Research teams from around the world have made remarkable progress in the past decade in elucidating the genomic landscape of most types of childhood cancer. A decade ago it was possible to hope that targetable oncogenes, such as activated tyrosine kinases, might be identified in a high percentage of childhood cancers. However, it is now clear that the genomic landscape of childhood cancer is highly varied, and in many cases is quite distinctive from that of the common adult cancers.
There are examples of genomic lesions that have provided immediate therapeutic direction, including the following:
For some cancers, the genomic findings have been highly illuminating in the identification of genomically defined subsets of patients within histologies that have distinctive biological features and distinctive clinical characteristics (particularly in terms of prognosis). In some instances, identification of these subtypes has resulted in early clinical translation as exemplified by the WNT subgroup of medulloblastoma. Because of its excellent outcome, the WNT subgroup will be studied separately in future medulloblastoma clinical trials so that reductions in therapy can be evaluated with the goal of maintaining favorable outcome while reducing long-term morbidity. However, the prognostic significance of the recurring genomic lesions for some other cancers remains to be defined.
A key finding from genomic studies is the extent to which the molecular characteristics of childhood cancers correlate with their tissue (cell) of origin. As with most adult cancers, mutations in childhood cancers do not arise at random, but rather are linked in specific constellations to disease categories. A few examples include the following:
Another theme across multiple childhood cancers is the contribution of mutations of genes involved in normal development of the tissue of origin of the cancer and the contribution of genes involved in epigenomic regulation.
Structural variations play an important role for many childhood cancers. Translocations resulting in oncogenic fusion genes or overexpression of oncogenes play a central role, particularly for the leukemias and sarcomas. However, for other childhood cancers that are primarily characterized by structural variations, functional fusion genes are not produced. Mechanisms by which these recurring structural variations have oncogenic effects have been identified for osteosarcoma (translocations confined to the first intron of TP53) and medulloblastoma (structural variants juxtapose GFI1 or GFI1B coding sequences proximal to active enhancer elements leading to transcriptional activation [enhancer hijacking]).[1,2] However, the oncogenic mechanisms of action for recurring structural variations of other childhood cancers (e.g., the segmental chromosomal alterations in neuroblastoma) need to be elucidated.
Understanding of the contribution of germline mutations to childhood cancer etiology is being advanced by the application of whole-genome and exome sequencing to cohorts of children with cancer. Estimates for rates of pathogenic germline mutations approaching 10% have emerged from studies applying these sequencing methods to childhood cancer cohorts.[3,4,5] In some cases, the pathogenic germline mutations are clearly contributory to the patient's cancer (e.g., TP53 mutations arising in the context of Li-Fraumeni syndrome), whereas in other cases the contribution of the germline mutation to the patient's cancer is less clear (e.g., mutations in adult cancer predisposition genes such as BRCA1 and BRCA2 that have an undefined role in childhood cancer predisposition).[4,5] The frequency of germline mutations varies by tumor type (e.g., lower for neuroblastoma and higher for osteosarcoma), and many of the identified germline mutations fit into known predisposition syndromes (e.g., DICER1 for pleuropulmonary blastoma, SMARCB1 and SMARCA4 for rhabdoid tumor and small cell ovarian cancer, TP53 for adrenocortical carcinoma and Li-Fraumeni syndrome cancers, RB1 for retinoblastoma, etc.). The germline contribution to the development of specific cancers is discussed in the disease-specific sections that follow.
Each section of this document is meant to provide readers with a brief summary of current knowledge about the genomic landscape of specific childhood cancers, an understanding that is critical in considering how to apply precision medicine concepts to childhood cancers.
Acute Lymphoblastic Leukemia (ALL)
Genomics of childhood ALL
The genomics of childhood ALL has been extensively investigated, and multiple distinctive subtypes have been defined on the basis of cytogenetic and molecular characterizations, each with its own pattern of clinical and prognostic characteristics. Figure 1 illustrates the distribution of ALL cases by cytogenetic/molecular subtype.
Figure 1. Subclassification of childhood ALL. Blue wedges refer to B-progenitor ALL, yellow to recently identified subtypes of B-ALL, and red wedges to T-lineage ALL. Reprinted from Seminars in Hematology, Volume 50, Charles G. Mullighan, Genomic Characterization of Childhood Acute Lymphoblastic Leukemia, Pages 314–324, Copyright (2013), with permission from Elsevier.
The genomic landscape of B-ALL is typified by a range of genomic alterations that disrupt normal B-cell development and, in some cases, by mutations in genes that provide a proliferation signal (e.g., activating mutations in RAS family genes or mutations/translocations leading to kinase pathway signaling). Genomic alterations leading to blockage of B-cell development include translocations (e.g., TCF3-PBX1 and ETV6-RUNX1), point mutations (e.g., IKZF1 and PAX5), and intragenic/intergenic deletions (e.g., IKZF1, PAX5, EBF, and ERG).
The genomic alterations in B-ALL tend not to occur at random, but rather to cluster within subtypes that can be delineated by biological characteristics such as their gene expression profiles. Cases with recurring chromosomal translocations (e.g., TCF3-PBX1, ETV6-RUNX1, and KMT2A [MLL]-rearranged ALL) have distinctive biological features and illustrate this point, as do the examples below of specific genomic alterations within distinctive biological subtypes:
Activating point mutations in kinase genes are uncommon in high-risk B-ALL. JAK genes are the primary kinases that are found to be mutated. These mutations are generally observed in patients with Ph-like ALL that have CRLF2 abnormalities, although JAK2 mutations are also observed in approximately 15% of children with Down syndrome ALL.[4,8,9] Several kinase genes and cytokine receptor genes are activated by translocations, as described below in the discussion of Ph+ ALL and Ph-like ALL. FLT3 mutations occur in a minority of cases (approximately 10%) of hyperdiploid ALL and KMT2A-rearranged ALL, and are rare in other subtypes.
Understanding of the genomics of B-ALL at relapse is less advanced than the understanding of ALL genomics at diagnosis. Childhood ALL is often polyclonal at diagnosis and under the selective influence of therapy, some clones may be extinguished and new clones with distinctive genomic profiles may arise. Of particular importance are new mutations that arise at relapse that may be selected by specific components of therapy. As an example, mutations in NT5C2 are not found at diagnosis, whereas specific mutations in NT5C2 were observed in 7 of 44 (16%) and 9 of 20 (45%) cases of B-ALL with early relapse that were evaluated for this mutation in two studies.[11,12]NT5C2 mutations are uncommon in patients with late relapse, and they appear to induce resistance to mercaptopurine (6-MP) and thioguanine. Another gene that is found mutated only at relapse is PRSP1, a gene involved in purine biosynthesis. Mutations were observed in 13.0% of a Chinese cohort and 2.7% of a German cohort, and were observed in patients with on-treatment relapses. The PRSP1 mutations observed in relapsed cases induce resistance to thiopurines in leukemia cell lines. CREBBP mutations are also enriched at relapse and appear to be associated with increased resistance to glucocorticoids.[11,14] With increased understanding of the genomics of relapse, it may be possible to tailor upfront therapy to avoid relapse or detect resistance-inducing mutations early and intervene before a frank relapse.
A number of recurrent chromosomal abnormalities have been shown to have prognostic significance, especially in B-ALL. Some chromosomal alterations are associated with more favorable outcomes, such as high hyperdiploidy (51–65 chromosomes) and the ETV6-RUNX1 fusion. Other alterations historically have been associated with a poorer prognosis, including the Ph chromosome (t(9;22)(q34;q11.2)), rearrangements of the KMT2A gene, hypodiploidy, and intrachromosomal amplification of the AML1 gene (iAMP21).
In recognition of the clinical significance of many of these genomic alterations, the 2016 revision of the World Health Organization classification of tumors of the hematopoietic and lymphoid tissues lists the following entities for B-ALL:
These and other chromosomal and genomic abnormalities for childhood ALL are described below.
High hyperdiploidy, defined as 51 to 65 chromosomes per cell or a DNA index greater than 1.16, occurs in 20% to 25% of cases of B-ALL, but very rarely in cases of T-ALL. Hyperdiploidy can be evaluated by measuring the DNA content of cells (DNA index) or by karyotyping. In cases with a normal karyotype or in which standard cytogenetic analysis was unsuccessful, interphase fluorescence in situ hybridization (FISH) may detect hidden hyperdiploidy.
High hyperdiploidy generally occurs in cases with clinically favorable prognostic factors (patients aged 1 to <10 years with a low white blood cell [WBC] count) and is an independent favorable prognostic factor.[17,18,19] Within the hyperdiploid range of 51 to 65 chromosomes, patients with higher modal numbers (58–66) appeared to have a better prognosis in one study. Hyperdiploid leukemia cells are particularly susceptible to undergoing apoptosis and accumulate higher levels of methotrexate and its active polyglutamate metabolites, which may explain the favorable outcome commonly observed in these cases.
While the overall outcome of patients with high hyperdiploidy is considered to be favorable, factors such as age, WBC count, specific trisomies, and early response to treatment have been shown to modify its prognostic significance.[21,22]
Patients with trisomies of chromosomes 4, 10, and 17 (triple trisomies) have been shown to have a particularly favorable outcome, as demonstrated by both Pediatric Oncology Group (POG) and Children's Cancer Group analyses of National Cancer Institute (NCI) standard-risk ALL. POG data suggest that NCI standard-risk patients with trisomies of 4 and 10, without regard to chromosome 17 status, have an excellent prognosis.
Chromosomal translocations may be seen with high hyperdiploidy, and in those cases, patients are more appropriately risk-classified on the basis of the prognostic significance of the translocation. For instance, in one study, 8% of patients with the Ph chromosome (t(9;22)(q34;q11.2)) also had high hyperdiploidy, and the outcome of these patients (treated without tyrosine kinase inhibitors) was inferior to that observed in non-Ph+ high hyperdiploid patients.
Certain patients with hyperdiploid ALL may have a hypodiploid clone that has doubled (masked hypodiploidy). These cases may be interpretable based on the pattern of gains and losses of specific chromosomes (hyperdiploidy with two and four copies of chromosomes rather than three copies). These patients have an unfavorable outcome, similar to those with hypodiploidy.
Near triploidy (68–80 chromosomes) and near tetraploidy (>80 chromosomes) are much less common and appear to be biologically distinct from high hyperdiploidy. Unlike high hyperdiploidy, a high proportion of near tetraploid cases harbor a cryptic ETV6-RUNX1 fusion.[28,29,30] Near triploidy and tetraploidy were previously thought to be associated with an unfavorable prognosis, but later studies suggest that this may not be the case.[28,30]
The genomic landscape of hyperdiploid ALL is characterized by mutations in genes of the receptor tyrosine kinase (RTK)/RAS pathway in approximately one-half of cases. Genes encoding histone modifiers are also present in a recurring manner in a minority of cases. Analysis of mutation profiles demonstrates that chromosomal gains are early events in the pathogenesis of hyperdiploid ALL.
B-ALL cases with fewer than the normal number of chromosomes have been subdivided in various ways, with one report stratifying on the basis of modal chromosome number into the following four groups:
Most patients with hypodiploidy are in the near-haploid and low-hypodiploid groups, and both of these groups have an elevated risk of treatment failure compared with nonhypodiploid cases.[27,32] Patients with fewer than 44 chromosomes have a worse outcome than do patients with 44 or 45 chromosomes in their leukemic cells. A number of studies have shown that patients with high minimal residual disease (MRD) (≥0.01%) after induction do very poorly, with 5-year event-free survival (EFS) rates ranging from 25% to 47%. Although hypodiploid patients with low MRD after induction fare better (5-year EFS, 64%–75%), their outcomes are still inferior to most children with other types of ALL.[33,34,35]
The recurring genomic alterations of near-haploid and low-hypodiploid ALL appear to be distinctive from each other and from other types of ALL. In near-haploid ALL, alterations targeting RTK signaling, RAS signaling, and IKZF3 are common. In low-hypodiploid ALL, genetic alterations involving TP53, RB1, and IKZF2 are common. Importantly, the TP53 alterations observed in low-hypodiploid ALL are also present in nontumor cells in approximately 40% of cases, suggesting that these mutations are germline and that low-hypodiploid ALL represents, in some cases, a manifestation of Li-Fraumeni syndrome. Approximately two-thirds of patients with ALL and germline pathogenic TP53 variants have hypodiploid ALL.
Fusion of the ETV6 gene on chromosome 12 to the RUNX1 gene on chromosome 21 is present in 20% to 25% of cases of B-ALL but is rarely observed in T-ALL. The t(12;21)(p13;q22) produces a cryptic translocation that is detected by methods such as FISH, rather than conventional cytogenetics, and it occurs most commonly in children aged 2 to 9 years.[38,39] Hispanic children with ALL have a lower incidence of t(12;21)(p13;q22) than do white children.
Reports generally indicate favorable EFS and overall survival (OS) in children with the ETV6-RUNX1 fusion; however, the prognostic impact of this genetic feature is modified by the following factors:[41,42,43,44,45]
In one study of the treatment of newly diagnosed children with ALL, multivariate analysis of prognostic factors found age and leukocyte count, but not ETV6-RUNX1, to be independent prognostic factors. It does not appear that the presence of secondary cytogenetic abnormalities, such as deletion of ETV6 (12p) or CDKN2A/B (9p), impacts the outcome of patients with the ETV6-RUNX1 fusion.[45,46]
There is a higher frequency of late relapses in patients with ETV6-RUNX1 fusions compared with other relapsed B-ALL patients.[41,47] Patients with the ETV6-RUNX1 fusion who relapse seem to have a better outcome than other relapse patients, with an especially favorable prognosis for patients who relapse more than 36 months from diagnosis. Some relapses in patients with t(12;21)(p13;q22) may represent a new independent second hit in a persistent preleukemic clone (with the first hit being the ETV6-RUNX1 translocation).[50,51]
The Ph chromosome t(9;22)(q34.1;q11.2) is present in approximately 3% of children with ALL and leads to production of a BCR-ABL1 fusion protein with tyrosine kinase activity (refer to Figure 2).
Figure 2. The Philadelphia chromosome is a translocation between the ABL1 oncogene (on the long arm of chromosome 9) and the BCR gene (on the long arm of chromosome 22), resulting in the fusion gene BCR-ABL1. BCR-ABL1 encodes an oncogenic protein with tyrosine kinase activity.
This subtype of ALL is more common in older children with B-ALL and high WBC count, with the incidence of the t(9;22)(q34.1;q11.2) increasing to about 25% in young adults with ALL.
Historically, the Ph chromosome t(9;22)(q34.1;q11.2) was associated with an extremely poor prognosis (especially in those who presented with a high WBC count or had a slow early response to initial therapy), and its presence had been considered an indication for allogeneic hematopoietic stem cell transplantation (HSCT) in patients in first remission.[25,52,53,54] Inhibitors of the BCR-ABL1 tyrosine kinase, such as imatinib mesylate, are effective in patients with Ph+ ALL. A study by the Children's Oncology Group (COG), which used intensive chemotherapy and concurrent imatinib mesylate given daily, demonstrated a 5-year EFS rate of 70% (± 12%), which was superior to the EFS rate of historical controls in the pre-tyrosine kinase inhibitor (imatinib mesylate) era.[56,57]
Rearrangements involving the KMT2A gene occur in approximately 5% of childhood ALL cases overall, but in up to 80% of infants with ALL. These rearrangements are generally associated with an increased risk of treatment failure.[58,59,60,61] The t(4;11)(q21;q23) is the most common rearrangement involving the KMT2A gene in children with ALL and occurs in approximately 1% to 2% of childhood ALL.[59,62]
Patients with the t(4;11)(q21;q23) are usually infants with high WBC counts; they are more likely than other children with ALL to have central nervous system (CNS) disease and to have a poor response to initial therapy. While both infants and adults with the t(4;11)(q21;q23) are at high risk of treatment failure, children with the t(4;11)(q21;q23) appear to have a better outcome than either infants or adults.[58,59] Irrespective of the type of KMT2A gene rearrangement, infants with leukemia cells that have KMT2A gene rearrangements have a worse treatment outcome than older patients whose leukemia cells have a KMT2A gene rearrangement.[58,59]
Whole-genome sequencing has determined that cases of infant ALL with KMT2A gene rearrangements have few additional genomic alterations, none of which have clear clinical significance. Deletion of the KMT2A gene has not been associated with an adverse prognosis.
Of interest, the t(11;19)(q23;p13.3) involving KMT2A and MLLT1/ENL occurs in approximately 1% of ALL cases and occurs in both early B-lineage and T-ALL. Outcome for infants with the t(11;19) is poor, but outcome appears relatively favorable in older children with T-ALL and t(11;19).
The t(1;19) occurs in approximately 5% of childhood ALL cases and involves fusion of the TCF3 gene on chromosome 19 to the PBX1 gene on chromosome 1.[66,67] The t(1;19) may occur as either a balanced translocation or as an unbalanced translocation and is the primary recurring genomic alteration of the pre-B–ALL immunophenotype (cytoplasmic immunoglobulin positive). Black children are relatively more likely than white children to have pre-B–ALL with the t(1;19).
The t(1;19) had been associated with inferior outcome in the context of antimetabolite-based therapy, but the adverse prognostic significance was largely negated by more aggressive multiagent therapies.[67,71] However, in a trial conducted by St. Jude Children's Research Hospital (SJCRH) on which all patients were treated without cranial radiation, patients with the t(1;19) had an overall outcome comparable to children lacking this translocation, with a higher risk of CNS relapse and a lower rate of bone marrow relapse, suggesting that more intensive CNS therapy may be needed for these patients.[72,73]
The t(17;19) resulting in the TCF3-HLF fusion occurs in less than 1% of pediatric ALL cases. ALL with the TCF3-HLF fusion is associated with disseminated intravascular coagulation and hypercalcemia at diagnosis. Outcome is very poor for children with the t(17;19), with a literature review noting mortality for 20 of 21 cases reported. In addition to the TCF3-HLF fusion, the genomic landscape of this ALL subtype was characterized by deletions in genes involved in B-cell development (PAX5, BTG1, and VPREB1) and by mutations in RAS pathway genes (NRAS, KRAS, and PTPN11).
Approximately 5% of standard-risk and 10% of high-risk pediatric B-ALL patients have a rearrangement involving DUX4 that leads to its overexpression.[5,6] The frequency in older adolescents (aged >15 years) is approximately 10%. The most common rearrangement produces IGH-DUX4 fusions, with ERG-DUX4 fusions also observed.DUX4-rearranged cases show a distinctive gene expression pattern that was initially identified as being associated with focal deletions in ERG,[75,76,77,78] and one-half to more than two-thirds of these cases have focal intragenic deletions involving ERG that are not observed in other ALL subtypes.[5,75]ERG deletions often appear to be clonal, but using sensitive detection methodology, it appears that most cases are polyclonal.IKZF1 alterations are observed in 20% to 40% of DUX4-rearranged ALL.[5,6]
ERG deletion connotes an excellent prognosis, with OS rates exceeding 90%; even when the IZKF1 deletion is present, prognosis remains highly favorable.[76,77,78] While DUX4-rearranged ALL has an overall favorable prognosis, there is uncertainty as to whether this applies to both ERG-deleted and ERG-intact cases. In a study of 50 patients with DUX4-rearranged ALL, patients with ERG deletion detected by genomic polymerase chain reaction (PCR) (n = 33) had a more favorable EFS rate of approximately 90% than did patients with intact ERG (n = 17), with an EFS rate of approximately 70%.
Gene fusions involving MEF2D, a transcription factor that is expressed during B-cell development, are observed in approximately 4% of childhood ALL cases.[79,80] Although multiple fusion partners may occur, most cases involve BCL9, which is located on chromosome 1q21, as is MEF2D.[79,81] The interstitial deletion producing the MEF2D-BCL9 fusion is too small to be detected by conventional cytogenetic methods. Cases with MEF2D gene fusions show a distinctive gene expression profile, except for rare cases with MEF2D-CSFR1 that have a Ph-like gene expression profile.[79,82]
The median age at diagnosis for cases of MEF2D-rearranged ALL in studies that included both adult and pediatric patients was 12 to 14 years.[79,80] For 22 children with MEF2D-rearranged ALL enrolled in a high-risk ALL clinical trial, the 5-year EFS rate was 72% (standard error, ± 10%), which was inferior to that for other patients.
ZNF384 is a transcription factor that is rearranged in approximately 4% to 5% of pediatric B-ALL cases.[79,83,84] Multiple fusion partners for ZNF384 have been reported, including ARID1B, CREBBP, EP300, SMARCA2, TAF15, and TCF3. Regardless of the fusion partner, ZNF384-rearranged ALL cases show a distinctive gene expression profile.[79,83,84]ZNF384 rearrangement does not appear to confer independent prognostic significance.[79,83,84] The immunophenotype of B-ALL with ZNF384 rearrangement is characterized by weak or negative CD10 expression, with expression of CD13 and/or CD33 commonly observed.[83,84] Cases of mixed phenotype acute leukemia (MPAL) (B/myeloid) that have ZNF384 gene fusions have been reported, [85,86] and a genomic evaluation of MPAL found that ZNF384 gene fusions were present in approximately one-half of B/myeloid cases.
This entity is included in the 2016 revision of the WHO classification of tumors of the hematopoietic and lymphoid tissues. The finding of t(5;14)(q31.1;q32.3) in patients with ALL and hypereosinophilia in the 1980s was followed by the identification of the IL3-IGH fusion as the underlying genetic basis for the condition.[88,89] The joining of the IGH locus to the promoter region of the IL3 gene leads to dysregulation of IL3 expression. Cytogenetic abnormalities in children with ALL and eosinophilia are variable, with only a subset resulting from the IL3-IGH fusion.
The number of cases of IL3-IGH ALL described in the published literature is too small to assess the prognostic significance of the IL3-IGH fusion. Diagnosis of cases of IL3-IGH ALL may be delayed because it can present with hypereosinophilia in the absence of cytopenias and circulating blasts.
iAMP21 is generally diagnosed using FISH and is defined by the presence of greater than or equal to five RUNX1 signals per cell (or ≥3 extra copies of RUNX1 on a single abnormal chromosome). It occurs in approximately 2% of B-ALL cases and is associated with older age (median, approximately 10 years), presenting WBC of less than 50 × 109 /L, a slight female preponderance, and high end-induction MRD.[92,93,94]
The United Kingdom Acute Lymphoblastic Leukaemia (UKALL) clinical trials group initially reported that the presence of iAMP21 conferred a poor prognosis in patients treated in the MRC ALL 97/99 trial (5-year EFS, 29%). In their subsequent trial (UKALL2003 [NCT00222612]), patients with iAMP21 were assigned to a more intensive chemotherapy regimen and had a markedly better outcome (5-year EFS, 78%). Similarly, the COG has reported that iAMP21 was associated with a significantly inferior outcome in NCI standard-risk patients (4-year EFS, 73% for iAMP21 vs. 92% in others), but not in NCI high-risk patients (4-year EFS, 73% vs. 80%). On multivariate analysis, iAMP21 was an independent predictor of inferior outcome only in NCI standard-risk patients. The results of the UKALL2003 and COG studies suggest that treatment of iAMP21 patients with high-risk chemotherapy regimens abrogates its adverse prognostic significance and obviates the need for HSCT in first remission.
Gene expression analysis identified two distinctive ALL subsets with PAX5 genomic alterations, termed PAX5alt and PAX5 p.Pro80Arg. The alterations in the PAX5alt subtype included rearrangements, sequence mutations, and focal intragenic amplifications.
PAX5alt. PAX5 rearrangements have been reported to represent 2% to 3% of pediatric ALL. More than 20 partner genes for PAX5 have been described, with PAX5-ETV6, the primary genomic alteration in dic(9;12)(p13;p13), being the most common gene fusion.
Intragenic amplification of PAX5 was identified in approximately 1% of B-ALL cases, and it was usually detected in cases lacking known leukemia-driver genomic alterations. Cases with PAX5 amplification show male predominance (66%), with most (55%) having NCI high-risk status. For a cohort of patients with PAX5 amplification diagnosed between 1993 and 2015, the 5-year EFS rate was 49% (95% confidence interval [CI], 36%–61%), and the OS rate was 67% (95% CI, 54%–77%), suggesting a relatively poor prognosis for this B-ALL subtype.
PAX5 p.Pro80Arg. PAX5 with a p.Pro80Arg mutation shows a gene expression profile distinctive from that of other cases with PAX5 alterations. Cases with PAX5 p.Pro80Arg appear to be more common in the adolescent and young adult (AYA) and adult populations (3%–4% frequency) than in children with NCI standard-risk or high-risk ALL (0.4% and 1.9% frequency, respectively). Outcome for the pediatric patients with PAX5 p.Pro80Arg and PAX5alt treated on a COG clinical trial appears to be intermediate (5-year EFS, approximately 75%).
BCR-ABL1–negative patients with a gene expression profile similar to BCR-ABL1–positive patients have been referred to as Ph-like.[99,100,101] This occurs in 10% to 20% of pediatric ALL patients, increasing in frequency with age, and has been associated with an IKZF1 deletion or mutation.[8,99,100,102,103]
Retrospective analyses have indicated that patients with Ph-like ALL have a poor prognosis.[4,99] In one series, the 5-year EFS for NCI high-risk children and adolescents with Ph-like ALL was 58% and 41%, respectively. While it is more frequent in older and higher-risk patients, the Ph-like subtype has also been identified in NCI standard-risk patients. In a COG study, 13.6% of 1,023 NCI standard-risk B-ALL patients were found to have Ph-like ALL; these patients had an inferior EFS compared with non–Ph-like standard-risk patients (82% vs. 91%), although no difference in OS (93% vs. 96%) was noted. In one study of 40 Ph-like patients, the adverse prognostic significance of this subtype appeared to be abrogated when patients were treated with risk-directed therapy on the basis of MRD levels.
Many of the remaining cases of Ph-like ALL have been noted to have a series of translocations with a common theme of involvement of kinases, including ABL1, ABL2, CSF1R, JAK2, and PDGFRB.[4,102] Fusion proteins from these gene combinations have been noted in some cases to be transformative and have responded to tyrosine kinase inhibitors both in vitro and in vivo, suggesting potential therapeutic strategies for these patients. The prevalence of targetable kinase fusions in Ph-like ALL is lower in NCI standard-risk patients (3.5%) than in NCI high-risk patients (approximately 30%). Point mutations in kinase genes, aside from those in JAK1 and JAK2, are uncommon in Ph-like ALL cases.
Approximately 9% of Ph-like ALL cases result from rearrangements that lead to overexpression of a truncated erythropoietin receptor (EPOR). The C-terminal region of the receptor that is lost is the region that is mutated in primary familial congenital polycythemia and that controls stability of the EPOR. The portion of the EPOR remaining is sufficient for JAK-STAT activation and for driving leukemia development.
CRLF2. Genomic alterations in CRLF2, a cytokine receptor gene located on the pseudoautosomal regions of the sex chromosomes, have been identified in 5% to 10% of cases of B-ALL; they represent approximately 50% of cases of Ph-like ALL.[109,110,111] The chromosomal abnormalities that commonly lead to CRLF2 overexpression include translocations of the IGH locus (chromosome 14) to CRLF2 and interstitial deletions in pseudoautosomal regions of the sex chromosomes, resulting in a P2RY8-CRLF2 fusion.[8,106,109,110] These two genomic alterations are associated with distinctive clinical and biological characteristics.
The P2RY8-CRLF2 fusion is observed in 70% to 75% of pediatric patients with CRLF2 genomic alterations, and it occurs in younger patients (median age, approximately 4 years vs. 14 years for patients with IGH-CRLF2).[112,113]P2RY8-CRLF2 occurs not infrequently with established chromosomal abnormalities (e.g., hyperdiploidy, iAMP21, dic(9;20)), while IGH-CRLF2 is generally mutually exclusive with known cytogenetic subgroups. CRLF2 genomic alterations are observed in approximately 60% of patients with Down syndrome ALL, with P2RY8-CRLF2 fusions being more common than IGH-CRLF2 (approximately 80% vs. 20%).[110,112]
IGH-CRLF2 and P2RY8-CRLF2 commonly occur as an early event in B-ALL development and show clonal prevalence. However, in some cases they appear to be a late event and show subclonal prevalence. Loss of the CRLF2 genomic abnormality in some cases at relapse confirms the subclonal nature of the alteration in these cases.[112,115]
CRLF2 abnormalities are strongly associated with the presence of IKZF1 deletions. Other recurring genomic alterations found in association with CRLF2 alterations include deletions in genes associated with B-cell differentiation (e.g., PAX5, BTG1, EBF1, etc.) and cell cycle control (CDKN2A), as well as genomic alterations activating JAK-STAT pathway signaling (e.g., IL7R and JAK mutations).[4,106,107,110,116]
Although the results of several retrospective studies suggest that CRLF2 abnormalities may have adverse prognostic significance in univariate analyses, most do not find this abnormality to be an independent predictor of outcome.[106,109,110,117,118] For example, in a large European study, increased expression of CRLF2 was not associated with unfavorable outcome in multivariate analysis, while IKZF1 deletion and Ph-like expression signatures were associated with unfavorable outcome. Controversy exists about whether the prognostic significance of CRLF2 abnormalities should be analyzed on the basis of CRLF2 overexpression or on the presence of CRLF2 genomic alterations.[117,118]
IKZF1 deletions, including deletions of the entire gene and deletions of specific exons, are present in approximately 15% of B-ALL cases. Less commonly, IKZF1 can be inactivated by deleterious point mutations.
Cases with IKZF1 deletions tend to occur in older children, have a higher WBC count at diagnosis, and are therefore, more common in NCI high-risk patients than in NCI standard-risk patients.[2,100,116,119] A high proportion of Ph-like cases have a deletion of IKZF1,[3,116] and ALL arising in children with Down syndrome appears to have elevated rates of IKZF1 deletions.IKZF1 deletions are also common in cases with CRLF2 genomic alterations and in Ph-like ALL.[76,99,116]
Multiple reports have documented the adverse prognostic significance of an IKZF1 deletion, and most studies have reported that this deletion is an independent predictor of poor outcome in multivariate analyses.[76,99,100,103,116,121,122,123,124,125,126,127]; [Level of evidence: 2Di] However, the prognostic significance of IKZF1 may not apply equally across ALL biological subtypes, as illustrated by the apparent lack of prognostic significance in patients with ERG deletion.[76,77,78] Similarly, the prognostic significance of the IKZF1 deletion also appeared to be minimized in a cohort of COG patients with DUX4-rearranged ALL and with ERG transcriptional dysregulation that frequently occurred by ERG deletion. The Associazione Italiana di Ematologia e Oncologia Pediatrica (AIEOP)–Berlin-Frankfurt-Münster (BFM) group reported that IKZF1 deletions were significant adverse prognostic factors only in B-ALL patients with high end-induction MRD and in whom co-occurrence of deletions of CDKN2A, CDKN2B, PAX5, or PAR1 (in the absence of ERG deletion) were identified.
There are few published results of changing therapy on the basis of IKZF1 gene status. The Malaysia-Singapore group published results of two consecutive trials. In the first trial (MS2003), IKZF1 status was not considered in risk stratification, while in the subsequent trial (MS2010), IKZF1-deleted patients were excluded from the standard-risk group. Thus, more IKZF1-deleted patients in the MS2010 trial received intensified therapy. Patients with IKZF1-deleted ALL had improved outcomes in MS2010 compared with patients in MS2003, but interpretation of this observation is limited by other changes in risk stratification and therapeutic differences between the two trials.[Level of evidence: 2A]
T-ALL is characterized by genomic alterations leading to activation of transcriptional programs related to T-cell development and by a high frequency of cases (approximately 60%) with mutations in NOTCH1 and/or FBXW7 that result in activation of the NOTCH1 pathway. In contrast to B-ALL, the prognostic significance of T-ALL genomic alterations is less well-defined. Cytogenetic abnormalities common in B-lineage ALL (e.g., hyperdiploidy, 51–65 chromosomes) are rare in T-ALL.[132,133]
Notch pathway signaling is commonly activated by NOTCH1 and FBXW7 gene mutations in T-ALL, and these are the most commonly mutated genes in pediatric T-ALL.[131,134]NOTCH1-activating gene mutations occur in approximately 50% to 60% of T-ALL cases, and FBXW7-inactivating gene mutations occur in approximately 15% of cases, with the result that approximately 60% of cases have Notch pathway activation by mutations in at least one of these genes.
The prognostic significance of NOTCH1/FBXW7 mutations may be modulated by genomic alterations in RAS and PTEN. The French Acute Lymphoblastic Leukaemia Study Group (FRALLE) and the Group for Research on Adult Acute Lymphoblastic Leukemia groups reported that patients having mutated NOTCH1/FBXW7 and wild-type PTEN/RAS constituted a favorable-risk group while patients with PTEN or RAS mutations, regardless of NOTCH1/FBXW7 status, have a significantly higher risk of treatment failure.[136,137] In the FRALLE study, 5-year cumulative incidence of relapse and disease-free survival (DFS) were 50% and 46% for patients with mutated NOTCH1/FBXW7 and mutated PTEN/RAS versus 13% and 87% for patients with mutated NOTCH1/FBXW7 and wild-type PTEN/RAS. The overall 5-year DFS in the FRALLE study was 73%, and additional research is needed to determine whether the same prognostic significance for NOTCH1/FBXW7 and PTEN/RAS mutations will apply to current treatment regimens, which produce overall 5-year DFS rates that approach 90%.
Multiple chromosomal translocations have been identified in T-ALL that lead to deregulated expression of the target genes. These chromosome rearrangements fuse genes encoding transcription factors (e.g., TAL1/TAL2, LMO1 and LMO2, LYL1, TLX1, TLX3, NKX2-I, HOXA, and MYB) to one of the T-cell receptor loci (or to other genes) and result in deregulated expression of these transcription factors in leukemia cells.[131,132,139,140,141,142,143] These translocations are often not apparent by examining a standard karyotype, but can be identified using more sensitive screening techniques, including FISH or PCR. Mutations in a noncoding region near the TAL1 gene that produce a super-enhancer upstream of TAL1 represent nontranslocation genomic alterations that can also activate TAL1 transcription to induce T-ALL.
Translocations resulting in chimeric fusion proteins are also observed in T-ALL.
Early T-cell precursor ALL cytogenetics/genomics
Detailed molecular characterization of early T-cell precursor ALL showed this entity to be highly heterogeneous at the molecular level, with no single gene affected by mutation or copy number alteration in more than one-third of cases. Compared with other T-ALL cases, the early T-cell precursor group had a lower rate of NOTCH1 mutations and significantly higher frequencies of alterations in genes regulating cytokine receptors and RAS signaling, hematopoietic development, and histone modification. The transcriptional profile of early T-cell precursor ALL shows similarities to that of normal hematopoietic stem cells and myeloid leukemia stem cells.
Studies have found that the absence of biallelic deletion of the TCR-gamma locus (ABD), as detected by comparative genomic hybridization and/or quantitative DNA-PCR, was associated with early treatment failure in patients with T-ALL.[154,155] ABD is characteristic of early thymic precursor cells, and many of the T-ALL patients with ABD have an immunophenotype consistent with the diagnosis of early T-cell precursor phenotype.
Mixed phenotype acute leukemia (MPAL) cytogenetics/genomics
|NOS = not otherwise specified.|
|a Adapted from Béné MC: Biphenotypic, bilineal, ambiguous or mixed lineage: strange leukemias! Haematologica 94 (7): 891-3, 2009.Obtained from Haematologica/the Hematology Journal websitehttp://www.haematologica.org.|
|Acute undifferentiated leukemia||Acute leukemia that does not express any marker considered specific for either lymphoid or myeloid lineage|
|Mixed phenotype acute leukemia with t(9;22)(q34;q11.2);BCR-ABL1(MPAL withBCR-ABL1)||Acute leukemia meeting the diagnostic criteria for mixed phenotype acute leukemia in which the blasts also have the (9;22) translocation or theBCR-ABL1rearrangement|
|Mixed phenotype acute leukemia with t(v;11q23);KMT2A(MLL) rearranged (MPAL withKMT2A)||Acute leukemia meeting the diagnostic criteria for mixed phenotype acute leukemia in which the blasts also have a translocation involving theKMT2Agene|
|Mixed phenotype acute leukemia, B/myeloid, NOS (B/M MPAL)||Acute leukemia meeting the diagnostic criteria for assignment to both B and myeloid lineage, in which the blasts lack genetic abnormalities involvingBCR-ABL1orKMT2A|
|Mixed phenotype acute leukemia, T/myeloid, NOS (T/M MPAL)||Acute leukemia meeting the diagnostic criteria for assignment to both T and myeloid lineage, in which the blasts lack genetic abnormalities involvingBCR-ABL1orKMT2A|
|Mixed phenotype acute leukemia, B/myeloid, NOS—rare types||Acute leukemia meeting the diagnostic criteria for assignment to both B and T lineage|
|Other ambiguous lineage leukemias||Natural killer–cell lymphoblastic leukemia/lymphoma|
|a Adapted from Arber et al.|
|b Strong defined as equal to or brighter than the normal B or T cells in the sample.|
|Myeloid lineage||Myeloperoxidase (flow cytometry, immunohistochemistry, or cytochemistry);or monocytic differentiation (at least two of the following: nonspecific esterase cytochemistry, CD11c, CD14, CD64, lysozyme)|
|T lineage||Strongb cytoplasmic CD3 (with antibodies to CD3 epsilon chain);or surface CD3|
|B lineage||Strongb CD19 with at least one of the following strongly expressed: CD79a, cytoplasmic CD22, or CD10;or weak CD19 with at least two of the following strongly expressed: CD79a, cytoplasmic CD22, or CD10|
The classification system for MPAL includes two entities that are defined by their primary molecular alteration: MPAL with BCR-ABL1 translocation and MPAL with KMT2A rearrangement. The genomic alterations associated with the MPAL, B/myeloid, NOS (B/M MPAL) and MPAL, T/myeloid, NOS (T/M MPAL) entities are distinctive, as described below:
Gene polymorphisms in drug metabolic pathways
Patients with mutant phenotypes of TPMT (a gene involved in the metabolism of thiopurines such as mercaptopurine) appear to have more favorable outcomes, although such patients may also be at higher risk of developing significant treatment-related toxicities, including myelosuppression and infection.[162,163] Patients with homozygosity for TPMT variants associated with low enzymatic activity tolerate only very low doses of mercaptopurine (approximately 10% of the standard dose) and are treated with reduced doses of mercaptopurine to avoid excessive toxicity. Patients who are heterozygous for this mutant enzyme gene generally tolerate mercaptopurine without serious toxicity, but they do require more frequent dose reductions for hematologic toxicity than do patients who are homozygous for the normal allele.[164,165]
Germline variants in NUDT15 that reduce or abolish activity of this enzyme also lead to diminished tolerance to thiopurines.[164,166] The variants are most common in East Asians and Hispanics, and they are rare in Europeans and Africans. Patients homozygous for the risk variants tolerate only very low doses of mercaptopurine, while patients heterozygous for the risk alleles tolerate lower doses than do patients homozygous for the wild-type allele (approximately 25% dose reduction on average), but there is broad overlap in tolerated doses between the two groups.[164,167]
Gene polymorphisms may also affect the expression of proteins that play central roles in the cellular effects of anticancer drugs. As an example, patients who are homozygous for a polymorphism in the promoter region of CEP72 (a centrosomal protein involved in microtubule formation) are at increased risk of vincristine neurotoxicity.
Genome-wide polymorphism analysis has identified specific single nucleotide polymorphisms associated with high end-induction MRD and risk of relapse. Polymorphisms of interleukin-15, as well as genes associated with the metabolism of etoposide and methotrexate, were significantly associated with treatment response in two large cohorts of ALL patients treated on SJCRH and COG protocols. Polymorphic variants involving the reduced folate carrier and methotrexate metabolism have been linked to toxicity and outcome.[170,171] While these associations suggest that individual variations in drug metabolism can affect outcome, few studies have attempted to adjust for these variations; it is unknown whether individualized dose modification on the basis of these findings will improve outcome.
(Refer to the PDQ summary on Childhood Acute Lymphoblastic Leukemia Treatment for information about the treatment of childhood ALL.)
Acute Myeloid Leukemia (AML)
Molecular features of acute myeloid leukemia
Genetic analysis of leukemia blast cells (using both conventional cytogenetic methods and molecular methods) is performed on children with AML because both chromosomal and molecular abnormalities are important diagnostic and prognostic markers.[174,175,176,177,178] Clonal chromosomal abnormalities are identified in the blasts of about 75% of children with AML and are useful in defining subtypes with both prognostic and therapeutic significance.
Detection of molecular abnormalities can also aid in risk stratification and treatment allocation. For example, mutations of NPM and CEBPA are associated with favorable outcomes while certain mutations of FLT3 portend a high risk of relapse, and identifying the latter mutations may allow for targeted therapy.[179,180,181,182]
The 2016 revision to the World Health Organization (WHO) classification of myeloid neoplasms and acute leukemia emphasizes that recurrent chromosomal translocations in pediatric AML may be unique or have a different prevalence than in adult AML. The pediatric AML chromosomal translocations that are found by conventional chromosome analysis and those that are cryptic (identified only with fluorescence in situ hybridization or molecular techniques) occur at higher rates than in adults. These recurrent translocations are summarized in Table 3.[16,173] Table 3 also shows, in the bottom three rows, additional relatively common recurrent translocations observed in children with AML.[173,176,177,183]
|Gene Fusion Product||Chromosomal Translocation||Prevalence in Pediatric AML (%)|
|a Cryptic chromosomal translocation.|
|CBFB-MYH11||inv(16)(p13.1;q22) or t(16;16)(p13.1;q22)||4–9|
The genomic landscape of pediatric AML cases can change from diagnosis to relapse, with mutations detectable at diagnosis dropping out at relapse and, conversely, with new mutations appearing at relapse. In a study of 20 cases for which sequencing data were available at diagnosis and relapse, a key finding was that the variant allele frequency at diagnosis strongly correlated with persistence of mutations at relapse. Approximately 90% of the diagnostic variants with variant allele frequency greater than 0.4 persisted to relapse, compared with only 28% with variant allele frequency less than 0.2 (P < .001). This observation is consistent with previous results showing that presence of a mutation in the FLT3 gene resulting from internal tandem duplications (ITD) predicted for poor prognosis only when there was a high FLT3 ITD allelic ratio.
Specific recurring cytogenetic and molecular abnormalities are briefly described below. The abnormalities are listed by those in clinical use that identify patients with favorable or unfavorable prognosis, followed by other abnormalities. The nomenclature of the 2016 revision to the WHO classification of myeloid neoplasms and acute leukemia is incorporated for disease entities where relevant.
Genetic abnormalities associated with a favorable prognosis
Genetic abnormalities associated with a favorable prognosis include the following:
Both RUNX1-RUNX1T1 and CBFB-MYH11 subtypes commonly show mutations in genes that activate receptor tyrosine kinase signaling (e.g., NRAS, FLT3, and KIT); NRAS and KIT are the most commonly mutated genes for both subtypes. The prognostic significance of activating KIT mutations in adults with CBF AML has been studied with conflicting results. A meta-analysis found that KIT mutations appear to increase the risk of relapse without an impact on OS for adults with RUNX1-RUNX1T1 AML.KIT mutations are often subclonal in children and adults with CBF AML;[193,194] and in adults with RUNX1-RUNX1T1 AML, higher KIT-mutant allele ratio appears to be associated with higher risk of treatment failure.[189,193] The prognostic significance of KIT mutations in pediatric CBF AML remains unclear; some studies have found no impact of KIT mutations on outcome,[195,196,197] while other studies have reported a higher risk of treatment failure when KIT mutations are present.[194,198,199,200,201]
Utilization of quantitative reverse transcriptase–polymerase chain reaction (RT-PCR) for PML-RARA transcripts has become standard practice. Quantitative RT-PCR allows identification of the three common transcript variants and is used for monitoring response on treatment and early detection of molecular relapse. Other much less common translocations involving the retinoic acid receptor alpha can also result in APL (e.g., t(11;17)(q23;q21) involving the PLZF gene).[206,207] Identification of cases with the t(11;17)(q23;q21) is important because of their decreased sensitivity to tretinoin.
Studies of children with AML suggest a lower rate of occurrence of NPM1 mutations in children compared with adults with normal cytogenetics. NPM1 mutations occur in approximately 8% of pediatric patients with AML and are uncommon in children younger than 2 years.[179,180,214,215]NPM1 mutations are associated with a favorable prognosis in patients with AML characterized by a normal karyotype.[179,180,215] For the pediatric population, conflicting reports have been published regarding the prognostic significance of an NPM1 mutation when a FLT3 ITD mutation is also present. One study reported that an NPM1 mutation did not completely abrogate the poor prognosis associated with having a FLT3 ITD mutation,[179,216] but other studies showed no impact of a FLT3 ITD mutation on the favorable prognosis associated with an NPM1 mutation.[173,180,215]
CEBPA mutations occur in 5% to 8% of children with AML and have been preferentially found in the cytogenetically normal subtype of AML with FAB M1 or M2; 70% to 80% of pediatric patients have double-mutant alleles, which is predictive of a significantly improved survival, similar to the effect observed in adult studies.[181,223] Although both double-mutant and single-mutant alleles of CEBPA were associated with a favorable prognosis in children with AML in one large study, a second study observed inferior outcome for patients with single CEBPA mutations. However, very low numbers of children with single-allele mutants were included in these two studies (only 13 total patients), which makes a conclusion regarding the prognostic significance of single-allele CEBPA mutations in children premature. In newly diagnosed patients with double-mutant CEBPA AML, germline screening should be considered in addition to usual family history queries, because 5% to 10% of these patients are reported to have a germline CEBPA mutation.CSF3R mutations commonly occur in a subset of patients with mutant CEBPA AML, and when present, may be associated with increased risk of treatment failure but without an impact on OS.
GATA1 mutations confer increased sensitivity to cytarabine by down-regulating cytidine deaminase expression, possibly providing an explanation for the superior outcome of children with Down syndrome and M7 AML when treated with cytarabine-containing regimens.
Genetic abnormalities associated with an unfavorable prognosis
Genetic abnormalities associated with an unfavorable prognosis include the following:
In the past, patients with del(7q) were also considered to be at high risk of treatment failure, and data from adults with AML support a poor prognosis for both del(7q) and monosomy 7. However, outcome for children with del(7q), but not monosomy 7, appears comparable to that of other children with AML.[177,234] The presence of del(7q) does not abrogate the prognostic significance of favorable cytogenetic characteristics (e.g., inv(16) and t(8;21)).[175,234]
Chromosome 5 and 7 abnormalities appear to lack prognostic significance in AML patients with Down syndrome who are aged 4 years and younger.
Abnormalities involving MECOM can be detected in some AML cases with other 3q abnormalities and are also associated with poor prognosis.
The prognostic significance of FLT3 ITD is modified by the presence of other recurring genomic alterations. The prevalence of FLT3 ITD is increased in certain genomic subtypes of pediatric AML, including those with the NUP98-NSD1 fusion gene, of which 80% to 90% have FLT3 ITD.[246,247] Approximately 15% of patients with FLT3 ITD have NUP98-NSD1, and patients with both FLT3 ITD and NUP98-NSD1 have a poorer prognosis than do patients who have FLT3 ITD without NUP98-NSD1. For patients who have FLT3 ITD, the presence of either WT1 mutations or NUP98-NSD1 fusions is associated with poorer outcome (EFS rates below 25%) than for patients who have FLT3 ITD without these alterations. Conversely, when FLT3 ITD is accompanied by NPM1 mutations, the outcome is relatively favorable and is similar to that of pediatric AML cases without FLT3 ITD.
For APL, FLT3 ITD and point mutations occur in 30% to 40% of children and adults.[242,244,248,249] Presence of the FLT3 ITD mutation is strongly associated with the microgranular variant (M3v) of APL and with hyperleukocytosis.[248,250,251,252] It remains unclear whether FLT3 mutations are associated with poorer prognosis in patients with APL who are treated with modern therapy that includes tretinoin and arsenic trioxide.[249,251,253,254,255,256]
Activating point mutations of FLT3 have also been identified in both adults and children with AML, although the clinical significance of these mutations is not clearly defined. Some of these point mutations appear to be specific to pediatric patients.
Other genetic abnormalities observed in pediatric AML
Other genetic abnormalities observed in pediatric AML include the following:
The most common translocation, representing approximately 50% of KMT2A-rearranged cases in the pediatric AML population, is t(9;11)(p22;q23), in which the KMT2A gene is fused with MLLT3 gene. The 2016 revision to the WHO classification defined AML with t(9;11)(p21.3;q23.3); MLLT3-KMT2A as a distinctive disease entity. However, more than 50 different fusion partners have been identified for the KMT2A gene in patients with AML.
The median age for 11q23/KMT2A-rearranged cases in children is approximately 2 years, and most translocation subgroups have a median age at presentation of younger than 5 years. However, significantly older median ages are seen at presentation of pediatric cases with t(6;11)(q27;q23) (12 years) and t(11;17)(q23;q21) (9 years).
Outcome for patients with de novo AML and KMT2A gene rearrangement is generally reported as being similar to that for other patients with AML.[175,176,259,260] However, as the KMT2A gene can participate in translocations with many different fusion partners, the specific fusion partner appears to influence prognosis, as demonstrated by a large international retrospective study evaluating outcome for 756 children with 11q23- or KMT2A-rearranged AML. For example, cases with t(1;11)(q21;q23), representing 3% of all 11q23/KMT2A-rearranged AML, showed a highly favorable outcome, with a 5-year event-free survival (EFS) rate of 92%.
While reports from single clinical trial groups have variably described more favorable prognosis for patients with AML who have t(9;11)(p21.3;q23.3)/MLLT3-KMT2A, the international retrospective study did not confirm the favorable prognosis for this subgroup.[175,176,259] An international collaboration evaluating pediatric AMKL patients observed that the presence of t(9;11), which was seen in approximately 5% of AMKL cases, was associated with an inferior outcome compared with other AMKL cases.
KMT2A-rearranged AML subgroups that appear to be associated with poor outcome include the following:
In a study of approximately 2,000 children with AML, the CBFA2T3-GLIS2 fusion was identified in 39 cases (1.9%), with a median age at presentation of 1.5 years, and with all cases observed in children younger than 3 years. Approximately one-half of cases had M7 megakaryoblastic morphology, and 29% of patients were black or African American (exceeding the 12.8% frequency in patients lacking the fusion). The CBFA2T3-GLIS2 fusion was an independent prognostic factor for both OS and EFS. The OS rate at 5 years was 22% for patients with CBFA2T3-GLIS2 versus 63% for fusion-negative patients. CBFA2T3-GLIS2 leukemia cells have a distinctive immunophenotype (initially reported as the RAM phenotype),[276,277] with high CD56, dim or negative expression of CD45 and CD38, and a lack of HLA-DR expression.
An international collaborative retrospective study of 51 t(1;22) cases reported that patients with this abnormality had a 5-year EFS rate of 54.5% and an OS rate of 58.2%, similar to the rates for other children with AMKL. In another international retrospective analysis of 153 cases with non–Down syndrome AMKL who had samples available for molecular analysis, the 4-year EFS rate for patients with t(1;22) was 59% and the OS rate was 70%, significantly better than AMKL patients with other specific genetic abnormalities (CBFA2T3/GUS2, NUP98/KDM5A, KMT2A rearrangements, monosomy 7).
A substantial proportion of infants diagnosed with t(8;16) AML in the first month of life show spontaneous remission, although AML recurrence may occur months to years later.[284,285,286,287] These observations suggest that a watch and wait policy could be considered in cases of t(8;16) AML diagnosed in the neonatal period if close long-term monitoring can be ensured.
The NUP98-NSD1 fusion gene, which is often cytogenetically cryptic, results from the fusion of NUP98 (chromosome 11p15) with NSD1 (chromosome 5q35).[246,247,263,293] This alteration occurs in approximately 4% to 7% of pediatric AML cases.[16,183,246,263,294]
The NUP98-KDM5A fusion gene results from the fusion of the NUP98 gene with the KDM5A gene, which results from a cytogenetically cryptic translocation, t(11;12)(p15;p13). Approximately 2% of pediatric AML patients have NUP98-KDM5A, and these cases tend to present at a young age (median age, 3 years).
The prognostic significance of activating KIT mutations in adults with CBF AML has been studied with conflicting results. A meta-analysis found that KIT mutations appear to increase the risk of relapse without an impact on OS for adults with RUNX1-RUNX1T1 AML.KIT mutations are often subclonal in children and adults with CBF AML;[193,194] in adults with RUNX1-RUNX1T1 AML, higher KIT mutant–allele ratio appears to be associated with higher risk of treatment failure in adults with RUNX1-RUNX1T1 AML.[189,193] The prognostic significance of KIT mutations in pediatric CBF AML remains unclear; some studies found no impact of KIT mutations on outcome,[195,196,197] while other studies reported a higher risk of treatment failure when KIT mutations were present.[194,198,199,200,201]
In children with AML, WT1 mutations are observed in approximately 10% of cases.[305,306] Cases with WT1 mutations are enriched among children with normal cytogenetics and FLT3 ITD, but are less common among children younger than 3 years.[305,306] AML cases with NUP98-NSD1 are enriched for both FLT3 ITD and WT1 mutations. In univariate analyses, WT1 mutations are predictive of poorer outcome in pediatric patients, but the independent prognostic significance of WT1 mutation status is unclear because of its strong association with FLT3 ITD and its association with NUP98-NSD1.[246,305,306] The largest study of WT1 mutations in children with AML observed that children with WT1 mutations in the absence of FLT3 ITD had outcomes similar to that of children without WT1 mutations, while children with both WT1 mutation and FLT3 ITD had survival rates less than 20%.
In a study of children with refractory AML, WT1 was overrepresented compared with a cohort who did achieve remission (54% [15 of 28 patients] vs. 15%).
Mutations in IDH1 and IDH2 are rare in pediatric AML, occurring in 0% to 4% of cases.[310,319,320,321,322,323] There is no indication of a negative prognostic effect for IDH1 and IDH2 mutations in children with AML.
Activating mutations in CSF3R are also observed in patients with severe congenital neutropenia. These mutations are not the cause of severe congenital neutropenia, but rather arise as somatic mutations and can represent an early step in the pathway to AML. In one study of patients with severe congenital neutropenia, 34% of patients who had not developed a myeloid malignancy had CSF3R mutations detectable in peripheral blood neutrophils and mononuclear cells, while 78% of patients who had developed a myeloid malignancy showed CSF3R mutations. A study of 31 patients with severe congenital neutropenia who developed AML or MDS observed CSF3R mutations in approximately 80% of patients, and also observed a high frequency of RUNX1 mutations (approximately 60%), suggesting cooperation between CSF3R and RUNX1 mutations for leukemia development within the context of severe congenital neutropenia.
(Refer to the PDQ summary on Childhood Acute Myeloid Leukemia/Other Myeloid Malignancies Treatment for information about the treatment of childhood AML.)
Juvenile Myelomonocytic Leukemia (JMML)
Molecular features of JMML
The genomic landscape of JMML is characterized by mutations in one of five genes of the RAS pathway: NF1, NRAS, KRAS, PTPN11, and CBL.[327,328,329] In a series of 118 consecutively diagnosed JMML cases with RAS pathway–activating mutations, PTPN11 was the most commonly mutated gene, accounting for 51% of cases (19% germline and 32% somatic) (refer to Figure 3). Patients with mutated NRAS accounted for 19% of cases, and patients with mutated KRAS accounted for 15% of cases. NF1 mutations accounted for 8% of cases, and CBL mutations accounted for 11% of cases. Although mutations among these five genes are generally mutually exclusive, 4% to 17% of cases have mutations in two of these RAS pathway genes,[327,328,329] a finding that is associated with poorer prognosis.[327,329]
The mutation rate in JMML leukemia cells is very low, but additional mutations beyond those of the five RAS pathway genes described above are observed.[327,328,329] Secondary genomic alterations are observed for genes of the transcriptional repressor complex PRC2 (e.g., ASXL1 was mutated in 7%–8% of cases). Some genes associated with myeloproliferative neoplasms in adults are also mutated at low rates in JMML (e.g., SETBP1 was mutated in 6%–9% of cases).[327,328,329,330]JAK3 mutations are also observed in a small percentage (4%–12%) of JMML cases.[327,328,329,330] Cases with germline PTPN11 and germline CBL mutations showed low rates of additional mutations (refer to Figure 3). The presence of mutations beyond disease-defining RAS pathway mutations is associated with an inferior prognosis.[327,328]
A report describing the genomic landscape of JMML found that 16 of 150 patients (11%) lacked canonical RAS pathway mutations. Among these 16 patients, 3 were observed to have in-frame fusions involving receptor tyrosine kinases (DCTN1-ALK, RANBP2-ALK, and TBL1XR1-ROS1). These patients all had monosomy 7 and were aged 56 months or older. One patient with an ALK fusion was treated with crizotinib plus conventional chemotherapy and achieved a complete molecular remission and proceeded to allogeneic bone marrow transplantation.
Figure 3. Alteration profiles in individual JMML cases. Germline and somatically acquired alterations with recurring hits in the RAS pathway and PRC2 network are shown for 118 patients with JMML who underwent detailed genetic analysis. Blast excess was defined as a blast count ≥10% but <20% of nucleated cells in the bone marrow at diagnosis. Blast crisis was defined as a blast count ≥20% of nucleated cells in the bone marrow. NS, Noonan syndrome. Reprinted by permission from Macmillan Publishers Ltd: Nature Genetics (Caye A, Strullu M, Guidez F, et al.: Juvenile myelomonocytic leukemia displays mutations in components of the RAS pathway and the PRC2 network. Nat Genet 47 : 1334-40, 2015), copyright (2015).
Prognosis (genomic and molecular factors)
Several genomic factors affect the prognosis of patients with JMML, including the following:
(Refer to the PDQ summary on Childhood Acute Myeloid Leukemia/Other Myeloid Malignancies Treatment for information about the treatment of JMML.)
Myelodysplastic Syndromes (MDS)
Molecular features of myelodysplastic syndromes (MDS)
Pediatric MDS are associated with a distinctive constellation of genetic alterations compared with MDS arising in adults. In adults, MDS often evolves from clonal hematopoiesis and is characterized by mutations in TET2, DNMT3A, and TP53. In contrast, mutations in these genes are rare in pediatric MDS, while mutations in GATA2, SAMD9/SAMD9L, SETBP1, ASXL1, and RAS/MAPK pathway genes are observed in subsets of pediatric MDS cases.[333,334]
A report of the genomic landscape of pediatric MDS described the results of whole-exome sequencing for 32 pediatric patients with primary MDS and targeted sequencing for another 14 cases. These 46 cases were equally divided between refractory cytopenia of childhood and MDS with excess blasts (MDS-EB). The results from the report include the following:
A second report described the application of a targeted sequencing panel of 105 genes to 50 pediatric patients with MDS (refractory cytopenia of childhood = 31 and MDS-EB = 19) and was enriched for cases with monosomy 7 (48%).[333,334]SAMD9 and SAMD9L were not included in the gene panel. The second report described the following results:
Patients with germline GATA2 mutations, in addition to MDS, show a wide range of hematopoietic and immune defects as well as nonhematopoietic manifestations. The former defects include monocytopenia with susceptibility to atypical mycobacterial infection and DCML deficiency (loss of dendritic cells, monocytes, and B and natural killer lymphoid cells). The resulting immunodeficiency leads to increased susceptibility to warts, severe viral infections, mycobacterial infections, fungal infections, and human papillomavirus–related cancers. The nonhematopoietic manifestations include deafness and lymphedema. Germline GATA2 mutations were studied in 426 pediatric patients with primary MDS and 82 cases with secondary MDS who were enrolled in consecutive studies of the European Working Group of MDS in Childhood (EWOG-MDS). The study had the following results:
SAMD9 and SAMD9L germline mutations are both associated with pediatric MDS cases in which there is an additional loss of all or part of chromosome 7.[338,339] In 2016, SAMD9 was identified as the cause of the MIRAGE syndrome (myelodysplasia, infection, restriction of growth, adrenal hypoplasia, genital phenotypes, and enteropathy), which is associated with early-onset MDS with monosomy 7. Subsequently, mutations in SAMD9L were identified in patients with ataxia pancytopenia syndrome (ATXPC; OMIM 159550). SAMD9 and SAMD9L mutations were also identified as the cause of myelodysplasia and leukemia syndrome with monosomy 7 (MLSM7; OMIM 252270), a syndrome first identified in phenotypically normal siblings who developed MDS or AML associated with monosomy 7 during childhood.
(Refer to the PDQ summary on Childhood Acute Myeloid Leukemia/Other Myeloid Malignancies Treatment for information about the treatment of childhood MDS.)
Mature B-cell Lymphoma
The mature B-cell lymphomas include Burkitt lymphoma, diffuse large B-cell lymphoma, and primary mediastinal B-cell lymphoma.
Genomics of Burkitt lymphoma/leukemia
The malignant cells show a mature B-cell phenotype and are negative for the enzyme terminal deoxynucleotidyl transferase. These malignant cells usually express surface immunoglobulin, most bearing a clonal surface immunoglobulin M with either kappa or lambda light chains. A variety of additional B-cell markers (e.g., CD19, CD20, CD22) are usually present, and most childhood Burkitt lymphomas/leukemias express CD10.
Burkitt lymphoma/leukemia expresses a characteristic chromosomal translocation, usually t(8;14) and more rarely t(8;22) or t(2;8). Each of these translocations juxtaposes the MYC oncogene and immunoglobulin (IG) locus regulatory elements, resulting in the inappropriate expression of MYC, a gene involved in cellular proliferation.[2,3,4] The presence of one of the variant translocations t(2;8) or t(8;22) does not appear to affect response or outcome.
Mapping of IGH-translocation breakpoints demonstrated that IG-MYC translocations in sporadic Burkitt lymphoma most commonly occur through aberrant class-switch recombination and less commonly through somatic hypermutation; translocations resulting from aberrant variable, diversity, and joining (VDJ) gene segment recombinations are rare. These findings are consistent with a germinal center derivation of Burkitt lymphoma.
While MYC translocations are present in all Burkitt lymphoma, cooperating genomic alterations appear to be required for lymphoma development. Some of the more commonly observed recurring mutations that have been identified in Burkitt lymphoma in pediatric and adult cases are listed below. The clinical significance of these mutations for pediatric Burkitt lymphoma remains to be elucidated.
A study that compared the genomic landscape of endemic Burkitt lymphoma with the genomics of sporadic Burkitt lymphoma found the expected high rate of Epstein-Barr virus (EBV) positivity in endemic cases, with much lower rates in sporadic cases. There was general similarity between the patterns of mutations for endemic and sporadic cases and for EBV-positive and EBV-negative cases; however, EBV-positive cases showed significantly lower mutation rates for selected genes/pathways, including SMARCA4, apoptosis, CCND3, and TP53.
Cytogenetic evidence of MYC rearrangement is the gold standard for diagnosis of Burkitt lymphoma/leukemia. For cases in which cytogenetic analysis is not available, the World Health Organization (WHO) has recommended that the Burkitt-like diagnosis be reserved for lymphoma resembling Burkitt lymphoma/leukemia or with more pleomorphism, large cells, and a proliferation fraction (i.e., MIB-1 or Ki-67 immunostaining) of 99% or greater. BCL2 staining by immunohistochemistry is variable. The absence of a translocation involving the BCL2 gene does not preclude the diagnosis of Burkitt lymphoma/leukemia and has no clinical implications.
Burkitt-like lymphoma with 11q aberration was added as a provisional entity in the 2017 revised WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues. In this entity, MYC rearrangement is absent, and the characteristic chromosome 11q finding (detected cytogenetically and/or with copy-number DNA arrays) is 11q23.2-q23.3 gain/amplification and 11q24.1-qter loss.[15,16] Most patients present in the adolescent and young adult age range with localized nodal disease, and outcomes appear favorable in the small number of cases identified. Cases show a very high proliferative index and can show a focal starry sky pattern. The mutational landscape of Burkitt-like lymphoma with 11q aberration is distinct from that of Burkitt lymphoma; mutations commonly observed in Burkitt lymphoma (e.g., ID3, TCF3, and CCND3) are uncommon in Burkitt-like lymphoma with 11q aberration. Conversely, mutations in GNA13 appear to be common (up to 50%) in patients with Burkitt-like lymphoma with 11q aberration and are less common in patients with Burkitt lymphoma.
(Refer to the PDQ summary on Childhood Non-Hodgkin Lymphoma Treatment for information about the treatment of childhood Burkitt lymphoma/leukemia.)
Diffuse large B-cell lymphoma
Genomics of diffuse large B-cell lymphoma
The World Health Organization (WHO) classification system categorizes diffuse large B-cell lymphoma on the basis of molecular characteristics into the germinal center B-cell subtype and the activated B-cell subtype, with the remaining classes being classified as diffuse large B-cell lymphoma, not otherwise specified (NOS).
Diffuse large B-cell lymphoma in children and adolescents differs biologically from diffuse large B-cell lymphoma in adults in the following ways:
Large B-cell lymphoma with IRF4 rearrangement (LBCL-IRF4) was added as a provisional entity in the 2017 revision of the WHO classification of lymphoid neoplasms.
High-grade B-cell lymphoma, NOS is defined as a clinically aggressive B-cell lymphoma that lacks MYC plus BCL2 and/or BCL6 rearrangements and that does not meet criteria for diffuse large B-cell lymphoma, NOS or Burkitt lymphoma.
(Refer to the PDQ summary on Childhood Non-Hodgkin Lymphoma Treatment for information about the treatment of childhood diffuse large B-cell lymphoma.)
Primary mediastinal B-cell lymphoma
Genomics of primary mediastinal B-cell lymphoma
Primary mediastinal B-cell lymphoma was previously considered a subtype of diffuse large B-cell lymphoma, but is now a separate entity in the most recent World Health Organization (WHO) classification. These tumors arise in the mediastinum from thymic B-cells and show a diffuse large cell proliferation with sclerosis that compartmentalizes neoplastic cells.
Primary mediastinal B-cell lymphoma can be very difficult to distinguish morphologically from the following types of lymphoma:
Primary mediastinal B-cell lymphoma has a distinctive gene expression profile compared with diffuse large B-cell lymphoma; however, its gene expression profile has features similar to those seen in Hodgkin lymphoma.[31,32] Primary mediastinal B-cell lymphoma is also associated with a distinctive constellation of chromosomal aberrations compared with other NHL subtypes. Because primary mediastinal B-cell lymphoma is primarily a cancer of adolescents and young adults, the genomic findings are presented without regard to age.
(Refer to the PDQ summary on Childhood Non-Hodgkin Lymphoma Treatment for information about the treatment of childhood primary mediastinal B-cell lymphoma.)
As opposed to pediatric acute lymphoblastic leukemia, chromosomal abnormalities and the molecular biology of pediatric lymphoblastic lymphoma are not well characterized. The Berlin-Frankfurt-Münster group reported that loss of heterozygosity at chromosome 6q was observed in 12% of patients and NOTCH1 mutations were seen in 60% of patients, but NOTCH1 mutations are rarely seen in patients with loss of heterozygosity at 6q.[43,44]
(Refer to the PDQ summary on Childhood Non-Hodgkin Lymphoma Treatment for information about the treatment of childhood lymphoblastic lymphoma.)
Anaplastic Large Cell Lymphoma
While the predominant immunophenotype of anaplastic large cell lymphoma is mature T-cell, null-cell disease (i.e., no T-cell, B-cell, or natural killer-cell surface antigen expression) does occur. The World Health Organization (WHO) classifies anaplastic large cell lymphoma as a subtype of peripheral T-cell lymphoma.
All anaplastic large cell lymphoma cases are CD30-positive. More than 90% of pediatric anaplastic large cell lymphoma cases have a chromosomal rearrangement involving the ALK gene. About 85% of these chromosomal rearrangements will be t(2;5)(p23;q35), leading to the expression of the fusion protein NPM-ALK; the other 15% of cases are composed of variant ALK translocations. Anti-ALK immunohistochemical staining pattern is quite specific for the type of ALK translocation. Cytoplasm and nuclear ALK staining is associated with NPM-ALK fusion protein, whereas cytoplasmic staining only of ALK is associated with the variant ALK translocations, as shown in Table 4.
|Gene Fusion||Partner Chromosome Location||Frequency of Gene Fusion|
|a Adapted from Tsuyama et al.|
In adults, ALK-positive anaplastic large cell lymphoma is viewed differently from other peripheral T-cell lymphomas because prognosis tends to be superior. Also, adult ALK-negative anaplastic large cell lymphoma patients have an inferior outcome compared with patients who have ALK-positive disease. In children, however, this difference in outcome between ALK-positive and ALK-negative disease has not been demonstrated. In addition, no correlation has been found between outcome and the specific ALK-translocation type.[50,51,52]
In a European series of 375 children and adolescents with systemic ALK-positive anaplastic large cell lymphoma, the presence of a small cell or lymphohistiocytic component was observed in 32% of patients and was significantly associated with a high risk of failure in the multivariate analysis, controlling for clinical characteristics (hazard ratio, 2.0; P = .002). The prognostic implication of the small cell variant of anaplastic large cell lymphoma was also shown in the COG-ANHL0131 (NCT00059839) study, despite a different chemotherapy backbone.
(Refer to the PDQ summary on Childhood Non-Hodgkin Lymphoma Treatment for information about the treatment of childhood anaplastic large cell lymphoma.)
Pediatric-type Follicular Lymphoma
Pediatric-type follicular lymphoma appears to be molecularly distinct from follicular lymphoma that is more commonly observed in adults. The pediatric type lacks BCL2 rearrangements; BCL6 and MYC rearrangements are also not present. The TNFSFR14 mutations are common in pediatric-type follicular lymphoma, and they appear to occur with similar frequency in adult follicular lymphoma.[53,54] However, MAP2K1 mutations, which are uncommon in adults, are observed in as many as 43% of pediatric-type follicular lymphomas. Other genes (e.g., MAPK1 and RRAS) have been found to be mutated in cases without MAP2K1 mutations, suggesting that the MAP kinase pathway is important in the pathogenesis of pediatric-type follicular lymphoma.[55,56] Translocations of the immunoglobulin locus and IRF4, mutations in IRF8, and abnormalities in chromosome 1p have also been observed in pediatric-type follicular lymphoma.[27,53,57]
(Refer to the PDQ summary on Childhood Non-Hodgkin Lymphoma Treatment for information about the treatment of pediatric-type follicular lymphoma.)
Central nervous system (CNS) tumors include pilocytic astrocytomas and other astrocytic tumors, diffuse astrocytic tumors, brain stem gliomas, CNS atypical teratoid/rhabdoid tumors, medulloblastomas, nonmedulloblastoma embryonal tumors, and ependymomas.
The terminology of the 2016 World Health Organization (WHO) Classification of Tumors of the Central Nervous System is used below. The 2016 WHO CNS classification incorporates genomic features in addition to histology, and it includes multiple changes from the previous 2007 WHO classification. Of particular relevance for childhood brain cancers is the new entity diffuse midline glioma, H3 K27M-mutant, which includes diffuse intrinsic pontine glioma (DIPG) with the H3 K27M mutation and other high-grade gliomas of the midline with the H3 K27M mutation. Other examples of molecularly defined entities discussed below are RELA-fusion–positive ependymoma, WNT-activated and SHH-activated medulloblastoma, and embryonal tumor with multilayered rosettes, C19MC-altered.
Pilocytic Astrocytomas and Other Astrocytic Tumors
Molecular features of low-grade gliomas
Pilocytic and diffuse astrocytomas
Genomic alterations involving activation of BRAF and the ERK/MAPK pathway are very common in sporadic cases of pilocytic astrocytoma, a type of low-grade glioma.
BRAF activation in pilocytic astrocytoma occurs most commonly through a BRAF-KIAA1549 gene fusion, producing a fusion protein that lacks the BRAF regulatory domain.[2,3,4,5,6] This fusion is seen in most infratentorial and midline pilocytic astrocytomas, but is present at lower frequency in supratentorial (hemispheric) tumors.[2,3,7,8,9,10,11,12]
Presence of the BRAF-KIAA1549 fusion predicted a better clinical outcome (progression-free survival [PFS] and overall survival [OS]) in one report that described children with incompletely resected low-grade gliomas. However, other factors such as CDKN2A deletion, whole chromosome 7 gain, and tumor location may modify the impact of the BRAF mutation on outcome.; [Level of evidence: 3iiiDiii] Progression to high-grade glioma is rare for pediatric low-grade glioma with the BRAF-KIAA1549 fusion.
BRAF activation through the BRAF-KIAA1549 fusion has also been described in other pediatric low-grade gliomas (e.g., pilomyxoid astrocytoma).[10,11] Other genomic alterations in pilocytic astrocytomas that can activate the ERK/MAPK pathway (e.g., alternative BRAF gene fusions, RAF1 rearrangements, RAS mutations, and BRAF V600E point mutations) are less commonly observed.[3,5,6,16]
BRAF V600E point mutations are occasionally observed in pilocytic astrocytoma; the mutations are also observed in nonpilocytic pediatric low-grade gliomas, including ganglioglioma, desmoplastic infantile ganglioglioma, and approximately two-thirds of pleomorphic xanthoastrocytomas.[18,19,20]
Studies have observed the following:
Activating mutations in FGFR1, PTPN11, and NTRK2 fusion genes have also been identified in noncerebellar pilocytic astrocytomas. In pediatric grade II diffuse astrocytomas, the most common alterations reported (up to 53% of tumors) are rearrangements in the MYB family of transcription factors.[24,25]
Angiocentric gliomas typically arise in children and young adults as cerebral tumors presenting with seizures.
Two reports in 2016 identified MYB gene alterations as being present in almost all cases diagnosed as angiocentric glioma, with QKI being the primary fusion partner in cases where fusion-partner testing was possible.[26,27] While angiocentric gliomas most commonly occur supratentorially, brain stem angiocentric gliomas with MYB-QKI fusions have also been reported.[28,29]
Astroblastomas are defined histologically as glial neoplasms composed of GFAP-positive cells and contain astroblastic pseudorosettes that often demonstrate sclerosis. Astroblastomas are diagnosed primarily in childhood through young adulthood.
The following studies have described genomic alterations associated with astroblastoma:
These reports suggest that the histologic diagnosis of astroblastoma encompasses a heterogeneous group of genomically defined entities; astroblastomas with MN1 fusions represent a distinctive subset of histologically diagnosed cases.
Neurofibromatosis type 1 (NF1)
Children with NF1-associated low-grade gliomas often have tumors in the optic pathway that are not biopsied. In a series of pediatric patients (n = 17; median age, 10 years) with NF1-associated low-grade gliomas in which tissue was collected and subjected to whole-exome sequencing, the number of mutations was very low (median, 6 per case). Germline NF1 mutations were observed in 88% of patients, and the most common somatic alteration was loss of heterozygosity for NF1, with a smaller number of cases showing inactivating mutations in the second NF1 allele. CDKN2A loss was observed in 1 of 17 patients (6%). Alterations in TP53 and ATRX were not observed among the 17 pediatric patients with NF1-associated low-grade gliomas. Activating BRAF genomic alterations are uncommon in pilocytic astrocytoma and other low-grade gliomas occurring in children with NF1.[9,35]
Most children with tuberous sclerosis have a germline mutation in one of two tuberous sclerosis genes (TSC1 or TSC2). Either of these mutations results in activation of the mammalian target of rapamycin (mTOR) complex 1. These children are at risk of developing subependymal giant cell astrocytomas, cortical tubers, and subependymal nodules. Because subependymal giant cell astrocytomas are driven by mTOR activation, mTOR inhibitors are active agents that can induce tumor regression in children with these tumors.
(Refer to the PDQ summary on Childhood Astrocytomas Treatment for information about the treatment of low-grade childhood astrocytomas.)
Diffuse Astrocytic Tumors
This category includes, among other diagnoses, diffuse astrocytomas (grade II) and pediatric high-grade gliomas (anaplastic astrocytoma [grade III], glioblastoma [grade IV], and diffuse midline glioma, H3 K27M-mutant (grade IV]).
For pediatric diffuse astrocytomas (grade II), rearrangements in the MYB family of transcription factors (MYB and MYBL1) are the most commonly reported genomic alteration.[24,25,27] Other alterations observed include FGFR1 alterations (primarily duplications involving the tyrosine kinase domain),[25,27]BRAF alterations, NF1 mutations, and RAS family mutations.[24,25]IDH1 mutations, which are the most common genomic alteration in adult diffuse astrocytomas, are uncommon in children with diffuse astrocytomas and, when present, are observed almost exclusively in older adolescents.[24,37]
Anaplastic astrocytomas and glioblastomas
Molecular features of high-grade gliomas
Subgroups identified using DNA methylation patterns
Pediatric high-grade gliomas can be separated into distinct subgroups on the basis of epigenetic patterns (DNA methylation), and these subgroups show distinguishing chromosome copy number gains/losses and gene mutations in the tumor.[41,42,43] Particularly distinctive subtypes of pediatric high-grade gliomas are those with recurring mutations at specific amino acids in histone genes, and together these account for approximately one-half of pediatric high-grade gliomas.
The following pediatric high-grade glioma subgroups were identified on the basis of their DNA methylation patterns, and they show distinctive molecular and clinical characteristics:
Pediatric glioblastoma multiforme high-grade glioma patients whose tumors lack both histone mutations and IDH1 mutations represent approximately 40% of pediatric glioblastoma multiforme cases.[43,47] This is a heterogeneous group, with higher rates of gene amplifications than other pediatric high-grade glioma subtypes. The most commonly amplified genes are PDGFRA, EGFR, CCND/CDK, and MYC/MYCN;[41,42] MGMT promoter methylation rates are low in this group. One report divided this group into three subtypes. The subtype characterized by high rates of MYCN amplification showed the poorest prognosis, while the subtype characterized by TERT promoter mutations and EGFR amplification showed the most favorable prognosis. The third group was characterized by PDGFRA amplification.
High-grade gliomas in infants
Infants and young children with high-grade gliomas appear to have tumors with distinctive molecular characteristics when compared with tumors of older children and adults with high-grade gliomas. An indication of this difference was noted with the application of DNA methylation analysis to pediatric high-grade tumors, which found that approximately 7% of pediatric patients with a histologic diagnosis of high-grade glioma had tumors with methylation patterns more closely resembling those of low-grade gliomas. Ten of 16 infants (younger than 1 year) with a high-grade glioma diagnosis were in this methylation array–defined group. The 5-year survival rate for patients in this report diagnosed at younger than 1 year exceeded 60%, while the 5-year survival rate for patients aged 1 to 3 years and older was less than 20%.
Two studies of the molecular characteristics of high-grade gliomas in infants and young children have further defined the distinctive nature of tumors arising in children younger than 1 year. A key finding from both studies is the importance of gene fusions involving tyrosine kinases (e.g., ALK, NTRK1, NTRK2, NTRK3, and ROS1) in patients in this age group. Both studies also found that infants with high-grade gliomas whose tumors have these gene fusions have survival rates much higher than those of older children with high-grade gliomas.[48,49]
The first study presented data for 118 children younger than 1 year with a low-grade or high-grade glioma diagnosis who had tumor tissue available for genomic characterization. Approximately 75% of the cases were classified as low grade, but the diminished utility of histologic classification in this age group was illustrated by the relatively low OS rate for the low-grade cohort (71%) and the relatively favorable survival for the high-grade cohort (55%). Rates of surgical resection were higher for patients with high-grade tumors, a result of many of the low-grade tumors occurring in midline locations while the high-grade tumors were found in supratentorial locations; this finding may also help to explain the relative outcomes for the two groups. Genomic characterization divided the infant glioma population into the following three groups, the first of which included patients with high-grade gliomas:
The second study focused on tumors from children younger than 4 years with a pathological diagnosis of WHO grades II, III, and IV gliomas, astrocytomas, or glioneuronal tumors. Among the 191 tumors studied that met inclusion criteria, 61 had methylation profiles consistent with glioma subtypes that occur in older children (e.g., IDH1, diffuse midline glioma K27M-mutant, subependymal giant cell astrocytoma, pleomorphic xanthoastrocytoma, etc.). The remaining 130 cases were termed the intrinsic set and were the focus of additional molecular characterization:
Secondary high-grade glioma
Childhood secondary high-grade glioma (high-grade glioma that is preceded by a low-grade glioma) is uncommon (2.9% in a study of 886 patients). No pediatric low-grade gliomas with the BRAF-KIAA1549 fusion transformed to a high-grade glioma, whereas low-grade gliomas with the BRAF V600E mutations were associated with increased risk of transformation. Seven of 18 patients (approximately 40%) with secondary high-grade glioma had BRAF V600E mutations, with CDKN2A alterations present in 8 of 14 cases (57%).
Neurofibromatosis type 1 (NF1)
High-grade gliomas can arise in children with NF1, although low-grade gliomas are much more common. When a high-grade tumor occurs, it is most often in adulthood. Genomic characterization of 23 patients with NF1-associated high-grade gliomas (median age, 38.8 years; 5 patients younger than 18 years) showed higher rates of mutations compared with NF1 patients who had low-grade gliomas (21.5 vs. 6 mutations, respectively). The vast majority of patients showed NF1 germline mutations, with either loss of heterozygosity or with an inactivating mutation in the second NF1 allele. In contrast to NF1-associated low-grade gliomas, genomic alterations associated with high-grade gliomas were common (CDKN2A [58%], ATRX [38%], and TP53 [29%]).
(Refer to the PDQ summary on Childhood Astrocytomas Treatment for information about the treatment of high-grade childhood astrocytomas.)
Neuronal and Mixed Neuronal-Glial Tumors
Molecular features of neuronal and mixed neuronal-glial tumors
Neuronal and mixed neuronal-glial tumors are generally low-grade tumors, with an exception of the grade III anaplastic gangliogliomas. The histologies recognized by the 2016 WHO classification include the following:
Dysembryoplastic neuroepithelial tumor (DNET)
DNET presents in children and adults, with the median age at diagnosis in mid-to-late adolescence. It is characterized histopathologically by the presence of columns of oligodendroglial-like cells and cortical ganglion cells floating in mucin. The temporal lobe is the most common location, and it is associated with drug-refractory epilepsy.[51,52]
FGFR1 alterations have been reported in 60% to 80% of DNETs, and include FGFR1 activating point mutations, internal tandem duplication of the kinase domain, and activating gene fusions.[27,53,54]BRAF mutations are uncommon in DNET.
DNET of the septum pellucidum
Septal DNET generally presents with symptoms related to obstructive hydrocephalus.[55,56] Septal DNET has an indolent clinical behavior, with most tumors not requiring treatment other than surgery. In a single-institution series that incorporated other literature-reported cases, the median age at presentation was in the adolescent age range.
Mutations that are common in low-grade gliomas (e.g., BRAF V600E) and in cortical DNETs (FGFR1 mutations) are uncommon in septal DNET.[56,57,58] Instead, mutations in PDGFRA at the K385 residue typify most cases of septal DNET.
A report of the molecular characterization of 18 septal DNETs showed that 14 had a PDGFRA mutation, with all but one being a mutation at the K385 residue, which is in the extracellular region of PDGFRA that mediates the receptor-receptor interaction required for dimerization and activation upon binding of PDGFs. Among the remaining four cases, three had FGFR1 mutations in line with those observed in cortical DNET. A second report observed PDGFRA mutations at K385 in each of four cases of septal DNET. Combined, the two reports indicate that septal DNET is a distinct entity characterized by a stereotypic anatomic location and, in most cases, a PDGFRA mutation. Low-grade glioneuronal tumors with the K385 PDGFRA mutation have also been identified as arising in the corpus callosum and periventricular white matter of the lateral ventricle, leading to the proposal that myxoid glioneuronal tumor, PDGFRA p.K385-mutant be considered as a distinct central nervous system (CNS) tumor entity.
Ganglioglioma presents during childhood and into adulthood. It most commonly arises in the cerebral cortex and is associated with seizures, but also presents in other sites, including the spinal cord.[51,60]
The unifying theme for the molecular pathogenesis of ganglioglioma is genomic alterations leading to MAPK pathway activation.[27,61]BRAF alterations are observed in approximately 50% of ganglioglioma cases, with V600E being by far the most common alteration; however, other BRAF mutations and gene fusions are also observed. Other less commonly altered genes in ganglioglioma include KRAS, FGFR1/2, RAF1, NTRK2, and NF1.[27,61]
Desmoplastic infantile astrocytomas (DIA) and desmoplastic infantile gangliogliomas (DIG)
DIA and DIG most often present in the first year of life and show a characteristic imaging appearance in which a contrast-enhancing solid nodule accompanies a large cystic component.[62,63] DIG is more common than DIA, and by methylation array analysis, both diagnoses cluster together. Survival outcome is generally favorable with surgical resection.
The most commonly observed genomic alterations in DIA and DIG are BRAF mutations involving V600; gene fusions involving kinase genes are observed less frequently.
Papillary glioneuronal tumor
Papillary glioneuronal tumor is a low-grade biphasic neoplasm with astrocytic and neuronal differentiation that primarily arises in the supratentorial compartment. The median age at presentation is in the early 20s, but it can be observed during childhood through adulthood.
The primary genomic alteration associated with papillary glioneuronal tumor is a gene fusion, SLC44A1-PRKCA, that is associated with the t(9:17)(q31;q24) translocation.[67,68] In one study of 28 cases diagnosed histologically as papillary glioneuronal tumor using methylation arrays, 11 of the cases clustered in a distinctive methylation class, while the remaining cases showed methylation profiles typical for other tumor entities. Molecular analysis of the cases in the distinctive methylation cluster showed that all of them had the SLC44A1-PRKCA gene fusion except for a single case with a NOTCH1-PRKCA gene fusion. This suggests that molecular methods for identifying the presence of a PRKCA fusion are less susceptible to misclassification in diagnosing papillary glioneuronal tumor than are morphology-based methods.
Rosette-forming glioneuronal tumor (RGNT)
RGNT presents in adolescents and adults, with tumors generally located infratentorially, although tumors can arise in mesencephalic or diencephalic regions. The typical histological appearance shows both a glial component and a neurocytic component arranged in rosettes or perivascular pseudorosettes. Outcome for patients with RGNT is generally favorable, consistent with the WHO grade I designation.
DNA methylation profiling shows that RGNT has a distinct epigenetic profile that distinguishes it from other low-grade glial/glioneuronal tumor entities. A study of 30 cases of RGNT observed FGFR1 hotspot mutations in all analyzed tumors. In addition, PIK3CA activating mutations were concurrently observed in 19 of 30 cases (63%). Missense or damaging mutations in NF1 were identified in 10 of 30 cases (33%), with 7 tumors having mutations in FGFR1, PIK3CA, and NF1. The co-occurrence of mutations that activate both the MAPK pathway and the PI3K pathway makes the mutation profile of RGNT distinctive among astrocytic and glioneuronal tumors.
Diffuse leptomeningeal glioneuronal tumor (DLGNT)
DLGNT is a rare CNS tumor that has been characterized radiographically by leptomeningeal enhancement on magnetic resonance imaging (MRI) that may involve the posterior fossa, brain stem region, and spinal cord. Intraparenchymal lesions, when present, typically involve the spinal cord; localized intramedullary glioneuronal tumors without leptomeningeal dissemination and with histomorphologic, immunophenotypic, and genomic characteristics similar to DLGNT have been reported.
DLGNT showed a distinctive epigenetic profile on DNA methylation arrays, and unsupervised clustering of array data applied to 30 cases defined two subclasses of DLGNT: methylation class (MC)-1 (n = 17) and MC-2 (n = 13). Of note, many of the array-defined cases had originally been diagnosed as other entities (e.g., primitive neuroectodermal tumors, pilocytic astrocytoma, and anaplastic astrocytoma). Patients with DLGNT-MC-1 were diagnosed at an earlier age than were patients with DLGNT-MC-2 (5 years vs. 14 years, respectively). The 5-year overall survival rate was higher for patients with DLGNT-MC-1 than for those with DLGNT-MC-2 (100% vs. 43%, respectively). Genomic findings from the 30 methylation array–defined DLGNT cases are provided below:
Extraventricular neurocytoma is histologically similar to central neurocytoma, consisting of small uniform cells that demonstrate neuronal differentiation, but it arises in the brain parenchyma rather than in association with the ventricular system. It presents during childhood through adulthood.
In a study of 40 tumors histologically classified as extraventricular neurocytoma and subjected to methylation array analysis, only 26 formed a separate cluster distinctive from reference tumors of other histologies. Among cases with an extraventricular neurocytoma methylation array classification for which genomic characterization could be performed, 11 of 15 (73%) showed rearrangements affecting members of the FGFR family, with FGFR1-TACC1 being the most common alteration.
Diffuse Midline Glioma, H3 K27M-Mutant (Including Diffuse Intrinsic Pontine Gliomas [DIPGs])
The diffuse midline glioma, H3 K27M-mutant, category includes tumors previously classified as DIPG; most of the data is derived from experience with DIPG. This category also includes gliomas with the H3 K27M mutation arising in midline structures such as the thalamus.
Genomics of DIPGs
The genomic characteristics of DIPGs appear to differ from those of many other pediatric high-grade gliomas of the cerebrum and from those of adult high-grade gliomas. The molecular and clinical characteristics of DIPGs align with those of other midline high-grade gliomas, with a specific H3 K27M mutation in histone H3.3 (H3F3A) or H3.1 (HIST1H3B and HIST1H3C), which led the World Health Organization to group these tumors together into a single entity, called diffuse midline glioma, H3 K27M-mutant.
In one report of 64 children with thalamic tumors, 50% of high-grade gliomas (11 of 22) had an H3 K27M mutation, and approximately 10% of tumors with low-grade morphological characteristics (5 of 42) had an H3 K27M mutation. Five-year overall survival (OS) was only 6% (1 of 16). In another study that included 202 children with glioblastoma, 68 of the tumors were midline (primarily thalamic) and had an H3 K27M mutation. Five-year OS for this group was only 5%, which was significantly inferior to the survival rates of the remaining patients in the study.
A number of chromosomal and genomic abnormalities have been reported for DIPG, including the following:
An autopsy study that examined multiple tumor sites (primary, contiguous, and metastatic) in seven DIPG patients found that the H3 K27M mutation was invariably present, supporting its role as a driver mutation for DIPG.
Patients with H3.1 K27M mutations have a longer median survival (15 months) than do patients with H3.3 K27M mutations (10.4 months).
An autopsy study that examined multiple tumor sites (primary, contiguous, and metastatic) in seven DIPG patients found that PDGFRA amplification was variably present across these sites, suggesting that this change is a secondary genomic alteration in DIPG.
The gene expression profile of DIPG differs from that of non–brain stem pediatric high-grade gliomas, further supporting a distinctive biology for this subset of pediatric gliomas. Pediatric H3 K27M-mutant tumors rarely show O-6-methylguanine-DNA methyltransferase (MGMT) promoter methylation, which explains the lack of efficacy of temozolomide when it was tested in patients with DIPG.
(Refer to the Genomic Alterations section in the PDQ summary on Childhood Astrocytomas Treatment for more information about the genetics of low-grade gliomas.)
(Refer to the PDQ summary on Childhood Brain Stem Glioma Treatment for information about the treatment of childhood brain stem gliomas.)
Central Nervous System (CNS) Atypical Teratoid/Rhabdoid Tumors (AT/RT)
AT/RT was the first primary pediatric brain tumor in which a candidate tumor suppressor gene, SMARCB1 (previously known as INI1 and hSNF5), was identified.SMARCB1 is genomically altered in most rhabdoid tumors, including CNS, renal, and extrarenal rhabdoid malignancies. Loss of SMARCB1/SMARCA4 staining is a defining marker for AT/RT. Additional genomic alterations (mutations and gains/losses) in other genes are very uncommon in patients with SMARCB1-associated AT/RT. Less commonly, SMARCA4-negative (with retained SMARCB1) tumors have been described. No other genes are recurrently mutated in AT/RT.[89,90,91]
SMARCB1 is a component of a switch (SWI) and sucrose non-fermenting (SNF) adenosine triphosphate–dependent chromatin-remodeling complex. Rare familial cases of rhabdoid tumors expressing SMARCB1 and lacking SMARCB1 mutations have also been associated with germline mutations of SMARCA4/BRG1, another member of the SWI/SNF chromatin-remodeling complex.[93,94]
The 2016 WHO classification defines AT/RT by the presence of either SMARCB1 or SMARCA4 alterations. Tumors with histological features of AT/RT that lack these genomic alterations are termed CNS embryonal tumor with rhabdoid features.
Despite the absence of recurring genomic alterations beyond SMARCB1 (and, more rarely, other SWI/SNF complex members), biologically distinctive subsets of AT/RT have been identified.[95,96] The following three distinctive subsets of AT/RT were identified through the use of DNA methylation arrays for 150 AT/RT tumors and gene expression arrays for 67 AT/RT tumors:
In addition to somatic mutations, germline mutations in SMARCB1 have been reported in a substantial subset of AT/RT patients.[87,98] A study of 65 children with rhabdoid tumors found that 23 (35%) had germline mutations and/or deletions of SMARCB1. Children with germline alterations in SMARCB1 presented at an earlier age than did sporadic cases (median age, approximately 5 months vs. 18 months) and were more likely to present with synchronous, multifocal tumors. One parent was found to be a carrier of the SMARCB1 germline abnormality in 7 of 22 evaluated cases showing germline alterations, with four of the carrier parents being unaffected by SMARCB1-associated cancers. This indicates that AT/RT shows an autosomal dominant inheritance pattern with incomplete penetrance.
Gonadal mosaicism has also been observed, as evidenced by families in which multiple siblings are affected by AT/RT and have identical SMARCB1 alterations, but both parents lack a SMARCB1 mutation/deletion.[99,100] Screening for germline SMARCB1 mutations in children diagnosed with AT/RT may provide useful information for counseling families on the genetic implications of their child's AT/RT diagnosis.
Loss of SMARCB1 or SMARCA4 protein expression has therapeutic significance, because this loss creates a dependence of the cancer cells on EZH2 activity. Preclinical studies have shown that some AT/RT xenograft lines with SMARCB1 loss respond to EZH2 inhibitors with tumor growth inhibition and occasional tumor regression.[102,103] In a study of the EZH2 inhibitor tazemetostat, objective responses were observed in adult patients whose tumors had either SMARCB1 or SMARCA4 loss (non-CNS malignant rhabdoid tumors and epithelioid sarcoma). (Refer to the Treatment of Recurrent Childhood CNS Atypical Teratoid/Rhabdoid Tumor section of this summary for more information.)
(Refer to the PDQ summary on Childhood Central Nervous System Atypical Teratoid/Rhabdoid Tumors Treatment for information about the treatment of childhood CNS atypical teratoid/rhabdoid tumors.)
Molecular subtypes of medulloblastoma
Multiple medulloblastoma subtypes have been identified by integrative molecular analysis.[105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122] Since 2012, the general consensus is that medulloblastoma can be molecularly separated into at least four core subtypes, including WNT-activated, sonic hedgehog (SHH)–activated, group 3, and group 4 medulloblastoma. However, different regions of the same tumor are likely to have other disparate genetic mutations, adding to the complexity of devising effective molecularly targeted therapy. These subtypes remain stable across primary and metastatic components.[124,125]
The 2016 World Health Organization (WHO) classification has endorsed this consensus by adding the following categories for molecularly defined medulloblastoma:
WNT tumors are medulloblastomas with aberrations in the WNT signaling pathway and represent approximately 10% of all medulloblastomas. WNT medulloblastomas show a WNT signaling gene expression signature and beta-catenin nuclear staining by immunohistochemistry. They are usually histologically classified as classic medulloblastoma tumors and rarely have a large cell/anaplastic appearance. WNT medulloblastomas generally occur in older patients (median age, 10 years) and are infrequently metastasized at diagnosis.
CTNNB1 mutations are observed in 85% to 90% of WNT medulloblastoma cases, with APC mutations detected in many of the cases that lack CTNNB1 mutations. Patients with WNT medulloblastoma whose tumors have APC mutations often have Turcot syndrome (i.e., germline APC mutations). In addition to CTNNB1 mutations, WNT medulloblastoma tumors show 6q loss (monosomy 6) in 80% to 90% of cases. While monosomy 6 is observed in most medulloblastoma patients younger than 18 years at diagnosis, it appears to be much less common (approximately 25% of cases) in patients older than 18 years.[126,128]
The WNT subset is primarily observed in older children, adolescents, and adults and does not show a male predominance. The subset is believed to have brain stem origin, from the embryonal rhombic lip region. WNT medulloblastomas are associated with a very good outcome in children, especially in individuals whose tumors have beta-catenin nuclear staining and proven 6q loss and/or CTNNB1 mutations.[120,130,131]
Medulloblastoma, SHH-activated andTP53-mutant and medulloblastoma, SHH-activated andTP53-wildtype
SHH tumors are medulloblastomas with aberrations in the SHH pathway and represent approximately 25% of medulloblastoma cases. SHH medulloblastomas are characterized by chromosome 9q deletions; desmoplastic/nodular histology; and mutations in SHH pathway genes, including PTCH1, PTCH2, SMO, SUFU, and GLI2.
Heterozygous deleterious germline mutations in the GPR161 gene were identified in approximately 3% of cases of SHH medulloblastoma.GPR161 is an inhibitor of SHH signaling. Median age at diagnosis for GPR161-mutated cases was 1.5 years. Loss of heterozygosity (LOH) at the GPR161 locus was noted in all tumors, with tumors from five of six patients showing copy-neutral LOH of chromosome 1q (on which GPR161 resides).
Mutations in the third nucleotide (r.3A>G) of the U1 spliceosomal small nuclear RNAs (snRNAs) are highly specific for SHH medulloblastoma.[133,134] U1 snRNA r.3A>G mutations are observed in virtually all cases of SHH medulloblastoma in adults, in approximately one-third of cases in children and adolescents, and are absent in infant cases. U1 snRNA mutations disrupt RNA splicing, leading to inactivation of tumor-suppressor genes (e.g., PTCH1) and activation of oncogenes (e.g., GLI2). The significance of U1 snRNA r.3A>G mutations in specific SHH medulloblastoma subtypes is described below.
SHH medulloblastomas show a bimodal age distribution and are observed primarily in children younger than 3 years and in older adolescence/adulthood. The tumors are believed to emanate from the external granular layer of the cerebellum. The heterogeneity in age at presentation maps to distinctive subsets identified by further molecular characterization, as follows:
A second report that used DNA methylation arrays also identified two subtypes of SHH medulloblastoma in young children. One of the subtypes contained all of the cases with SMO mutations, and it was associated with a favorable prognosis. The other subtype had most of the SUFU mutations, and it was associated with a much lower progression-free survival (PFS) rate. PTCH1 mutations were present in both subtypes.
The outcome for patients with nonmetastatic SHH medulloblastoma is relatively favorable for children younger than 3 years and for adults. Young children with the MBEN histology have a particularly favorable prognosis.[137,138,139,140,141] Patients with SHH medulloblastoma at greatest risk of treatment failure are children older than 3 years whose tumors have TP53 mutations, often with co-occurring GLI2 or MYCN amplification and large cell/anaplastic histology.[126,135,142]
The 2016 WHO classification identifies SHH medulloblastoma with a TP53 mutation as a distinctive entity (medulloblastoma, SHH-activated and TP53-mutant). Approximately 25% of SHH-activated medulloblastoma cases have TP53 mutations, with a high percentage of these cases also showing a TP53 germline mutation (9 of 20 in one study). These patients are commonly between the ages of 5 years and 18 years and have a worse outcome (overall survival at 5 years, <50%). The tumors often show large cell anaplastic histology.
The WHO classification combines group 3 and group 4 medulloblastoma cases into a single entity, partly on the basis of the absence of immediate clinical impact for this distinction. Group 3 medulloblastoma represents approximately 25% of medulloblastoma cases, while group 4 medulloblastoma represents approximately 40% of medulloblastoma cases.[126,128] Both group 3 and group 4 medulloblastoma patients are predominantly male.[114,125] Group 3 and group 4 medulloblastomas can be further subdivided on the basis of characteristics such as gene expression and DNA methylation profiles, but the optimal approach to their subdivision is not established.[126,127]
Various genomic alterations are observed in group 3 and group 4 medulloblastomas; however, no single alteration occurs in more than 10% to 20% of cases. Genomic alterations include the following:
Group 3 patients with MYC amplification or MYC overexpression have a poor prognosis, with fewer than 50% of these patients surviving 5 years after diagnosis. This poor prognosis is especially true in children younger than 4 years at diagnosis. However, patients with group 3 medulloblastoma without MYC amplification who are older than 3 years have a prognosis similar to that of most patients with non-WNT medulloblastoma, with a 5-year PFS rate higher than 70%.[143,145]
Group 4 medulloblastomas occur throughout infancy and childhood and into adulthood. The prognosis for group 4 medulloblastoma patients is similar to that for patients with other non-WNT medulloblastomas and may be affected by additional factors such as the presence of metastatic disease, chromosome 11q loss, and chromosome 17p loss.[118,119,126,142] One study found that group 4 patients with either chromosome 11 loss or gain of chromosome 17 were low risk, regardless of metastases. In cases lacking both of these cytogenetic features, metastasis at presentation differentiated between high and intermediate risk.
For group 3 and group 4 standard-risk patients (i.e., without MYC amplification or metastatic disease), the gain or loss of whole chromosomes appears to connote a favorable prognosis. This finding was derived from the data of 91 patients with non-WNT/non-SHH medulloblastoma enrolled in the SIOP-PNET-4 (NCT01351870) clinical trial and was confirmed in an independent group of 70 children with non-WNT/non-SHH medulloblastoma treated between 1990 and 2014. Chromosomal abnormalities include the following:
The classification of medulloblastoma into the four major subtypes will likely be altered in the future.[126,127,146,147] Further subdivision within subgroups based on molecular characteristics is likely because each of the subgroups is further molecularly dissected, although the studies are nearing consensus as data from multiple independent studies are merged. As an example, using complementary bioinformatics approaches, concordance was analyzed between multiple large published cohorts, and a more unified subgrouping was described. For children with group 3 and group 4 medulloblastomas, eight distinct subgroups were determined by DNA methylation clustering. Specific subgroups had different prognoses.[118,128,135,148]
Whether the classification for adults with medulloblastoma has a predictive ability similar to that for children is unknown.[119,121] In one study of adult medulloblastoma, MYC oncogene amplifications were rarely observed, and tumors with 6q deletion and WNT activation (as identified by nuclear beta-catenin staining) did not share the excellent prognosis seen in pediatric medulloblastomas, although another study did confirm an excellent prognosis for WNT-activated tumors in adults.[119,121]
(Refer to the PDQ summary on Childhood Medulloblastoma and Other Central Nervous System Embryonal Tumors Treatment for information about the treatment of childhood medulloblastoma.)
Nonmedulloblastoma Embryonal Tumors
This section describes the genomic characteristics of embryonal tumors other than medulloblastoma and atypical teratoid/rhabdoid tumor. The 2016 WHO classification removed the term primitive neuroectodermal tumors (PNET) from the diagnostic lexicon. This change resulted from the recognition that many tumors previously classified as CNS PNETs have the common finding of amplification of the C19MC region on chromosome 19. These entities included ependymoblastoma, embryonal tumors with abundant neuropil and true rosettes (ETANTR), and some cases of medulloepithelioma. The 2016 WHO classification now categorizes tumors with C19MC amplification as embryonal tumor with multilayered rosettes (ETMR), C19MC-altered. Tumors previously classified as CNS PNETs are now termed CNS embryonal tumor, NOS, with the recognition that tumors in this category will likely be classified by their defining genomic lesions in future editions of the WHO classification.
Molecular subtypes of nonmedulloblastoma embryonal tumors
A study applying unsupervised clustering of DNA methylation patterns for 323 nonmedulloblastoma embryonal tumors found that approximately one-half of these tumors diagnosed as nonmedulloblastoma embryonal tumors showed molecular profiles characteristic of other known pediatric brain tumors (e.g., high-grade glioma and atypical teratoid/rhabdoid tumor [AT/RT]). This observation highlights the utility of molecular characterization to assign this class of tumors to their appropriate biology-based diagnosis.
Among the same collection of 323 tumors diagnosed as nonmedulloblastoma embryonal tumors, molecular characterization identified genomically and biologically distinctive subtypes, including the following:
ETMRs are defined at the molecular level by high-level amplification of the microRNA cluster C19MC and by a gene fusion between TTYH1 and C19MC.[149,152,153] This gene fusion puts expression of C19MC under control of the TTYH1 promoter, leading to high-level aberrant expression of the microRNAs within the cluster. The World Health Organization (WHO) allows histologically similar tumors without C19MC alteration to be classified as ETMR, not otherwise specified (NOS).
The contribution of DNA methylation profiling to correctly diagnose supratentorial embryonal tumors was demonstrated in a clinical trial of patients with supratentorial primitive neuroectodermal tumors of the CNS (CNS-PNET) and pineoblastoma. For the pineoblastoma cases, there was high concordance between the diagnosis made by methylation profiling and the diagnosis made by central pathology review diagnosis (26 of 29). However, for the remaining 31 patients, the diagnosis made by methylation profiling was high-grade glioma in 18 patients, AT/RT in 2 patients, and RELA-fusion–positive ependymoma in 2 patients. Adjudication of discrepancies between the diagnosis made by central review pathology and the diagnosis made by methylation profiling was in favor of methylation profiling in the ten cases that were re-examined.
Medulloepithelioma with the classic C19MC amplification is considered an ETMR, C19MC-altered (refer to the ETMR information above). However, when a tumor has the histological features of medulloepithelioma, but without a C19MC amplification, it is identified as a histologically discrete tumor within the WHO classification system and called medulloepithelioma.[157,158] Medulloepithelioma tumors are rare and tend to arise most commonly in infants and young children. Medulloepitheliomas, which histologically recapitulate the embryonal neural tube, tend to arise supratentorially, primarily intraventricularly, but may arise infratentorially, in the cauda, and even extraneurally, along nerve roots.[157,158]
(Refer to the PDQ summary on Childhood Medulloblastoma and Other Central Nervous System Embryonal Tumors Treatment for information about the treatment of childhood PNETs.)
Pineoblastoma, which was previously conventionally grouped with embryonal tumors, is now categorized by the World Health Organization (WHO) as a pineal parenchymal tumor. Given that therapies for pineoblastoma are quite similar to those utilized for embryonal tumors, the previous convention of including pineoblastoma with the central nervous system (CNS) embryonal tumors is followed here. Pineoblastoma is associated with germline mutations in both the RB1 gene and in DICER1, as described below:
(Refer to the PDQ summary on Childhood Medulloblastoma and Other Central Nervous System Embryonal Tumors Treatment for information about the treatment of childhood pineoblastoma.)
Molecular Subgroups of Ependymoma
Molecular characterization studies have identified nine molecular subgroups of ependymoma, six of which predominate in childhood. The subgroups are determined by their distinctive DNA methylation and gene expression profiles and by their distinctive spectrum of genomic alterations (refer to Figure 4).[166,167,168,169]
Subependymoma—whether supratentorial, infratentorial, or spinal—accounts for the remaining three molecular variants, and it is rarely, if ever, seen in children.
Figure 4. Graphical summary of key molecular and clinical characteristics of ependymal tumor subgroups. Schematic representation of key genetic and epigenetic findings in the nine molecular subgroups of ependymal tumors as identified by methylation profiling. CIN, Chromosomal instability. Reprinted from Cancer Cell, Volume 27, Kristian W. Pajtler, Hendrik Witt, Martin Sill, David T.W. Jones, Volker Hovestadt, Fabian Kratochwil, Khalida Wani, Ruth Tatevossian, Chandanamali Punchihewa, Pascal Johann, Juri Reimand, Hans-Jorg Warnatz, Marina Ryzhova, Steve Mack, Vijay Ramaswamy, David Capper, Leonille Schweizer, Laura Sieber, Andrea Wittmann, Zhiqin Huang, Peter van Sluis, Richard Volckmann, Jan Koster, Rogier Versteeg, Daniel Fults, Helen Toledano, Smadar Avigad, Lindsey M. Hoffman, Andrew M. Donson, Nicholas Foreman, Ekkehard Hewer, Karel Zitterbart, Mark Gilbert, Terri S. Armstrong, Nalin Gupta, Jeffrey C. Allen, Matthias A. Karajannis, David Zagzag, Martin Hasselblatt, Andreas E. Kulozik, Olaf Witt, V. Peter Collins, Katja von Hoff, Stefan Rutkowski, Torsten Pietsch, Gary Bader, Marie-Laure Yaspo, Andreas von Deimling, Peter Lichter, Michael D. Taylor, Richard Gilbertson, David W. Ellison, Kenneth Aldape, Andrey Korshunov, Marcel Kool, and Stefan M. Pfister, Molecular Classification of Ependymal Tumors across All CNS Compartments, Histopathological Grades, and Age Groups, Pages 728–743, Copyright (2015), with permission from Elsevier.
Posterior fossa A ependymoma (PF-EPN-A)
The most common posterior fossa ependymoma subgroup is PF-EPN-A and is characterized by the following:
A study that included over 600 cases of PF-EPN-A used methylation array profiling to divide this population into two distinctive subgroups, PFA-1 and PFA-2. Gene expression profiling suggested that these two subtypes may arise in different anatomic locations in the hindbrain. Within both of the PFA-1 and PFA-2 groups, distinctive minor subtypes could be identified, suggesting the presence of heterogeneity. Additional study will be required to define the clinical significance of these subtypes.
Posterior fossa B ependymoma (PF-EPN-B)
The PF-EPN-B subgroup is less common than the PF-EPN-A subgroup, representing 15% to 20% of all posterior fossa ependymomas in children. PF-EPN-B is characterized by the following:
Supratentorial ependymomas withRELAfusions (ST-EPN-RELA)
ST-EPN-RELA is the largest subset of pediatric supratentorial ependymomas and is characterized by gene fusions involving RELA,[179,180] a transcriptional factor important in NF-κB pathway activity. ST-EPN-RELA is characterized by the following:
Supratentorial ependymomas withYAP1fusions (ST-EPN-YAP1)
ST-EPN-YAP1 is the second, less common subset of supratentorial ependymomas and has fusions involving YAP1 on chromosome 11. ST-EPN-YAP1 is characterized by the following:
Supratentorial ependymomas without RELA or YAP1 fusions (on chromosome 11) are an undefined entity, and it is unclear what these samples represent. By DNA methylation analysis, these samples often cluster with other entities such as high-grade gliomas and embryonal tumors; caution should be taken when diagnosing a supratentorial ependymoma that does not harbor a fusion involving chromosome 11.[30,181]
(Refer to the PDQ summary on Childhood Ependymoma Treatment for information about the treatment of childhood ependymoma.)
Genomic abnormalities related to hepatoblastoma include the following:
Similar mutations have been found in many types of cancer, including hepatocellular carcinoma. These mutations render NFE2L2 insensitive to KEAP1-mediated degradation, leading to activation of the NFE2L2-KEAP1 pathway, which activates resistance to oxidative stress and is believed to confer resistance to chemotherapy.
To date, these genetic mutations have not been used to select therapeutic agents for investigation in clinical trials.
Genomic abnormalities related to hepatocellular carcinoma include the following:
To date, these genetic mutations have not been used to select therapeutic agents for investigation in clinical trials.
(Refer to the PDQ summary on Childhood Liver Cancer Treatment for information about the treatment of liver cancer.)
Molecular Features of Osteosarcoma
The genomic landscape of osteosarcoma is distinctive from that of other childhood cancers. It is characterized by an exceptionally high number of structural variants with relatively small numbers of single nucleotide variants compared with many adult cancers.[1,2]
Key observations regarding the genomic landscape of osteosarcoma are summarized below:
Figure 6. Circos plots of osteosarcoma cases from the National Cancer Institute's Therapeutically Applicable Research to Generate Effective Treatments (TARGET) project. The red lines in the interior circle connect chromosome regions involved in either intra- or inter-chromosomal translocations. Osteosarcoma is distinctive from other childhood cancers because it has a large number of intra- and inter-chromosomal translocations. Credit: National Cancer Institute.
Genetic predisposition to osteosarcoma
Several germline mutations are associated with susceptibility to osteosarcoma; Table 5 summarizes the syndromes and associated genes for these conditions. The frequency of pathogenic or likely pathogenic cancer-susceptibility gene variants in individuals with osteosarcoma is 28% greater than that in control groups.
Mutations in TP53 are the most common germline alterations associated with osteosarcoma. Mutations in this gene are found in approximately 70% of patients with Li-Fraumeni syndrome (LFS), which is associated with increased risk of osteosarcoma, breast cancer, various brain cancers, soft tissue sarcomas, and other cancers. While rhabdomyosarcoma is the most common sarcoma arising in patients aged 5 years and younger with TP53-associated LFS, osteosarcoma is the most common sarcoma in children and adolescents aged 6 to 19 years. One study observed a high frequency of young osteosarcoma cases (age <30 years) carrying a known LFS-associated or likely LFS-associated TP53 mutation (3.8%) or rare exonic TP53 variant (5.7%), with an overall TP53 mutation frequency of 9.5%. Another study observed germline TP53 mutations in 7 of 59 osteosarcoma cases (12%) subjected to whole-exome sequencing. Other groups have reported lower rates (3%–7%) of TP53 germline mutations in patients with osteosarcoma.[6,7]
|AML = acute myeloid leukemia; IL-1 = interleukin-1; MDS = myelodysplastic syndrome; RANKL = receptor activator of nuclear factor kappa beta ligand; TNF = tumor necrosis factor.|
|a Adapted from Kansara et al.|
|Bloom syndrome||Rare inherited disorder characterized by short stature and sun-sensitive skin changes. Often presents with a long, narrow face, small lower jaw, large nose, and prominent ears.||15q26.1||BLM||DNA helicase|
|Diamond-Blackfan anemia||Inherited pure red cell aplasia. Patients at risk for MDS and AML. Associated with skeletal abnormalities such as abnormal facial features (flat nasal bridge, widely spaced eyes).||Ribosomal proteins||Ribosome production[10,11]|
|Li-Fraumeni syndrome||Inherited mutation inTP53gene. Affected family members at increased risk of bone tumors, breast cancer, leukemia, brain tumors, and sarcomas.||17p13.1||P53||DNA damage response|
|Paget disease||Excessive breakdown of bone with abnormal bone formation and remodeling, resulting in pain from weak, malformed bone.||18q21-qa22||LOH18CR1||IL-1/TNF signaling; RANKL signaling pathway|
|Retinoblastoma||Malignant tumor of the retina. Approximately 66% of patients are diagnosed by age 2 years and 95% of patients by age 3 years. Patients with heritable germ cell mutations at greater risk of subsequent neoplasms.||13q14.2||RB1||Cell-cycle checkpoint|
|Rothmund-Thomson syndrome (also called poikiloderma congenitale)[15,16]||Autosomal recessive condition. Associated with skin findings (atrophy, telangiectasias, pigmentation), sparse hair, cataracts, small stature, and skeletal abnormalities. Increased incidence of osteosarcoma at a younger age.||8q24.3||RECQL4||DNA helicase|
|Werner syndrome||Patients often have short stature and in their early twenties, develop signs of aging, including graying of hair and hardening of skin. Other aging problems such as cataracts, skin ulcers, and atherosclerosis develop later.||8p12-p11.2||WRN||DNA helicase; exonuclease activity|
Refer to the following PDQ summaries for more information about these genetic syndromes:
(Refer to the PDQ summary on Osteosarcoma and Undifferentiated Pleomorphic Sarcoma of Bone Treatment for information about the treatment of osteosarcoma.)
Molecular Features of Ewing Sarcoma
The detection of a translocation involving the EWSR1 gene on chromosome 22 band q12 and any one of a number of partner chromosomes is the key feature in the diagnosis of Ewing sarcoma (refer to Table 6). The EWSR1 gene is a member of the TET family [TLS/EWS/TAF15] of RNA-binding proteins. The FLI1 gene is a member of the ETS family of DNA-binding genes. Characteristically, the amino terminus of the EWSR1 gene is juxtaposed with the carboxy terminus of the STS family gene. In most cases (90%), the carboxy terminus is provided by FLI1, a member of the family of transcription factor genes located on chromosome 11 band q24. Other family members that may combine with the EWSR1 gene are ERG, ETV1, ETV4 (also termed E1AF), and FEV. Rarely, TLS, another TET family member, can substitute for EWSR1. Finally, there are a few rare cases in which EWSR1 has translocated with partners that are not members of the ETS family of oncogenes. The significance of these alternate partners is not known.
Besides these consistent aberrations involving the EWSR1 gene at 22q12, additional numerical and structural aberrations have been observed in Ewing sarcoma, including gains of chromosomes 2, 5, 8, 9, 12, and 15; the nonreciprocal translocation t(1;16)(q12;q11.2); and deletions on the short arm of chromosome 6. Trisomy 20 may be associated with a more aggressive subset of Ewing sarcoma.
Three papers have described the genomic landscape of Ewing sarcoma and all show that these tumors have a relatively silent genome, with a paucity of mutations in pathways that might be amenable to treatment with novel targeted therapies.[23,24,25] These papers also identified mutations in STAG2, a member of the cohesin complex, in about 15% to 20% of the cases, and the presence of these mutations was associated with advanced-stage disease. CDKN2A deletions were noted in 12% to 22% of cases. Finally, TP53 mutations were identified in about 6% to 7% of cases and the coexistence of STAG2 and TP53 mutations is associated with a poor clinical outcome.[23,24,25]
Figure 7 below from a discovery cohort (n = 99) highlights the frequency of chromosome 8 gain, the co-occurrence of chromosome 1q gain and chromosome 16q loss, the mutual exclusivity of CDKN2A deletion and STAG2 mutation, and the relative paucity of recurrent single nucleotide variants for Ewing sarcoma.
Figure 7. A comprehensive profile of the genetic abnormalities in Ewing sarcoma and associated clinical information. Key clinical characteristics are indicated, including primary site, type of tissue, and metastatic status at diagnosis, follow-up, and last news. Below is the consistency of detection of gene fusions by RT-PCR and whole-genome sequencing (WGS). The numbers of structural variants (SV) and single-nucleotide variants (SNV) as well as indels are reported in grayscale. The presence of the main copy-number changes, chr 1q gain, chr 16 loss, chr 8 gain, chr 12 gain, and interstitial CDKN2A deletion is indicated. Listed last are the most significant mutations and their types. For gene mutations, "others" refers to: duplication of exon 22 leading to frameshift (STAG2), deletion of exon 2 to 11 (BCOR), and deletion of exons 1 to 6 (ZMYM3). Reprinted from Cancer Discovery, Copyright 2014, 4 (11), 1342–53, Tirode F, Surdez D, Ma X, et al., Genomic Landscape of Ewing Sarcoma Defines an Aggressive Subtype with Co-Association of STAG2 and TP53 mutations, with permission from AACR.
Ewing sarcoma translocations can all be found with standard cytogenetic analysis. A more rapid analysis looking for a break apart of the EWS gene is now frequently done to confirm the diagnosis of Ewing sarcoma molecularly. This test result must be considered with caution, however. Ewing sarcomas that utilize the TLS translocations will have negative tests because the EWSR1 gene is not translocated in those cases. In addition, other small round tumors also contain translocations of different ETS family members with EWSR1, such as desmoplastic small round cell tumor, clear cell sarcoma, extraskeletal myxoid chondrosarcoma, and myxoid liposarcoma, all of which may be positive with a EWS fluorescence in situ hybridization (FISH) break-apart probe. A detailed analysis of 85 patients with small round blue cell tumors that were negative for EWSR1 rearrangement by FISH with an EWSR1 break-apart probe identified eight patients with FUS rearrangements. Four patients who had EWSR1-ERG fusions were not detected by FISH with an EWSR1 break-apart probe. The authors do not recommend relying solely on EWSR1 break-apart probes for analyzing small round blue cell tumors with strong immunohistochemical positivity for CD99.
Undifferentiated small blue round cell sarcomas with the EWSR1-NFATC2 fusion have been studied with DNA methylation profiling; this revealed a homogeneous methylation cluster for these sarcomas with EWSR1-NFATC2 fusions, which clearly segregated them from the more common form of Ewing sarcoma with EWS-ETS translocations.
Small round blue cell tumors of bone and soft tissue that are histologically similar to Ewing sarcoma but do not have rearrangements of the EWSR1 gene have been analyzed and translocations have been identified. These include BCOR-CCNB3, CIC-DUX4, and CIC-FOX4.[29,30,31,32] The molecular profile of these tumors is different from the profile of EWS-FLI1 translocated Ewing sarcoma, and limited evidence suggests that they have a different clinical behavior. In almost all cases, the patients were treated with therapy designed for Ewing sarcoma on the basis of the histologic and immunohistologic similarity to Ewing sarcoma (refer to the Undifferentiated Round Cell Sarcomas With BCOR-CCNB3 Rearrangements and Undifferentiated Round Cell Sarcomas With CIC-DUX4 Rearrangements sections of this summary for more information). There are too few cases associated with each translocation to determine whether the prognosis for these small round blue cell tumors is distinct from the prognosis of Ewing sarcoma of similar stage and site.[29,30,31,32]
Some undifferentiated round cell sarcomas are characterized by paracentric inversion of chromosome X and a BCOR-CCNB3 rearrangement; alternative BCOR partners, including MAML3 and ZC3H7B, have also been reported. Despite clinical pathologic similarities to Ewing sarcoma, these tumors are biologically different by expression profiling and single-nucleotide polymorphism array analysis. (Refer to the Undifferentiated Round Cell Sarcomas With BCOR-CCNB3 Rearrangements section of this summary for more information about the treatment of this disease.)
Other undifferentiated round cell sarcomas are characterized by a CIC-DUX4 fusion resulting from a recurrent t(4;19) or t(10;19) and are the most common EWSR1-FUS fusion–negative undifferentiated round cell sarcomas. (Refer to the Undifferentiated Round Cell Sarcomas With CIC-DUX4 Rearrangements section of this summary for more information about the treatment of this disease.)
Genome-wide association studies have identified susceptibility loci for Ewing sarcoma at 1p36.22, 10q21, and 15q15.[35,36,37] Deep sequencing through the 10q21.3 region identified a polymorphism in the EGR2 gene, which appears to cooperate with and magnify the enhanced activity of the gene product of the EWSR1-FLI1 fusion that is seen in most patients with Ewing sarcoma. The polymorphism associated with the increased risk is found at a much higher frequency in whites than in blacks or Asians, possibly contributing to the epidemiology of the relative infrequency of Ewing sarcoma in the latter populations. Three new susceptibility loci have been identified at 6p25.1, 20p11.22, and 20p11.23.
|TET Family Partner||Fusion With ETS-like Oncogene Partner||Translocation||Comment|
|a These partners are not members of theETSfamily of oncogenes.|
|EWS||EWSR1-FLI1||t(11;22)(q24;q12)||Most common; ~85% to 90% of cases|
|EWSR1-ERG||t(21;22)(q22;q12)||Second most common; ~10% of cases|
(Refer to the PDQ summary on Ewing Sarcoma Treatment for information about the treatment of Ewing sarcoma.)
Genomics of rhabdomyosarcoma
The embryonal and alveolar histologies have distinctive molecular characteristics that have been used for diagnostic confirmation, and may be useful for assigning risk group, determining therapy, and monitoring residual disease during treatment.[38,39,40,41,42]
Embryonal histology with anaplasia: Anaplasia has been reported in a minority of children with rhabdomyosarcoma, primarily arising in children with the embryonal subtype who are younger than 10 years.[48,49] Rhabdomyosarcoma with nonalveolar anaplastic morphology may be a presenting feature for children with Li-Fraumeni syndrome and germline TP53 mutations. Among eight consecutively presenting children with rhabdomyosarcoma and TP53 germline mutations, all showed anaplastic morphology. Among an additional seven children with anaplastic rhabdomyosarcoma and unknown TP53 germline mutation status, three of the seven children had functionally relevant TP53 germline mutations. The median age at diagnosis of the 11 children with TP53 germline mutation status was 40 months (range, 19–67 months).
For the diagnosis of alveolar rhabdomyosarcoma, a FOXO1 gene rearrangement may be detected with good sensitivity and specificity using either fluorescence in situ hybridization or reverse transcription–polymerase chain reaction.
The alveolar histology that is associated with the PAX7 gene in patients with or without metastatic disease appears to occur at a younger age and may be associated with longer event-free survival rates than those associated with PAX3 gene rearrangements.[54,55,56,57,58,59] Patients with alveolar histology and the PAX3 gene are older and have a higher incidence of invasive tumor (T2). Around 22% of cases showing alveolar histology have no detectable PAX gene translocation.[42,52]
In addition to FOXO1 rearrangements, alveolar tumors are characterized by a lower mutational burden than are fusion-negative tumors, with fewer genes having recurring mutations.[46,47]BCOR and PIK3CA mutations and amplification of MYCN, MIR17HG, and CDK4 have also been described.
For congenital/infantile spindle cell rhabdomyosarcoma, a study reported that 10 of 11 patients showed recurrent fusion genes. Most of these patients had truncal primary tumors, and no paratesticular tumors were found. Novel VGLL2 rearrangements were observed in seven patients (63%), including the VGLL2-CITED2 fusion in four patients and the VGLL2-NCOA2 fusion in two patients. Three patients (27%) harbored different NCOA2 gene fusions, including TEAD1-NCOA2 in two patients and SRF-NCOA2 in one patient. All fusion-positive congenital/infantile spindle cell rhabdomyosarcoma patients with available long-term follow-up were alive and well, and no patients developed distant metastases. Further study is needed to better define the prevalence and prognostic significance of these gene rearrangements in young children with spindle cell rhabdomyosarcoma.
In older children and adults with spindle cell/sclerosing rhabdomyosarcoma, a specific MYOD1 mutation (p.L122R) has been observed in a large proportion of patients.[61,62,63,64] Activating PIK3CA mutations are seen in about one-half of the cases, and 60% of these cases have pure sclerosing morphology. The presence of the MYOD1 mutation is associated with an increased risk of local and distant failure.[61,62,63] In one study that included 15 children with MYOD1-mutant tumors, the most common primary site was the head and neck region. These patients had sclerosing spindle or mixed histology, and 10 of 15 patients died of disease despite aggressive multimodal therapy.
These findings highlight the important differences between embryonal and alveolar tumors. Data demonstrate that PAX-FOXO1 fusion–positive alveolar tumors are biologically and clinically different from fusion-negative alveolar tumors and embryonal tumors.[42,67,68,69,70] In a study of Intergroup Rhabdomyosarcoma Study Group patients, which captured an entire cohort from a single prospective clinical trial, the outcome for patients with translocation-negative alveolar rhabdomyosarcoma was better than that observed for translocation-positive patients. The outcome was similar to that seen in patients with embryonal rhabdomyosarcoma and demonstrated that fusion status is a critical factor for risk stratification in pediatric rhabdomyosarcoma.
Genome-wide methylation assays can accurately identify PAX3 and PAX7 fusion–positive rhabdomyosarcomas, as well as wild-type and RAS mutant fusion–negative tumors.
(Refer to the PDQ summary on Childhood Rhabdomyosarcoma Treatment for information about the treatment of childhood rhabdomyosarcoma.)
Genomics of Langerhans Cell Histiocytosis
Studies published in 1994 showed clonality in Langerhans cell histiocytosis (LCH) using polymorphisms of methylation-specific restriction enzyme sites on the X-chromosome regions coding for the human androgen receptor, DXS255, PGK, and HPRT.[1,2] Analysis of lesions with single-system or multisystem disease showed a proliferation of LCH cells from a single clone. The discovery of recurring genomic alterations (primarily BRAF V600E) in LCH (see below) confirmed the clonality of LCH in children.
Pulmonary LCH in adults was initially reported to be nonclonal in approximately 75% of cases, while an analysis of BRAF mutations showed that 25% to 50% of adult lung LCH patients had evidence of BRAF V600E mutations.[3,4] Another study of 26 pulmonary LCH cases found that 50% had BRAF V600E mutations and 40% had NRAS mutations. Approximately the same number of mutations are polyclonal as are monoclonal. It has not yet been determined whether clonality and BRAF pathway mutations are concordant in the same patients, which might suggest a reactive rather than a neoplastic condition in smoker's lung LCH and a clonal neoplasm in other types of LCH.
Figure 8. Courtesy of Rikhia Chakraborty, Ph.D. Permission to reuse the figure in any form must be obtained directly from Dr. Chakraborty.
The theory for the genomic basis of LCH was advanced by a 2010 report of an activating mutation of the BRAF oncogene (V600E) that was detected in 35 of 61 cases (57%). Multiple subsequent reports have confirmed the presence of BRAF V600E mutations in 50% or more of LCH cases in children.[7,8,9] Other BRAF mutations that result in signal activation have been described.[8,10]ARAF mutations are infrequent in LCH but, when present, can also lead to RAS-MAPK pathway activation.
The RAS-MAPK signaling pathway (refer to Figure 8) transmits signals from a cell surface receptor (e.g., a growth factor) through the RAS pathway (via one of the RAF proteins [A, B, or C]) to phosphorylate MEK and then the extracellular signal-regulated kinase (ERK), which leads to nuclear signals affecting cell cycle and transcription regulation. The V600E mutation of BRAF leads to continuous phosphorylation, and thus activation, of MEK and ERK without the need for an external signal. Activation of ERK occurs by phosphorylation, and phosphorylated ERK can be detected in virtually all LCH lesions.[6,12]
Because RAS-MAPK pathway activation can be detected in all LCH cases, but not all cases have BRAF mutations, the presence of genomic alterations in other components of the pathway was suspected. The following genomic alterations were identified:
Studies support the universal activation of ERK in LCH; ERK activation in most cases is explained by BRAF and MAP2K1 alterations.[6,12,14] Altogether, these mutations in the MAP kinase pathway account for nearly 90% of the causes of the universal activation of ERK in LCH.[6,12,14]
The presence of the BRAF V600E mutation in blood and bone marrow was studied in a series of 100 patients, 65% of whom tested positive for the BRAF V600E mutation by a sensitive quantitative polymerase chain reaction technique. Circulating cells with the BRAF V600E mutation could be detected in all high-risk patients and in a subset of low-risk multisystem patients. The presence of circulating cells with the mutation conferred a twofold increased risk of relapse. In a similar study that included 48 patients with BRAF V600E–mutated LCH, the BRAF V600E allele was detected in circulating cell-free DNA in 100% of patients with risk-organ–positive multisystem LCH, 42% of patients with risk-organ–negative LCH, and 14% of patients with single-system LCH.
The myeloid dendritic cell origin of LCH was confirmed by finding CD34-positive stem cells with the mutation in the bone marrow of high-risk patients. In those with low-risk disease, the mutation was found in more mature myeloid dendritic cells, suggesting that the stage of cell development at which the somatic mutation occurs is critical in defining the extent of disease in LCH. LCH is now generally considered to represent a myeloid neoplasm.
Clinical implications of the described genomic findings include the following:
Several case reports and two case series have demonstrated the efficacy of BRAF inhibitors for the treatment of LCH in children.[27,28,29,30,31,32] However, the long-term role of this therapy is complicated because most patients will relapse when the inhibitors are stopped.
(Refer to the PDQ summary on Langerhans Cell Histiocytosis Treatment for information about the treatment of childhood LCH.)
Molecular features of neuroblastoma
Children with neuroblastoma can be subdivided into subsets with different predicted risks of relapse on the basis of clinical factors and biological markers at the time of diagnosis.
Genomic analyses of localized, resectable non-MYC amplified neuroblastoma cases from two consecutive European studies and a North American cohort revealed a different prognostic impact of tumor genomics depending on patient age (<18 months or >18 months). Patients were treated with surgery alone regardless of a tumor residuum.[Level of evidence: 3iii]
Key genomic characteristics of high-risk neuroblastoma that are discussed below include the following:
Segmental chromosomal aberrations
Segmental chromosomal aberrations, found most frequently in 1p, 1q, 3p, 11q, 14q, and 17p, are best detected by comparative genomic hybridization and are seen in most high-risk and/or stage 4 neuroblastoma tumors.[4,5,6,7,8] Among all patients with neuroblastoma, a higher number of chromosome breakpoints (i.e., a higher number of segmental chromosome aberrations) correlated with the following:[4,5,6,7,8][Level of evidence: 3iiD]
An international collaboration studied 556 patients with high-risk neuroblastoma and identified two types of segmental copy number aberrations that are associated with extremely poor outcome. Distal 6q losses were found in 6% of patients and were associated with a 10-year survival rate of only 3.4%; amplifications of regions not encompassing the MYCN locus, in addition to MYCN amplification, were detected in 18% of the patients and were associated with a 10-year survival rate of 5.8%.
In a study of children older than 12 months who had unresectable primary neuroblastomas without metastases, segmental chromosomal aberrations were found in most, and older children were more likely to have them and to have more of them per tumor cell. In children aged 12 to 18 months, the presence of segmental chromosomal aberrations had a significant effect on EFS but not on OS. However, in children older than 18 months, there was a significant difference in OS between children with segmental chromosomal aberrations (67%) and children without segmental chromosomal aberrations (100%), regardless of tumor histology.
In all stages of disease, amplification of the MYCN gene strongly predicts a poorer prognosis, in both time to tumor progression and OS, in almost all multivariate regression analyses of prognostic factors.[1,2] Within the localized-tumor MYCN-amplified cohort, patients with hyperdiploid tumors have better outcomes than do patients with diploid tumors. However, patients with hyperdiploid tumors with MYCN amplification or any segmental chromosomal aberrations do relatively poorly compared with patients with hyperdiploid tumors without MYCN amplification.
In a Children's Oncology Group study of MYCN copy number in 4,672 patients with neuroblastoma, the following results were reported:
Most unfavorable clinical and pathobiological features are associated, to some degree, with MYCN amplification; in a multivariable logistic regression analysis of 7,102 patients in the International Neuroblastoma Risk Group (INRG) study, pooled segmental chromosomal aberrations and gains of 17q were poor prognostic features even when not associated with MYCN amplification. However, another poor prognostic feature, segmental chromosomal aberrations at 11q, are almost entirely mutually exclusive of MYCN amplification.[15,16]
In a cohort of 6,223 patients from the INRG database with known MYCN status, the OS hazard ratio (HR) associated with MYCN amplification was 6.3 (95% confidence interval [CI], 5.7–7.0; P < .001). The greatest adverse prognostic impact of MYCN amplification for OS was in the youngest patients (<18 months: HR, 19.6; ≥18 months: HR, 3.0). Patients whose outcome was most impacted by MYCN status were those with otherwise favorable features, including age younger than 18 months, high mitosis karrhyohexis index, and low ferritin.[Level of evidence: 3iiiA]
Intratumoral heterogeneous MYCN amplification (hetMNA) refers to the coexistence of MYCN-amplified cells as a cluster or as single scattered cells and non-MYCN–amplified tumor cells. HetMNA has been reported infrequently and can occur spatially within the tumor as well as between the tumor and the metastasis at the same time or temporally during the disease course. The International Society of Paediatric Oncology Europe Neuroblastoma (SIOPEN) biology group investigated the prognostic significance of this neuroblastoma subtype. Tumor tissue from 99 patients identified as having hetMNA and diagnosed between 1991 and 2015 was analyzed to elucidate the prognostic significance of MYCN-amplified clones in otherwise non-MYCN–amplified neuroblastomas. Patients younger than 18 months showed a better outcome in all stages compared with older patients. The genomic background correlated significantly with relapse frequency and OS. No relapses occurred in cases of only numerical chromosomal aberrations. This study suggests that hetMNA tumors have to be evaluated in the context of the genomic tumor background in combination with the clinical pattern, including the patient's age and disease stage. Future studies are needed in patients younger than 18 months who have localized disease with hetMNA.
Exonic mutations in neuroblastoma
Multiple reports have documented that a minority of high-risk neuroblastomas have a low incidence of recurrently mutated genes. The most commonly mutated gene is ALK, which is mutated in approximately 10% of patients (see below). Other genes with even lower frequencies of mutations include ATRX, PTPN11, ARID1A, and ARID1B.[19,20,21,22,23,24,25] As shown in Figure 9, most neuroblastoma cases lack mutations in genes that are altered in a recurrent manner.
Figure 9. Data tracks (rows) facilitate the comparison of clinical and genomic data across cases with neuroblastoma (columns). The data sources and sequencing technology used were whole-exome sequencing (WES) from whole-genome amplification (WGA) (light purple), WES from native DNA (dark purple), Illumina WGS (green), and Complete Genomics WGS (yellow). Striped blocks indicate cases analyzed using two approaches. The clinical variables included were sex (male, blue; female, pink) and age (brown spectrum). Copy number alterations indicates ploidy measured by flow cytometry (with hyperdiploid meaning DNA index >1) and clinically relevant copy number alterations derived from sequence data. Significantly mutated genes are those with statistically significant mutation counts given the background mutation rate, gene size, and expression in neuroblastoma. Germline indicates genes with significant numbers of germline ClinVar variants or loss-of-function cancer gene variants in our cohort. DNA repair indicates genes that may be associated with an increased mutation frequency in two apparently hypermutated tumors. Predicted effects of somatic mutations are color coded according to the legend. Reprinted by permission from Macmillan Publishers Ltd: Nature Genetics (Pugh TJ, Morozova O, Attiyeh EF, et al.: The genetic landscape of high-risk neuroblastoma. Nat Genet 45 (3): 279-84, 2013), copyright (2013).
The ALK gene provides instructions for making a cell surface receptor tyrosine kinase, expressed at significant levels only in developing embryonic and neonatal brains; ALK is the exonic mutation found most commonly in neuroblastoma. Germline mutations in ALK have been identified as the major cause of hereditary neuroblastoma. Somatically acquired ALK-activating exonic mutations are also found as oncogenic drivers in neuroblastoma.
The presence of an ALK mutation correlates with significantly poorer survival in high-risk and intermediate-risk neuroblastoma patients. ALK mutations were examined in 1,596 diagnostic neuroblastoma samples and the following results were observed:
In a study that compared the genomic data of primary diagnostic neuroblastomas originating in the adrenal gland (n = 646) with that of neuroblastomas originating in the thoracic sympathetic ganglia (n = 118), 16% of thoracic tumors harbored ALK mutations.
Small-molecule ALK kinase inhibitors such as crizotinib (added to conventional therapy) are being tested in patients with newly diagnosed high-risk neuroblastoma and activated ALK (COG ANBL1531).
Genomic evolution of exonic mutations
There are limited data regarding the genomic evolution of exonic mutations from diagnosis to relapse for neuroblastoma. Whole-genome sequencing was applied to 23 paired diagnostic and relapsed neuroblastoma tumor samples to define somatic genetic alterations associated with relapse, while a second study evaluated 16 paired diagnostic and relapsed specimens. Both studies identified an increased number of mutations in the relapsed samples compared with the samples at diagnosis; this has been confirmed in a study of neuroblastoma tumor samples sent for next-generation sequencing.
In addition, three relapse samples showed structural alterations involving MAPK pathway genes consistent with pathway activation, so aberrations in this pathway were detected in 18 of 23 (78%) relapse samples. Aberrations were found in ALK (n = 10), NF1 (n = 2), and one each in NRAS, KRAS, HRAS, BRAF, PTPN11, and FGFR1. Even with deep sequencing, 7 of the 18 alterations were not detectable in the primary tumor, highlighting the evolution of mutations presumably leading to relapse and the importance of genomic evaluations of tissues obtained at relapse.
In a deep-sequencing study, 276 neuroblastoma samples (comprised of all stages and from patients of all ages at diagnosis) underwent very deep (33,000X) sequencing of just two amplified ALK mutational hot spots, which revealed 4.8% clonal mutations and an additional 5% subclonal mutations, suggesting that subclonal ALK gene mutations are common. Thus, deep sequencing can reveal the presence of mutations in tiny subsets of neuroblastoma tumor cells that may be able to survive during treatment and grow to constitute a relapse.
Genomic alterations promoting telomere lengthening
Lengthening of telomeres, the tips of chromosomes, promotes cell survival. Telomeres otherwise shorten with each cell replication, resulting eventually in the lack of a cell's ability to replicate. Low-risk neuroblastoma tumors have little telomere lengthening activity. Aberrant genetic mechanisms for telomere lengthening have been identified in high-risk neuroblastoma tumors.[19,20,31] Thus far, the following three mechanisms, which appear to be mutually exclusive, have been described:
Additional biological factors associated with prognosis
MYC and MYCN expression
Immunostaining for MYC and MYCN proteins on a restricted subset of 357 undifferentiated/poorly differentiated neuroblastoma tumors demonstrated that elevated MYC/MYCN protein expression is prognostically significant. Sixty-eight tumors (19%) highly expressed the MYCN protein, and 81 were MYCN amplified. Thirty-nine tumors (10.9%) expressed MYC highly and were mutually exclusive of high MYCN expression; in the MYC-expressing tumors, MYC or MYCN gene amplification was not seen. Segmental chromosomal aberrations were not examined in this study.
Neurotrophin receptor kinases
Expression of neurotrophin receptor kinases and their ligands vary between high-risk and low-risk tumors. TrkA is found on low-risk tumors, and absence of its ligand NGF is postulated to lead to spontaneous tumor regression. In contrast, TrkB is found in high-risk tumors that also express its ligand, BDNF, which promotes neuroblastoma cell growth and survival.
Immune system inhibition
Anti-GD2 antibodies, along with modulation of the immune system to enhance the antibody's antineuroblastoma activity, are often used to help treat neuroblastoma. The clinical effectiveness of one such antibody led to U.S. Food and Drug Administration approval of dinutuximab. The patient response to immunotherapy may, in part, be caused by variation in immune function among patients. One anti-GD2 antibody, termed 3F8, used for treating neuroblastoma exclusively at one institution, utilizes natural killer cells to kill the neuroblastoma cells. However, the natural killer cells can be inhibited by the interaction of HLA antigens and killer immunoglobulin receptor (KIR) subtypes.[35,36] This finding was confirmed and expanded by an analysis of outcomes for patients treated on the national randomized COG-ANBL0032 (NCT00026312) study with the anti-GD2 antibody dinutuximab combined with granulocyte-macrophage colony-stimulating factor and interleukin-2. The study found that certain KIR/KIR-ligand genotypes were associated with better outcomes for patients who were treated with immunotherapy.[Level of evidence: 1A] The presence of inhibitory KIR/KIR ligands was associated with a decreased effect of immunotherapy. Thus, the patient's immune system genes help determine response to immunotherapy for neuroblastoma. Additional studies are needed to determine whether this immune system genotyping can guide patient selection for certain immunotherapies.
(Refer to the PDQ summary on Neuroblastoma Treatment for information about the treatment of neuroblastoma.)
Retinoblastoma is a tumor that occurs in heritable (25%–30%) and nonheritable (70%–75%) forms. Heritable disease is defined by the presence of a germline mutation of the RB1 gene. This germline mutation may have been inherited from an affected progenitor (25% of cases) or may have occurred in a germ cell before conception or in utero during early embryogenesis in patients with sporadic disease (75% of cases). The presence of positive family history or bilateral or multifocal disease is suggestive of heritable disease.
Heritable retinoblastoma may manifest as unilateral or bilateral disease. The penetrance of the RB1 mutation (laterality, age at diagnosis, and number of tumors) is probably dependent on concurrent genetic modifiers such as MDM2 and MDM4 polymorphisms.[1,2] All children with bilateral disease and approximately 15% of patients with unilateral disease are presumed to have the heritable form, even though only 25% have an affected parent.
Children with heritable retinoblastoma tend to be diagnosed at a younger age than are children with the nonheritable form of the disease. It was thought that unilateral retinoblastoma in children younger than 1 year raises concern for the presence of heritable disease, whereas older children with a unilateral tumor are more likely to have the nonheritable form of the disease. However, in a retrospective single-institution report of 182 patients with unilateral retinoblastoma, patients with a positive genetic result (n = 32) were diagnosed at a mean age of 26 months, and patients without genetic results were diagnosed at a mean age of 22 months (P = .31).
The genomic landscape of retinoblastoma is driven by alterations in RB1 that lead to biallelic inactivation.[5,6] A rare cause of RB1 inactivation is chromothripsis, which may be difficult to detect by conventional methods.
Other recurring genomic changes that occur in a small minority of tumors include BCOR mutation/deletion, MYCN amplification, and OTX2 amplification.[5,6,7] A study of 1,068 unilateral nonfamilial retinoblastoma tumors reported that a small percentage of cases (approximately 3%) lacked evidence of RB1 loss. Approximately one-half of these cases with no evidence of RB1 loss (representing approximately 1.5% of all unilateral nonfamilial retinoblastoma) showed MYCN amplification. The functional status of the retinoblastoma protein (pRb) is inferred to be inactive in retinoblastoma with MYCN amplification. This suggests that inactivation of RB1 by mutation or inactive pRb is a requirement for the development of retinoblastoma, independent of MYCN amplification.
Genetic counseling is recommended for all patients with retinoblastoma. (Refer to the Genetic Counseling section of the PDQ summary on Retinoblastoma Treatment for more information.)
(Refer to the PDQ summary on Retinoblastoma Treatment for information about the treatment of retinoblastoma.)
Molecular Features of Wilms Tumor
Wilms tumor may arise during embryogenesis on the background of an otherwise genomically normal kidney, or it may arise from nongermline somatic genetic precursor lesions residing in histologically and functionally normal kidney tissue. Hypermethylation of H19, a known component of a subset of Wilms tumors, is a very common genetic abnormality found in these normal-appearing areas of precursor lesions.
One study performed genome-wide sequencing, mRNA and miRNA expression, DNA copy number, and methylation analysis on 117 Wilms tumors followed by targeted sequencing of 651 Wilms tumors. The tumors were selected for either favorable histology (FH) Wilms that had relapsed or those with diffuse anaplasia. The study showed the following: