CpG island methylator phenotype classification improves risk assessment in pediatric T-cell acute lymphoblastic leukemia

Abstract

Current intensive treatment of pediatric T-cell acute lymphoblastic leukemia (T-ALL) has substantial side effects, highlighting a need for novel biomarkers to improve risk stratification. Canonical biomarkers, such as genetics and immunophenotype, are largely not used in pediatric T-ALL stratification. This study aimed to validate the prognostic relevance of DNA methylation CpG island methylator phenotype (CIMP) risk stratification in 2 pediatric T-ALL patient cohorts: the Nordic Society of Paediatric Haematology (NOPHO) ALL2008 T-ALL study cohort (n = 192) and the Dutch Childhood Oncology Group (DCOG) ALL-10/ALL-11 validation cohorts (n = 156). Both cohorts revealed that combining CIMP classification at diagnosis with measurable residual disease (MRD) at treatment day 29 (D29) or 33 (D33) significantly improved outcome prediction. The poor prognosis subgroup, characterized by CIMP low/D29 or D33 MRD ≥ 0.1%, had a cumulative incidence of relapse (pCIR5yr) of 29% and 23% and overall survival (pOS5yr) of 59.7% and 65.4%, in NOPHO and DCOG, respectively. Conversely, a good prognosis subgroup was also identified representing CIMP high/D29 or D33 MRD < 0.1% with pCIR5yr of 0% and 3.4% and pOS5yr of 98.2% and 94.8%, in NOPHO and DCOG, respectively. For NOPHO, MRD was also evaluated on D15, and the relapse prediction accuracy of CIMP/D29 MRD (0.79) and CIMP/D15 MRD (0.75) classification was comparable, indicating potential for earlier stratification. The evaluation of the biology behind the CIMP subgroups revealed associations with transcriptome profiles, genomic aberrations, and mitotic history, suggesting distinct routes for leukemia development. In conclusion, integrating MRD assessment with the novel CIMP biomarker has the potential to improve risk stratification in pediatric T-ALL and guide future therapeutic decisions.

Graphical abstract

Figure 1.

CIMP, MRD, WBC, and outcome in T-ALL in the study of NOPHO ALL2008 cohort. (A) Heat map of the 1091 CIMP panel CpGs in n = 192 NOPHO ALL2008 T-ALL samples at diagnosis, classified as CIMP low (n = 76) and CIMP high (n = 116) in CIMP% order. Healthy sorted CD34⁺ and CD3⁺ T cells and lymph node samples are revealed for comparison. (B) Distribution of the outcomes in 172 patients in relation to D29 MRD, WBC, and CIMP methylation %. MRD is log₁₀ transformed, and WBC is zero scaled at 50 and log₁₀ transformed. Yellow lines represent MRD 0.1%, WBC 50, and CIMP 40% cutoffs. (C) Alluvial plot depicting the timeline of patients from CIMP subgroups at diagnosis to the D15 MRD and D29 MRD until outcomes. In panels B and C, only the 172 patients with complete CIMP, WBC, and MRD data are illustrated. NS, not stratified due to IF.

Figure 2.

CIMP, D29/D33 MRD (0.1%), and survival in the study of NOPHO ALL2008 and validation DCOG ALL-10/ALL-11 cohorts. (A-D) pEFS and (E-H) pOS probabilities in pediatric patients with T-ALL stratified by CIMP and MRD. In the NOPHO cohort, samples were grouped by CIMP (n = 116 CIMP high and n = 76 CIMP low), at D29 MRD (<0.1%, n = 128 and ≥0.1%, n = 101), or CIMP and D29 MRD combined subgroups (CIMP low/D29 MRD ≥ 0.1%, n = 38; CIMP low/D29 MRD < 0.1%, n = 35; CIMP high/D29 MRD ≥ 0.1%, n = 42; and CIMP high/D29 MRD < 0.1%, n = 57). In the validation DCOG cohort, CIMP and D33 MRD were combined (CIMP low/D33 MRD ≥ 0.1%, n = 26; CIMP low/D33 MRD < 0.1%, n = 27; CIMP high/D33 MRD ≥ 0.1%, n = 26; and CIMP high/D33 MRD < 0.1%, n = 58).

Figure 2.

CIMP, D29/D33 MRD (0.1%), and survival in the study of NOPHO ALL2008 and validation DCOG ALL-10/ALL-11 cohorts. (A-D) pEFS and (E-H) pOS probabilities in pediatric patients with T-ALL stratified by CIMP and MRD. In the NOPHO cohort, samples were grouped by CIMP (n = 116 CIMP high and n = 76 CIMP low), at D29 MRD (<0.1%, n = 128 and ≥0.1%, n = 101), or CIMP and D29 MRD combined subgroups (CIMP low/D29 MRD ≥ 0.1%, n = 38; CIMP low/D29 MRD < 0.1%, n = 35; CIMP high/D29 MRD ≥ 0.1%, n = 42; and CIMP high/D29 MRD < 0.1%, n = 57). In the validation DCOG cohort, CIMP and D33 MRD were combined (CIMP low/D33 MRD ≥ 0.1%, n = 26; CIMP low/D33 MRD < 0.1%, n = 27; CIMP high/D33 MRD ≥ 0.1%, n = 26; and CIMP high/D33 MRD < 0.1%, n = 58).

Figure 3.

CIMP, MRD 0.1%, and relapse in the study of NOPHO ALL2008 and validation DCOG ALL-10/ALL-11 cohorts. Proportion of outcomes in 172 NOPHO (A) and 137 DCOG (D) pediatric patients with T-ALL; outcomes were relapse (n = 23/n = 18), DCR (n = 12/n = 8), SMN (n = 2/n = 1), and nonresponders (n = NA/n = 1) with CR (n = 135/n = 109) as reference, respectively. Samples were stratified by CIMP and MRD at D29/D33 MRD combined subgroups and (B,E) corresponding risk of relapse tables. Forest plots reveal the accuracy (estimated as [true positives + true negatives]/total) of D15 MRD and D29/D33 MRD above 0.1/0.01% cutoffs and CIMP-low status, alone or in combination, for differentiating relapse from CR in (C) the NOPHO and (F) the DCOG cohorts. NA, not available.

Figure 4.

Genome-wide multiomics biology phenotype of CIMP subgroups in the NOPHO ALL2008 cohort. Unsupervised principal component analysis of (A) genome-wide DNA methylation (740 067 CpGs, n = 128 EPICv.1-analyzed T-ALL), (B) genomic alterations (294 unique CNVs, n = 128 T-ALL), and (C) whole-genome RNA expression (19 043 protein-coding genes, n = 108 T-ALL). (D) Average β-values of the array CpGs (740 067 CpGs, n = 128 EPIC-analyzed T-ALL) separated into CIMP subgroups and controls. (E) Average number of CNVs in CIMP-low (n = 51) vs CIMP-high (n = 77) T-ALL. (F) Common fusions in T-ALL and a relevant CNV in T-ALL (chr9) and 3 significantly different CNVs between CIMP low and CIMP high in 108 T-ALLs. The control cells in panels A, C, and D consisted of CD34⁺, CD3⁺, and lymph node, respectively.

Figure 5.

Methylome phenotype of the CIMP-low and CIMP-high subgroups in the NOPHO ALL2008 cohort. (A) Differentially methylated CpGs between CIMP-low and CIMP-high leukemias (n = 192 450K+EPICv.1 array-analyzed samples). β-values from EPICv.1 arrays (n = 128 T-ALL and 743 434 CpGs) were merged with 450K arrays (n = 64 T-ALL, 410 369 CpGs), whereafter 380 829 CpGs remained for differential methylation analysis. Of 380 829 CpGs, 3 CpGs had higher mean β-values in CIMP low and 7627 CpGs had higher mean β-values in CIMP high. (B) CGI distribution of the merged array data, the CIMP panel, and the DM-CpGs. (C) Heat map of the β-values of the 7630 DM-CpGs between CIMP subgroups in 192 T-ALL and 3 reference cells. (D) Gene expression (in VST counts) of the corresponding 1046 unique genes annotated to the DM-CpGs in 108 T-ALL and 3 reference cells. Reference cells in panels C and D are CD34⁺, CD3⁺, and lymph node. VST, variance-stabilizing transformation.

Figure 6.

Transcriptome phenotype of CIMP-low and CIMP-high subgroups in the NOPHO ALL2008 cohort. (A) Heat map revealing the relative RNA expression of n = 213 DEG between CIMP-low (n = 43) and CIMP-high (n = 65) T-ALL samples; genes hierarchical clustering highlights a cluster 1 with higher expression in CIMP low (134 genes) and a cluster 2 with higher expression in CIMP high (79 genes). (B,C) Heat maps presenting the mean β-values of CpGs located on the promoter region (B) or on the gene body (C) of the DEG visualized in the same gene and sample order as in panel A. (D) Pearson correlation between clusters 1 and 2 gene RNA expression and their mean promoter region DNA methylation. (E) GO analysis of DEGs with higher expression in CIMP low (n = 134 genes, cluster 1) and CIMP high (n = 79 genes, cluster 2), respectively. (F) Volcano plot with highlighted n = 213 DEG in n = 108 T-ALL. (G) RNA expression (in VST counts) heat map using hierarchical clustering of 108 T-ALL samples, revealing 40 selected genes relevant in T-ALL biology or differentially expressed between CIMP-low and CIMP-high subgroups. Sorted CD34⁺ and CD3⁺ T cells and lymph node are illustrated as normal cell types in the heat maps. DEG, differentially expressed gene; GO, gene ontology.

Figure 7.

CIMP percentage association with DNA methyltransferase gene expression, chronological and cellular age in the NOPHO ALL2008 cohort. Correlation of CIMP methylation percentage and (A) DNA methyltransferase 3 alpha (DNMT3A) and (B) beta (DNMT3B) gene expression (in VST counts) in 108 T-ALL patient samples and 3 control cells (CD34⁺, CD3⁺, and lymph node). Correlation of CIMP methylation percentage and (C) chronological age of peripheral blood leukocytes in healthy children (n = 78, aged 1-16 years) and diagnostic T-ALL patient samples (n = 192). (D) Horvath biological age and (E) EpiTOC mitotic clock. Spearman correlation rho and P values were estimated among T-ALL samples only, not including healthy samples. CIMP percentage is defined as the percentage of CIMP panel CpGs with a β-value >0.4.