|Home | About | Journals | Submit | Contact Us | Français|
Embryonal brain tumors are a heterogeneous group of neoplasms united by the presence of poorly differentiated stem-like cells. Molecular details are increasingly being used to separate them into biologically and clinically meaningful groups. For medulloblastoma, integrated mRNA expression profiling and DNA analysis by a number of research groups defines four to six distinctive molecular variants. A subset with prominent Wnt activity is associated with good clinical outcomes and classic histology. Medulloblastomas showing a Hedgehog gene expression signature are frequently of the desmoplastic/nodular subtype. Interestingly, Hedgehog activity is found in tumors arising either in infants or older teenagers and adults. The association of clinically aggressive medulloblastoma with MYC expression, large cell/anaplastic change and high levels of photoreceptor differentiation transcripts has also been noted in several studies. Immunohistochemical analysis of just one or two genes per molecular medulloblastoma variant may be sufficient for accurate classification, and this would be of great practical utility if validated. Advances have also been made in the classification of CNS Primitive Neuroectodermal Tumors (PNET), as several groups have identified an amplicon at chromosome 19q13.41–42, which appears to define a unique PNET subtype associated with prominent true rosettes, young age and very poor outcomes.
The protean qualities of embryonal brain tumors have fascinated clinicians and scientists for almost a century. While undifferentiated cells form the bulk of these lesions, many show clear differentiation along neuronal, glial or (less frequently) other lineages (49). Questions regarding the cells which give rise to these embryonal neoplasms, and how their appearance and origins might relate to therapeutic response and clinical outcome, have been debated since the initial central nervous system (CNS) tumor classification was proposed. Such issues have been the subject of extensive clinical, translational and basic research over the last decade, and some longstanding questions in the field are beginning to be answered through the analysis of key genes and the mRNA transcripts and proteins they encode. This review examines progress in the molecular analysis of medulloblastoma and other embryonal brain tumors. While some early work will be presented, the focus will be on recent research addressing how molecular markers can be used to more precisely classify embryonal brain tumors. The relationship between molecular classification, demographic details and clinical prognosis will also be examined.
Molecular changes are generally evaluated in relation to specific tumor subtypes, and the examination of this interplay has helped to promote more accurate classification. The current World Health Organization (WHO) scheme was last revised in 2007, and changes were made to both the medulloblastoma and PNET categories based in part on molecular data (49). Medulloblastoma represents the largest group of embryonal brain tumors. By definition, they arise in the cerebellum, and in addition to the general “classic” medulloblastoma designation (Figure 1A), four variants are recognized in the current WHO scheme. Desmoplastic/Nodular Medulloblastomas are distinctive due to their round, oval or elongated “pale islands” comprised of more differentiated cells that most often resemble neurocytes or small mature neurons with variable neuropil formation (Figure 1B). These nodules are surrounded by more primitive and proliferative tumor cells enmeshed in a reticulin-rich stroma. When nodules become extensive or confluent, with minimal proliferative internodular tissue, the tumors are known as Medulloblastoma with Extensive Nodularity, a rare variant seen primarily in infants with favorable prognosis (Figure 1C). Large Cell Medulloblastomas are distinctive due to their generally discohesive growth, abundant apoptotic and mitotic figures, large nuclei, vesicular chromatin, and prominent nucleoli. Anaplastic Medulloblastoma share these increases in nuclear size, proliferation and apoptosis, but are comprised of more pleomorphic, angular or molded nuclei with dark chromatin, rather than rounded, discohesive cells. Since these latter two variants can coexist in a single specimen (Figure 1D), and there are many shared clinicopathologic features, the combined term “large cell/anaplastic medulloblastoma” is often used. Myogenic or melanocytic differentiation is also occasionally encountered in medulloblastomas, but these are no longer regarded as distinct entities in the current WHO classification since they can be encountered as secondary patterns within several of the major variants already described (49).
The remaining embryonal tumors include the Atypical Teratoid/Rhabdoid Tumor (AT/RT) and a rather mixed group of CNS Primitive Neuroectodermal Tumors (CNS PNET). AT/RTs are characterized microscopically by a variable presence of small cell/embryonal, mesenchymal and epithelial features, with the embryonal component commonly predominating (Figure 1E)(49, 64). Rhabdoid cells are often, but not always, encountered, such that a high index of suspicion is required, particularly in infants. The CNS PNET group encompasses a heterogeneous collection of non-cerebellar medulloblastoma-like lesions united by the presence of undifferentiated or poorly differentiated neuroepithelial cells with varying capacity for differentiation along neuronal, glial or other lineages. The 2007 WHO classification lists ICD codes for CNS PNET NOS (Figure 1F), CNS Neuroblastoma, CNS Ganglioneuroblastoma, Medulloepithelioma, and Ependymoblastoma (49), but the group remains in a state of flux. For example, it has been suggested that ependymoblastoma does not represent a distinct lesion (42), and that CNS PNET containing true rosettes and extensive neuronal differentiation (Figure 1G, H) are in fact a newer distinct entity most frequently referred to as “Embryonal Tumor With Abundant Neuropil and True Rosettes” (ETANTR). The molecular and pathological data related to these issues will be discussed in more detail below.
Initial attempts to stratify embryonal CNS neoplasms based on molecular features used what would now be considered basic techniques: standard metaphase karyotype analysis, immunohistochemistry, fluorescence in situ hybridization (FISH), single gene sequencing, Southern blotting, etc. Nevertheless, they greatly assisted in determining some basic molecular similarities and differences between the various tumor subtypes. More recently, advanced techniques such as array-based comparative genomic hybridization (CGH) and expression profiling of mRNA and miRNA, often used in an integrated fashion and coupled to sophisticated biostatistical analyses, have allowed us to begin to further tease apart various tumor subgroups.
Early attempts to subclassify medulloblastoma yielded a number of interesting observations, and began to associate specific molecular groups with histopathological subtype. Medulloblastoma arising in the context of Turcot and Gorlin syndromes suggested roles for Wnt and Hedgehog signaling respectively (37, 62, 68, 74). Sporadic medulloblastomas associated with Wnt activation were discovered by directed sequencing of the pathway members, CTNNB1 (beta-catenin), AXIN and APC (4, 15, 39, 76, 77), and by immunohistochemical analysis of the nuclear translocation of beta-catenin which marks its activation (22). Tumors with Wnt activation tend to be of the “classic” histological subtype, and often also show loss of chromosome 6 (72). Another subset of sporadic tumors with mutations activating the Hedgehog pathway have also been identified, and these are often, though not always, of the nodular/desmoplastic subtype (60, 62, 74). Finally, amplicons containing either the c-myc (MYC) or N-myc (MYCN) loci were also identified (2, 3, 38, 66), and were frequently associated with large cell/anaplastic medulloblastoma subtypes (2, 21, 31). It was also shown that loss of chromosome 17p and gain of 17q, frequently in the form of an isochromosome 17q, was present in a third or more of medulloblastoma (8, 34, 66).
Gene expression profiling using oligonucleotide microarrays, as well as array-based CGH, have helped to confirm and extend these early studies. First generation expression analyses by Pomeroy and colleagues identified distinct signatures for supratentorial CNS PNET, AT/RT, and medulloblastoma (61). These studies also found unique profiles for classic and nodular/desmoplastic medulloblastoma, and confirmed the linkage of the latter variant to Hedgehog pathway activation. Chromosomal changes identified using array CGH have also been used to partition medulloblastomas via unsupervised clustering. Mendrzyk and colleagues divided 47 cases into two groups, and found that large cell/anaplastic histology, chromosome 17q gain, and amplification of MYC, MYCN and/or CDK6 were significantly associated with what they called “Group II” (52).
Four gene expression studies have begun to more accurately define the medulloblastoma landscape in terms of molecular variants. The first, by Thompson and colleagues, used Affymetrix U133av2 arrays to profile 46 medulloblastoma samples (72). Unsupervised clustering partitioned the cases into five subgroups, which they designated A to E (Figure 2). The B and D groups were found to have expression signatures characteristic of Wnt and Hedgehog pathway activation, and had predominantly classic and nodular/desmoplastic histology, respectively. Anaplasia was most common in the “E” group. Activating mutations in CTNNB1 were identified in tumors of the Wnt group, and were associated with monosomy 6, while most of the Hedgehog group had mutations in pathway members PTCH1 or SUFU (72).
A subsequent study by Kool et al. examined 62 medulloblastoma, and also divided them into five subsets (44). Array CGH studies were performed on 52 of the cases, allowing correlation with chromosomal alterations. A “Wnt” group (A) comprised 15% of the cases and was associated with monosomy 6, CTNNB1 mutations and good outcome (Figure 2). An additional 24% of cases formed group B and showed signs of Hedgehog activation; many of these tumors had loss of chromosome 9q and/or mutations in PTCH1. Groups C and D showed signs of neuronal differentiation, as well as expression of both glutamate and GABA receptor families. Chromosome 17 changes were also concentrated in these tumors. In groups D and E, the expression of retinal genes and other signs of photoreceptor differentiation were notable.
Northcott and colleagues analyzed gene expression and chromosomal alterations in 103 primary medulloblastoma (54). Their unsupervised hierarchical clustering yielded four classes of tumors, including the Wnt and Hedgehog associated clusters identified in prior studies, as well as groups they named C and D. The Hedgehog group showed YAP1 gene amplification, consistent with recent reports linking this pathway to Hippo/YAP1 signaling (27). The C and D groups were both associated with expression of axon guidance genes and the oncogenes OTX2 and FOXG1B, as well as isochromosome 17q (Figure 2). Others have also identified OTX2 overexpression and/or gene amplification in a defined subset of medulloblastoma outside the Wnt and Hedgehog groups (1, 12, 16, 17). The Northcott Group C was further associated with phototransduction, glutamate signaling, increased MYC expression and MYC gene amplification. Pathway analysis of the Group D signature revealed signs of neuronal semaphorin signaling, G-protein coupled receptors, and cyclic AMP signaling. A unique feature of this study was the development of single gene class predictors based on immunohistochemical analysis. The Wnt, Hedgehog, C and D groups could be uniquely recognized on tissue microarrays by antibodies to DKK1, SFRP1, NPR3 and KCNA1, respectively. This approach will need to be further validated, but shows great promise as a means to classify cases in routine pathological practice.
The fourth study, by Cho and colleagues, involved the largest number of cases and included both mRNA and miRNA expression profiling as well as SNP array analysis of chromosomal changes (13). Unsupervised cluster analysis based on the mRNA profiles of 194 tumors divided the medulloblastomas into 6 subsets, including Wnt (c6) and Hedgehog (c3) groups similar to those identified in other studies (Figure 1). Amplification of the GLI2 locus was noted in the Hedgehog group. Group c1 was characterized by high MYC expression and copy number gains, as well as increased expression of photoreceptor associated transcripts and GABRA5. Many of these features were prominent in group c5 as well, but the MYC-related signature was felt to be significantly more pronounced in c1. OTX2 amplifications, in contrast, were concentrated in group c5. Neuronal differentiation markers, including high levels of the glutamatergic markers GRM1 and GRM8, were found in both Groups c2 and c4. Interestingly group c4 also showed some degree of photoreceptor differentiation as well as GABRA5 expression and a MYC signature. Immunostaining for the photoreceptor/GABAergic and neuronal/glutamatergic markers CRX and GRM8 suggested that this mixed expression phenotype was due to distinct subpopulations of cells in tumors of this group. Amplification of the MYCN locus and low level DNA copy gains including MYC were relatively common in group c4, while many c2 tumors showed few chromosomal alterations, and some had only i17q. miRNA profiles were also analyzed, and the miR-183~96~182 cluster was associated with groups expressing photoreceptor genes (c1 and c5). Additional associations included increased miR-592 in groups c2 and c4, increased miR-199b, miR-378, miR-28, and miR-625 as well as decreased miR-135a/b, miR-338, miR-124, miR-138, in group c3, and increased miR-23~27~24 cluster in group c6.
The demographics of these newly discovered molecular subgroups are also intriguing. In the Kool et. al. study, patients below 3 years of age fell into the B (Hedgehog), D and E groups (44). In the paper by Northcott and colleagues, the Hedgehog group could be partitioned into tumors arising in infants less than 3 years of age, and adults over 16, while group C tumors with high MYC all occurred in children 10 year old or less (54). Cho and colleagues also found a bimodal age distribution for the Hedgehog group, which included the highest percentage of patients under 3 years of age and all of the older adults (13). Interestingly, while medulloblastoma overall are more common in males, in both of these studies the Wnt and Hedgehog cohorts were predominantly female (13, 54).
Other researchers have also recently examined the relationship between patient age and medulloblastoma molecular changes. Korshunov and colleagues examined 146 adult medulloblastoma specimens using array CGH and/or FISH (47). They found that CDK6 amplification, 10q loss, and 17q gain were present and prognostically significant in adult cases, while MYC gene changes were less frequent than in children.
While AT/RTs were initially recognized as neoplasms with a distinctive microscopic appearance, they were soon found to feature loss of chromosome 22 as a common molecular change (64). It was subsequently shown that mutation or deletion of the SMARCB1 gene (also known as INI1 or hSNF5) at 22q11 is a defining molecular change in this tumor (7, 73). SMARCB1 encodes a member of the SWI/SNF ATP-dependant remodeling complex (reviewed in (70)). Families with an inherited disposition to AT/RT due to genomic SMARCB1 changes have been described (40), and in tumor registries as many as 10 of 41 patients with CNS AT/RT have been found to harbor germline alterations of the locus (45), consistent with the hereditary form known as rhabdoid tumor predisposition syndrome. In such cases, family members should be tested for mutations in addition to the patient, in order to assess genetic risks. Recently, inactivation by germline mutation and loss of heterozygosity of SMARCA4 was documented in two sisters with rhabdoid tumors lacking SMARCB1 alterations, indicating that more than one member of the SWI/SNF complex can participate in formation of these lesions (67).
Immunohistochemical assays of SMARCB1/INI1 protein are now in widespread clinical diagnostic use, and loss of protein expression is found in almost all cases (41). The utility of tumor-specific molecular markers such as INI1 protein loss is exemplified by studies in which investigators were able to reclassify PNET or choroid plexus carcinomas as AT/RT (43), or to identify INI1-immunonegative medulloblastoma and PNET lacking rhabdoid features which had clinical characteristics consistent with AT/RT (36). Based on mRNA expression profiling comparisons, Claudin 6 has recently been suggested as an additional marker which can distinguish AT/RT from other brain tumors (9). This novel marker has the potential to be particularly useful, since like INI1 it can be assessed using immunohistochemistry.
Molecular studies have played a key role in resolving disputes on the distinction between CNS PNET and medulloblastoma, which otherwise can have similar histological appearances. However, the former have significantly fewer chromosome 17 alterations than medulloblastoma, and the presence of isochromosome 17q is very rare (49, 51, 58). In contrast, loss of the CDKN2A locus is significantly more common in CNS PNET (51, 58). The analysis of broad mRNA expression profiles (61), as well as smaller groups of transcripts associated with neural differentiation, also supports the notion that these are distinct entities (59).
More recent molecular studies have begun to subdivide CNS PNET into different variants. Analysis of a single ETANTR at the German Cancer Research Center identified amplification of the 19q13.42 locus including a cluster encoding miR-372 and miR-373, and these microRNAs were overexpressed as compared to normal cerebellum (57). In a larger study, Li and colleagues examined 45 CNS PNET, and reported that amplification at 19q13.41 is present in approximately a quarter of tumors (48). The authors demonstrated that these gains centered on a 1 megabase region encoding the miRNA clusters C19MC and miR-371~373. Overexpression of miR-517c and 520g from the C19MC cluster was prominent in PNET, while in this study miR-371 and miR-373 levels were modest. When miR-517c and 520g were introduced into human neural stem cells, they promoted oncogenic cellular properties (48).
In both of these initial studies, the amplicon at 19q13.41–42 was strongly associated with PNETs showing a particular histological appearance. The case reported by Pfister and colleagues was diagnosed as an ETANTR, a variant which is defined by the presence of abundant neuropil and multilayered rosettes with well-defined lumens (30, 57). Li and colleagues separate their cases into “classic” PNET and those with “variant” histology, the latter distinguished by “ependymal or ependymoblastic differentiation with distinct rosette structures” (48). The image they use to illustrate this variant group appears to be consistent with the diagnosis of ETANTR. Of the 45 cases they examined, 8 of the 10 with variant histology had the amplified 19q13.42 locus, compared to 3 of 35 classic PNETs, a statistically significant difference. Interestingly, their “variant” PNETs arose in younger patients and were particularly aggressive (48), both features previously associated with ETANTR (18, 30). However, Li et al felt that while these associations were significant, the molecular changes were not specific for a single histological appearance.
It is not yet clear to what degree the 19q13.41–42 amplicon and associated distinctive histopathological and clinical factors define a unique neoplastic entity. In an attempt to further address this issue, Korshunov and colleagues used FISH to examine a number of tumors with “ependymoblastic” rosettes, and were able to identify 19q13.42 amplicons in 93% of the cases, including 19 ETANTRs and 18 ependymoblastomas (46). All but two of these tumors occurred in children three years of age or less, and clinical outcomes were worse than in other CNS PNET. They proposed that CNS PNET with ependymoblastic rosettes may comprise a single diagnostic entity, and note that they have analyzed over 300 pediatric brain tumors of various other types without detecting a 19q13.42 amplification (46).
It remains to be determined by what name these lesions will be known – ETANTR, ependymoblastoma, a combination of the two, or some new term. Judkins and Ellison recently reviewed the history of the ependymoblastoma, as well as 14 cases from their institutional files (42). They concluded that the majority of tumors with ependymoblastic rosettes belong to the distinctive ETANTR group, and suggest that ependymoblastoma is not a discrete entity. Adopting their point of view would have the benefit of simplification via subtraction, and it does seem that most (if not all) PNET with 19q13 gains have similar, if not identical, genetic, clinical and histological features. They also propose changing the name of these lesions to ETANER, with “ependymoblastic” replacing “true” as a descriptor for the rosettes. However, because cells in “ependymoblastic” rosettes do not seem to be primarily ependymal (or even glial) precursors, this term seems a bit of a misnomer. The name “Embryonal tumor with multilayered rosettes” (ETMR) has also been recently suggested (56).
Significant progress has been made in our ability to separate medulloblastoma patients into subgroups with better or worse prospects for long term survival. The quest for prognostic markers has been driven in part by a desire to reduce the amount of treatment given to patients with intrinsically less aggressive tumors, as radiation and the other therapies used in pediatric brain tumor patients are associated with significant long-term neurocognitive side effects (11, 29). Early studies largely focused on using single factors to predict outcome, but more recently the large datasets described above have allowed increasingly integrated approaches. Nevertheless, many simple markers have stood “the test of time”, and they can be more straightforward to implement in clinical practice.
The list of single molecular factors associated with improved outcomes in medulloblastoma patients is relatively short. Prominent examples associated with longer survival include increased TrkC mRNA expression (35, 65, 69) and nuclear translocation of beta-catenin indicating Wnt pathway activation (14, 24–26). Markers prognostic of worse outcomes include MYC gene amplification and/or overexpression (2, 20, 24, 38), chromosome 17 aberrations including 17p loss and i17q formation (5, 32, 55), and strong p53 immunoreactivity or TP53 mutation (19, 71, 75). A potential strategy for stratifying medulloblastoma patients based on age, Wnt and Hedgehog pathway status, histopathological subtype and MYC gene amplification has recently been proposed by David Ellison (23).
Several early microarray-based studies also sought to identify molecular prognostic markers in medulloblastoma. Comparison of expression profiles from metastatic and non-metastatic tumors identified the PDGFRA and RAS/MAPK pathways as being associated with aggressive disease (50). Some have questioned this finding (33), although other groups have subsequently documented amplification of PDGFRA in medulloblastoma and an association with poor outcomes (10, 51). In another seminal early study, Pomeroy and colleagues used a supervised learning approach to analyze expression profiles and associated clinical outcomes in 60 medulloblastomas, and developed an 8-gene predictor set that included a number of loci encoding ribosomal proteins (61). Neben and colleagues identified increased levels of MYC, STK15 and 52 other genes as being associated with shorter survival (53). The same group later used array CGH to define a high-risk group of patients, and documented amplification of CDK6, MYC, and MYCN as well as chromosome 17 alterations in this group (52).
The studies summarized in Figure 2 also attempted to identify molecular prognostic factors in medulloblastoma, and have in general, validated the markers described above. In the paper by Thompson and colleagues, metastatic tumors and large cell/anaplastic histology were most frequent in group E, but survival data were not reported (72). Metastatic disease at diagnosis was associated with groups C, D and E in the Kool et al. study (44). Their group E showed the tightest association with metastasis, and increased expression of ribosomal protein encoding genes similar to those previously identified by Pomeroy and colleagues in aggressive medulloblastoma were noted (44, 61). In the paper by Northcott and colleagues it was group C with high MYC expression which was significantly associated with decreased progression-free and overall survival (54). As noted above, they were also able to define this group based on NPR3 immunoreactivity, and in an analysis of 286 medulloblastoma classified by immunohistochemistry the NPR3-positive group also had the worst prognosis. The molecular group “c1” characterized by Cho and colleagues as having the most MYC gene amplification and transcriptional enrichment of photoreceptor pathways was tightly associated with lower event-free and overall survival (13). Upregulation of the miR-183~96~182 cluster was also found in these tumors. Patients in the c3 category defined by Hedgehog activity also experienced somewhat worse outcomes in this particular study, which is at odds with some earlier reports that desmoplastic/nodular tumors are associated with longer survival (49).
Additional work will be needed to validate the broad microarray-based prognostic classification described above, as specific discriminating gene sets can vary significantly between the four research groups. In their discussion Cho and colleagues note that the c1 and c5 groups could potentially be combined into a single class, but argue that this simpler classification would result in overtreatment of the c5 tumors. This contrasts to a degree with the other papers in which clustering yielded fewer groups, but this could be due to the size of the cohorts analyzed in the various studies. It will also be important to determine if these groupings can help predict response to a given type of therapy. For example, it would seem likely that only tumors with a “high Hedgehog” signal would respond to agents targeting that pathway, but little data directly addressing this issue exists, and it is possible that even the lower levels of Hedgehog pathway activity seen in tumors of other classes are functionally significant.
AT/RTs as a group have a very poor outcome, although recent studies using high-dose chemotherapy suggest that at least a subset of cases can benefit from intensive regimens (28). The emergence of patients with longer survival will allow the search for molecular prognostic and predictive markers to begin in earnest. In one recent study, germline SMARCB1 mutations were associated with multicentric CNS disease and shorter survival (Kordes et al, 2010).
Nuclear beta-catenin immunostaining was noted in 36% of the 33 CNS PNET examined in a recent study, but in these tumors the association of Wnt activation with improved outcome was not as clear as in medulloblastoma (63). In another study, polysomies of chromosomes 2 and 8 as well as large cell/anaplastic changes were significantly linked to worse outcomes in CNS PNET (6). Interestingly, amplification of the MYCN locus was identified in 50% of the 22 pediatric cases examined, and cases with either MYC or MYCN copy number gain showed a trend towards worse outcome (p =0.11). Finally, as noted above, tumors with chromosome 19q13.41–42 amplifications and/or ETANTR-like histology are associated with very poor clinical outcomes (30, 46, 48).
The analysis of relatively large tumor cohorts using increasingly sophisticated microarray platforms has generated at least four molecular medulloblastoma classification schemes. As highlighted in Figure 1, many similarities exist between these, but the alignments are not perfect and specific gene sets used as classifiers vary between studies. This may be due to differences in composition of the various patient cohorts, as rare genetic events can be differentially represented due to chance. For example, the frozen tumor material studied by Northcott and colleagues included only three medulloblastoma with MYC gene amplification, and these were scattered over all but one of their four categories. However, when they examined several hundred additional paraffin embedded cases using immunohistochemistry and FISH, MYC amplicons were tightly restricted to the aggressive group C tumors. A similar association between MYC amplification and aggressive tumor subtype was found by Cho et al (13). This highlights the need to examine relatively large numbers of cases, particularly when studying rare events, before drawing firm conclusions.
Hopefully it will be possible to resolve the differences between the various studies and generate a workable consensus molecular classification that stratifies patients for optimized use of current “standard” therapies, while also incorporating biomarkers potentially predictive of response to newer “targeted” therapies. Achieving consensus on such issues will be complicated by the ongoing need to evaluate a deluge of new data generated by emerging technologies such as high-throughput analysis of epigenetic chromatin marks, miRNA expression profiling and next generation sequencing. The use of relatively simple tests as surrogates for more complex molecular data, such as the limited panel of immunohistochemical markers used by Northcott and colleagues (54), is one approach which may make practical translation more feasible.
For CNS PNET, we must decide how to incorporate 19q13.41–42 amplification into our diagnostic armamentarium. The central question is whether DNA gains at this locus define a unique CNS PNET variant. Tumors with the 19q13.41–42 amplicon generally show both the prominent neuropil and multilayered rosettes of an ETANTR, or at a minimum the rosettes, which can also be seen in ependymoblastoma (46, 48). However, a few published cases lack either the characteristic molecular or pathological features, raising questions about the specificity of the molecular change (48). Comparisons to AT/RT may be helpful in making sense of these outliers. For example, as INI1 loss became easier to analyze, a number of tumors previously considered to be PNET, choroid plexus carcinoma or medulloblastoma were incorporated into the AT/RT diagnostic class. These cases lacked some or all of the “teratoid” (variably mixed epithelioid and spindled components) and rhabdoid features of “classic” AT/RT, due to either limited sampling or true heterogeneity within the AT/RT spectrum. Sampling and heterogeneity may also explain the spectrum of molecular and histological features in CNS PNET with rosettes and 19q13.41–42 amplification, while the association between young age, very poor clinical outcome and distinctive histology in most cases argues strongly that this represents a unique tumor entity. What to call it remains unclear. The recently proposed “embryonal tumor with multilayered rosettes” (ETMR) is pleasing in its relative brevity and inclusive simplicity, although those who follow the field may suffer from “acronym whiplash” due to the frequent changes in nomenclature.
The authors’ embryonal brain tumor research is supported by R01 NS055089.