DNA methylation profiling for molecular classification of neuroblastoma

Abstract

Neuroblastoma is a heterogeneous disease where patient stratification is critical for prognosis and treatment decisions and where it recently has been suggested that the presence of telomere maintenance mechanisms (TMM) should be considered in risk stratification. We investigated the utility of DNA methylation-based classification for neuroblastoma diagnostics by analysing 303 tumours samples from two cohorts. We show that of the total number of cases, an average of 90% of the samples classified as neuroblastoma, while 66% also achieved confident classification into the three NB subclasses: "MYCN-type", "ALT/TERT TMM positive" and "TMM negative". The tumours classified as MYCN-type showed genomic amplification of MYCN (MNA); however, some MYCN-type cases lacked evident MNA, suggesting that epigenetic states might be influenced by other factors such as activating ALK mutations. Survival analysis indicated similar poor survival probabilities for patients classified as TMM positive or MYCN type, distinct from the inferior survival of TMM-negative cases. All cases, with complementary genomic data available, associated with TMM positivity also presented features associated with telomere lengthening mechanisms, including TERT or ATRX alteration. However, some tumours positive for these features, especially TERT rearrangement, classified as MYCN type rather than TMM positive, indicating that MNA and other mechanisms introduce a methylation pattern that supersede or overlap with pattern imposed by TERT. Chromosomal copy number alterations (CNAs) characteristic of methylation subclasses were identified, including 1p deletion and 17q gain in MYCN type and combinations of 11q loss, 3p loss, 7q gain, and 17q gain in TMM-positive cases, highlighting the potential of the methylation arrays to replace SNP arrays for prognostic genomic assessments. Our study demonstrates that DNA methylation-based classification stratifies neuroblastoma into clinically relevant subgroups, aiding diagnostic and prognostic decisions, although discrepancy between genomic features and methylation classification does occur. The interplay between genomic alterations and methylation patterns could give clues into the discrepancy and underscores the complexity of neuroblastoma biology and the need for further research and validation of clinical outcomes of the patients in the respective subclasses.

Fig. 1

Classification of NB tumours using the MNP Classifier. A The hierarchy tree of classes for peripheral NBs in the MNP classifier v12.5. B, C Classification of the local (n = 90) and the TARGET cohort (n = 213) at the Superfamily (B), and at the Subclass level (C), grouped using the recommended Calibrated Score (CS) cut off level ≥ 0.9. D Subclass association of tumours that received a CS < 0.9, and ≥ 0.3 in the local cohort (top panel, n = 28), and in the TARGET cohort (lower panel; n = 67). CS, Calibrated Score; Cra. Par., the superfamily “Cranial and Paraspinal Nerve Tumours”; TRT, the molecular class “Teratoma (novel)”; SCHW, the molecular class “Schwannoma”

Fig. 2

Overview of TMM positive classified, or associated, tumours, and of additional tumours with genomic features associated with telomere maintenance mechanism (TMM) in the local cohort. The group of the cases with genomic alterations associated with TMM, that did not classify as such, consisted of nine samples that classified into (n = 3) or associated with (CS 0.4–0.77, n = 6) the MYCN type whereof five cases did present MNA. One TERT positive sample associated with TMM neg (CS 0.72). Red, has feature; orange, positive for c-circles but pre-existing prior amplification; and black line indicates where data are not available. SV, structural variant; CS, classification score

Fig. 3

Evaluation of copy number alterations (CNA) estimated from DNA methylation data. A Exemplified genome plot with estimated CNA from R package conumee 2.0 of a tumour that shows signs of MYCN amplification. B Conumee-estimated relative MYCN copy number estimations are higher in MYCN-amplified (MNA), than in not MNA tumours, in both the local (top panel) and in the TARGET (lower panel) cohort. C. Tumours with methylation-based molecular classes of CS ≥ 0.9 are plotted against conumee-estimated relative MYCN amplification, in the local cohort (top panel) and in the TARGET cohort (lower panel). In both cohorts, MYCN status was lacking for one tumour, respectively. MNA, MYCN-amplified; ***, P value < 0.001

Fig. 4

Summary copy number prediction plots of neuroblastoma subclasses in the TARGET cohort. Percentage distribution of copy number alterations in samples that classified into TMM positive (n = 47), MYCN type (n = 52) and TMM negative (n = 38) with CS ≥ 0.9 in TARGET cohort

Fig. 5

Overall survival probability for primary tumours that classify at CS ≥ 0.9 using the methylation-based classification. A Kaplan–Meier plot from the survminer package shows difference in survival probability between the classes MYCN type, and TMM NEG, in the local cohort (P = 0.00017). B. Kaplan–Meier plot from the survminer package shows differences in survival probability between the classes TMM NEG versus both MYCN type (P < 0.0001), and TMM POS (P < 0.0001), in the TARGET cohort

Fig. 6

Methylation-based classification in paired tumour samples from same patient. A Nine paired tumour samples were evaluated in the local cohort. For three patients the primary tumour was analysed using both the 450 K and the EPIC array. For these pair 7 to 9, the 450 K-based classification is shown as left bar, and the EPIC-based classification is shown as the right bar. B Nine pairs of primary and relapsed tumours were evaluated in the TARGET cohort. Dotted lines show CS of 0.9. Terat., the molecular class “Teratoma (novel)”