The evaluation of central nervous system tumors increasingly relies on molecular genetic methods to aid in classification, offer prognostic information, and predict response to therapy. Available assays make it possible to assess genetic losses, amplifications, translocations, mutations, or the expression levels of specific gene transcripts or proteins. Current molecular diagnostics frequently use a panel-based approach and whole genome analysis, and generally rely either on DNA sequencing or on hybridization-based methodologies, such as those used in cytogenomic microarrays. In some cases, immunohistochemistry can be used as a surrogate for genetic analysis when the mutation of interest consistently results in overexpression or underexpression of a known protein product. In surgical neuropathology practice, the diagnostic workup of diffuse gliomas, medulloblastomas, low-grade circumscribed gliomas, as well as other diseases, now routinely incorporates the results of genomic studies. Here we summarize our institution's current approach to diagnostic surgical neuropathology, using these contemporary molecular diagnostic applications.
The practice of surgical pathology is ever dynamic, be it due to the recognition of novel diagnostic entities, adaptation to evolving treatment algorithms, or the utilization of newly developed diagnostic tools. In surgical neuropathology, the description of several clinically relevant and readily assessed molecular, cytogenetic, and epigenetic hallmarks of central nervous system (CNS) neoplasia is driving a sea change in practice with the goal of increasing diagnostic accuracy and reproducibility as well as improving therapeutic outcomes. On a more practical level, recognition of these hallmarks has led to their incorporation as World Health Organization (WHO) diagnostic criteria in some instances.1,2 With these cardinal abnormalities in mind, we in Emory Neuropathology (Emory University, Atlanta, Georgia) are incorporating broad-spectrum genomic tumor analysis alongside traditional morphologic assessment in our diagnostic approach. This shift has already offered several lessons that point toward future directions for the field.
Glial neoplasms that diffusely infiltrate the surrounding CNS parenchyma constitute a category of uniformly deadly tumors with widely variable degrees of clinical aggression. Classification of these tumors based upon histologic analysis has led to the traditional designation of 3 groups based primarily on nuclear and cytoplasmic morphology: oligodendroglioma, astrocytoma, and mixed tumors termed as oligoastrocytoma. Histologic features used to grade these neoplasms have included mitotic activity, tumor necrosis, and microvascular proliferation.3 More recently, the description and study of IDH1/2 (isocitrate dehydrogenase 1/2) mutations in a significant portion of infiltrating gliomas in adults has led to the steady realization that IDH-mutant gliomas form a group of genetically, morphologically, and behaviorally distinct neoplasms found within each of the previously established morphologic categories of disease.4,5 In short, IDH mutations do not merely carry prognostic import for gliomas, they are diagnostic hallmarks of a specific class of infiltrating glioma primarily found in adults and clinically distinguished by a more indolent behavioral phenotype.6–8
Furthermore, the characterization of IDH-mutant tumors has provided a new context in which to interpret other genetic alterations. Part of the current conceptual paradigm of gliomagenesis holds that IDH1/2 mutations are early events in glioma formation and are followed by the acquisition of additional characteristic genetic lesions important for diagnosis and therapy. Other (IDH nonmutant; “wild-type”) infiltrating gliomas never acquire these hallmark IDH mutations and instead follow distinct genetic and behavioral paths of progression.9,10 For an IDH-mutant tumor, the most important subsequent change “downstream” of an IDH mutation is an unbalanced whole-arm translocation of 1p and 19q with loss of the derivative chromosome, resulting in the prototypical “1p/19q codeletion” event long associated with oligodendroglial tumors.11,12 The results of numerous recent investigations have led to the recognition of 1p/19q codeletion as definitional of an oligodendroglioma in the setting of an IDH-mutant glioma, and an IDH-mutant neoplasm lacking a 1p/19q codeletion is best characterized as an infiltrating astrocytoma.6,13 IDH-mutant astrocytomas also tend to show inactivating alterations in TP53 and ATRX, findings that can assist in their molecular diagnosis as well.6,13,14 Other alterations in IDH-mutant gliomas have also been noted as markers of progression, including mutations in KRAS and PIK3CA and amplifications of PDGFRA and MYC.15,16 More recently, progression of IDH-mutant gliomas has been tied to changes in global DNA methylation patterns, specifically development of an increasingly hypomethylated epigenome that correlates with activating alterations in cell cycle genes.10,17
In the case of IDH wild-type infiltrating gliomas, most fall into 2 broad molecular categories.18 One, a “pediatric” variant of infiltrating glioma, is indeed most commonly diagnosed in pediatric populations and is characterized primarily by the presence of mutations in genes encoding histone variants such as H3F3A. One such mutation, H3F3A K27M, has been described in midline, brainstem, and spinal cord astrocytomas in children and adults; tumors with this mutation may show considerable histologic variation yet typically are associated with a poor prognosis.19–21 The other broad category of IDH wild-type infiltrating gliomas includes the most common overall type of infiltrating glioma: the “primary” glioblastoma, an aggressive infiltrating glial neoplasm.3,18,22 Neoplasms in this group can display a bevy of individual cytogenetic and molecular alterations but most characteristically show gain of chromosome 7 (including the EGFR locus) and loss of chromosome 10 (including PTEN). Other common alterations include mutations and/or deletions in the tumor suppressors CDKN2A, RB1, and TP53 as well as mutations and/or amplifications in receptor tyrosine kinase genes such as EGFR or PDGFRA; in combination with histopathologic analysis, these findings can assist in accurate diagnosis and classification.1,22–24 An extensive literature also exists tying the finding of MGMT promoter methylation to a more indolent clinical course and durable response to alkylating chemotherapy.25 In addition to the aforementioned common abnormalities, an increasing number of low-frequency events representing potential therapeutic targets have been described in glioblastoma, most notably fusions of FGFR3 and TACC3.26 While these tumors are classically detected as WHO grade IV lesions, recent investigations have demonstrated that neoplasms sharing molecular phenotypes in this category may present as WHO grade II, III, or IV tumors and still imply similar clinical outcomes.6
At Emory, we have implemented a targeted gene-array panel to identify many of the most common, clinically significant, and diagnostically meaningful alterations found in infiltrating gliomas. Our approach (Figure 1) begins with histologic assessment of the neoplasm, which, in combination with clinical and radiographic data, allows for the diagnosis of an infiltrating glioma and provides some insight as to the potential molecular background of the tumor. We then test for IDH mutations through a combination of immunohistochemistry for the most common IDH1 mutation (p.R132H)27 and a more comprehensive mutation panel to capture a broader range of IDH1 and IDH2 mutations. For IDH-mutant tumors, we next address the question of a 1p/19q codeletion as well as the presence of an ATRX or TP53 mutation.13 We check for the 1p/19q codeletion through the use of a cytogenomic microarray, specifically, a molecular inversion probe (MIP) designed for use on formalin-fixed, paraffin-embedded (FFPE) tissue. We favor this approach over fluorescence in situ hybridization given the array's broader scope that allows verification of the presence of the true whole-arm codeletion as well as assessment for other abnormalities. We currently assess ATRX and TP53 through immunohistochemistry28 (Figure 2, A through D), though we plan soon to augment this relatively imprecise method (particularly in the case of p53) with a broader-range mutation panel. In combination, these assays provide the capability of diagnosing an oligodendroglioma in the setting of an IDH-mutant, 1p/19q codeleted infiltrating glioma. In the case of an IDH-mutant, 1p/19q intact, ATRX/TP53-mutant infiltrating glioma, we can diagnose an IDH-mutant astrocytoma.6 For infiltrating gliomas that are IDH wild-type, we use the findings of chromosome 7 gain and 10 loss (Figure 3, A and B) alongside other more focal changes listed above (such as EGFR amplification) to classify the lesion in a category akin to a primary glioblastoma.1 We also assess for H3F3 mutations to identify pediatric-type infiltrating gliomas among the IDH wild-type tumors; currently, we are using an immunostain for the H3F3A K27M mutant protein29 but will likely transition to a broader-range mutation panel capable of assessing other mutations in this gene. With classification and subtyping complete, MGMT promoter methylation is also assessed via methylation-specific polymerase chain reaction. Future development of our testing algorithm will include an expansion of mutation and fusion testing through a more comprehensive sequencing-based platform.
POORLY INFILTRATIVE GLIAL AND GLIONEURONAL TUMORS
The “poorly infiltrative” gliomas and glioneuronal tumors are a morphologically diverse group of neoplasms that typically form discrete masses within the CNS that can often be surgically excised. Clinically, the behavior of these tumors is more indolent than that of their infiltrating cousins, leading to the designation of this group by some as “low-grade neuroepithelial tumors.” 30 A large number of diagnostic entities fall into this category, including pilocytic astrocytoma (PA), ganglioglioma, pleomorphic xanthoastrocytoma (PXA), rosette-forming glioneuronal tumor (RGNT), and dysembryoplastic neuroepithelial tumor (DNET). Further complicating the morphologic interpretation of these neoplasms is the fact that a small proportion of cases that have the genetic features of poorly infiltrative tumors will show diffuse infiltration of the brain parenchyma, though they may still behave in an indolent fashion.30,31 In light of the morphologic variety of these neoplasms, their potential for morphologic mimicry of more aggressive infiltrating tumors, as well as the relative rarity of several of these entities, molecular analysis provides very useful diagnostic assistance in their workup. Even more compellingly, some of these pertinent molecular findings can identify potential therapeutic targets.32,33
As with many low-grade neoplasms, these tumors often do not show a large number of genetic anomalies. Sometimes cases will display only a single abnormality, be it a mutation or rearrangement, in a quiescent background.30,34 This reality stands as a contrast to that seen in the infiltrating gliomas, where many cases will exhibit considerable genomic complexity.6,22,24 Many of the diagnostic entities within the poorly infiltrative tumor category exhibit alterations in the MAPK signaling pathway, most commonly in the BRAF gene.30 Activating BRAF V600E mutations have been noted in a large portion of PXAs as well as a significant minority of gangliogliomas, PAs, and DNETs.35,36 Fusion events that produce a constitutively active BRAF are noted in a portion of PAs and in other tumor types.37–40 The best-known example is the BRAF:KIAA1549 fusion typical of PAs found in the posterior fossa (Figure 4).38,41 These genomic findings not only carry implications for tumor classification, but also may assist in directing therapy. An assortment of pharmacologic inhibitors of mutant and nonmutant BRAF exists, and published experience in their application to CNS neoplasia is expanding.32 Targeted BRAF inhibition may soon become an important component of therapy for poorly infiltrative gliomas, particularly for those tumors that are incurable by standard means.
Other recurrent abnormalities noted in PAs, DNETs, pediatric oligodendroglial tumors, and RGNTs include mutations, tyrosine kinase duplications, and fusion events in FGFR1.30,42,43 Some cases of RGNT have been described with PIK3CA mutations.44 NTRK2 fusion events have been reported in PAs.37 Additionally, alterations in MYB and MYBL1, primarily fusion events, have been observed in angiocentric gliomas and low-grade astrocytomas.30,45 Clinical trials for targeted therapy addressing some of these various abnormalities are ongoing or in development.30,46
Given the large number of diagnostic entities with overlapping histologic and molecular features in this category, the diagnostic approach is less algorithmic than for infiltrating gliomas. Our method involves first ruling out more biologically aggressive diseases. Clinical history, radiographic assessment, and histologic findings provide a wealth of information in this regard, but molecular diagnostic data can prove extremely helpful as well. The discovery of an IDH mutation or H3 mutation in a glial tumor argues strongly against the diagnosis of a poorly infiltrative glioma.2 Identification of gain of chromosome 7, loss of chromosome 10, focal deletion of CDKN2A, and amplification of an oncogene such as EGFR would also militate against the diagnosis of a poorly infiltrative glioma.13,22 Following consideration of an alternative diagnostic category, we then use a combination of mutation and DNA copy number testing to identify cardinal mutations or chromosomal rearrangements involving such genes as BRAF, FGFR1, MYB/L1, NTRK2, and PIK3CA. In most cases, documentation of 1 of these mutations or rearrangements does not indicate a specific diagnosis; rather, the finding reinforces the interpretation of the lesion as a low-grade or poorly infiltrative glioma, gives some guidance as to likely diagnosis, and in some instances provides potential therapeutic options. Our current testing pipeline does not allow for detection of all of these abnormalities, but the MIP array can show copy number changes that may suggest the presence of fusion events. Some fusions, such as the tandem duplication that results in most BRAF:KIAA1549 events, can also be detected with the MIP array.47,48 Other lower-incidence fusion events are more reliably assessed by fluorescence in situ hybridization, polymerase chain reaction, or sequencing-based assays. Future directions for our testing approach include expanded mutation analysis that will allow us to detect more of the aforementioned characteristic mutations. Over time, we plan to incorporate an RNA-based “fusion panel” that should provide the capability of confirming a wider range of fusions than currently possible.
EMBRYONAL CNS NEOPLASMS
This broad category of CNS neoplasia typically occurs in pediatric patients and is best characterized by the presence of an “undifferentiated” or “primitive” neoplastic component that looks most similar to cells found in the developing fetus.2,3,49 Though the overarching category of “embryonal neoplasm” applies to all diagnoses within this group, further subclassification has produced a sometimes-bewildering array of diagnoses based upon location in the CNS as well as morphologic findings.2,3 Further complicating matters is the wide variation in response to therapy among these tumors. All of the embryonal tumors are WHO grade IV; however, some respond very well to treatment, while others carry dire prognoses. For instance, a medulloblastoma is often curable, whereas a morphologically similar tumor above the tentorium has a high potential for mortality.50,51 Fortunately, studies have shown that many embryonal CNS neoplasms can be reliably classified or subclassified through broad-spectrum analyses of gene expression, epigenetic profiles, mutational background, and cytogenetic findings.52–56 The advent of this multimodal testing approach has provided new avenues with which to differentiate these neoplasms from a clinical and diagnostic standpoint.
For example, the atypical teratoid/rhabdoid tumor (AT/RT) displays a degree of morphologic plasticity that may be frustrating for the diagnostician. Nonetheless, this tumor type is readily defined via the elucidation of SMARCB1 protein loss or related mutations or deletions in SMARCB1, a gene (previously widely known as INI1) involved in the SWI/SNF chromatin remodeling complex.57,58 While a small number of AT/RTs show retained SMARCB1 expression, these tumors instead typically harbor mutations or deletions in another SWI/SNF gene, SMARCA4. Furthermore, these much rarer AT/RTs will often demonstrate loss of BRG1, the protein product of SMARCA4.59 The strong association of AT/RT with either SMARCB1 or SMARCA4 abnormalities has led to WHO criteria requiring demonstration of at least SMARCB1 or BRG1 protein loss for the diagnosis.2 Recognition of the AT/RT is important for reasons beyond its potential to stump the pathologist: AT/RTs exhibit aggressive behavior and require a chemotherapeutic regimen very much different from that used to address other embryonal neoplasms.60,61 Fortunately, immunohistochemistry provides a reliable indicator of SMARCB1 and BRG1 abnormalities in most AT/RTs.2,57,59
Another category of embryonal neoplasms recently defined by a genomic abnormality was previously divided into 3 separate entities. This trio of uniformly aggressive embryonal neoplasms, including the embryonal tumor with abundant neuropil and true rosettes, the ependymoblastoma, and the medulloepithelioma, are histologically united by the shared feature of lumen-forming, multilayered, “ependymoblastic” rosettes. Molecular and cytogenetic studies have likewise shown that these 3 histologic tumor types all harbor a characteristic focal chromosomal gain/amplification of a group of oncogenic microRNAs at 19q13.42 (C19MC) along with overexpression of the RNA regulatory protein LIN28A. These unifying features have led to the consensus that all 3 diagnoses be joined as a single entity, the embryonal tumor with multilayered rosettes (ETMR).2,53,54 Although many of these tumors can be recognized morphologically, given the serious prognostic implications of this diagnosis it is best confirmed by demonstrating the characteristic chromosomal alteration or LIN28A overexpression.53,62
In the case of medulloblastomas, gene expression and methylation profiling have allowed the placement of individual tumors into 1 of at least 4 separate clinically relevant subgroups: those with mutations in the Wnt pathway (“group 1”); those driven by sonic hedgehog (SHH) pathway mutations (“group 2”); those generally characterized by MYC overexpression (“group 3”); and those not readily falling into other categories (“group 4”).55 Group 1 tumors are typically marked by indolent behavior, so much so that some have proposed that they merit less aggressive therapy.55,56 Though group 2 tumors exhibit an overall more aggressive phenotype than their group 1 brethren, a variety of Shh pathway inhibitors have been developed that may provide targeted therapies for group 2 tumors, particularly in a clinical trial format.56 Group 3 tumors portend an overall poor prognosis, and group 4 tumors are associated with outcomes intermediate between groups 1 and 3.55
Some other embryonal neoplasms are less well defined, particularly those constituting the miasmatic category of “CNS-primitive neuroectodermal tumors.2 ” Some within this group are now recognized as embryonal-appearing forms of other established entities.63–66 For example, a portion of these tumors harbor similar genetic changes to infiltrating gliomas, and it is well recognized that some infiltrating gliomas can harbor a poorly differentiated “embryonal component,” a group referred to as “infiltrating glioma with primitive neuronal component” or “glioblastoma with primitive neuronal component.2 ” The detection of combined infiltrating gliomas and embryonal neoplasms may require careful histologic analysis to observe the 2 components, along with molecular data that can support their related nature. For instance, discovery of an IDH mutation in an embryonal-appearing neoplasm should prompt a search for an adjacent or antecedent IDH-mutant infiltrating glioma.65 Similar cases in pediatric patients may contain mutations in genes encoding histone variants common in pediatric glioblastomas as described above. Such cases, although histopathologically neuroblastic, behave clinically as glioblastomas, lacking propensity to metastasize through the CSF.64 Other seemingly embryonal CNS tumors have been shown to harbor genetic features typical of ependymomas, such as fusions involving the RELA gene on chromosome 11.67–69
Our clinical workup of embryonal CNS neoplasms is still rooted in histologic and clinicopathologic information, with important guidance provided by genomic data. We appraise tumor morphology, location, and genomics together to arrive at a unified diagnosis, and we use morphologic and genomic information as applicable to address clinically relevant tumor subclassification. As a rule, we work to exclude cardinal features of specific diagnostic entities that are not distinctly bound by location or morphology. Chief among this group is the AT/RT. In our practice, all embryonal neoplasms are tested for the expression of SMARCB1 by immunohistochemistry, as demonstration of loss of nuclear expression of SMARCB1 in an embryonal CNS neoplasm is diagnostic of the notorious masquerader AT/RT in most scenarios.2,57 Furthermore, AT/RTs show a characteristic cytogenetic picture: they usually exhibit a noncomplex karyotype with most abnormalities concentrated at the location of SMARCB1 on 22q70 ; we use our MIP array to provide information on both the tumor cytogenetic makeup as well as any focal changes on 22q (Figure 5). If a suspicious case has retained SMARCB1 and displays a compatible cytogenomic picture, assessment of BRG1 expression is sought.59 Another embryonal neoplasm that is not tied to a specific location is the ETMR. As stated previously, ETMRs are typified by a focal gain at 19q13.42,54 and they can also be diagnosed by documenting LIN28A overexpression by immunohistochemistry.62 In our practice, we rely on the MIP array to detect the focal gain of 19q13.42 rather than LIN28A expression.53 Finally, the detection of embryonal-appearing neoplasms genetically related to other categories of CNS neoplasia requires careful histologic and genomic analysis. For instance, combined infiltrating gliomas and embryonal neoplasms may be revealed by thorough sampling and morphologic observation to observe the 2 components, along with molecular data that can support their related nature. Along these lines, discovery of an IDH mutation in an embryonal-appearing neoplasm may indicate the existence of an adjacent or antecedent IDH-mutant infiltrating glioma.65 In each of these instances, we use genomic tools to provide us with data to inform our morphologic eye and assist in identification of disease-defining abnormalities.
In the case of medulloblastomas, we use genomic information to assist with subclassification and prognostication. After histopathologic confirmation of a medulloblastoma, we use the MIP array to assess for prognostically significant chromosomal abnormalities, such as MYC or MYCN amplification or loss of 17p.71–73 Cytogenomic data from the MIP array are also used along with immunostains for β-catenin, GAB1, and YAP1 in order to gain additional subclassification data.55,74 Within group 1, Wnt pathway–driven medulloblastomas commonly display “classic” histology though they may also show anaplastic/large cell morphology. These tumors usually demonstrate nuclear β-catenin expression by immunohistochemistry, reflecting an underlying mutation in the CTNNB1 gene.55,56,74 Group 1 tumors will also display YAP1 expression by immunohistochemistry.74,75 Most tumors in this group also exhibit monosomy 6, which the MIP array displays well, often showing it as a lone cytogenetic abnormality.56,75 Identification of group 2, SHH pathway–driven medulloblastomas is aided by histology in that nodular/desmoplastic medulloblastomas almost always fall into the group 2 category; however, group 2 tumors may exhibit any of the defined histologic subtypes of medulloblastoma. Many group 2 tumors show both GAB1 and YAP1 expression by immunohistochemistry.55,74,75 Group 2 tumors are associated with numerous cytogenetic abnormalities that may include changes at regions encompassing SHH-related genes such as PTCH1, GLI2, NOTCH1, SMO, and SUFU.55,75–78 Additionally, the presence of TP53 mutations in group 2 medulloblastomas is highly relevant clinically: those with TP53 mutations tend to occur in an older patient population, are more frequently associated with germline TP53 mutations, and portend shorter survival.79 Groups 3 and 4 medulloblastomas display a wide variety of changes but usually lack the immunoprofile and genetic features that unify tumors from groups 1 and 2.55,75 MYC amplification is concentrated in group 3 tumors, though this finding is not seen in most cases.55 The classic cytogenetic hallmark of medulloblastomas—isochromosome 17q—is found most commonly in group 3 and 4 tumors.55 However, in many cases our current approach in subtyping medulloblastomas does not allow for definite differentiation of a group 3 from a group 4 tumor. Hence, we focus on separating tumors in groups 1 and 2 from those in groups 3 and 4, and also work to separate group 1 from group 2 tumors. Though we currently use a combination of histologic, immunohistochemical, molecular, and cytogenetic means to aid us in this subclassification, we are mindful of the powerful capabilities of methylation-based arrays and gene expression analysis to make these distinctions in a more straightforward manner.55,56,77 Further, it is important to consider that national and international clinical trials are moving toward stipulating the mechanisms through which medulloblastoma subclassification can be accomplished.56 With all of this in mind, we are evaluating the inclusion of broad-spectrum methylation analysis in our workup. Helpfully, this type of methylation data would assist us not only in the subclassification of medulloblastomas, but also in the classification of many of the other aforementioned embryonal neoplasms (not to mention a wide variety of other CNS neoplasia).18,66,77
FURTHER APPLICATION OF GENOMIC DATA IN SURGICAL NEUROPATHOLOGY
As illustrated by the above examples, we use broad-spectrum molecular and cytogenetic information in the diagnosis and workup of many of the primary CNS neoplasms encountered in our clinical service. However, the use of these analyses is not limited to the circumstances described above. Rather, these approaches have repeatedly revealed unanticipated features of numerous lesions that partly or entirely altered the diagnostic interpretation, clinical workup, or both. A sample of these instances follows.
A young woman presented with a nonenhancing, partly cystic imaging abnormality in the temporal lobe. Open biopsy disclosed a banal proliferation of astrocytic cells that show marked infiltration of the surrounding neuropil (Figure 6). No distinct features to indicate a pilocytic astrocytoma were present, and no particular perivascular orientation of tumor cells was noted. No mitotic figures were seen and a MIB-1 immunostain revealed a low (1%) proliferative rate. Potential diagnoses included an infiltrating astrocytoma, an angiocentric glioma, and a diffuse variant of a low-grade astrocytoma such as a pilocytic astrocytoma. Molecular and cytogenetic analysis showed a BRAF V600E mutation but no IDH mutation or copy number abnormalities in the neoplasm. Altogether, the data best fit the classification of the lesion as a rare BRAF-mutant infiltrating astrocytoma,80 suggesting that it might be responsive to BRAF-targeted therapy.
A middle-aged man presented with a remote history of medulloblastoma, status post resection and chemoradiation, and a slowly growing nonenhancing lesion at the resection site (Figure 7, A and B). A biopsy revealed disorganized but well-differentiated glial and neuronal elements. Focally, atypical astrocytic-appearing cells were seen, but no clear morphologic evidence of residual or recurrent medulloblastoma was present. Immunohistochemistry for glial fibrillary acidic protein, synaptophysin, and neurofilament highlighted the glial and neuronal components, respectively, and a MIB-1 immunostain revealed a vanishingly low (<1%) proliferative rate. Diagnostic considerations included an infiltrating astrocytoma (potentially related to radiation), a reactive lesion related to previous therapy, and a medulloblastoma exhibiting a maturation-like phenomenon like that which is most commonly seen in neuroblastomas. We performed a MIP array on both the lesion in question and the patient's original medulloblastoma, and showed that each displayed identical copy number abnormalities despite the markedly divergent clinical, radiographic, and morphologic pictures. This information strongly indicated that the 2 lesions shared a common origin despite their different clinicopathologic profiles. The data together indicated a medulloblastoma with maturation, a finding rarely documented in medulloblastomas.81
A young child presented with an enhancing dura-based mass adjacent to the cerebellum. Upon resection, the lesion was confidently classified as a meningioma, and clear cell features were noted. Given the odd morphologic appearance and the atypical clinical scenario, final diagnosis was held in anticipation of molecular and cytogenetic data. The MIP array showed a loss of heterozygosity on chromosome 17q, encompassing SMARCE1 among numerous other genes. In light of the recent description of a tumor syndrome typified by spinal clear cell meningiomas in pediatric patients sharing germline mutations in SMARCE1, a recommendation was placed to consider genetic counseling and germline testing for the patient.82,83 Follow-up targeted mutation analysis showed a germline SMARCE1 mutation, consistent with the diagnosis of the abovementioned tumor syndrome.
A young man presented with a partly enhancing, poorly demarcated mass in the temporal lobe. Surgical excision revealed a cellular glial neoplasm with numerous astrocytic and oligodendroglial-appearing cells displaying scattered mitotic figures. Scattered throughout the lesion were numerous multinucleated, atypical neurons (Figure 8). Immunohistochemistry for glial fibrillary acidic protein, synaptophysin, and neurofilament illuminated the expected glial and neuronal constituents. The differential diagnosis based upon this information included a glioneuronal tumor such as a ganglioglioma, a poorly infiltrative glioma such as a PXA, and a more traditional infiltrating glioma. Immunostaining for IDH1 p.R132H mutant protein was positive, and an immunostain for ATRX showed loss of expression in neoplastic nuclei, suggestive of an ATRX mutation. The IDH1 mutation was confirmed by molecular testing, and MIP array testing was negative for a 1p/19q codeletion event. Instead of a glioneuronal tumor or poorly infiltrative glioma, we diagnosed the lesion as an IDH-mutant anaplastic astrocytoma, significantly affecting the treatment approach for this patient.13
CLINICAL AND TECHNICAL CONSIDERATIONS
As can be appreciated from the above discussion, we view molecular and cytogenetic analysis as a critical adjunct to morphologic evaluation and an integral part of our diagnostic workup in surgical neuropathology. Furthermore, we undertake efforts to ensure that, as part of our diagnostic algorithm, we gather sufficient intact tumor-derived genetic material to facilitate the relevant testing. Though some of these assays can be performed by using tissue-sparing, highly sensitive immunohistochemical methods,27 many methods require more substantial amounts of tissue and of tumor content to ensure a successful test and reliable result.84 These requirements may necessitate changes in the traditional approaches applied to diagnosis and workup.
First, the need for adequate viable tissue for complete workup should be clearly communicated at the time of biopsy or excision. The exact amount of tissue needed will vary between institutions and testing platforms, but surgeons and pathologists should at least be aware of the potential necessity of additional tissue, particularly at the time of frozen section. Second, similar communication is necessary between the pathologist and the person performing gross examination and histologic sampling (if that person is not the pathologist). If possible, grossly viable, nonnecrotic tissue should be gathered together in tissue blocks to maximize tumor percentage.84 As long as no fresh tissue is needed for testing, tissue should be transferred into formalin as quickly as possible. We prefer FFPE-based testing platforms to allow for histologic evaluation of all tissues, selection of optimal areas to test, and ready integration into standard pathologic practice. Finally, solutions that damage genetic material (acid decalcifiers in particular) are to be avoided.
At the time of initial hematoxylin-eosin slide review, tissue management should also be pursued with genomic testing in mind. Situations in which tissue is limited may require a special approach. In our laboratory, we use unstained recut slides as our source for most genetic material, and on cases with limited tissue we will typically request multiple unstained recuts at the same time as we request our immunostains to minimize potential tissue loss by cutting all needed sections in a single instance. As needed, we may also request unstained slides from multiple tumor blocks, or focus our immunostains on a single block with lower tumor cellularity, while retaining a higher cellularity block for molecular testing. Obviously, microscopic examination is critical to making these distinctions and decisions, a fact that in our experience again significantly favors the use of FFPE-based testing platforms for solid tumors such as those in the CNS.84 These platforms also allow for the direct comparison of morphologic findings with molecular and cytogenetic findings.
In our experience, molecular and cytogenetic testing usually requires an additional week to 10 days for results, depending on the laboratory's caseload and testing schedule. During this time, detailed morphologic and immunohistochemical investigation can be accomplished that serves as a basis with which to interpret the molecular, cytogenetic, and epigenetic data. At our institution we hold a consensus conference on a weekly basis, attended by anatomic pathologists, cytogeneticists, molecular pathologists, laboratory technicians, and trainees, where we discuss the implications of the wide variety of molecular findings that may be present in each individual case. This conference, and the communication that it fosters, has been extremely useful in producing anatomic and molecular diagnostic reports that are in agreement with, and build upon, one another. It also provides a forum to troubleshoot cases that pose difficulties from a technical or interpretive standpoint.
Once testing and interpretation are complete, clear communication is necessary with our clinical colleagues involved in treatment and follow-up. In our experience, multidisciplinary tumor conferences serve as the best cross-specialty mechanism to discuss all of the pathologic findings and their meaning for patient diagnosis and future management.85 On an even more practical level, open communication with colleagues from other services allows for shared understanding of the potential delays in diagnostic report preparation necessitated by the need to obtain molecular and cytogenetic data.
While many useful review publications address the burgeoning field of molecular diagnostics in surgical neuropathology, herein we attempted to provide a current view of the algorithms we use at Emory to integrate multiple modalities into neuropathologic diagnosis. Numerous platforms and mechanisms exist to obtain the type of molecular, cytogenetic, and epigenetic information that is now shaping diagnostic surgical neuropathology, and much work remains to standardize diagnostic algorithms and testing approaches across institutions. Future directions at Emory include an expanded mutation panel, better identification of fusion events relevant to diagnosis or treatment, and potential utilization of DNA methylation arrays as have been well documented elsewhere. Regardless, what appears clear to us is that the information garnered from these types of analyses is becoming essential to the provision of diagnoses that are both more accurate and more reproducible.
The authors have no relevant financial interest in the products or companies described in this article.