Pilocytic astrocytoma (PA) is the most common pediatric brain tumor and the most frequent childhood glioma. The most classic histologic appearance of PA is that of a biphasic astrocytic neoplasm with alternating compact and loose arrangements of cells with elongated “piloid” processes admixed with Rosenthal fibers and eosinophilic granular bodies. Unfortunately, PA is well known for its diagnostically problematic ability to mimic a wide variety of other neoplasms, including high grade infiltrative gliomas and others with significantly different biologic potential, prognosis, and applicable treatment options (Fig. 16.4).

Though most are sporadic, there is a well-established association of PA and Neurofibromatosis Type 1 (NF1) with germline mutations of the NF1 gene [59]. The resultant loss of the NF1 gene product neurofibromin leads to activation of the mitogen activated protein kinase (MAPK) signaling pathway. More recently, a number of groups have identified a low-level gain at chromosome 7q34 resulting in a tandem duplication of the BRAF gene in most sporadic PAs [60, 61]. This duplication results in a KIAA1549-BRAF fusion that is nearly specific to PAs and absent or exceedingly rare in other low grade or high grade infiltrative astrocytomas [62–65]. KIAA1549-BRAF fusion is more frequent in PAs arising in the cerebellum (up to 80 %) compared to those arising in the brainstem, and hypothalamic/optic pathway (approximately 60 %) or the cerebral hemispheres (14–19 %) [62, 65–67]. KIAA1549-BRAF fusion detection is particularly helpful in differentiating those uncommon cerebellar PAs with a diffuse pattern of growth (so-called “diffuse variant”) from WHO grade II diffuse astrocytomas in this region [68]. BRAF fusion detection also may be helpful in distinguishing PAs with atypical radiographic and/or histologic features from pleomorphic xanthoastrocytoma (PXA) or malignant glioma [69]. KIAA1549-BRAF fusion therefore represents a diagnostically useful molecular marker for differentiating PA from potential diagnostic mimics [70, 71]. The prognostic or therapeutic predictive utility of KIAA1549-BRAF fusion is not yet established, though a number of clinical trials for pediatric PAs utilizing drugs targeting the MAPK pathway are ongoing [67, 72]. It should be noted that in contrast to pediatric PAs, KIAA1549-BRAF fusion is less commonly detected in PAs arising in the adult population [73] and has been encountered in occasional adult diffuse gliomas, particularly oligodendrogliomas, often together with IDH mutations and 1p/19q codeletions [74].

A number of approaches have been employed in the clinical testing of BRAF fusions. RT-PCR may be used to directly amplify the fusion transcript. Though this procedure is quite specific, assessment is limited to predetermined regions and requires sufficient RNA material from tumor samples [62, 63, 72]. Pyrosequencing methods for detecting BRAF copy number gains compared favorably with RT-PCR methodology and can provide a sensitive, cost-effective alternative to RT-PCR [75]. Fresh or FFPE tissue samples may be used for either of these techniques; however, FISH is at present the best method for testing of the KIAA1549-BRAF fusion in FFPE. Detection of BRAF duplication can be accomplished by determining the ratio of signals of a fluorescent probe that includes the centromeric end of BRAF versus a ploidy control probe (i.e., CEP7); a ratio of test: control probe ≥ 1.15 and > 20 % of cells with relative BRAF gain would be considered positive. This method is considered indirect evidence of the KIAA1549-BRAF fusion and would typically be visualized as nuclei containing 2 CEP7 control signals, 2 BRAF signals, and a third smaller BRAF signal adjacent to one of the larger BRAF signals [67]. An additional FISH method employs separately labeled probes targeting BRAF and KIAA1549 genes, typically tagged red and green respectively. The presence of nuclei with one fused red/green signal and a normal pair of red and green signals is considered positive for a KIAA1549-BRAF fusion [76–78]. Both FISH methods allow for direct visualization of BRAF alterations in tumor cells and can be used on small tumor samples. The indirect FISH technique is more difficult to interpret, particularly in FFPE tissue samples in which truncation of small signals could result in false negatives and more interobserver variability.

In addition to KIAA1549-BRAF fusions, activating mutations of BRAF have been detected in a variety of pediatric glial tumors. The most common alteration is the V600E mutation (BRAF ^V600E) caused by a point mutation at codon 600 which results in a substitution of glutamic acid for valine. This mutation has been described in a wide variety of tumors outside of the CNS and is found most frequently in melanoma, lung, and colon cancers [79]. Although BRAF ^V600E mutations may be found in a small subset of PAs, particularly in extra-cerebellar examples, they are exceedingly more common in other pediatric low grade gliomas [80]. Up to 75 % of pleomorphic xanthoastrocytomas (PXA) have a BRAF ^V600E mutation, including those with anaplastic features [80–83]. Nearly all temporal lobe PXAs harbor this mutation compared with 50 % of non-temporal examples [84]. Similarly, this alteration is present in up to 40 % of gangliogliomas and subependymal astrocytomas as well as a significant proportion of dysembryoplastic neuroepithelial tumors (DNET) [81, 82].

Despite its significant histologic pleomorphism, PXA is less biologically aggressive than pleomorphic infiltrative gliomas, thus earning it a WHO grade II designation. From a clinical perspective, it is imperative to differentiate PXA from other gliomas as treatment and prognosis are significantly different. Demonstration of the BRAF ^V600E mutation serves as a useful diagnostic marker in this setting. Although BRAF ^V600E mutation may be encountered in either PA or PXA, detection of KIAA1549-BRAF fusion would confirm the diagnosis of PA as this alteration is not present in PXA. When glioblastoma is in the differential diagnosis, the presence of a BRAF ^V600E mutation would strongly favor PXA as this mutation is rare in GBM [80, 81, 85, 86]. Combined BRAF ^V600E and cyclin-dependent kinase inhibitor 2A (CDKN2A also known as P16) loss may be detected in up to 50 % of PXAs, but this signature is not observed in rhabdoid or giant cell glioblastomas, both potential histologic mimics [84, 86]. Unfortunately, BRAF ^V600E is not useful in separating PXA from epithelioid GBM as 50 % of the later may have this mutation [86]. Additionally, recent studies have indicated that pediatric GBMs have a slightly higher rate of BRAF ^V600E compared to GBMs in adults (12 versus 8 %) [87–89] (Fig. 16.5).

The prognostic utility of BRAF ^V600E mutational analysis is uncertain [82]. In one study, BRAF ^V600E PXAs had a significantly longer overall survival compared with those lacking the mutation [83]. In contradistinction, others have found BRAF ^V600E mutation to be associated with an increased risk of malignant progression in pediatric low grade gliomas [89, 90]. Additionally, there is some evidence that suggests BRAF ^V600E may represent an important biomarker of targeted therapy response. Recent reports have shown that treatment of BRAF-mutated anaplastic PXA and recurrent GBM with vemurafenib, a BRAF inhibitor, may be effective [91, 92]. However, additional prospective studies/clinical trials are necessary to determine whether BRAF mutational analysis will serve a key clinical prognostic and/or predictive role in the workup of these glial tumors.

The gold standard for BRAF mutation analysis is direct sequencing; however, this technique is complex, labor-intensive, and requires specialized equipment not universally available in all clinical laboratories. Development of BRAFV600E-mutant specific monoclonal antibodies has provided a cost-effective alternative to sequencing techniques for routine clinical testing [93, 94]. The VE1 clone of the BRAFV600E-mutant specific antibody provides strong cytoplasmic staining with 100 % specificity and sensitivity versus standard molecular methods for detecting BRAF^V600E mutation [93]. This IHC method is able to accurately demonstrate BRAF^V600E in a variety of tumor types including gliomas and is amenable to very small biopsy samples, which might otherwise be inadequate for sequencing analysis. Interpretation is straightforward, though careful validation is important and care should be taken to not over interpret weak nonspecific staining.

Although glioblastoma (GBM) represents the most common primary malignant CNS tumor in adults, it is uncommon in children. Pediatric GBMs share a similar histologic appearance and aggressive biologic behavior as their adult counterparts; however, their anatomic distribution is notably different. In addition to the typical cerebral locations of adult GBMs, pediatric high grade gliomas may also arise within deep grey structures as well as the brainstem in the form of diffuse intrinsic pontine glioma (DIPG). Pediatric GBMs harbor a significantly different set of genetic alterations than those of adult tumors with minimal overlap. Specifically, mutations in the H3 histone family 3A (H3F3A)–α-thalassemia/mental retardation syndrome X-linked (ATRX)–death-domain associated protein (DAXX) chromatin remodeling pathway are present in 44 % of pediatric GBM [95, 96]. The H3 histone family 3A (H3F3A) gene mutation leads to an amino acid substitution at either K27 or G34. These mutations are present in up to 60 % of DIPGs and 33 % of non-brainstem pediatric GBMs [95]. Similarly, mutations in histone cluster 1, H3b (HIST1H3B), an isoform of H3F3A, have also been documented in nearly 20 % of DIPGs [97]. Thus, approximately 80 % of DIPGs will harbor some form of H3-related mutation [97]. Interestingly, the K27 and G34 mutations correspond to distinct pediatric GBM subgroups, with the majority of K27 mutations found in DIPGs and deep midline tumors and a mix of K27 and G34 mutations found in non-brainstem pediatric GBMs [96–98]. Mutations involving ATRX (α-thalassemia/mental retardation syndrome X-linked) and DAXX (death-domain associated protein), both gene products involved in the chromatin remodeling complex requisite for H3F3A incorporation at telomeres, have been identified in 31 % of pediatric GBMs as well as in all tumors harboring H3F3A G34 mutations [95].

From a diagnostic standpoint, H3F3A mutations have been found to be specific to GBMs arising in children and young adults and have not been found in other CNS tumors. A variety of other pediatric CNS tumors may pose a significant diagnostic challenge given the multitude of histologic appearances noted in pediatric GBMs. These include PXA, PA, medulloblastoma, primitive neuroectodermal tumor (PNET) and AT/RT. Demonstration of H3F3A (or ATRX or DAXX) mutations may assist in securing an accurate tumor classification and allow for appropriate treatment. Gene sequencing techniques are the gold standard of testing for these mutations. However, IHC is much more cost-effective and more suitable for clinical detection. Unfortunately, a commercially available mutant specific antibody has yet to be developed. A single large multi-institutional study utilized antibodies against oligodendrocyte lineage transcription factor 2 (OLIG2), forkhead box G1 (FOXG1), and IDH1^R132H to classify these tumors into specific mutation defined GBM subgroups. In this study, there was excellent correlation between sequencing-derived mutation status and the IHC subgroups with H3F3A K27 corresponding to OLIG2+/FOXG1−, and H3F3A G34 corresponding to OLIG2−/FOXG1+ staining [96]. Given an observed difference in overall survival for these three mutation-defined subgroups (K27 associated with shortened survival, versus prolonged survival with G34 and particularly with IDH1 ^R132H) [98, 99], assessment for these IHC markers may well become routine practice in the workup of pediatric GBM.

Medulloblastoma represents the most common childhood malignant CNS tumor. Traditionally their classification has been based entirely on histopathologic features with classic, large cell, anaplastic, desmoplastic nodular, and extensively nodular as the accepted recognized variants [100]. Unfortunately, this classification alone does not provide accurate prognostic information as there is still significant variability of outcomes in these histologic groups. Genome-wide molecular characterization studies have revealed four molecular subgroups of medulloblastomas each with their own molecular signature as well as prognostic and treatment implications [78, 106].

The WNT group is the best characterized of the subgroups, though it accounts for the smallest proportion of medulloblastomas overall (approximately 10 %) [107]. Germline mutations of adenomatous polyposis coli (APC), a wingless-type MMTV integration site family (WNT) pathway inhibitor, predispose patients with Turcot syndrome to medulloblastomas. Conversely, sporadic medulloblastomas of this group bear catenin (cadherin-associated protein) beta 1 (CTNNB1) somatic mutations. WNT medulloblastomas are rare in infants, have classic histology with rare anaplastic examples, and are characterized by nuclear β-catenin staining by immunohistochemistry. Monosomy 6 and dickkopf WNT signaling pathway inhibitor 1 (DKK1) positive immunostaining are also typically present. Of the four subgroups, WNT medulloblastomas have the best prognosis with uncommon metastasis and prolonged survival [106–108]. These patients may also be more amendable to treatment de-escalation, though this option needs further investigation.

Medulloblastomas in the Sonic Hedgehog (SHH) signaling pathway subgroup have demonstrable mutations of patched 1 (PTCH1), suppressor of fused homolog (SUFU), smoothened frizzled class receptor (SMO), or amplification of GLI family zinc finger 1 or 2 (GLI1/GLI2) (2120). Germline PTCH1 mutations are also associated with the development of medulloblastomas in the context of Gorlin syndrome. SHH medulloblastomas are most common in infants and adults, but are rare in children. The vast majority of nodular desmoplastic medulloblastomas belong to this group, although approximately half of the SHH subgroup are of some other histology. SHH medulloblastomas can be accurately identified by immunopositivity for secreted fizzles-related protein (SFRP1) or GRB2-associated binding protein 1 (GAB1), as well as deletion of chromosome 9p detectable by FISH [106–108]. Importantly, a recent large multinational study determined that SHH medulloblastomas can be further stratified into three prognostically significant risk groups: GLI2 amplification alone or 14q loss combined with M+ (positive for metastatic disease) status correlates with a high risk group, whereas those lacking all of these markers constitute a low risk group having a prognosis similar to the WNT group. The remainder fall into a standard risk group with intermediate prognosis similar to group 4 [109] (Fig. 16.7).

The molecular alterations and pathways involved in subgroup 3 are less well characterized than the WNT or SHH subgroups. Group 3 medulloblastomas are more common in male infants and children and are exceedingly rare in adults. Although many have a classic histology, the majority of anaplastic/large cell medulloblastomas cluster in this subgroup often with metastatic disease at presentation. Originally classified based upon a specific transcriptional profile, accurate categorization can be accomplished through demonstration of immunopositivity for natriuretic peptide receptor C (NPR3) [106–108]. Like SHH group tumors, Group 3 medulloblastomas can be further subclassified into prognostically significant risk groups. Tumors with detectable v-myc avian myelocytomatosis viral oncogene homolog (MYC) amplification, Iso17q [78], or evidence of metastatic disease at presentation are considered high risk, with frequent tumor recurrence and short survival. In contradistinction, tumors lacking any of those three features are considered standard risk with prognoses similar to group 4 tumors [106, 109, 110] (Fig. 16.8).

Like Group 3 tumors, Group 4 medulloblastomas were originally defined by their unique transcription profile. Most are of classic histology arising preferentially in children with a male predominance. Detection of potassium voltage-gated channel, shaker-related subfamily member 1 (KCNA1) by immunohistochemistry can reliably classify Group 4 medulloblastomas [106–108]. Group 4 tumors can be further stratified into prognostically significant risk groups based on additional molecular testing and clinical parameters. The medulloblastomas of this group with demonstrable whole chromosome 11 loss (Chr11 loss) or chromosome 17 gain (Chr17 gain) fall into a low risk stratum with excellent prognosis and prolonged survival even in the presence of aggressive histology (anaplastic/large cell) or M+ disease. Conversely, those Group 4 patients presenting with metastases but without these chromosomal abnormalities would otherwise be considered as high risk. The remainder fall into a standard risk group with intermediate prognosis similar to SHH [109].

Determination of subgroup affiliation together with selective cytogenetic workup for risk stratification will provide for accurate determination of individual patient prognosis and risk-directed treatment options useful to neuro-oncologists. Accurate medulloblastoma subgroup determination can be made via immunohistochemistry, with positivity for Beta-catenin and DKK1 in the WNT group, SFRP1 in the SHH group, NPR3 in Group 3, and KCNA1 in the Group 4 tumors [107, 108]. Further risk-stratification in SHH, Group 3, and Group 4 tumors can be accomplished by employing interphase FISH assays using the following cytogenetic biomarkers: GLI2 and 14q in SHH group tumors, MYC, 17p, and 17q in Group 3 tumors, and chromosome 11, 17p, and 17q in Group 4 tumors [109]. 17p and 17q FISH probes are paired together for analysis of SHH and Group 3 tumors; nuclei with iso17q would contain three 17q signals with two 17p signals, whereas Chr17 loss would show loss of single 17p and 17q signals. Alternately, these cytogenetic copy number alterations could be accurately assessed in a single array CGH assay; however, this testing platform is fairly cost prohibitive for widespread clinical implementation.

Atypical teratoid/rhabdoid tumor (AT/RT) is a rare type of CNS embryonal tumor that exhibits extremely aggressive behavior and early metastases [100]. It most frequently arises in children under 5 years of age and is resistant to conventional medulloblastoma therapies [111, 112]. Accurate diagnosis of AT/RT from other CNS tumors is paramount in appropriately directing treatment options. Unfortunately, AT/RT is well known for its ability to display a wide array of histomorphologic features, and the diagnostic “rhabdoid cells” with eosinophilic cytoplasmic filamentous inclusions may be difficult to find. When embryonal small round blue cells predominate, these tumors may be easily confused with medulloblastoma or CNS PNET. Adding to the potential diagnostic dilemma, meningioma and glioblastoma can display a rhabdoid phenotype. The vast majority of AT/RTs have been found to harbor alterations of chromosome 22q11.2 with the SMARCB1 (also known as INI1, BAF47, hSNF5) gene being mapped to this region. Deletion and/or mutation of SMARCB1 has been detected in the majority of AT/RTs (up to 75 %) having demonstrable monosomy 22 or 22q deletion [111–113]; whereas these alterations are generally not detected in PNETs and medulloblastoma [114–116].

The presence of an SMARCB1 alteration can be assessed by a variety of molecular techniques. FISH analysis has predominately relied upon “laboratory developed” probes targeting the SMARCB1 locus and thus, testing has only been available at a few specialized centers [117]. This type of FISH analysis provides a rapid assessment for SMARCB1 deletions on FFPE samples; however, up to 30 % of AT/RTs have no detectable deletions necessitating SMARCB1 mutational analysis. Immunohistochemical staining for the SMARCB1 (INI1/BAF47) protein has demonstrated a universal loss of nuclear expression for SMARCB1 in AT/RT with excellent correlation between this immunohistochemical assay and other molecular techniques [118, 119]. Nuclear staining with anti-BAF47 likewise accurately differentiates rhabdoid variants of glioblastoma and meningioma from AT/RT [120]. Therefore, this loss of nuclear expression of SMARCB1 identified by IHC is considered the diagnostic hallmark of AT/RT, which can be easily incorporated into most clinical laboratory settings.

Malignant progression in meningiomas appears to be associated with the sequential accumulation of various genetic abnormalities with common alterations in chromosomes 1p and 14q [121–132]. Allelic losses on chromosome 1p have been shown to be of prognostic significance with alterations being more common in higher grade meningiomas. Additionally, 1p deletions are also associated with increased risk of recurrence/progression, although the significance of this finding is less clear when results are adjusted for histologic grade [121–137]. Likewise, loss of chromosome 14q is associated with poor prognostic features such as higher grade lesions and increased risk of recurrence. Three year survivals for patients diagnosed with meningiomas with 14q deletions compared to those with wild type lesions were 20.5 % vs 68.6 % respectively [121]. 14q deletions, when present in grade I meningiomas, are also associated with a higher risk of recurrence, 17 % vs 50 % recurrence in deleted tumors. 1p and 14q deletions used along with chromosome 22q loss, which is a well-known frequent occurrence in meningiomas, can be diagnostically useful for differentiating meningiomas from other dural-based tumors such as hemangiopericytoma, metastatic carcinoma, superficial glioblastoma and gliosarcoma [123, 125, 138–143]. There is currently no known predictive value for these markers.