genomic analyses have stimulated a large number of discovery studies on etiological and clinical endpoints in acute lymphoblastic leukemia (ALL) to provide new avenues for preventive strategies, diagnostics and treatment. Recently, deletion of IKZF1 (IKZF1del) was described as a poor prognostic factor in pediatric ALL (Mullighan CG et al., 2012). In our trial AIEOP-BFM ALL 2000 patients with IKZF1del had a lower 5-year event-free survival (EFS; 0.69±0.05 vs. 0.85±0.01; P<0.0001) compared to those without, mainly due to a higher cumulative incidence of relapses (CIR; 0.21±0.04 vs. 0.10±0.01; P=0.001) (Dörge P et al., 2013). IKZF1delwas an independent prognostic factor in this modern protocol. However, on its own, and even in association with minimal residual disease (MRD) levels, its relatively mild prognostic strength limited incorporation in clinical stratification strategies so far.

The present study was undertaken to evaluate the prognostic effect of concurrent recurrent genetic aberrations in association with IKZF1del to further refine a high-risk genetic signature for pediatric ALL treated on AIEOP-BFM protocols. For this purpose, we studied a cohort of 1100 German patients with B-cell precursor ALL treated on AIEOP-BFM ALL 2000 with available characterization of genetic aberrations at initial diagnosis by MLPA (MLPA SALSA kit P335; MRC-Holland) and a PCR assay for detection of ERG deletions (ERGdel; Zaliova M et al., 2014). Validation of results was conducted by use of an independent cohort of 417 Italian patients from the same trial. BCR/ABL1-positive patients were excluded from outcome analysis.

When single marker analyses were conducted, IKZF1del was the strongest determinator of outcome in the discovery as well as the validation cohort. When IKZF1del was analyzed in combination with PAX5, CDKN2A, CDKN2B, and PAR1 deletions, patients with an additional deletion to that of IKZF1 had the worst EFS and highest CIR in absence of ERGdel and, consequently, were grouped as IKZF1plus (definition: presence of IKZF1del and at least an additional deletion in PAX5, CDKN2A, CDKN2B or PAR1 in the absence of ERGdel). This group comprised 6% of B-lineage ALL patients and had a very poor clinical outcome: 5y-EFS 50%±0.06 compared to 86%±0.01 in IKZF1plus-negatives (p<0.0001); 5y-CIR 45%±0.06 compared to 11%±0.01 (p<0.0001) and was an independent prognostic factor in multivariate analyses including MRD, slow early response, prednisone response, ETV6/RUNX1 status, and WBC (≥100.000/µl) (hazard ratio for an event: 3.39; 95% CI 2.09 – 5.48; p<0.0001). Surprisingly, stratified analysis by MRD demonstrated that the effect of IKZF1plus was restricted to those patients still carrying MRD loads of at least 10E-4 after induction treatment: standard-risk group 5y-EFS 94%±0.06 compared to 38%±0.09 in intermediate-risk and 27%±0.13 in high-risk patients (p<0.0001); standard-risk group 5y-CIR 6%±0.10 compared to 62%±0.10 in intermediate risk and 55%±0.17 in high-risk patients (p<0.0001). Hierarchical clustering of gene expression profiles of IKZF1plus patients did not demonstrate association specific to the different MRD risk groups. Similarly, analysis of whole exome and transcriptome data from material at initial diagnosis in a very restricted number of patients could not demonstrate insights into the observed differences.

Newly identified very poor prognostic ALL subgroup – termed IKZF1plus – represented worse outcome compared to that of IKZF1del or others as a sole marker. The differential prognostic effect of IKZF1plus at different MRD levels without currently discernable molecular explanations suggest a quantitative mechanism with higher levels of leukemic cell burden during the early treatment phases predisposing to evolution of a treatment resistant leukemic clone under exposure towards genotoxic chemotherapeutic agents. Potential explanations for these differences in treatment response may be undetected sub-clones already present at diagnosis or underlying germline genetic variation associated with treatment response. Further analyses will be required to better understand this phenomenon. However, due to its strength, the definition of IKZF1plus is likely to aid in the practical implementation of newly detected markers for risk stratification in childhood ALL in a clinical setting.

Support: EU FP7 (ENCCA, TRANSCALL), IGA MZ NT/13170-4