A set of genes that regulate cell proliferation predicts treatment outcome in childhood acute lymphoblastic leukemia

Abstract

To identify novel predictors of outcome in childhood acute lymphoblastic leukemia (ALL), we analyzed gene expression in the leukemic cells of 187 children with newly diagnosed ALL and compared the findings with minimal residual disease (MRD) results obtained on day 19 of remission induction treatment. Genes that showed a significant relationship to MRD were then tested for their capacity to predict leukemic relapse in an independent cohort of 99 patients. We identified 674 probe sets that were associated with MRD on day 19 (P < .006); 40 of the identified genes predicted relapse (P < .03). Among these, 14 showed independent prognostic significance after adjustment for age, leukocyte count at diagnosis, and genetic subtype. More than half of the 40 genes and nearly all of the 14 genes were functionally related, as indicated by their roles in the regulation of cell proliferation. Underexpression of genes promoting cell proliferation was associated with resistance to chemotherapy. The biologic processes regulated by the genes we identified appear to be key determinants of the early cytoreductive response to remission induction therapy and subsequent clinical outcome in childhood ALL. Incorporation of the expression levels of these genes into existing strategies of risk classification could improve clinical management.

Figure 1

Genes significantly associated with minimal residual disease (MRD) on day 19 and with leukemic relapse. Among the 674 probe sets that were significantly associated with MRD on day 19 in 187 patients (P < .006), 44 were significantly associated with leukemic relapse (P < .03) in an independent cohort of 99 patients, regardless of whether they were analyzed as a continuous or categoric variable. Bars indicate the fold increase/decrease in hazard per unit increase in expression of each gene, calculated with the proportional hazards regression model. Black bars correspond to genes that were independent predictors of outcome after adjustment for known prognostic features of childhood acute lymphoblastic leukemia (ALL).

Figure 2

Cumulative incidence of leukemic relapse according to the expression of CKS1B, FLJ20105, BUB3, and H2AFZ, as detected by probe sets 219650_at, 201897_s_at, 201457_x_at, and 200853_at on the Affymetrix 133A gene chip, respectively. Solid lines represent patients with low gene expression and dashed lines those with expression at or above the median value; 10-year cumulative incidence of relapse (± SD) is indicated. In a multivariate analysis, each of these genes was a stronger predictor of leukemic relapse than any competing covariate included in the model.

Figure 3

Clustering by gene-expression profiling and principal component analysis of 286 diagnostic samples of childhood ALL in relation to risk of relapse. Signals represent the 15 probe sets (14 genes) that were associated with MRD on day 19, and independently predicted outcome. (A,B) Heat map and principal component analysis. Clusters are indicated in blue and red. (C) Cumulative incidence of relapse among the 99 patients enrolled on Total XIII belonging to each cluster (P < .001).

Figure 4

Correlation between expression of CASP8AP2 and that of 3 of the 14 independent predictors of leukemic relapse. Plots show signal intensities as measured by the Affymetrix 133A gene chip. Spearman correlation coefficients are shown.

Figure 5

Cumulative incidence of leukemic relapse according to the expression of CKS1B, FLJ20105, and H2AFZ among 66 patients with low or intermediate levels of CASP8AP2 expression. Solid lines represent patients with low gene expression and dashed lines those with expression above the median value; 10-year cumulative incidence of relapse (± SD) is indicated.

Figure 6

Biologic processes influenced by the 40 gene products associated with MRD on day 19 and with leukemic relapse