Inactivation of CDKN2A in Childhood Acute Lymphoblastic Leukemia (ALL) Occurs Principally by Deletion and Is Strongly Correlated with Cytogenetic Subgroups.

Genetic alterations play a key role in the leukemogenesis of childhood ALL. Inactivation of CDKN2A (p16), a tumour suppressor gene located at 9p21, can occur by deletion, methylation or mutation. Published reports are inconsistent in terms of incidence and mode of inactivation. We report a comprehensive analysis of CDKN2A inactivation in 1230 diagnostic and 101 relapse samples, including 46 matched diagnostic and relapse pairs, from 1285 children with ALL. Using data from cytogenetics (CC) (n=1088), FISH (n=1209), SNP arrays (SNPA) (n=106), CGH arrays (aCGH) (n=106), dHPLC (n=48) and methylation specific-PCR (MSP) (n=96) we have assessed the mode and frequency of CDKN2A inactivation. Mutation or methylation of CDKN2A was rare occurring in 1 patient each (2% and 1% respectively). In contrast, CDKN2A deletion was highly prevalent. The frequencies of deletion detected by the different methodologies were: CC 166 (15%), FISH 335 (28%), SNPA 17 (16%) and aCGH 35 (33%). The proportion of biallelic deletions also varied: CC 15 (9%), FISH 174 (52%) and aCGH 15 (65%). This variation was directly related to the resolution of each technique with a high degree of concordance across samples investigated by >1 method. Analysis of 50 deletions by aCGH showed that the size of the deletion ranged from 0.03Mb to 39.1Mb with a mean of 14.8Mb. Furthermore, analysis of 15 biallelic deletions demonstrated that they comprised one large deletion (mean size 23.3Mb) and a second much smaller deletion (mean size 1.4Mb). In addition, SNPA revealed copy number neutral LOH in 8 (8%) cases, but only once in association with a CDKN2A mutation. At diagnosis CDKN2A inactivation by any method was noted in 329 (27%) patients which was not different from that observed at relapse [25 (25%)]. However, the frequencies of CDKN2A inactivation and biallelic deletion were significantly greater in T-ALL compared with B cell precursor (BCP) ALL: 135/269 (50%) v 190/918 (21%) (p<0.001) and 83/135 (61%) v 82/190 (43%) (p=0.001), respectively. Within BCP-ALL, older patients (10+ yrs) were more likely to have CDKN2A inactivation compared to younger patients (<10 yrs) whereas the reverse was true in T-ALL: 52/196 (27%) v 138/722 (19%) (p=0.023) and 53/126 (42%) v 82/143 (57%) (p=0.012). Among 46 matched samples CDKN2A inactivation was retained (n=8), lost (n=3) or gained (n=6). The frequency of CDKN2A inactivation was strongly correlated with cytogenetics. Lower frequencies were observed among high hyperdiploid [31/302 (10%) p<0.001] and ETV6-RUNX1 patients [36/236 (15%) p=0.02] with higher frequencies among those with t(9;22) [11/19 (58%) p<0.001], t(1;19) [10/25 (40%) p=0.019] and other abnormalities [92/226 (41%) p<0.001]. In conclusion, we have confirmed the importance of CDKN2A inactivation in childhood ALL and demonstrated that by far the most prevalent method of inactivation is deletion. While it is clear that loss of CDKN2A acts as a cooperating mutation in childhood ALL it is strongly correlated with age, phenotype and genotype. The observation that it is negatively correlated with good risk cytogenetic subgroups may explain why it has been inconsistently associated with a poor outcome. The discovery of copy number neutral LOH at 9p with no evidence of CDKN2A inactivation suggests the presence of another tumour suppressor gene or oncogene in this region.