Genomic Analysis of Adult B-ALL Identifies Changes in Copy Number Profile at Relapse and IKZF1/CDKN2A Co-Deletion at Diagnosis As a Marker of Shorter Survival

B lymphoblastic leukemia (B-ALL) in adults has a higher risk of relapse and lower long-term survival than pediatric B-ALL, but data regarding prognostic biomarkers are much more limited for adult patients.  Microarray-based genome-wide profiling studies in pediatric B-ALL patients have revealed recurrent abnormalities in B-cell development and cell cycle regulation.  IKZF1 alterations convey a negative prognostic impact in pediatric B-ALL, but their significance is not well characterized in adult B-ALL.  CDKN2A alterations have been associated with a poorer prognosis in adult Ph+ ALL, possibly by mediating resistance to targeted therapy.  The copy number landscape of adult B-ALL has not been fully assessed and is likely distinct from its pediatric counterpart.  In addition, the copy number changes in relapsed adult B-ALL, including comparison with initial diagnostic samples, have not been characterized.  We identified 70 adult B-ALL patients (median age 45 years, range 18-83) from 1998-2013 at three institutions.  DNA was isolated from formalin-fixed, paraffin-embedded (FFPE) diagnostic bone marrow clots, as well as relapse samples when available, and assessed with the OncoScan FFPE Express genome-wide single nucleotide polymorphism (SNP) assay (Affymetrix).  Copy number alteration (CNA) and loss of heterozygosity (LOH) analysis was performed using Nexus Software V7 (Biodiscovery) and in-house coding.  For cases with adequate DNA, IKZF1 sequencing was also performed.  Clinical data included age, gender, hematologic laboratory values at presentation, CSF involvement, receipt of allogeneic transplant, cytogenetic profile, presence of t(9;22), event-free survival (EFS), and overall survival (OS).   Recurrent deletions in the diagnostic samples were noted at several loci, including CDKN2A (48.6%), IKZF1 (40%), ICR (38.6%), PAX5 (24.2%), BTG1 (17.1%), ETV6 (15.7%), BTLA (14.3%), RB1 (5.7%), EBF1 (5.7%), LEF1 (2.9%), and TCF3 (2.9%).  Recurrent gains were identified at the following loci: ERG (30%), ETS2 (21.4%), MYB (20%), UBASH3B (20%), PTEN (20%), PRKCH (18.6%), CDK6 (17.1%), and ETV6 (12.9%).  In addition, LOH was observed in all loci with recurrent CNAs named above.  27 samples obtained at the time of relapse were screened for the recurrent CNAs identified in the diagnostic samples.  The cumulative incidence of CNAs at these genes was lower in the relapse cohort than the initial diagnostic cohort, although there was a higher frequency of CNAs at PAX5, BTLA, ETS2, RB1, LEF1, PTEN, and TCF3 in the relapse samples.  16 patients had samples available at both the time of diagnosis and time of relapse.  The average number of CNAs at diagnosis and relapse was the same, but analysis of paired samples revealed frequent changes at the previously defined loci of CNAs.  There were a number of new CNAs at these loci in relapse samples, but there was also a similar incidence of reversion to the copy neutral state in loci that previously had CNAs in the diagnostic samples.  Sequencing data was available on 26 diagnostic samples for the IKZF1 gene.  Only two samples had mutations corresponding to amino acid changes.  One of these two had a heterozygous deletion at IKZF1, indicating that the mutated isoform was the only one expressed.  11 other samples had a SNP without a corresponding amino acid change.  14 relapse samples were sequenced, two of which harbored a point mutation in IKZF1.  Neither of these samples had a CNA at this locus.  10 of the remaining 12 relapse samples had a SNP without a corresponding amino acid change.  When correlated with outcomes, no individual CNA heralded a significant prognostic impact in the entire cohort or in subgroup analyses stratified by presence of t(9;22) for either EFS or OS.  However, the combination of both CDK2NA and IKZF1 deletions (26%) correlated with a significantly worse OS than having only one or neither of these deletions (both vs. CDKN2A only:  p=0.028, both vs. IKZF1 only:  p=0.027, both vs. neither deleted:  p=0.048).  Age was the only other covariate significant in univariate analyses for OS, yet IKZF1/CDKN2A co-deletion remained significant in multivariate analysis adjusting for age.  Adult B-ALL demonstrates recurrent copy number changes, and the pattern of CNAs is often different at relapse than at diagnosis indicating clonal evolution.  When identified in diagnostic samples, co-deletion of CDKN2A/IKZF1 is a negative prognostic marker in adult B-ALL.