Abstract

Accurate risk stratification constitutes the fundamental paradigm of treatment in acute lymphoblastic leukemia (ALL), allowing the intensity of therapy to be tailored to the patient’s risk of relapse. We have recently identified 2 known (E2A-PBX1, MLL) and 6 novel cluster groups with varying prognostic significance within a high risk ALL cohort (COG P9906) using gene expression profiling. Key genes predicted membership in these novel cluster groups, including MUC4, GPR110, IGJ (poorest outcome cluster) and CENTG2 and PTPRM (most favorable outcome cluster). The poor outcome cluster had a 4-year relapse free survival (RFS) of 20.9%, while the favorable outcome group had an RFS of 94.7% -- both significantly different from the RFS of 61% for the overall cohort. In the current study, we perform quantitative RT- PCR for expression of these genes (MUC4, GPR110, IGJ, CENTG2 and PTPRM) in tandem with gene expression microarray analysis on a new cohort (n=85) of high risk ALL patients enrolled in COG 0232 to

  • validate the findings of our prior microarray classification, and

  • identify a manageable subset of discriminatory genes amenable to RT-PCR analysis, thus permitting the detection of these prognostically significant groups in a clinical setting.

Evaluation of PBX1 expression by RT-PCR was included for further validation. RT-PCR assays were designed to simulate as closely as possible testing conditions consistent with a clinical assay; thus, RT-PCR was performed using 10 ng of RNA from each sample, run in duplicate in a quantitative TaqMan assay, and normalized to EEF2 expression. The expression of the analyzed genes by RT-PCR showed excellent correlation with the data generated by expression microarray analysis; using receiver operator curve (ROC) analysis and employing the microarray clusters as a gold standard, the analyzed genes produced areas under the curve (AUC) ranging from 0.88 to 1.0 (a finding characteristic of useful clinical tests) and highly desirable sensitivity and specificity for the identification of poor and favorable outcome groups (see table). Cluster analysis using only RT-PCR expression of these six genes reproduced the significant cluster assignments determined by the microarray analysis. RT-PCR expression of the genes showed a wide dynamic range, a valuable feature for a potential clinical assay. Moreover, using a decision tree approach, we established a step-wise algorithm using the RT-PCR data that permits accurate classification of patients into the poor, favorable and neutral outcome groups. Finally, PBX1 was significantly expressed exclusively in patients found to have E2A-PBX1 by cytogenetic or standard PCR analyses. These results confirm our prior microarray findings and demonstrate that the previously-described prognostic subgroups within high-risk pediatric ALL may be identified in a real-time, clinical setting with a robust quantitative RT-PCR assay for expression of as few as two or three genes. Identification of these prognostic subgroups may improve risk classification in ALL, enhance therapeutic targeting, and thereby improve overall outcome.

Gene Prognostic group Area under ROC curve p value for ROC curve Sensitivity (%) Specificity (%) 
MUC4 Poor 0.88 <0.0001 87.5 92.2 
GPR110 Poor 0.89 <0.0001 100 80.5 
IGJ Poor 0.89 <0.0002 87.5 87 
CENTG2 Good 0.97 <0.0001 100 83.3 
PTPRM Good 0.97 <0.0001 85.7 100 
PBX1 E2A-PBX 1.00 <0.0001 100 100 
Gene Prognostic group Area under ROC curve p value for ROC curve Sensitivity (%) Specificity (%) 
MUC4 Poor 0.88 <0.0001 87.5 92.2 
GPR110 Poor 0.89 <0.0001 100 80.5 
IGJ Poor 0.89 <0.0002 87.5 87 
CENTG2 Good 0.97 <0.0001 100 83.3 
PTPRM Good 0.97 <0.0001 85.7 100 
PBX1 E2A-PBX 1.00 <0.0001 100 100 

Disclosures: No relevant conflicts of interest to declare.

Author notes

Corresponding author