Outcome modeling with CRLF2, IKZF1, JAK, and minimal residual disease in pediatric acute lymphoblastic leukemia: a Children's Oncology Group study

Abstract

As controversy exists regarding the prognostic significance of genomic rearrangements of CRLF2 in pediatric B-precursor acute lymphoblastic leukemia (ALL) classified as standard/intermediate-risk (SR) or high-risk (HR), we assessed the prognostic significance of CRLF2 mRNA expression, CRLF2 genomic lesions (IGH@-CRLF2, P2RY8-CRLF2, CRLF2 F232C), deletion/mutation in genes frequently associated with high CRLF2 expression (IKZF1, JAK, IL7R), and minimal residual disease (MRD) in 1061 pediatric ALL patients (499 HR and 562 SR) on COG Trials P9905/P9906. Whereas very high CRLF2 expression was found in 17.5% of cases, only 51.4% of high CRLF2 expressors had CRLF2 genomic lesions. The mechanism underlying elevated CRLF2 expression in cases lacking known genomic lesions remains to be determined. All CRLF2 genomic lesions and virtually all JAK mutations were found in high CRLF2 expressors, whereas IKZF1 deletions/mutations were distributed across the full cohort. In multivariate analyses, NCI risk group, MRD, high CRLF2 expression, and IKZF1 lesions were associated with relapse-free survival. Within HR ALL, only MRD and CRLF2 expression predicted a poorer relapse-free survival; no difference was seen between cases with or without CRLF2 genomic lesions. Thus, high CRLF2 expression is associated with a very poor outcome in high-risk, but not standard-risk, ALL. This study is registered at www.clinicaltrials.gov as NCT00005596 and NCT00005603.

Figure 1

Expression profiles of CRLF2 by NCI risk group. CRLF2 expression by NCI risk group is shown along with the corresponding IKZF1 deletions and mutations, CRLF2 lesions and mutations, and JAK mutations. (A-B) All samples for NCI HR and SR, respectively, with the red box indicating the cases with “high” CRLF2 expression. The threshold for high was set at the point where the CRLF2 lesion with the lowest expression (ΔC_t = 8.08) was found. (C-D) Enlarged regions within the boxes for NCI HR and NCI SR, respectively, to help illustrate the patterns of lesions and mutations within each. Black dot represents CRLF2 intensity; yellow circle, IGH@-CRLF2; red square, P2RY8-CRLF2; blue diamond, JAK mutations; yellow triangle, IKZF1 deletions or mutations; and green circle, CRLF2 F232C mutation.

Figure 2

Variables independently predictive of outcome in full cohort. After stepwise multivariate analysis, 4 variables were independently correlated with outcome among the cases tested: (A) NCI risk group, (B) MRD, (C) high CRLF2 expression, and (D) dIKZF1 status. Their correlations with outcome are depicted in Kaplan-Meier survival plots, stratified by NCI risk group. Black lines indicate NCI SR; and red lines, NCI HR. The hazard ratios and log-rank P values were calculated separately for the 2 NCI risk groups and are shown in black (NCI SR) and red (NCI HR) at the bottom of each plot.

Figure 3

Variables independently predictive of outcome in NCI HR. After stepwise multivariate analysis, only 2 variables were independently correlated with outcome among the NCI HR cases. Their univariate impacts are shown: (A) MRD and (B) high CRLF2 expression. (C) The performance of the final model of these 2 variables. (D) The model split by its 4 categories to illustrate the relative impact of the variables.

Figure 4

Impact of CRLF2 lesions on outcome among patients with high CRLF2 expression. The Kaplan-Meier survival curves are shown for NCI HR (red) and SR (black) patients according to their CRLF2 lesion status. Cases without CRLF2 lesions have equivalently poor outcome to those with lesions, regardless of NCI risk group.

Figure 5

Relationships of CRLF2 expression range and outcome in NCI HR. The expression of CRLF2 was partitioned into ranges, and the survival curves of each were plotted. Given the small number of very high expressing cases, patients with a ΔC_t ≤ 3.0 were grouped together. Likewise, patients with a ΔC_t > 9 were grouped together because there was no distinguishable difference in outcome between any of the ranges in these low expressing cases.