Something old and something new about molecular diagnostics in gliomas

Synopsis

Progress in our understanding of the molecular biology of neoplasms has been driven by remarkable improvements in molecular biology techniques. This has created a rapidly moving field in which even subspecialists struggle to keep abreast of the current literature. Nowhere is this more clearly demonstrated than in neuro-oncology, wherein molecular diagnostics can now wring more clinically useful information out of very small biopsies than ever before. Herein the biologic and practical aspects of four key molecular biomarkers in gliomas are discussed, including two that have been known for some time (1p/19q codeletion and EGFR amplification) as well as two whose relevance was discovered via advanced whole-genome assays (IDH1/2 mutations and BRAF alterations).

Histologic review of the most common gliomas

Gliomas are the most common primary brain tumors in both children and adults, though in children the posterior fossa is the most frequent site whereas adult gliomas are much more likely to develop in the supratentorium. The WHO grading scheme ranges from I to IV, with an accompanying increase in severity.^{1, 2}

Pilocytic astrocytomas

Grade I gliomas are usually pilocytic astrocytomas (PAs), which are the most common primary brain tumors in children. About two-thirds are located in the cerebellum but they can occur anywhere in the neuraxis. Key histologic characteristics include biphasic architecture with or without microcysts, bland nuclei, Rosenthal fibres, and eosinophilic granular bodies [virtual slide #1]. Yet the appearance of these tumors can vary greatly, creating diagnostic overlap with other gliomas. In keeping with their WHO grade I designation, most are curable if gross total resection can be achieved.³ However, tumors arising in harder-to-reach areas like the diencephalon (thalamus and hypothalamus) and myelencephalon (brainstem) usually cannot be completely removed and often recur, with worse outcome more likely.

Pilomyxoid astrocytoma

Pilomyxoid astrocytoma (PMA) is an uncommon grade II variant of PA, usually arising in the diencephalon and featuring a looser arrangement of spindly bland tumor cells with a bluish-myxoid background and variable degrees of perivascular orientation [virtual slide #2]. Again, though, there is considerable overlap with PA morphology. In fact, it is possible that many PMAs are simply less “mature” PAs that, because they arise in the biologically sensitive diencephalon, do not have enough time to develop more classic PA histology before being resected.⁴

Infiltrative glioma

Unlike grade I PAs, the most common grades II—IV gliomas are, by definition, diffusely infiltrative. As a result, surgical debulking can only extend survival. Even radiation and chemotherapy, usually reserved for grades III and IV tumors, are not curative. If a glioma shows signs of being infiltrative (e.g. spreading along neurons and capillaries) yet lacks mitoses, necrosis, and microvascular proliferation, it is WHO grade II. If the tumor nuclei are generally angulated and irregular, the glioma is classified as an astrocytoma, with a median survival of about 5 years.¹ Grade II oligodendrogliomas feature round nuclei [virtual slide #3] and have a median survival double that of grade II astrocytomas. All grade II tumors are still considered “low grade” even though they are technically incurable, simply because death is delayed. As a general principle, grade II tumors do not enhance with contrast-inducing agents on radiology. The presence of calcifications should not be used as a discriminator between astrocytomas and oligodendrogliomas; likewise, the absence of “chickenwire” capillaries and “fried egg” artifact do not rule out an oligodendroglioma. The term “oligoastrocytoma” refers to gliomas with mixed features, but while the entity is recognized by the WHO it suffers from poor interobserver consistency.

Mitoses in glioma

The presence of mitoses in a glioma raises the WHO grade to III, though the exact number of mitoses required per microscopic area has never been firmly established. These tumors usually show at least some contrast enhancement on scans. Grade III anaplastic astrocytomas [virtual slide #4] have a median survival of 2 to 3 years, while anaplastic oligodendrogliomas [virtual slide #5] are slightly longer at 3 to 4 years.¹ A key diagnostic point is that, while necrosis and/or microvascular proliferation bump an anaplastic astrocytoma up to a grade IV glioblastoma, they do not necessarily increase the WHO grade of an anaplastic oligodendroglioma. There is some controversy about this, though, especially in anaplastic gliomas with necrosis and mixed astrocytic/oligodendroglial morphology.⁵ Such tumors are sometimes termed “glioblastomas with oligodendroglial component” (GBM-O) or “mixed oligoastrocytoma grade IV” (MOA IV).

Glioblastoma

Glioblastomas (formerly called “glioblastoma multiforme” or “GBM”), are WHO grade IV and have a median survival of 9 to 12 months.¹ Unfortunately this is also the most common grade in adults, comprising about 70% of all gliomas. In addition to the aforementioned criteria for grades II and III astrocytomas, GBMs show microvascular proliferation and necrosis, the latter often surrounded by a pseudopalisade of glioma cells [virtual slide #6]. As befits the traditional “multiforme” appellation, though, there is a great deal of microscopic heterogeneity, including gliosarcoma [virtual slide #7], giant cell variant [virtual slide #8], and granular cell variant [virtual slide #9], among others.

Whereas progression to a higher grade is extremely uncommon in PAs, it is the rule in diffusely infiltrative gliomas. Yet most GBMs seem to arise de novo and thus are called “primary” GBMs, in contrast to “secondary” GBMs that developed from known grade II and III astrocytomas.

Glioma molecular biomarkers

The explosion of data in the field of glioma molecular biomarkers has rendered a comprehensive review of all noteworthy markers unwieldy. Instead, discussion is focused on four high-yield genetic biomarkers:

• Two that have been known to exist for a long time:

  • 1p/19q codeletion

  • EGFR amplification

• Two that were discovered only recently as the result of whole genome assays and sophisticated bioinformatics:

  • IDH1/2 mutation

  • BRAF fusion/mutation

Careful use of both “old” and “new” biomarkers greatly enhances the practice of surgical neuropathology (Table 1).

Table 1.

Key genetic alterations and their use in glioma diagnostics. 1p/19q codeletion, EGFR amplification, IDH1/2 mutations, and BRAF fusion or V600E mutation are all useful in resolving diagnostic dilemmas and/or refining patient prognosis in gliomas. Each test must be applied judiciously in specific contexts for maximal efficacy, and must always be correlated with light microscopic analysis and radiology. With the exception of BRAF, none of these tests are useful in pre-adolescent patients.

Genetic alteration	Main gliomas	Diagnostic value	Prognostic value	Predictive value	Most common methods
1p/19q codeletion	~80% of grade II and III oilgodendrogliomas	Differentiates most oligos from astrocytic gliomas and oligo mimickers (e.g neurocytoma, clear cell ependymoma, small cell GBM)	Longer survival	No specific therapy to date, but associated with better response to adjuvant therapy	FISH, PCR-based LOH
EGFR amplification	~40% of GBMs	Often detects scattered tumor cells in undersampled GBMs; differentiates small cell GBM from AO	Controversial	None	FISH
IDH1/2 mutation	~80% of grades II–III astrocytomas and oligodendrogliomas; over 90% of secondary GBMs	Presence strongly suggests an infiltrative glioma; negative in nonneoplastic glioma mimickers and noninfiltrative gliomas (e.g. PA, non-glial tumors)	Longer survival, may even trump WHO grade in some cases	No specific therapy to date, but associated with better response to adjuvant therapy	PCR and sequencing, IHC
BRAF fusion or V600E mutation	>75% of PAs (mostly fusions), ~50% PMA, 80% PXA (mostly V600E); 25% GG (mostly V600E)	Fusion suggests a PA or PMA; mutation suggests PXA or GG (though not a perfect discriminator from PAs or diffusely infiltrative gliomas)	May be a favorable marker, but still uncertain	No specific therapy to date; anti-MEK and anti-BRAF V600E clinical trials ongoing	fusions: PCR breakpoint analysis, FISH V600E: PCR and sequencing, IHC

GBM = glioblastoma; AO = grade III anaplastic oligodendroglioma; PA = pilocytic astrocytoma; PMA = pilomyxoid astrocytoma; PXA = pleomorphic xanthoastrocytoma; GG = ganglioglioma; FISH = fluorescence in situ hybridization; LOH = loss of heterozygosity; IHC = immunohistochemistry.

1p/19q: Some standardization required

Codeletion of the short arm of chromosome 1 and the long arm of chromosome 19 has been known to be a marker of oligodendroglial morphology for nearly 20 years.^{6, 7} This codeletion is the result of an unbalanced translocation between the two chromosomes with subsequent loss of der(1;19)(p10;q10).^{8, 9} Despite our karyotype-induced temptation to think of them as separated by vast distances, chromosomes 1 and 19 are actually close together in the nonrandom organization of the nucleus, at least partially explaining why this translocation appears so frequently.

Codeletion and adjuvant therapy

Over the past two decades an abundance of research has shed more light on the clinical and biological implications of 1p/19q codeletion. For example, numerous studies have shown that oligodendrogliomas with 1p/19q codeletion show improved response to adjuvant radiochemotherapy and longer survival, with the survival difference being much stronger in grade III anaplastic oligodendrogliomas than in grade II oligodendrogliomas.^10–16 Interestingly, though, the codeletion may impart such a survival difference only if the tumor is treated with adjuvant therapy—i.e. 1p/19q codeletion may not make a difference if the tumor is not challenged with post-surgical radiation and/or chemotherapy.¹⁷ An oligodendroglioma that shows textbook-type morphology such as round nuclei, delicate branching vasculature, and microcalcifications is more likely to contain the codeletion, as opposed to gliomas with only a few features suggestive of oligodendroglioma.^{18, 19} Yet even in gliomas that contain both oligodendroglial and astrocytic features (so-called “oligoastrocytomas”), those with the codeletion tend to behave more like oligodendrogliomas,^20–22 whereas strong nuclear p53 staining is more suggestive of astrocytic differentiation.²³

Codeletion and reclassification

However, this author’s experience suggests that histologically unequivocal GBMs should not be reclassified based solely on the presence of 1p/19q codeletion, because those tumors will still behave like GBMs and, through extensive genomic instability, produce false-positive results (manuscript under review). Likewise, though isolated loss of 1p has been suggested to be a weakly favorable prognostic factor in GBM,²⁴ this has not proven to be a consistent finding²⁵ (manuscript under review) and is not worth testing for in an otherwise unequivocal GBM.

Codeletion for differentiation among tumors

Testing for 1p/19q codeletion also can help differentiate between oligodendroglioma and its common mimickers, dysembryoplastic neuroepithelial tumor (DNET), clear cell ependymoma, and central neurocytoma, none of which have 1p/19q codeletion.^{26, 27} However, some case reports have suggested an overlap between oligodendrogliomas and neurocytomas since 1p/19q codeletion is occasionally seen in rare oligo-like tumors that also show unequivocal signs of neurocytic differentiation.^28–30

Gliomas with 1p/19q codeletion tend to have a proneural expression profile, MGMT promoter methylation, and IDH1/2 mutations.^31–33 In particular, it has been demonstrated that all gliomas with true whole-arm 1p/19q codeletion should also have IDH1/2 mutations.^{31, 34} Personal experience supports this finding; now this author recommends interrogating IDH1/2 status first, then considering additional 1p/19q testing only if an IDH1/2 mutation is detected. Coexistence of EGFR amplification and 1p/19q codeletion is extremely rare,^{16, 35, 36} to the point where such cases are probably just false-positive for true whole-arm codeletion. 1p/19q codeletion is about half as frequent in oligodendrogliomas arising in temporal lobe tumors versus those in the frontal, parietal, or occipital lobes, and has been reported to be especially uncommon in tumors from the insula,^{16, 37, 38} though the latter assertion has been disputed.³⁹ All these observations are applicable only to the adult patient population, since pediatric oligodendrogliomas rarely have 1p/19q codeletion and, even when apparent codeletion is found, it is not clear whether it has any prognostic value in this context.^{40, 41}

Tests for 1p/19q codeletion

There are many ways to test for 1p/19q codeletion in routine clinical samples; two of the most common are fluorescence in situ hybridization (FISH) and polymerase chain reaction (PCR)-based microsatellite loss of heterozygosity (LOH).

FISH

FISH uses fluorophore-labeled DNA probes to bind targeted regions in tumor nuclei. In the case of 1p/19q, 1p and 19q obviously have to be interrogated, and 1q and 19p are included as internal ploidy controls. The most commonly used probe pairs are commercially available, targeting 1p36/1p25 and 19q13/19p13 [Figure 1]. Because only two unstained slides are needed (one slide per probe pair), and only 60 to 100 tumor nuclei need to be counted to get a reliable score, FISH can be done on very small tissue biopsies—a feature of great importance in neuropathology. Also, a corresponding H&E section can be used to focus analysis on a specific area of interest. Patient-matched nonneoplastic DNA is not needed as an internal control. And FISH is well-suited to detecting polysomy of chromosomes 1 and 19, which has been shown to correlate with shorter progression-free survival in anaplastic oligodendrogliomas that have relative 1p/19q codeletion.⁴²

Figure 1.

Testing for 1p/19q codeletion. (A) Microsatellites spanning 1p and 19q are targeted via PCR-based LOH analysis, whereas the most widely used FISH probes bind to 1p36 and 19q13 (red) and 1q25 and 19p13 (green). (B) Under the microscope, DAPI-counterstained nuclei in a codeleted oligodendroglioma show the same signal pattern for both probe pairs: 2 green signals but only 1 red signal.

There are, however, some drawbacks to FISH in assessing 1p/19q status. Because the goal is to find relative deletions of 1p and 19q, and tissue sections need to be 4–6 microns thick for FISH to work even though nuclei are usually greater than 6 microns in diameter, it follows that partial loss of some tumor nuclei, including their DNA, is inevitable. This issue is addressed by the use of unrelated nonneoplastic brain tissue (e.g. postmortem white matter) as a negative validation control; any case showing fewer 1p and 19q signals than the controls is considered codeleted. Even this approach has problems, though, as there are no universally-adopted cutoff criteria for how many cells have to show codeletion in “true” 1p/19q codeletion.⁴³ For that matter, even the scoring parameters vary between institutions. Some incorporate 1p/1q and 19q/19p signal ratios into the analysis, whereas others base it solely on the percent of tumor cells showing deletion of 1p and 19q signals relative to 1q and 19p signals. Because FISH probes cannot be more than about 1 Mb to hybridize properly, it is impractical to accurately interrogate all of 1p and 19q by FISH. Thus, only small representative areas can be targeted. Prior studies indicated that 1p36 and 19q13 are minimally-deleted regions in diffuse gliomas,^{44, 45} which is why these loci were initially targeted by commercial developers of FISH probes. However, since it is whole-arm codeletion that has prognostic value, these loci end up being sensitive but not as specific as one would like.⁴⁶ Even if optimal probes and interpretation criteria were known, 1p/19q FISH would still be unreliable for proving a glioma in the case of a biopsy where only scattered atypical cells are present, as “dilution” with nonneoplastic nuclei would render the data unreliable.

Based on our current outcomes-based work, it appears that if one wishes to adopt signal ratios as the interpretation criteria, 1p36/1q25 and 19q13/19p13 should each be less than 0.75 for optimal prognostic stratification and concordance with PCR-based microsatellite LOH (see below).⁴⁶ If using percent of tumor nuclei with relative deletion, at least 40% should show relative loss of both 1p36 and 19q13. As mentioned above, IDH1/2 mutation status is a useful prescreening test; IDH1/2 wild-type cases practically never have true whole-arm codeletion. Also, 10q copy number testing can act as a check against 1p/19q FISH, since the presence of 10q deletion should raise suspicion of a false-positive codeletion.⁴⁶

PCR-based analysis of LOH

The other most commonly used tool for testing 1p/19q status is PCR-based analysis of LOH in DNA repeats known as microsatellites, which are scattered throughout the genome and are used as surrogate markers for nearby regions of interest. Microsatellites vary in the number of repeats they contain, and when both copies of a DNA locus contain the same microsatellite with two different lengths, the PCR products will migrate at different rates. Such a locus is then said to be informative. If a known informative locus shows only one PCR product size, the other is assumed to have been lost. Like FISH, PCR-based LOH can be done on formalin-fixed, paraffin-embedded (FFPE) tissue. Unlike FISH, it is quite feasible to interrogate many regions of DNA at once, allowing for a more complete view of 1p and 19q (Figure 1A). This reduces the risk of scattered interstitial deletions producing false-positive data, a risk especially notable in higher-grade gliomas. Criteria for interpretation of LOH data must be strict, calling codeletion only if all or nearly all of the microsatellites on both 1p and 19q show LOH. While FISH and PCR-based LOH generally show over 90% concordance,^{45, 47–49} concomitant prospective testing of gliomas with both assays showed PCR-based LOH is better at prognostic stratification in anaplastic oligodendrogliomas.⁴⁶

Although PCR-based microsatellite LOH analysis is more specific than FISH for true 1p/19q codeletion, it too has its drawbacks. For one thing, it is harder to accurately target subregions on a piece of tissue for analysis, and thus is more susceptible to dilution with nonneoplastic DNA. For optimal results it requires control patient-matched nonneoplastic DNA, usually in the form of white blood cells. Both assays are fairly labor intensive, but PCR-based LOH tends to be more expensive because it needs more sophisticated equipment. Thus, right now FISH is more widely used. In the near future, use of single nucleotide polymorphism (SNP) arrays and array comparative genomic hybridization (aCGH) will probably become more popular since they can, in a single test, scan not just 1p and 19q but also other loci of interest like 7p12 (EGFR), 9p21 (CDKN2A), and 10q.

Targeted therapy for 1p/19q-codeleted glioma

From a clinical perspective there is no targeted therapy for 1p/19q-codeleted gliomas to date—that is, no particular chemotherapy or course of radiation has proven to be specifically more effective in 1p/19q-codeleted gliomas, just that such tumors respond better to any adjuvant therapy compared to their 1p/19q-intact brethren. Thus, the decision as to whether to administer adjuvant therapy, and what kind of therapy to give, is generally predicated more on the WHO grade and location of the tumor, extent of tumor resection, and overall condition of the patient. 1p/19q status is of interest to the clinicians mostly because they want to anticipate how well their patients will respond to treatment.

EGFR: A new twist on an old friend

Amplification of the epidermal growth factor receptor (EGFR) gene on 7p12 has been known to occur in gliomas since the 1980s, preceding even 1p/19q codeletion as a recognized phenomenon in gliomas.⁵⁰ The transmembrane receptor tyrosine kinase triggers multiple pathways known to promote cell growth, migration, and survival,^51–55 and its overexpression broadly correlates with increased WHO grade.

EGFR amplification in diagnosis

From a diagnostic standpoint EGFR amplification occurs in about 40% of all GBMs, specifically primary GBMs but not secondary GBMs.^{1, 56–58} Amplification is more common in patients 40 years or older, and is only very rarely seen in pediatric high-grade gliomas.^{59, 60} Nearly half of all GBMs with EGFR amplification (and less than 10% without amplification) contain a truncated variant of the gene, encoding a protein that lacks the extracellular domain and is therefore constitutively active (EGFRvIII).^61–63

Because the majority of gliomas with EGFR amplification are histologically GBMs, in cases where the tissue biopsy is not fully representative of the actual tumor, detecting amplification is usually tantamount to a diagnosis of GBM.^{21, 64} Yet around 15 to 25% of grade III anaplastic astrocytomas also show EGFR amplification,^{62, 65} and while some studies suggest such tumors will behave more like GBMs,^{66, 67} others do not.⁶² Thus, it is not entirely clear whether EGFR amplification in an otherwise grade III glioma always represents an undersampled GBM. The same, however, cannot be said for gliomas that only meet grade II histologic requirements yet have amplification; such cases behave far worse than true grade II gliomas and can be assumed to be undersampled high grade gliomas, especially if the radiologic findings suggest high grade features like contrast enhancement.⁴⁹

Another feature of EGFR amplification is its tight association with astrocytic gliomas—to wit, tumors that look like oligodendrogliomas but have EGFR amplification lack 1p/19q codeletion and are extremely aggressive, meriting the term “small cell GBM” instead of oligodendroglioma.^{16, 32, 36, 68, 69} Parenthetically, EGFR amplification and p53 mutation are also nearly mutually exclusive in gliomas,²⁵ although rare cases of dual amplification and mutation have been documented.⁷⁰

Prognostic significance of EGFR amplification

While testing for EGFR amplification clearly aids the diagnostic side of neuropathology, the prognostic significance of EGFR amplification is ambiguous if one has already reached an unequivocal histologic diagnosis of GBM. Some studies suggest there is no survival difference, while others show worse survival; a few have even suggested an association with longer survival.^{25, 63, 67, 71–73} Part of the problem may be the common practice of pooling all EGFR-amplified tumors into one large group for analysis, even though the degree of amplification varies widely between tumors. For example, if one is determining EGFR status by FISH (the most common method), the number of EGFR signals and chromosome 7 centromeric enumeration probe (CEP7) signals is expressed as an EGFR:CEP7 ratio. But although most consider a ratio greater than 2 to constitute amplification, the ratio in any given case can easily exceed 20 or more. One might expect a worse prognosis with a higher ratio, but the reality may not be so simple: in a series of 542 GBMs at the University of Pittsburgh, GBMs with EGFR:CEP7 ratios over 20 had longer median survival compared to non-amplified GBMs, which in turn fared better than tumors with EGFR:CEP7 ratios between 2 to 20.⁷⁴ At first blush such a finding seems counterintuitive, it has been shown that higher ratios of EGFR:CEP7 correlate with more fragments of extrachromosomal DNA in the tumor nuclei (a.k.a. double minutes), whereas lower levels of amplification tend to be interstitial repeats of the gene within an otherwise more stable chromosome.⁷⁵ Since there is a precedence for “excessive” cancer genomic instability being more prognostically favorable than “moderate” instability,⁷⁶ perhaps high-amplifiers simply have more fragile genomes and are thus more sensitive to genotoxic therapy. Of note, the size of the 7p12 amplicon can also vary greatly from one GBM to another, sometimes including co-amplified “passenger genes” like LANCL2 and ECOP that themselves could greatly modulate tumor biology.^{77, 78} However, analysis using The Cancer Genome Atlas Data showed no correlation between amplification of either gene and survival.⁷⁴

Tests for EGFR amplification

As implied, the most commonly used assay to test for EGFR amplification is FISH, which is particularly advantageous when attempting to spot just a few scattered infiltrating amplified cells in an otherwise under-representative biopsy. Even if there is no reason to suspect EGFR amplification in a particular case, the CEP7 in the EGFR probe cocktail can act as a test for trisomy 7, which helps differentiate between astrocytoma and reactive gliosis.⁷⁹

Worth noting are the occasional cases of GBM that show only scattered cells with EGFR amplification yet, by both histology and CEP7 hyperploidy, clearly contain many neoplastic cells (Figure 2). Recent work has shown that such cases have heterogeneous amplification of other receptor tyrosine kinases besides EGFR, including MET and PDGFRA.⁸⁰ Such patterns of amplification could explain why anti-EGFR therapies have been disappointing in EGFR-amplified GBMs.

Figure 2.

Unusual EGFR amplification pattern in a GBM. By FISH, EGFR amplification in tumor nuclei is readily appreciated as an excess of red EGFR probe signals relative to green centromeric enumeration probe 7 signals. However, in this case the surrounding nuclei are also neoplastic, as evidenced by trisomy 7 (arrowheads).

Before doing EGFR FISH an effective initial screen is EGFR immunohistochemistry; gliomas that are negative-to-weak do not need FISH since amplification in such cases is extremely rare.⁸¹ This same immunostain is also quite useful in oligodendrogliomas, with moderate-to-strong expression being an unfavorable prognostic marker in grade II oligodendrogliomas and, paradoxically, a favorable marker in grade III anaplastic oligodendrogliomas even after adjusting for 1p/19q status.^81–83

Targeted therapy for glioblastoma with EGFR amplification

As mentioned, anti-EGFR drugs have been tried in GBMs.^84–86 Results have been mixed thus far, not only because of the heterogeneous nature of receptor tyrosine kinase amplification, but also in part because alterations in other downstream proteins can thwart EGFR inhibition (e.g. loss of PTEN and/or high pAkt).^{85, 87} As of now there is no data to suggest that GBMs with or without EGFR amplification should be treated differently. For the time being, then, EGFR testing is solely for diagnostic (and perhaps prognostic) purposes.

IDH1/2: Metabolism’s never been hotter

One of the most recent and exciting events in neuro-oncology has been the discovery of mutations in isocitrate dehydrogenase types 1 and 2 (IDH1/2) in gliomas, found via whole-genome sequencing in GBMs.⁸⁸ Both IDH1 and IDH2 convert isocitrate to alpha-ketoglutarate, in the process reducing nicotinamide adenine dinucleotide phosphate to NADPH. Mutations in these enzymes are clinically remarkable for many reasons.

• First, IDH1 or IDH2 mutations are present in about 80% of astrocytomas and oligodendrogliomas, and in 15% of GBMs (practically all of which are secondary GBMs).^88–90

• Second, they are always point mutations, restricted to the arginine codons encoding isocitrate binding sites for both enzymes (R132 in IDH1 and R172 in IDH2, with very rare R100, R49, and G97 mutations in IDH1).^{91, 92}

• Third, only one of the two alleles is mutated in a tumor, with the other allele remaining wild-type.

• Fourth, the similarly high rate of IDH1/2 mutation in both astrocytic and oligodendroglial tumors, and its tight association with TP53 mutations in the former and 1p/19q codeletion in the latter, suggests that it is an early step in gliomagenesis, perhaps in a glial cell precursor (Figure 3).^{31, 88, 93–96}

• Fifth, these mutations appear to be relatively specific to gliomas versus most other solid neoplasms,^{89, 97, 98} though about half of all cartilaginous tumors, some intrahepatic cholangiocarcinomas, a minority of melanomas and thyroid carcinomas, and very rare prostate and colon cancers also contain IDH1/2 mutations.^99–105

• Finally, high-grade gliomas that harbor IDH1/2 mutations are less aggressive than their grade-matched wild-type counterparts.^{88, 89, 95, 106–109} In fact, IDH1/2 mutation status may be the most powerful independent molecular prognostic marker identified to date in high-grade gliomas,^{22, 110} though whether it is the most powerful marker has been disputed.^{96, 111} One study even showed that grade IV GBMs with IDH1/2 mutations fared better than grade III anaplastic astrocytomas without the mutation, and that this mutation, preferentially occurring in younger adults, accounts for the well-known correlation of younger patient age with longer survival.¹¹² In contrast, 15 to 30% of acute myeloid leukemias have IDH1/2 mutations, with either no effect or an adverse effect on prognosis.^{101, 113–118}

Figure 3.

IDH1/2 mutations are early events in gliomas. The presence of IDH1 (or IDH2) mutations in most WHO grade II and III astrocytomas and oligodendrogliomas suggests that they arise in glial progenitors, with p53 mutation or 1p/19q codeletion occurring at a later step.

IDH1/2 mutations are practically always heterozygous, meaning that there is always one wild-type allele in the tumor. Likewise, mutation of both IDH1 and IDH2 in the same tumor is exceedingly rare.¹¹⁹ These findings indicate a dominant gain-of-function, which was proven when studies showed that mutant IDH1 enzyme no longer performs its regular function, but instead converts alpha-ketoglutarate to D-2-hydroxyglutarate (D-2-HG), which is normally present in only trace amounts but is greatly increased in mutant gliomas.^120–122 While the full consequences of D-2-HG are still being characterized (and beyond the scope of this discussion), one noteworthy effect appears to be promotion of a hypermethylator profile by competitively inhibiting alpha-ketoglutarate-dependent demethylases.^123–128

Interestingly, there exist very rare organic acidurias that usually contain defects in D- or L-2-HG dehydrogenase, with consequent elevation of D-2-HG or L-2-HG, respectively. Both types of acidurias include seizures and severe retardation in their phenotypes, but only L-2-HG aciduria patients have developed brain tumors, and not all are gliomas.¹²⁹ Even more provocative was the finding that D-2-HG aciduria is sometimes caused by a germline IDH2 mutation at R140, with the same biochemical result as the R172 mutation, yet to date no D-2-HG aciduria patients have been diagnosed with any kind of cancer¹³⁰ other than chondromatosis.¹³¹ Along those lines, it is remarkable that somatic mosaic mutations in IDH1/2 have recently been shown to be present in Ollier disease and Mafucci syndrome.¹³²

IDH1/2 mutations for differentiation among tumors

The role of IDH1/2 mutations and D-2-HG in gliomas will obviously be the subject of intense investigation for quite some time, but testing for the mutation has immediate practical relevance in surgical neuropathology. Because of its relative specificity in gliomas, it can be used to differentiate between reactive gliosis and an undersampled diffusely infiltrative glioma,^{79, 133–135} as well as to distinguish an incurable diffusely infiltrative glioma from a potentially curable mimicker like ganglioglioma or dysembryoplastic neuroepithelial tumor.^{136, 137} It has been proposed to help distinguish between a diffusely infiltrative glioma and a PA, the latter being negative for IDH1/2 mutations but usually positive for BRAF mutations or rearrangements (see below).¹³⁸ However, since IDH1/2 mutations are rare in diffuse gliomas from preadolescent patients,¹³⁹ and the rate of BRAF fusions in PAs to declines with patient age,¹⁴⁰ the utility of dual IDH/BRAF testing to differentiate the two is unclear. Another topic of ambiguity is the prognostic power of IDH1/2 mutations in grade II gliomas; some have shown the mutations to be favorable markers,¹⁴¹ but others have not.^{96, 142–144} Considering that true grade I gliomas do not have IDH1/2 mutations, one wonders whether some gliomas that appear to be grade II histologically but are IDH1/2 wild-type might actually represent “overgraded” grade I gliomas. If so, inclusion of such cases in survival analyses of wild-type grade II gliomas would mask the prognostic effects of IDH1/2 mutations in true diffusely infiltrative grade II gliomas.¹⁴⁵ At this point, it remains a matter of speculation.

Tests for IDH1/2 mutation

The gold standard for IDH1/2 mutation testing is obviously PCR amplification and sequencing of the tumor, which is quite feasible in routine FFPE tissues.¹³⁴ But while PCR and conventional Sanger sequencing will detect as little as 20% mutant alleles in a specimen, real-time PCR with fluorescence melting curve analysis (RT-PCR/FMCA) is faster and twice as sensitive.¹³⁵ The latter technique is predicated on the fact that even just one base pair mismatch between the fluorescent probe and the target sequence will lower the PCR product melting temperature, which can be detected by a change in fluorescence. In the case of IDH-mutant gliomas, two melting peaks will be produced, a higher peak for the wild-type allele and a lower peak for the mutant allele.

However, the development of an R132H-specific IDH1 mutant antibody has already rendered upfront PCR/sequencing unnecessary, to be used only if the immunostain is negative or equivocal. This is because about 95% of all IDH1/2 mutations are in IDH1, and among those over 90% are arginine-to-histidine R132H, making an R132H-specific antibody an excellent screening test.^{137, 146–148} So far a couple of monoclonal antibodies exist, with comparable efficacy.¹⁴⁹ One that is commercially available is very specific and robust in FFPE tissues, although at times the background staining can be rather high, including occasionally prominent edge artifact. Also, some cells can show spurious reactivity, most notably neurons and red blood cells (Figure 4); such artifacts are generally readily discerned from true positivity. In the author’s experience, a true positive reaction in an immunoperoxidase system should be solid brown staining, including clear visualization of tumor cell processes. (In fact, this author never appreciated just how long glioma cell processes could grow in vivo before looking at R132H antibody results.) Sometimes staining heterogeneity is seen, where some areas are nicely positive while other areas of obvious tumor are negative; such cases are still always positive for the R132H IDH1 mutation on sequencing.¹⁴⁹

Figure 4.

R132H IDH1 immunohistochemistry. (A) In a 30 year-old male with a right frontal brain lesion, superficial cortical tissue is nondiagnostic for infiltrating glioma. (B) R132H IDH1 immunostain, however, highlights scattered tumor cells with long processes. (C) Deeper areas of the same tissue showed unequivocal anaplastic astrocytoma, which was also positive for R132H IDH1 (D) as well as having strong nuclear p53 immunostaining (E). In a different case, neurons show a fair amount of nonspecific staining (arrowheads) but are still readily distinguished from infiltrating glioma cells that are strongly positive for R132H IDH1.

The practical aspects of screening all gliomas and suspected gliomas via R132H IDH1 antibodies are obvious:

1. Faster turnaround time

2. Lower cost

3. Ability to detect just a few scattered positive cells that even the most sensitive PCR tests would miss

Thus, upfront R132H IDH1 immunostaining is now done routinely on all our gliomas, including suboptimal biopsies suspected of harboring glioma cells. Any case that is immunonegative or equivocal is reflex-tested for less common IDH1 and IDH2 mutations by RT-PCR/FMCA. And although IDH1/2-mutations are often thought to be restricted to adult patients, as mentioned, they can be seen in some gliomas from pediatric patients.¹³⁹ Testing for IDH1/2 mutations is therefore advisable on all infiltrative gliomas or suspected gliomas, especially in patients approximately age 14 years and up.

BRAF rearrangement and V600E in pilocytic astrocytomas: Six of one, half dozen of the other?

Just like the discovery of IDH1/2 mutations, cutting-edge molecular technologies have shed additional light on PA biology. Conventional comparative genomic hybridization showed no large cytogenetic abnormalities in most of these tumors,¹⁵⁰ but array comparative genomic hybridization identified a biologically active, low-level copy number gain of the BRAF gene on 7q34 in a significant proportion of PAs.^{151, 152} BRAF is part of the raf/mil family of serine/threonine protein kinases. It is a proto-oncogene that participates in the MAPK/ERK signaling pathway, which regulates cellular differentiation, proliferation, and migration, among other things.¹⁵³ In PAs the BRAF copy gain is due to a tandem duplication producing a KIAA1549:BRAF fusion protein with loss of the Ras-binding domain on BRAF,^{154, 155} though a minority of cases can involve RAF1 fusion with SRGAP or FAM131B.^{154, 156, 157} Either way the Ras-binding domain in BRAF is lost, resulting in constitutive BRAF activity.

While constitutive BRAF activity is sufficient to induce PA-like tumors in xenografted mice, it eventually leads to tumor senescence.^158–161 In contrast, WHO grades II-IV gliomas also have some form of MAPK activation, but in those tumors it is accompanied by impairment of p53/Rb cell cycle pathways, which might explain why those higher-grade tumors undergo progression rather than senescence. In fact, loss of p16 in PAs may correlate with higher risk of aggressive behavior.^{158, 162}

BRAF fusions are present in nearly 70% of all PAs, compared to about 15% of all other low-grade gliomas in the differential (e.g. diffuse grade II astrocytoma, ganglioglioma, etc.).^{151, 152, 160, 163–167} They are hardly ever seen in high-grade pediatric gliomas.^{165, 167} There is a difference by location, as BRAF fusions are seen in approximately 75 to 80% of cerebellar PAs and 55% of noncerebellar (mostly supratentorial) PAs.^{82, 151, 152, 157, 160, 163–165, 167} The frequency of BRAF fusions appears to decline with patient age, from about 80% in the first decade to 50% in the second decade, with less than 10% of PAs in patients over 40 having the fusion.¹⁴⁰ About half of all pilomyxoid astrocytomas contain BRAF fusions,^{160, 163} underscoring the link between the two tumors but making a distinction based on BRAF fusion impossible.

Another BRAF alteration is the constitutively active V600E point mutation, previously known to be important in melanomas, colonic adenocarcinomas, and papillary thyroid carcinomas. It is present in roughly 15% of grade II–IV diffusely infiltrative pediatric astrocytomas^168–170 but less than 2% of comparable adult gliomas.^{169, 171–173} In both children and adults about 10% of PAs have the mutation (but only 2% of cerebellar tumors).^{152, 157, 160, 163, 165, 167, 169} Sometimes the mutation can occur in conjunction with a BRAF fusion in the same tumor.^{160, 162} Although present in tumors with PA-like morphology, V600E is more common in other tumors in the PA differential, including 25% of pediatric and adult gangliogliomas and 80% of pleomorphic xanthoastrocytomas.^{160, 167, 169, 173, 174}

Thus, over 75% of sporadic PAs have some sort of BRAF alteration, compared to about 40 to 50% of all other pooled tumors in the PA differential. Less than 10% of NF1-associated PAs contain a BRAF fusion or mutation, but such cases do occur.^{157, 160, 162, 175–177} From a diagnostics standpoint the presence of a BRAF fusion is therefore most suggestive of a low-grade glioma, especially a PA, but whether it can be used to definitively prove PA is still unclear. In contrast, detecting a V600E mutation is more associated with other tumors in the PA differential, but again cannot be used to definitively prove one type of tumor versus another.

From an outcome-based perspective, recent work showed that supratentorial low-grade pediatric gliomas with BRAF fusion usually behave as typical grade I PAs regardless of what they look like under the microscope or whether they are completely resected, whereas PAs without BRAF fusion are more likely to behave like grade II diffuse astrocytomas.¹⁶⁰ A similar trend was seen in PAs in the cerebellum.¹⁶⁴ In another study, 6 of 118 WHO grade II diffuse astrocytomas and oligodendrogliomas in children had a BRAF fusion; all 6 tumors were in the posterior fossa of pediatric patients and were extremely indolent despite their grade II microscopic appearance.¹⁶⁵ However, two studies that included pediatric low-grade gliomas from both the supratentorium and posterior fossa did not show BRAF fusion to be an independent prognostic variable on multivariate analysis.^{162, 178} Nevertheless, while there is not yet a consensus on the prognostic impact of BRAF fusion in gliomas, as of now it appears to be at least a neutral biomarker, and perhaps even a favorable marker in certain contexts. In contrast, data on BRAF V600E mutations and outcome are very sparse. Our recent work comparing both types of BRAF alterations in the same cohort of pediatric low-grade gliomas suggested that there is a trend toward divergence in prognosis—i.e. BRAF fusions tend to be associated with longer PFS, while BRAF V600E mutations suggest shorter PFS. However, both markers pale in comparison to tumor location, and by extension the degree of tumor resection, as independent prognostic variables.¹⁶²

Regarding BRAF-specific treatments, clinical trials aimed at blocking the downstream effects of BRAF fusion via MEK inhibition are still in accrual as of the time of this writing. Likewise, comparable anti V600E clinical trials in low grade pediatric gliomas are still in progress, but if animal model data and experience with melanomas is any indication, such therapies are worth pursuing.^{170, 179}

BRAF fusion and mutation detection

Detection of BRAF fusion is feasible via several strategies, including RT-PCR breakpoint analysis, fusion transcript RT-PCR assay, and a three-probe FISH cocktail that encompasses the entire BRAF gene (Figure 5A).^{180, 181} Using this cocktail, a normal cell will show two signals while a cell with BRAF rearrangement will also show a third (often smaller) signal near one of the other signals, corresponding to the additional activation segment and kinase domain ¹⁶⁴(Figure 5B). Such a pattern should be clearly seen in at least 10% of tumor cells^{138, 164} and will detect both common and uncommon rearrangements.¹⁸² A two-color probe cocktail containing a green probe (KIAA-1549) and red probe (BRAF RP4-726N20) has also been used, with a positive signal being yellow and indicating fusion of red and green signals.^{138, 160} Although there is good correlation between BRAF FISH, RT-PCR breakpoint analysis, and array CGH,¹⁶⁰ from personal experience the detection of BRAF fusion via FISH is visually challenging, especially if the tumor has high overall ploidy. At this time the optimal method is probably RT-PCR aimed at detecting breakpoints or fusion transcript.¹⁶⁵ FISH could still be used if there is not enough tissue for PCR analysis.

Figure 5.

BRAF rearrangement can be detected by FISH. (A) Schematic of a three-probe cocktail spanning the entire BRAF gene [adapted with permission from Ciampi et al,, J Clin Invest. Jan 2005;115(1):94–101]. (B) Abnormal BRAF is detected by the presence of an excess red signal relative to centromeric enumeration probe 7 (green).

Naturally the standard way to detect BRAF V600E point mutation is via PCR and sequencing, but recently the same group that developed the aforementioned R132H IDH1-specific antibody has produced a V600E BRAF-specific antibody.¹⁸³ While the antibody will not work in tissues subjected to surgical coagulation or pathologic freezing, and staining intensity varies greatly between mutated cases, it does appear promising as a first-line screening test.

When to test for BRAF

In summary, testing for BRAF fusion and mutation is recommended in cases where the differential is between a PA and one of its mimickers, especially if there is concern that the tumor could be a grade II–IV diffusely infiltrative glioma. The presence of the fusion or point mutation, however, might not unequivocally diagnose one type of tumor versus another, and is less useful in non-pediatric patients. Similarly, the absence of a BRAF abnormality does not rule out a grade I glioma. Reflex testing of all gliomas for BRAF alterations is therefore not yet recommended, though this could quickly change depending on additional data, especially clinical trials.

Final thoughts on molecular pathology for gliomas

Strictly speaking there is no predictive molecular biomarker in gliomas because, unlike Herceptin and HER2 in breast cancer, no therapy has yet been developed that is only effective against tumors with a particular marker. But in certain circumstances, testing for 1p/19q codeletion, EGFR amplification, IDH1/2 mutations, and/or BRAF alterations can help refine the patient’s diagnosis and prognosis (Table 1, Figure 6). Wise and cost-effective use of these markers is still predicated on histologic evaluation of the tissue, as is WHO grading. Yet based on the rates of EGFR amplification and IDH1/2 mutations in the adult population, as well as the frequencies of grades II–IV gliomas, tumor cells can now be detected in about 60% of suboptimally-biopsied diffusely infiltrative gliomas, even in cases where the diagnosis of glioma cannot be made based on light microscopy alone.^{49, 79, 133, 134}

Figure 6.

Example of a possible diagnostic algorithm incorporating molecular biomarkers in the workup of gliomas. Judicious use of 1p/19q testing, EGFR FISH, IDH1/2 mutation screening, and BRAF analysis requires a logical series of decisions based on clinical and histologic factors. In most circumstances the best use of molecular tests are to support a favored diagnosis or help the neuropathologist get “off the fence” in difficult cases. The tests can also refine prognosis or detect scattered glioma cells in undersampled tumors. Of note, patient age plays a large role in determining which tests are likely to help (see text for details). See Table 1 legend for abbreviations.

Array-based technologies that screen for copy number aberrations across the entire genome are likely to gain in popularity as their cost goes down and their ability to interpret FFPE tissues goes up.¹⁸⁴ FFPE-based platforms for screening mutations in multiple genes is also promising.¹⁸⁵ Implementation of both copy number and mutation platforms may someday obviate the need for parsimonious, sequential targeting of specific biomarkers. But for the foreseeable future, and especially when tissue is inadequate for arrays, microscopic analysis of the venerable H&E-stained slide will be essential for selective testing of specific molecular biomarkers, both old standbys like 1p/19q and EGFR as well as new ones like IDH1/2 and BRAF.