https://orcid.org/0000-0002-7267-1052
Abstract
MYC/MYCN are the most frequent oncogene amplifications in medulloblastoma (MB) and its primary biomarkers of high-risk (HR) disease. However, while many patients’ MYC(N)-amplified tumors are treatment-refractory, some achieve long-term survival. We therefore investigated clinicobiological heterogeneity within MYC(N)-amplified MB and determined its relevance for improved disease management.
We characterized the clinical and molecular correlates of MYC- (MYC-MB; n = 64) and MYCN-amplified MBs (MYCN-MB; n = 95), drawn from >1600 diagnostic cases.
Most MYC-MBs were molecular group 3 (46/58; 79% assessable) and aged ≥3 years at diagnosis (44/64 [69%]). We identified a “canonical” very high-risk (VHR) MYC-amplified group (n = 51/62; 82%) with dismal survival irrespective of treatment (11% 5-year progression-free survival [PFS]), defined by co-occurrence with ≥1 additional established risk factor(s) (subtotal surgical-resection [STR], metastatic disease, LCA pathology), and commonly group 3/4 subgroup 2 with a high proportion of amplified cells. The majority of remaining noncanonical MYC-MBs survived (i.e. non-group 3/group 3 without other risk features; 11/62 (18%); 61% 5-year PFS). MYCN survival was primarily related to molecular group; MYCN-amplified SHH MB, and group 3/4 MB with additional risk factors, respectively defined VHR and HR groups (VHR, 39% [35/89]; 20% 5-year PFS/HR, 33% [29/89]; 46% 5-year PFS). Twenty-two out of 35 assessable MYCN-amplified SHH tumors harbored TP53 mutations; 9/12 (75%) with data were germline. MYCN-amplified group 3/4 MB with no other risk factors (28%; 25/89) had 70% 5-year PFS.
MYC(N)-amplified MB displays significant clinicobiological heterogeneity. Diagnostics incorporating molecular groups, subgroups, and clinical factors enable their risk assessment. VHR “canonical” MYC tumors are essentially incurable and SHH-MYCN-amplified MBs fare extremely poorly (20% survival at 5 years); both require urgent development of alternative treatment strategies. Conventional risk-adapted therapies are appropriate for more responsive groups, such as noncanonical MYC and non-SHH-MYCN MB.
MYC(N)-amplified medulloblastoma is clinically and biologically heterogeneous.
“Canonical” MYC and SHH-MYCN are near incurable and require new approaches.
Remaining MYC(N) patients commonly survive and may be stratified for conventional therapies.
Medulloblastoma (MB) is among the most common malignant brain tumors of childhood. MYC(N) family amplifications (MYC, ~3%; MYCN, ~6% of tumors) are the primary molecular biomarkers of poor prognosis, high-risk (HR) disease, underpinning risk-stratified therapies in international, biomarker-driven clinical trials (e.g. SIOP-PNET5-MB, SIOP-HR-MB, and SJMB12). Previous clinical trial analyses have indicated outcome differences within MYC/N amplified MB; however, studies to understand this heterogeneity have previously been limited by their rarity. We assembled a cohort of 64 MYC and 95 MYCN-amplified tumors and established significant clinicobiological heterogeneity within both MYC and MYCN-amplified disease. Disease molecular group is the primary determinant of their clinical features, interacting with other risk factors to define prognosis. We identify, and proffer clinico-molecular risk stratification schema for, very HR tumor groups (“canonical” MYC and SHH/MYCN) in which current multimodal therapies are ineffective, and HR groups compatible with long-term survival. This heterogeneity must be considered diagnostically and has the potential to immediately impact clinical management.
Medulloblastoma (MB) is one of the most common malignant brain tumors of childhood. Approximately 30% of patients will die of their disease, while survivors commonly experience life-long disease and treatment-associated morbidities.1 Focal amplifications of MYC or MYCN are the most frequent oncogenic amplifications, and have been consistently associated with a poor prognosis across different clinical studies.2–8 This has led to their routine diagnostic assessment as the primary biomarkers of high-risk (HR) MB disease, underpinning risk-stratified therapies in international biomarker-driven clinical trials (e.g. SIOP-PNET5-MB; NCT02066220,9 SIOP-HR-MB; NCT pending,10 SJMB12; NCT01878617).
However, retrospective survival analyses of the SIOP-UKCCSG-PNET3 and HIT-SIOP-PNET4 trial cohorts demonstrated outcome differences within MYC(N)–amplified MB, suggesting MYC(N) amplification in the absence of other clinicopathological risk factors may not confer a poor prognosis,4,11 leading to such patients potentially incurring unnecessary side effects from intensified risk-adapted protocols. Conversely, MYC(N) amplification in conjunction with large-cell/anaplastic (LCA) histology has long been recognized to confer poor prognosis.6,12
MYC(N)-amplified MBs are molecularly heterogeneous, which may influence their clinical behavior. MB comprises 4 consensus molecular groups: WNT (MBWNT), SHH (MBSHH), and non-WNT/non-SHH (comprising groups 3 and 4 [MBGrp3, MBGrp4]).13–15MYC amplifications occur predominantly in MBGrp3 but are observed to a lesser extent in all other groups.7,16 In contrast, MYCN amplifications are mainly found in MBSHH and MBGrp4.16–19MYCN-amplified SHH MB is associated with TP53 mutation, commonly in the germline,20 chromothripsis, and a poor prognosis.17,21 Conversely, MYCN amplification does not associate with prognosis in MBGrp4.2,17,22 Indeed, MYCN-amplified group 4 MB with no other HR disease features were treated as standard risk in the SIOP-PNET5 clinical trial.9
Recent studies have identified further heterogeneity of potential prognostic significance to MYC(N)-amplified tumors; these include the identification of component molecular subgroups within each molecular group17,23–25 which are detected using DNA methylation microarray, alongside variations in the pattern and proportion of cells displaying MYC(N) amplification.4
Understanding differences in molecular pathology and clinical behavior within the MYC(N)-amplified group of MBs is thus essential to define their optimal clinical management. However, their relative rarity (~3% [MYC] and ~6% [MYCN] of all MBs) has limited investigations to small numbers (i.e. typical n < 10 per study) in clinical trials and research studies published to date. To address this, we assembled a retrospective cohort of 64 MYC and 95 MYCN-amplified tumors, derived from screening approximately 1600 MBs, representing the largest cohorts studied to date. We report a comprehensive characterization of their clinical features, molecular pathology, and survival outcomes, revealing significant clinically relevant heterogeneity, including very high-risk (VHR) tumor groups near-universally refractory to current therapies, and groups associated with significant long-term survival. These findings serve as a foundation to (i) immediately aid the clinical interpretation of contemporary molecular diagnostics, and (ii) inform the design of future clinical and research investigations, for this important tumor group.
Materials and Methods
Study Design and Participants
Tumor samples were provided by the UK CCLG (CCLG-approved biological study BS-2007–04); informed, written consent was obtained. Samples were also obtained from retrospective, previously published, international Heidelberg cohorts.26,27 The MYC-amplified cohort comprised 34 patients from the CCLG and 30 from Heidelberg; 57 patients in the MYCN-amplified cohort were drawn from the CCLG and 38 from Heidelberg. Tumor investigations were done with approval from Newcastle-North Tyneside Research Ethics Committee (reference 07/Q0905/71); all tumor material was collected in accordance with this approval.
No statistical methods were used to predetermine the sample size. We interrogated our retrospective tumor cohorts to identify patients with MYC-MB (n = 64) and MYCN-MB (n = 95). Amplification was identified by iFISH (fluorescence in situ hybridization) and/or copy number (CN) estimates from microarray (methylation or SNP6). The criteria for identifying MYC-MB and MYCN-MB by FISH have been described previously.4 Briefly, for each assessable tumor, 100–200 nonoverlapping nuclei were examined, enabling the proportion of amplified cells to be estimated. Individual cells were defined as amplified if the ratio of test probe:centromeric control ratio exceeded >4:1). Individual tumors were classed as amplified when they contained (1) amplification in ≥5% of cells and (2) evidence of cells with a “speckled” or “clumped” signal patterns consistent with double minute formation or homogeneously staining regions (Figure 1). Amplifications from the SNP6 array were called as previously described16; for calling amplifications from 450k/EPIC methylation arrays, CN was derived using conumee v4.2 and amplifications at the MYC/MYCN loci defined as being focal (<10 Mb), with amplitude >0.4.
Clinicopathological and molecular features of MYC-MB and MYCN-MB. (A) Interphase fluorescence in situ hybridization (iFISH) of a group 3 tumor showing high levels of MYC amplification (green) vs centromeric control (red) in the majority of cells. (B, C) Example of a MYC-amplified group 4 tumor with a mixture of MYC-amplified, MYC gained and normal cells at 40× and 100× magnification. Clinical, molecular, and cytogenetic features are shown for MYC-MB (D) and MYCN-MB (G), arranged by molecular group. Groups (SHH, red; group 3, yellow; group 4, green; unknown, gray) and subgroups are colored by convention. Missing data are shown in gray. Factors with significant enrichment in specific molecular groups are shown in bold text (<.05, Fisher’s Exact test). The relationship between amplified cell fraction and molecular group is shown for MYC-MB (E) and MYCN-MB (H). Age distribution is shown for MYC-MB (F) and MYCN-MB (I). For molecular groups with few members, individual data points are shown.
Amplification was identified using published criteria4 by iFISH (fluorescence in situ hybridization) and/or CN estimates (Illumina 450k/EPIC methylation or Affymetrix SNP6 microarrays16). For calling amplifications from 450k/EPIC methylation arrays, CN was derived using conumee v4.2, and amplifications at the MYC/MYCN loci were defined as being focal (<10 Mb), with amplitude >0.4.
Biological Characterization of MYC-MB and MYCN-MB
The principal molecular group was assigned using Illumina methylation arrays or by MS-MIMIC for low-quantity and/or poor-quality samples as previously described.17,28–30 For samples with 450k/EPIC arrays, the molecular group and group 3/4 subgroup were assigned using MNPv11b4 https://www.molecularneuropathology.org/mnp/.28 SHH subgroup was assessed as described.31
We assessed established MB clinical, pathological, and molecular features for their associations with MYC(N)-amplified disease. Histopathological variants were assigned according to WHO 2021 guidelines14; all tumors were centrally reviewed. Metastatic status was assigned using Chang’s criteria32; M0/1 disease (M−) was compared against M2/3 disease (M+). Tumors were designated subtotally resected (STR) if their postsurgical residuum exceeded 1.5 cm2.33
TP53 and additional MB mutations were identified as previously described.22,34 Chromosomal abnormalities and amplifications of GLI1 and GLI2 were identified by analysis of CN profiles and/or iFISH. Chromothripsis was inferred from SNP6 microarray-based DNA CN profiles, according to previously described criteria.20,23,35,36
Gene fusions were detected from RNA-seq data as previously described.37 Primer sequences for confirmation of fusion events by RT-PCR are shown in Supplementary Figure 4.
Statistical Analysis
Survival analysis was performed using progression-free survival (PFS), defined as the interval between diagnosis (i.e. date of surgery) and disease progression (defined as the time at which disease progression was confirmed by MRI). While PFS/OS was available for almost every MYC-amplified tumor (63/64 and 64/64 tumors, respectively), OS was less widely available (94/95 and 84/95 PFS/OS) for MYCN-amplified tumors. There was no significant difference between OS and PFS in either cohort (Supplementary Figure 1); consequently, we used PFS for subsequent survival analyses. Kaplan-Meier curves were plotted and differences in survival between groups were assessed using log-rank tests. Univariable Cox models were constructed for key disease features and proportionality of hazards confirmed by examining scaled Schoenfeld residuals. We assessed the prognostic utility of current molecular and clinical variables. Fisher’s exact and chi-squared tests were used to assess associations between categorical variables. ANOVA and t-tests were used to compare continuous variables between groups. Significant associations were defined as having P < .05. Statistical and bioinformatics analyses were done using R statistical environment (version 4.3.0),38 using the survival v3.5-5, and rms v6.7-0 packages.
Results
Detection of MYC(N)-Amplified Tumors
MYC-MB and MYCN-MB (n = 64 and 95, respectively; Table 1) were identified by iFISH and/or microarray analysis, with the majority (116/159; 73%) assessed by both methods. Despite strong concordance overall, some tumors with lower percentages of amplified cells by iFISH (MYC-MB tumors, n = 5/49 [10%] tumors with both iFISH and methylation-array derived call, 7%–25% cells amplified; MYCN-MB tumors, n = 7/62 (11%) tumors with both iFISH and methylation-array derived call, 7%–60% cells amplified), were not detectable by CN array. Thus, while assessment of MYC(N)-amplification is readily accessible from DNA methylation microarrays, superior sensitivity together with the reported clinical significance of lower amplification frequencies,4 mandates continued use of iFISH as the “gold standard” for clinical assessment.39
Demographics and clinical characteristics of MYC-MB and MYCN-MB cohorts.
| MYC (n = 64) | MYCN (n = 95) | P value | |
|---|---|---|---|
| Sex | |||
| Male | 39(61%) | 62(65%) | |
| Female | 25(39%) | 33(35%) | .62 |
| Age at diagnosis | |||
| Median, years (range) | 4.6 (0.9–15.8) | 8.0 (1.9–33.3) | |
| <2.99 | 20(31%) | 2 (2%) | |
| >3.00 | 44(69%) | 93(98%) | <.001 |
| Histopathological variant | |||
| LCA | 40(63%) | 33(35%) | |
| Classic | 23(36%) | 57(61%) | |
| DN | 1 (1%) | 4 (4%) | |
| MB-NOS | 0 | 1 | .002 |
| Metastatic stage | |||
| M− | 28(45%) | 27(71%) | |
| M+ | 34(55%) | 65(29%) | .06 |
| No data | 2 | ||
| Resection | |||
| STR | 20(32%) | 30(32%) | |
| GTR | 43(68%) | 64(68%) | 1 |
| No data | 1 | 1 | |
| Isochromosome17q | |||
| Present | 26(44%) | 41(46%) | |
| Absent | 32(56%) | 48(54%) | 1 |
| No data | 6 | 6 | |
| TP53 mutation | |||
| Present | 1 (2%) | 21(25%) | |
| Absent | 51(98%) | 62(75%) | <.001 |
| No data | 12 | 12 | |
| GLI1/2 amplification | |||
| Present | 0 (0%) | 11(15%) | |
| Absent | 55(100%) | 64(85%) | .002 |
| No data | 9 | 20 | |
| Molecular group | |||
| WNT | 1 (2%) | 0 (0%) | |
| SHH | 5 (9%) | 36(40%) | |
| Group 3 | 46(79%) | 10(11%) | |
| Group 4 | 6 (10%) | 45(49%) | <.001 |
| No data | 6 | 4 | |
| Subgroup—group 3/4 | |||
| 1 | 1 (2%) | 0 (0%) | |
| 2 | 22(54%) | 2 (7%) | |
| 3 | 4 (10%) | 1 (4%) | |
| 4 | 1 (2%) | 1 (4%) | |
| 5 | 7 (17%) | 14(52%) | |
| 6 | 2 (5%) | 5 (19%) | |
| 7 | 3 (7%) | 4 (15%) | |
| 8 | 1 (2%) | 0 (0%) | |
| No data | 12 | 27 | |
| Subgroup—SHH | |||
| 1 | 1 (50%) | 2 (6%) | |
| 2 | 1 (50%) | 3 (9%) | |
| 3 | 0 | 18(55%) | |
| 4 | 0 | 10(30%) | |
| No subgroup data | 3 | 3 | |
| Treatment | |||
| RTX alone at diagnosis | 5 (8%) | 5 (6%) | |
| CTX alone at diagnosis | 21(36%) | 4 (4%) | |
| RTX and CTX at diagnosis | 32(54%) | 83(90%) | |
| None | 1 (2%) | 0 (0%) | <.001 |
| No data | 5 | 3 | |
| Follow-up time | |||
| Median, years (range) | 6.2 (0.1–17) | 6.3 (0.1–14) | |
| MYC (n = 64) | MYCN (n = 95) | P value | |
|---|---|---|---|
| Sex | |||
| Male | 39(61%) | 62(65%) | |
| Female | 25(39%) | 33(35%) | .62 |
| Age at diagnosis | |||
| Median, years (range) | 4.6 (0.9–15.8) | 8.0 (1.9–33.3) | |
| <2.99 | 20(31%) | 2 (2%) | |
| >3.00 | 44(69%) | 93(98%) | <.001 |
| Histopathological variant | |||
| LCA | 40(63%) | 33(35%) | |
| Classic | 23(36%) | 57(61%) | |
| DN | 1 (1%) | 4 (4%) | |
| MB-NOS | 0 | 1 | .002 |
| Metastatic stage | |||
| M− | 28(45%) | 27(71%) | |
| M+ | 34(55%) | 65(29%) | .06 |
| No data | 2 | ||
| Resection | |||
| STR | 20(32%) | 30(32%) | |
| GTR | 43(68%) | 64(68%) | 1 |
| No data | 1 | 1 | |
| Isochromosome17q | |||
| Present | 26(44%) | 41(46%) | |
| Absent | 32(56%) | 48(54%) | 1 |
| No data | 6 | 6 | |
| TP53 mutation | |||
| Present | 1 (2%) | 21(25%) | |
| Absent | 51(98%) | 62(75%) | <.001 |
| No data | 12 | 12 | |
| GLI1/2 amplification | |||
| Present | 0 (0%) | 11(15%) | |
| Absent | 55(100%) | 64(85%) | .002 |
| No data | 9 | 20 | |
| Molecular group | |||
| WNT | 1 (2%) | 0 (0%) | |
| SHH | 5 (9%) | 36(40%) | |
| Group 3 | 46(79%) | 10(11%) | |
| Group 4 | 6 (10%) | 45(49%) | <.001 |
| No data | 6 | 4 | |
| Subgroup—group 3/4 | |||
| 1 | 1 (2%) | 0 (0%) | |
| 2 | 22(54%) | 2 (7%) | |
| 3 | 4 (10%) | 1 (4%) | |
| 4 | 1 (2%) | 1 (4%) | |
| 5 | 7 (17%) | 14(52%) | |
| 6 | 2 (5%) | 5 (19%) | |
| 7 | 3 (7%) | 4 (15%) | |
| 8 | 1 (2%) | 0 (0%) | |
| No data | 12 | 27 | |
| Subgroup—SHH | |||
| 1 | 1 (50%) | 2 (6%) | |
| 2 | 1 (50%) | 3 (9%) | |
| 3 | 0 | 18(55%) | |
| 4 | 0 | 10(30%) | |
| No subgroup data | 3 | 3 | |
| Treatment | |||
| RTX alone at diagnosis | 5 (8%) | 5 (6%) | |
| CTX alone at diagnosis | 21(36%) | 4 (4%) | |
| RTX and CTX at diagnosis | 32(54%) | 83(90%) | |
| None | 1 (2%) | 0 (0%) | <.001 |
| No data | 5 | 3 | |
| Follow-up time | |||
| Median, years (range) | 6.2 (0.1–17) | 6.3 (0.1–14) | |
PFS was available for 63/64 MYC-amplified patients and for 94/95 MYCN-amplified patients. P values from Fisher’s exact tests are shown. P values <.05 are shown in bold text.
Demographics and clinical characteristics of MYC-MB and MYCN-MB cohorts.
| MYC (n = 64) | MYCN (n = 95) | P value | |
|---|---|---|---|
| Sex | |||
| Male | 39(61%) | 62(65%) | |
| Female | 25(39%) | 33(35%) | .62 |
| Age at diagnosis | |||
| Median, years (range) | 4.6 (0.9–15.8) | 8.0 (1.9–33.3) | |
| <2.99 | 20(31%) | 2 (2%) | |
| >3.00 | 44(69%) | 93(98%) | <.001 |
| Histopathological variant | |||
| LCA | 40(63%) | 33(35%) | |
| Classic | 23(36%) | 57(61%) | |
| DN | 1 (1%) | 4 (4%) | |
| MB-NOS | 0 | 1 | .002 |
| Metastatic stage | |||
| M− | 28(45%) | 27(71%) | |
| M+ | 34(55%) | 65(29%) | .06 |
| No data | 2 | ||
| Resection | |||
| STR | 20(32%) | 30(32%) | |
| GTR | 43(68%) | 64(68%) | 1 |
| No data | 1 | 1 | |
| Isochromosome17q | |||
| Present | 26(44%) | 41(46%) | |
| Absent | 32(56%) | 48(54%) | 1 |
| No data | 6 | 6 | |
| TP53 mutation | |||
| Present | 1 (2%) | 21(25%) | |
| Absent | 51(98%) | 62(75%) | <.001 |
| No data | 12 | 12 | |
| GLI1/2 amplification | |||
| Present | 0 (0%) | 11(15%) | |
| Absent | 55(100%) | 64(85%) | .002 |
| No data | 9 | 20 | |
| Molecular group | |||
| WNT | 1 (2%) | 0 (0%) | |
| SHH | 5 (9%) | 36(40%) | |
| Group 3 | 46(79%) | 10(11%) | |
| Group 4 | 6 (10%) | 45(49%) | <.001 |
| No data | 6 | 4 | |
| Subgroup—group 3/4 | |||
| 1 | 1 (2%) | 0 (0%) | |
| 2 | 22(54%) | 2 (7%) | |
| 3 | 4 (10%) | 1 (4%) | |
| 4 | 1 (2%) | 1 (4%) | |
| 5 | 7 (17%) | 14(52%) | |
| 6 | 2 (5%) | 5 (19%) | |
| 7 | 3 (7%) | 4 (15%) | |
| 8 | 1 (2%) | 0 (0%) | |
| No data | 12 | 27 | |
| Subgroup—SHH | |||
| 1 | 1 (50%) | 2 (6%) | |
| 2 | 1 (50%) | 3 (9%) | |
| 3 | 0 | 18(55%) | |
| 4 | 0 | 10(30%) | |
| No subgroup data | 3 | 3 | |
| Treatment | |||
| RTX alone at diagnosis | 5 (8%) | 5 (6%) | |
| CTX alone at diagnosis | 21(36%) | 4 (4%) | |
| RTX and CTX at diagnosis | 32(54%) | 83(90%) | |
| None | 1 (2%) | 0 (0%) | <.001 |
| No data | 5 | 3 | |
| Follow-up time | |||
| Median, years (range) | 6.2 (0.1–17) | 6.3 (0.1–14) | |
| MYC (n = 64) | MYCN (n = 95) | P value | |
|---|---|---|---|
| Sex | |||
| Male | 39(61%) | 62(65%) | |
| Female | 25(39%) | 33(35%) | .62 |
| Age at diagnosis | |||
| Median, years (range) | 4.6 (0.9–15.8) | 8.0 (1.9–33.3) | |
| <2.99 | 20(31%) | 2 (2%) | |
| >3.00 | 44(69%) | 93(98%) | <.001 |
| Histopathological variant | |||
| LCA | 40(63%) | 33(35%) | |
| Classic | 23(36%) | 57(61%) | |
| DN | 1 (1%) | 4 (4%) | |
| MB-NOS | 0 | 1 | .002 |
| Metastatic stage | |||
| M− | 28(45%) | 27(71%) | |
| M+ | 34(55%) | 65(29%) | .06 |
| No data | 2 | ||
| Resection | |||
| STR | 20(32%) | 30(32%) | |
| GTR | 43(68%) | 64(68%) | 1 |
| No data | 1 | 1 | |
| Isochromosome17q | |||
| Present | 26(44%) | 41(46%) | |
| Absent | 32(56%) | 48(54%) | 1 |
| No data | 6 | 6 | |
| TP53 mutation | |||
| Present | 1 (2%) | 21(25%) | |
| Absent | 51(98%) | 62(75%) | <.001 |
| No data | 12 | 12 | |
| GLI1/2 amplification | |||
| Present | 0 (0%) | 11(15%) | |
| Absent | 55(100%) | 64(85%) | .002 |
| No data | 9 | 20 | |
| Molecular group | |||
| WNT | 1 (2%) | 0 (0%) | |
| SHH | 5 (9%) | 36(40%) | |
| Group 3 | 46(79%) | 10(11%) | |
| Group 4 | 6 (10%) | 45(49%) | <.001 |
| No data | 6 | 4 | |
| Subgroup—group 3/4 | |||
| 1 | 1 (2%) | 0 (0%) | |
| 2 | 22(54%) | 2 (7%) | |
| 3 | 4 (10%) | 1 (4%) | |
| 4 | 1 (2%) | 1 (4%) | |
| 5 | 7 (17%) | 14(52%) | |
| 6 | 2 (5%) | 5 (19%) | |
| 7 | 3 (7%) | 4 (15%) | |
| 8 | 1 (2%) | 0 (0%) | |
| No data | 12 | 27 | |
| Subgroup—SHH | |||
| 1 | 1 (50%) | 2 (6%) | |
| 2 | 1 (50%) | 3 (9%) | |
| 3 | 0 | 18(55%) | |
| 4 | 0 | 10(30%) | |
| No subgroup data | 3 | 3 | |
| Treatment | |||
| RTX alone at diagnosis | 5 (8%) | 5 (6%) | |
| CTX alone at diagnosis | 21(36%) | 4 (4%) | |
| RTX and CTX at diagnosis | 32(54%) | 83(90%) | |
| None | 1 (2%) | 0 (0%) | <.001 |
| No data | 5 | 3 | |
| Follow-up time | |||
| Median, years (range) | 6.2 (0.1–17) | 6.3 (0.1–14) | |
PFS was available for 63/64 MYC-amplified patients and for 94/95 MYCN-amplified patients. P values from Fisher’s exact tests are shown. P values <.05 are shown in bold text.
Molecular Groups and Subgroups
MYC-MB tumors were predominantly MBGrp3 (46/58, 79%; Table 1; Figure 1D), although appreciable numbers were also observed in MBSHH (n = 5, 9%) and MBGrp4 (n = 6, 10%). Within MYC-MBGrp3, subgroup 2 was most prevalent (22/41, 54% assessable tumors). MYCN-MB tumors were typically MBSHH and MBGrp4 (36/90 [40%] and 45/90 [49%], respectively). MYCN-MBSHH were primarily members of MBSHH subgroups 3 and 4; MYCN-MBGrp4 were subgroups 4–7 where data was available (Table 1; Figure 1G).
Clinicopathological Characteristics and Subclonal Amplification
Specific clinicopathological disease features were strongly associated with molecular group identity in both MYC and MYCN-MB (Table 1). Infants (<3.0 years) and younger children (3.0–4.99 years) predominated in MYC-MB (31% <3 years; median age at diagnosis 4.6 years). In contrast, only 2/95 (2%) patients with MYCN-MB were <3 years (Table 1; Figure 1G, I). The predominance of SHH subgroups 3 and 4 within MYCN-MBSHH was consistent with their noninfant age profile.40 Male sex was significantly enriched in MYC-MBGrp3 (33/46 [72%] MYC-MBGrp3 vs 2/12 [17%] MYC-MBnon-Grp3, P = .0008; Figure 1D) and also predominated in MYCN-MB (Figure 1G), regardless of molecular group (1.88:1 M:F ratio vs 1.5:1 typically observed disease wide1).
Most (52/63; 83%) MYC-MB presented with ≥1 additional clinicopathological risk factor (Figures 1D and 3F). The majority of MYC-MBGrp3 tumors had LCA pathology (38/46 [83%] MYC-MBGrp3 vs 2/12 [17%] MYC-MBnon-Grp3; P < .0001). Notably, there were no LCA MYC-MBGrp4 tumors (0/6; Figure 1D). In addition, MYC-MBGrp3 tumors were significantly enriched for metastatic disease compared to MYC-MBnon-Grp3 (30/44 [68%] vs 3/12 (25%); P = .018). The majority of MYC-MBGrp3 tumors exhibited high proportions of MYC-amplified cells by iFISH, in contrast to MYC-MBnon-Grp3 (mean cells amplified 74% vs 33%; P < .0001; Figure 1E). Albeit with small numbers of assessable tumors, subtotal resection (STR) was a feature of most (4/5) MYC-MBSHH tumors (Figure 1D).
Fewer (56/91; 62%) MYCN-MBs presented with ≥1 other clinicopathological risk factor (Figures 1G and 4F). MYCN-MBSHH were also strongly associated with LCA pathology (23/35 [66%] vs 10/54 [19%] in MYCN-MBGrp3/4, P < .0001, Figure 1G). MYCN-MBSHH similarly had a significantly increased proportion of amplified cells (mean 67% vs 54% in MYCN-MBGrp3/4; P = .04; Figure 1H). STR and M+ disease appeared equivalently distributed across MYCN-MBSHH and MYCN-MBGrp3/4 (Table 1).
Genomic Profiles
MYC-MB mutational (n = 22; Supplementary Figure 2) and CN profiles (n = 53; Figure 1D) were consistent with MBGrp3 and MBGrp4 more widely.17,25 Additional gene-specific driver mutations were uncommon (Supplementary Figure 2). In contrast, MYCN-MBSHH (n = 30 CN/n = 18 mutation profiles) and MYCN-MBGrp3/4 (n = 37 CN/n = 13 mutation) harbored distinct CN and mutation profiles (Figure 1G; Supplementary Figure 3). Within MYCN-MBSHH, 9q loss was common (14/30, 47%). In contrast, i17q was common (29/37, 78%) in MYCN-MBGrp3/4. TP53 mutations were detected in 22/35 (63%) of assessed MYCN-MBSHH tumors (missense, n = 16/19 with available information; frameshift, n = 3/19), but not in MYCN-MBGrp3/4 (P < .0001). The majority (18/22, 82%) of TP53 mutations within MYCN-MBSHH co-occurred with 17p loss (P = .00059) and most were germline (9/12 [75%] with available data). TP53 mutations occurred in all MYCN-MBSHH subgroups, but most prevalently in subgroups 3 and 4 (respectively, 10/18, 56% and 9/10, 90% assessable tumors). GLI2 (10/35, 29%) or GLI1 (1/35, 3%) amplifications were associated with MYCN-MBSHH, and only found in TP53 mutated tumors (P = .0045; Supplementary Figure 3).
Genomic Instability Patterns: Chromothripsis and RNA Fusion Transcripts
We next assessed patterns of CN, chromothripsis, and gene fusion events in our cohorts. Chromothripsis was common in both MYCN-MB (8/23 [35%] assessable tumors) and MYC-MB (6/11 [55%]), but its patterns and correlates were markedly different. In MYCN-MB, chromothripsis was found in both MYCN-MBSHH and MYCN-MBGrp4 (6/14 vs 2/9; P = .40), co-occurred with TP53 mutation in MYCN-MBSHH (6/8), and manifested in multiple chromosomes (Figure 2A; Supplementary Figure 4A). In contrast, chromothripsis in MYC-MB occurred in conjunction with 17p loss (6/6), without TP53 mutation (5/5 assessable), and was restricted to chromosome 8 (MYC at 8q24; Figure 2B).
Differential patterns of chromothripsis and fusion transcripts within MYC-MB and MYCN-MB cohorts. Type and frequency of RNA fusion transcripts in (A) MYCN and (B) MYC-amplified tumors with molecular group, subgroup, chromothriptic chromosomes, and TP53 mutated status indicated. Missing data are shown by an empty cell. Chromosomal distribution of chromothripsis is shown for 8 individual MYCN-amplified tumors (C) and 6 individual MYC-amplified tumors (D), with CN profiles from SNP6 array data (each tumor individually colored within each cohort). Circos plots show the distribution of RNA fusion transcripts in (E) MYCN (n = 15, data combined) and (F) MYC-amplified tumors (n = 12, data combined); interchromosomal fusions shown in blue and intrachromosomal fusions shown in red.
The RNA fusion transcript landscape (n = 27 tumors) further supported differential genomic instability patterns. Thirty-two putative oncogenic gene fusions were identified (n = 23, MYC-MB; n = 9, MYCN-MB). Seven out of 10 (MYC-MB) and 6/8 (MYCN-MB) gene fusion loci assessed validated successfully (RT-PCR/Sanger sequencing; Supplementary Figure 4). Consistent with chromothripsis patterns, MYCN-MB had fusions affecting many chromosomes (Figure 2A, C; Supplementary Figure 4), while MYC-MB exhibited intrachromosomal fusions only on chromosome 8 (Figure 2B, D). Fusions involved chromosomes showing evidence of chromothripsis or multiple segmental changes, (Figure 2E, F; Supplementary Figure 4), and genes within coamplified regions.
MYCN-MB fusion transcripts were unique to each tumor; 2 recurrent fusion-partner genes, DDX1 and NBAS (immediately upstream of MYCN41) were involved in fusions with each other and additional partners (MYCNOS) in 3 MYCN-MBGrp3/4 tumors, but fusion position and gene order were not conserved (Supplementary Figure 4B). In MYC-MB, fusion transcripts involving PVT1 were most common (7/12 MYC-MB tumors; Figure 2F; Supplementary Figure 4C), were exclusive to MYC-MB (vs 220 non-MYC-MB tumors37) and present in both MYC-MBGrp3 and MYC-MBGrp4.16
Outcome Differences in MYC(N)-Amplified MB: Clinical and Molecular Correlates
In MYC-MB, MYC-MBGrp3 had significantly worse survival than MYC-MBGrp4 (P = .010; Figure 3A; Supplementary Figure 5); MYC-associated disease progression and/or all relapses typically occurred rapidly within 2 years of initial diagnosis. However, long-term survivors were observed in all non-MBWNT groups. Survival was dismal within MYC-MBGrp3 subgroup 2, with 20/21 patients showing relapse or disease progression within 2 years of diagnosis (5-year PFS 5%; P = .054, Figure 3B; Supplementary Figure 5). Moreover, survival for MYC-MBGrp3 was not dependent on infant status (P = .08; Supplementary Figure 5E). The behavior of other MYC-MBGrp3/4 subgroups remains unclear, due to small sample numbers, however, subgroup 5 patients (n = 7) also showed rapid relapse and poor PFS, with 6/7 relapsing or progressing within 2 years of diagnosis. Likewise, LCA pathology conferred a significantly poorer prognosis (5-year PFS 6%; P = .0004, Figure 3C; Supplementary Figure 5). The LCA MYC-MB group comprised both infants (n = 13), most of whom (11/13 (86%)) received no upfront radiotherapy, and older children (n = 26, 22/25 of whom received high-dose radiotherapy); this latter group contained the only 2 long-term survivors (Figure 3C). M+ disease was also significantly associated with worse survival (P = .011, Figure 3D; Supplementary Figure 5), whereas subtotal resection was not prognostic (Figure 3E). The strongest univariable survival predictor within MYC-MB was the percentage of amplified cells (HR 11.9, 95%CI 3.01–47.3, P = .0004; Supplementary Figure 5), which was significantly higher in MYC-MBGrp3 (Figure 1E). Overall, MYC-MBGrp3 long-term survivors (i.e. ≥4 years postdiagnosis) were characterized by an absence of additional risk factors (i.e. STR/M+/LCA; Figure 3F).
Within MYCN-MB, MYCN-MBSHH was associated with very poor survival (5-year PFS 20%; P = .005, Figure 4A; Supplementary Figure 5); survival in all assessable SHH subgroups (3 and 4) was equivalently poor (Figure 4B). The median time to relapse for MYCN-MBSHH was 1.4 years (range 0.1–7.8). MYCN-MBGrp4 had significantly better outcomes than MYCN-MBSHH (5-year PFS 56% vs 20% respectively; P = .0005) and, while MYCN-MBGrp3 were less common (n = 9/90), this group had survival rates comparable with MYCN-MBGrp4 (5-year PFS 65%; P = .58; Figure 4A). Molecular features significantly associated with poorer prognosis in univariable analyses included the SHH group and strongly SHH-associated features (e.g. TP53 mutation, GLI1/2 amplification; Supplementary Figure 5); neither feature was associated with a significantly different PFS within the MYCN-MBSHH cohort (Supplementary Figure 5). The prognostic significance of HR disease features within MYCN-MB was molecular group dependent. The presence/absence of established risk features (M+, LCA, and STR) was prognostic within MYCN-MBGrp3/4 (Figure 4C–E); in contrast, the MYCN-MBSHH group had a poor outcome regardless of other HR features, defining a VHR group in its own right. Overall, long-term survivors (i.e. ≥4 years postdiagnosis) were characterized by MBGrp3/4 disease with an absence of additional risk factors (Figure 4F).
An additive interaction between MYC(N)-amplification and additional clinicopathological risk factors has been suggested previously.4 Patients with MYC-MB, but otherwise standard risk, achieved 5-year PFS of 61%; 5-year PFS reduced to 29% with one additional risk factor (M+/LCA/STR) with no long-term survivors harboring ≥2 additional risk factors (Figure 3F). Patients with MYCN-MB and no additional risk factors had 5-year PFS of 70%; where molecular group was known, all long-term survivors (≥4 years postdiagnosis; n = 13) were MBGrp3/4. In contrast, patients with one additional risk factor had 45% 5-year PFS (7/8 long-term survivors were MBGrp3/4), and patients with ≥2 additional risk factors had 21% 5-year PFS (Figure 4F).
Survival of patients with MYC-amplified tumors by clinical and molecular risk features. (A–F) Kaplan-Meier plots and at-risk tables are shown for MYC-amplified tumors. Where appropriate, the molecular group is indicated by filled circles adjacent to censor points for survivors with PFS ≥ 4 years; the molecular group is shown on inset pie charts. Certain MYC-amplified tumors lacked molecular group information and were omitted from the pie charts. Molecular group colors: SHH, red; group 3, yellow; group 4, green; N/A, gray.
Survival of patients with MYCN-amplified tumors by clinical and molecular risk features. (A–F) Kaplan-Meier plots and at-risk tables are shown for MYCN-amplified tumors. Where appropriate, the molecular group is indicated by filled circles adjacent to censor points for survivors with PFS ≥ 4 years; the molecular group is shown on inset pie charts. Certain MYCN-amplified tumors lacked molecular group information and were omitted from the pie charts. Molecular group colors: SHH, red; group 3, yellow; group 4, green; N/A, gray.
Cranio-Spinal Irradiation Is Ineffective in MYC-MB With Additional Risk Factors
Overall, receipt of upfront cranio-spinal irradiation (CSI) was associated with significantly improved survival in MYC-MB patients (5-year PFS 30% vs 9% in non-irradiated patients; P = .0008, Figure 5A). In the absence of additional HR features (metastasis, LCA, STR), MYC-MB was compatible with long-term survival (irradiated patients 5-year PFS 63%, Figure 5B); long-term survival was observed in a subset of MYC-MBGrp3 (Figure 5B). However, no or only marginal improvements in survival were observed following irradiation in patients with ≥1 additional risk factor (Figure 5C; Supplementary Figure 6). Each additional risk factor assessed was individually associated with poorer survival (Supplementary Figure 6); however, these risk factors frequently co-occurred (Figure 5D). Additionally, survival was not significantly different in patients receiving high-dose vs standard-dose chemotherapy (5-year PFS 11% and 24% for high and standard-dose chemotherapy patients, respectively, P = .12; Supplementary Figures 5 and 7). In infant patients only, 5-year PFS was 13% and 10% for high and standard-dose chemotherapy patients respectively, P = .68).
Cranio-spinal irradiation is ineffective in MYC-amplified tumors with additional established risk features. (A) Survival of MYC-amplified tumors by receipt of radiotherapy. (B) Survival of nonmetastatic, gross-totally resected, non-LCA MYC-amplified tumors, stratified by receipt of radiation. (C) Survival of MYC-amplified tumors with positivity for one or more HR features in addition to MYC amplification, stratified by receipt of radiation. Where appropriate, the molecular group is indicated by filled circles adjacent to censor points for survivors with PFS ≥ 4 years; the molecular group is shown as inset pie charts; certain MYC-amplified tumors lacked molecular group information and were omitted from pie charts. (D) Venn diagram summarizes co-occurrence of HR disease features in MYC-MB. Molecular group colors: SHH, red; group 3, yellow; group 4, green; N/A, gray.
Defining Risk in MYC(N)-Amplified Patients
Molecular group is critical to assess risk in MYC-MB. The presence of additional risk factors (≥1 of STR/M+/LCA) allocates the majority (51/62; 82%) to a VHR group with 11% 5-year PFS (Figure 6A, B). This group is mostly MBGrp3 (42/49; 86%), and predominantly MBGrp3/4 subgroup 2 (21/34; 62%) and 5 (5/34; 15%). Remaining patients where MYC amplification is the sole risk factor are HR (5-year PFS 61%), and mostly MBGrp4 (5/8; 63%).
Treatment stratification and survival groups within MYC-MB and MYCN-MB. (A, C) Suggested stratification for MYC-MB and MYCN-MB. (B, D) Risk stratification identifies VHR patient groups and groups compatible with longer-term survival. For each treatment group, Kaplan-Meier plots with risk tables are shown, with insets showing additional features of each group—the proportion of amplified cells, molecular group, and subgroup. SR, standard risk; HR, high risk; VHR, very high risk.
MYCN-MB can be assigned into standard risk, HR, and VHR groups (Figure 6C, D). VHR disease was defined by MBSHH (35/89 (39%) patients, 5-year PFS 20%). These were predominantly SHH subgroup 3 (53%; 17/32) and had a higher proportion of amplified cells (P = .04; vs standard/HR groups). High-risk disease was defined by MBGrp3/4 with positivity for ≥1 of STR/M+/LCA, encompassing 29/89 (33%) patients (5-year PFS 46%). MBGrp3/4 patients negative for STR/M+/LCA (25/89 (28%)) were standard risk (5-year PFS, 70%).
Discussion
MYC(N) family amplification is the key molecular biomarker of HR MB in current clinical practice. Our investigation of >150 MYC(N)-amplified tumors, drawn from >1600 diagnostic cases, reveals significant clinical and biological heterogeneity within both MYC and MYCN-amplified disease. Disease molecular group is the primary determinant of their clinical features and interacts with established risk factors and other features to define prognosis (Figure 6). These characteristics must be considered diagnostically and have the potential to immediately impact clinical management. Moreover, to avoid misdiagnosing patients, iFISH for oncogene amplification detection must remain mandatory, since methylation arrays frequently failed to detect amplifications, possibly as a consequence of intra-tumoral heterogeneity.
Our findings define a group of canonical MYC-amplified tumors (74%) which are group 3, display other HR features (e.g. LCA, M+, and STR) and have exceptionally poor prognosis (5-year PFS 11%). Noncanonical tumors (non-group 3 or group 3 with MYC as the sole HR feature) represent a notable group (26%); experience within our cohort indicates a better prognosis—approximately 60% achieve durable outcomes. Canonical tumors are most commonly subgroup 2 with a high percentage of amplified cells, whereas noncanonical tumors typically comprise other group 3/4 subgroups and have fewer amplified cells. Chromothripsis of chromosome 8 (MYC-harboring) and MYC-associated fusion genes are common features of all MYC-amplified tumors.
MYCN-amplified tumors distribute evenly between SHH and group 4; this subdivision is the primary determinant of their clinical behavior. MYCN-amplified SHH MB (40%) have dismal outcomes (5-year PFS 20%) and are commonly LCA and/or TP53mut (~75% germline); however, these factors do not appear to further influence prognosis. In contrast, MYCN-amplified group 4 MB (~50%) more commonly achieve long-term survival, and their prognosis appears equivalent to non-MYCN-amplified group 4 MB, with other established factors (e.g. M+) defining their risk. Clinical behavior of rarer MYCN-amplified group 3 MB (~10%) appears consistent with group 4. Chromothripsis of chromosome 2 (MYCN-harboring) was common, but, in contrast to MYC, frequently coinvolved other chromosomes and their transcriptomes contained both intra- and interchromosomal fusion genes.
We proffer a system for risk stratification of MYC(N)-amplified tumors (Figure 6), combining molecular groups and other risk factors. Associated markers (subgroup and levels of intra-tumoral amplification) further corroborate and secure these diagnoses. Most importantly, these enable the distinction of VHR tumor groups (canonical MYC and MYCN-amplified SHH) in which all current therapies (conventional chemotherapy and CSI) are ineffective. Relapse and/or progression are near-universal and novel therapeutic strategies should be urgently considered. Notably, additional driver mutations were rare in both groups (Supplementary Figures 2 and 3) and indirect targeting strategies (e.g. immune and/or cellular therapies, targeting of biological codependencies) will likely be required.42–44 In the absence of effective therapies, more palliative strategies could be considered for these VHR groups. We found no evidence to suggest group 4 MYCN-amplified and other rarer tumors lying outside these canonical groups share this VHR prognosis, indicating they may be stratified for conventional therapies using established risk markers.14
Assembly of this large, retrospective cohort has been essential to understand the clinical behavior of MYC(N)-amplified MB. We acknowledge the limitations of its retrospective multicenter nature and the potential to introduce bias in preselected cohorts. Moreover, due to issues of collinearity of HR disease features (Figure 5D) and cohort size, multivariable survival models were not assessable. However, by definition, equivalent investigations will not be achievable in contemporary international clinical trials (typically n = 300–400); such cases must therefore be carefully monitored and discussed within a multidisciplinary tumor board setting, based on the available evidence.
In summary, our investigations refine diagnosis and prognostication of MYC(N)-amplified MB, allowing the definition of canonical MYC-amplified and MYCN-amplified SHH patients which are essentially incurable using current therapies and require novel treatment strategies, alongside lower-risk subsets compatible with longer-term survival.
Supplementary material
Supplementary material is available online at Neuro-Oncology (https://academic.oup.com/neuro-oncology).
Funding
This work was supported by Cancer Research UK; The Brain Tumour Charity; Children with Cancer UK; and Great Ormond Street Children’s Charity. The research funders played no role in the research design, execution, analysis, interpretation, and reporting of this work.
Conflict of interest statement
S.M.P. reports receipt of consulting fees from BioSkryb. In addition, S.M.P. reports a patent “DNAMethylation based method for classifying tumor species (European Patent 16 710700.2).” No other author has any financial or non-financial interests to report.
Authorship statement
Conceptualization: E.C.S., J.C.L., D.W., S.B., and S.C.C; methodology: E.C.S., J.C.L., S.B., and S.C.C; formal analysis: E.C.S., J.C.L., M.D.; investigation: E.C.S., J.C.L., M.D., R.M.H., S.C., S.L.R., D.W., J.C., D.H., S.B., and S.C.C.; resources: S.B., M.K., T.M., A.K., S.M.P.; writing—original draft: E.C.S., J.C.L. and S.C.C.; writing—review & editing: E.C.S., M.D., S.L.R., D.W., D.H., M.K., T.M., S.M.P., S.B., S.C.C; visualization: E.C.S., M.D., J.C.L. All authors approved the final version of the manuscript.
Data availability
Data is available from the authors on reasonable request.
