A neuropathology-based approach to epilepsy surgery in brain tumors and proposal for a new terminology use for long-term epilepsy-associated brain tumors

1Virtually any brain tumor can cause seizures. This review will focus on a particular group of brain tumors that usually manifest with seizure onset during early life (mean age = 16.5 years; Table 1), and which present with a broad histopathological spectrum of low-grade glial and glio-neuronal phenotypes. Herein, we designate this group as long-term epilepsy-associated brain tumors (LEATs or epileptomas). As in any other patient with a brain tumor, LEATs are readily detectable with standardized neuroimaging techniques (Figs. 1, 2). In contrast to most gliomas, slow LEAT growth and low risk for malignant progression may not necessitate immediate surgical intervention, nor will surgical strategies help to achieve long-term seizure control when aiming only at gross tumor resection. A common goal for successful treatment of these patients is the accurate identification of the epileptogenic zone, which may or may not match with the MRI visible lesion. Advanced neurophysiological procedures including invasive EEG recordings may be needed in some patients, as discussed further below. However, weak agreement in the histopathological diagnosis of LEAT tumor entities challenges any meaningful interpretation of published patient series. A major reason for poor agreement is the large spectrum of morphological variants when reviewing routinely stained hematoxylin and eosin (HE) sections, sharing one or more of the following features: (1) LEATs have a histologically variable appearance consisting of dysplastic neuronal and neoplastically transformed glial elements, mostly classified as glio-neuronal tumors by the World Health Organization (WHO). (2) The vast majority of LEATs correspond to WHO Grade I. Reliable guidelines for the identification of tumors that carry a higher risk for recurrence and malignant progression are not available. (3) LEATs occur predominately in the temporal lobe and present with early seizure onset (with a mean age at seizure onset = 16.5 years). However, the average duration of epilepsy before surgical treatment adds another 12 years in our series of 1,551 patients (Table 1; Fig. 3b). (4) LEATs are likely to occur during brain development. As such, tumors can be associated with Focal Cortical Dysplasia (FCD ILAE Type IIIb), or small tumor satellites infiltrating the adjacent neocortex (see chapters below). Which tumor entity preferentially follows which of these peculiar patterns, and whether any of these features contribute to enhanced epileptogenicity are important and yet unanswered questions in need of clarification. (5) LEATs do not share molecular features typically observed in diffusely infiltrating gliomas, such as IDH1 mutations or 1p/19q deletions. In contrast, the oncofetal marker protein CD34 can be frequently identified and developmental genes are likely to be involved. Mutations in B-RAF or mammalian target of rapamycin (mTOR) signaling have been also identified as key features in this group of tumors (see chapters below).

2It is difficult to microscopically describe and histopathologically classify the broad spectrum of LEAT entities and their variants (Table 2) using standard eosin and hematoxylin (H&E) stains, which produce different histomorphological classification schemes of LEAT subtypes, i.e., the four different variants of a dysembryoplastic neuroepithelial tumor (DNT) with diffuse and simple forms, as well as complex and non-specific variants. Other strategies have tried to incorporate rare or unusual subtypes into well-introduced WHO tumor entities, such as the GG. As a matter of fact, meta-analysis of any published LEAT series is almost impossible and generates inconsistent results, despite adhering to the existing WHO classification system for brain tumors. In a previous review of eight international LEAT series (covering a total of 2,055 patients), we identified DNT and GG as the most prevalent entities. However, the reported percentages for both categories showed a huge variance between published patient cohorts ranging from 7 to 70 %. These figures suggested a geographical bias, by which same tumors were classified either as GG or DNT, reflecting differences in neuropathology schools rather than reliable histopathological signatures. Again, the controversial and enigmatic question whether malignant progression can occur in a DNT (comprising any of its described variants) is a good, albeit unpleasant, example.

3Increasing availability of surgical tumor specimens should open the possibility to better characterize the molecular signature of each LEAT variant along with their molecular pathogenesis and epileptogenic potential. Notwithstanding, such studies will require use of a reliable terminology and histopathological classification that can be reproduced by any other laboratory. Prospectively designed randomized controlled trials for LEAT treatment are necessary to provide class 1 evidence for any suggested biomarker and classification scheme, an important goal that has never been addressed or realized up to this day.

4The benefit of tailored resection strategies in patients with drug-resistant temporal lobe epilepsy is also recognized for the treatment of LEAT patients. The comprehensive database of the European Epilepsy Brain Bank (EEBB) currently includes 5,842 samples, of which one quarter is diagnosed as LEAT (Table 1). With mean epilepsy duration of 11.8 years (Table 1), many tumors appear to escape detection in patients with early seizure onset or are medically treated for a long period of time period before surgery is considered as “ultima ratio”. Due to the preferential localization of LEATs in the temporal lobe, the various consequences of long-term epilepsy on cognition as well as social development and behavior require careful consideration and seem to have already shifted the attitude of many epileptologists, neurologists and neurosurgeons toward earlier surgical intervention (Fig. 3b).

5Neuropathological examination and diagnosis rely on microscopical inspection of surgical brain specimens and follow the current WHO classification and grading scale (last revised in 2007). This classification scheme has proven useful for the prediction of the biological behavior of most gliomas and other CNS tumor entities. However, the broad spectrum of LEATs and their variable histomorphological features are not fully reflected within the current WHO grading system. Our EEBB collection of 1,551 LEAT cases included more than 15 different tumor entities in 709 female and 821 male patients (Table 2). Consequently, it is important to emphasize that proper neuropathological evaluation should be obtained from experienced centers. The need for such expertise is mandatory, given the considerable variability of histopathological phenotypes, which might result in “over-interpretation” of tumor progression leading to erroneous use of more aggressive therapeutic measures, even though most LEATs tend to have a very modest clinical behavior in the long run without bold risk of recurrence or malignant transformation. At the same time, some tumors with a histologically typical, “benign”, glio-neuronal phenotype have been reported to rapidly turn into malignancies, underscoring the need for reliable biomarkers that could be used to predict the biological behavior of each individual tumor.

6Given the broad histopathological spectrum of LEATs and the contradictory results published in the literature, we need to pursue better definitions and use standardized parameters for the neuropathological diagnosis of tumors associated with early onset epilepsy (Epileptomas). Such a proposal would also help reinforce research strategies toward a better understanding of the underlying biological nature (i.e., molecular pathogenesis), risk of malignant progression (i.e., predictive biomarkers), their epileptogenic potential (including presence or absence of FCD), and would allow for meaningful comparisons among published research studies and clinical patient series (i.e., epidemiological measures). Notwithstanding, there are no prospective, randomized, controlled trials of LEATs inpatients with epilepsy that demonstrate any better interrater agreement and superior predictive value of an alternative classification scheme. On the other hand, neuropathologists have already gone through these long and controversial discussions and disagreements 20 years ago with regard to the classification of malignant gliomas or medulloblastomas. As a result, the neuropathology community has now attained major achievements in specifying useful molecular diagnostic approaches and targeted therapies in these brain tumors. It seems that we are 20 years behind, when it comes to LEATs and associated epilepsies, and urgently need to shift our attention toward developing and implementing a new, clinically applicable consensus terminology for the neuropathological diagnosis of LEAT.

7Given that the spectrum of histopathological changes of most glio-neuronal tumors is highly variable and cannot be sufficiently characterized by H&E staining alone, we recommend the introduction and routine use of a selected panel of immunohistochemical stainings that would allow neuropathologists to generate an appropriate differential diagnosis of LEATs as well as their distinction from other gliomas, i.e., CD34 and MAP2 (Figs. 4, 5, 6). According to published experience with these immunohistochemical markers, we propose a terminology, which would allow us to classify the majority of tumors that fall under the entire LEAT spectrum. As an example, the frequently used term “Ganglioglioma” could be replaced by two new terms based on the presence or absence of the CD34 oncofetal class II epitope. The proposed terms are: basic (oncofetal) neuroepithelial tumor “BNET” with CD34 expression, compared to a (predominant) gangliocytic neuroepithelial tumor “GNET” without CD34 expression (Fig. 4). This new terminology would likely help avoid the confusion that characterizes previous reports, and which have included a broad histopathological spectrum of gangliogliomas and DNTs associated with various degrees of gangliocytic components. BNETs will always be devoid of MAP2 labeling within their glial cell component (Fig. 4; Table 3). The MAP2 epitope of interest is an alternatively spliced transient (embryonic) isoform including exon 13 (MAP2E). This isoform is likely to occur also in astroglial precursors and can be detected in the vast majority of diffusely infiltrating gliomas. We consider this as important evidence for excluding any semi-malignant tumor mimicking LEAT variants, i.e., if only fragmented surgical specimens are available for diagnostic review and evidence for entrapped preexisting neuronal subpopulations is difficult to obtain. Mutation-specific antibodies directed against the IDH1 enzyme are important markers highly recognized in diffuse gliomas, and should be also encountered in the histopathologic LEAT work-up panel to identify and differentiate diffuse glioma variants.

8The microscopic appearance of BNETs shares the broad spectrum of gangliogliomas described in the literature with dysplastic neuronal and neoplastically transformed glial components (Fig. 4). BNETs will frequently reveal calcification, lymphocytic infiltrates and protein droplets. The dysplastic neuronal component may be difficult to identify when searching for multinucleated neurons, which requires additional immunohistochemical measures, i.e., synatophysin, NeuN or MAP2. Few BNET variants can thereby present with a predominant glial cell population (astroglial or with clear cell morphology), which need to be distinguished from diffusely infiltrating astrocytomas or oligodendrogliomas using appropriate immunohistochemical markers, such as IDH1 and MAP2.

9The growth pattern of BNETs may be nodular but diffuse infiltration into adjacent neocortex and white matter is also frequently encountered. This growth pattern becomes even more evident using CD34 immunohistochemistry, as the class II CD34 epitope is not expressed in the adult normal mammalian central nervous system or any other neuroimmunological or neurodegenerative disease. In BNETs, CD34 immunolabeling presents with three common patterns (Fig. 4). Type I labels bulk tumor cells (Fig. 4e), Type II diffusely infiltrating tumor cells (Fig. 4h) and Type III what we called “tumor satellites”. These are small clusters or nodules of CD34 immunoreactive cells in remote cortical areas. These may not be detectable by H&E staining, and often cause misinterpretation as FCD Type IIIb. Such satellites can be immunohistochemically identified even if small or fragmented surgical specimens are available for histopathological review. However, described patterns for CD34 immunoreactivity may not be simultaneously present in each tumor specimen. Some BNET will reveal all three patterns, which usually require an anatomically well-preserved surgical specimen. On the other hand, BNETs may only show Type II or III patterns, in particular if fragmented specimens are available. However, their presence is sufficient to classify them as BNET. So far, no other disease condition or normal brain cell population has been shown to express the class II CD34 epitope. Unfortunately, the nature and origin of CD34 immunoreactive cells have not been clarified in BNETs, but most likely present neoplastically transformed neural precursor cells, although normal precursor cells or any other normal adult cell population in the human brain were never shown to express the CD34 class II epitope.

10A challenge for the differential diagnosis is to distinguish BNETs from other brain tumors expressing CD34, such as pleomorphic xanthoastrocytomas. BNETs may also share cytological features of pilocytic astrocytomas and oligodendrogliomas, which need careful consideration (Figs. 4, 5). Immunohistochemical stainings using IDH1 and MAP2 will help to clarify their differential diagnoses.

11The term “DNET” and its current WHO definition will remain unchanged in this proposal. As originally described by Daumas-Duport, DNETs exhibit the characteristic multinodular appearance and specific glio-neuronal element as most prominent histopathological features. These tumors should always be devoid of CD34 and IDH1 immunoreactivity (Fig. 5). MAP2E can be present in the neuronal and glial component of DNETs. Diffuse or non-specific forms, characterized by a clear cell morphology,, are not encountered in the proposed A–B–C scheme. Such entities will be classified according to their CD34 immunoreactivity profile, i.e., most tumors with diffuse oligodendroglial-like clear cell components will express CD34 and should be encountered into the “BNET” category. Complex forms of DNETs are already recognized by the 2007 WHO classification system, and will now be subsumed into the group of composite neuroepithelial tumors, “CNETs”. They are characterized by the co-existence of at least two distinct LEAT entities. Histopathological diagnosis should always include specification of tumor variants, i.e., CNET (DNET/BNET) (DNET/GNET) or any other possible configuration. However, composite tumors containing semi-malignant components, such as PXA, should be graded according to the highest degree of malignancy, and not classified as CNET.

12The term “ANET” is synonymous to the angiocentric glioma and was introduced into the WHO classification system in 2007. The term “ANET” and its definition will remain unchanged, and reserved for those LEATs that are characterized by the peculiar growth pattern, and ependymoma-like differentiation displaying intracellular perinuclear EMA-immunopositive dots. Up to now, there is no report of CD34 immunoreactivity or IDH1 mutations in patients with ANETs, which matches with our experience. However, our series of ANETs also comprise tumors with a distinct neuronal component as well as variable other cytological morphologies. We prefer, therefore, to re-introduce the original proposal of Lellouch–Tubiana and describe them as “neuroepithelial” tumors rather than pure “gliomas”. This proposal will need also further consideration by the WHO panel and finally awaits molecular diagnostic confirmation.

13We would like to propose two new A–B–C LEAT entities. The “INET” is an isomorphic neuroepithelial tumor previously described as isomorphic astrocytoma variant corresponding to WHO Grade I, which lacks any CD34 and glial MAP2 labeling and showed benign long-term outcome without recurrence or progression for more than 13 years. In addition to the initial description of six cases in 2004, the EEBB has collected 29 of these stereotypically composed low-grade tumors (Table 2; Fig. 6). Their histomorphological characteristic shares a monomorphous astroglial appearance with low cellular density. Nuclear atypia and mitosis are absent. The tumor matrix is fibrillary and reveals intense labeling with antibodies directed against GFAP. This tumor can infiltrate adjacent cortical and archicortical structures and is usually visible on MRI as non-contrast enhancing mass lesion with homogenous signal increase on T2- and signal decrease on T1-weighted images (Fig. 1).

14Any other epileptogenic neuroepithelial tumor with early seizure onset that does not match the proposed A–B–C terminology will be provisionally classified as “ENET—Epileptoma/NOS”. These tumors will most frequently share the histomorphological H&E spectrum of BNET but do not express CD34. They also lack other distinguishing histopathological signatures for their classification described here as ANET, DNET, GNET or INET.

15The current proposal did not address any criteria of atypia or anaplasia as it is meant to start an interdisciplinary discussion about strategies how to better define LEAT entities in the near future, i.e., upcoming WHO classification systems. Long-term follow-up studies, preferentially obtained from prospective observational multi-center trials, will be mandatory to validate the clinical benefits of such new definition, i.e., with respect to tumor recurrence, malignant progression and postsurgical seizure control. These data cannot be retrospectively obtained from our EEBB and will need careful reevaluation from submitting centers. In addition, large-scale molecular genetic studies should be carried out to clarify signatures distinguishing A–B–C tumors from any other brain tumor entity. As already emphasized, existing literature is not helpful due to the inconsistent use of histopathological LEAT terminology.

16A promising approach to overcome the difficult distinction of LEAT variants will be to identify specific molecular signatures, which is also helpful to clarify their histogenesis and epileptogenicity. Several recent studies have been performed and will be briefly summarized. However, the distinction between GG and DNT was often not used with same or comparable definitions and terminology (as discussed above) and all data will be summarized, therefore, as LEAT instead of what have been published to represent specific GG or DNT characteristics.

17Analysis of chromosomal copy number alterations (CNAs) in a cohort of 61 patients with LEAT (specified as GG without reference to CD34) revealed gains of chromosomes 7 and 5 as common aberrations, the last being more frequent in LEAT compared to astrocytomas WHO grade II. More recently, copy number profiling of a large cohort of 131 LEAT confirmed the occurrence of gains of chromosomes 5 and 7 in both GG and DNT subtypes, detecting different patterns of chromosomal alterations, including also somatic intra- and/or interchromosomal chromothripsis (Prawabo et al. personal communication). These observations suggest that LEATs share similar molecular features despite their large spectrum of morphological variants. However, this conclusion needs to be confirmed using any new proposal for tumor terminology. Loss of heterozygosity (LOH) of chromosomal arms 1p and 19q has been repetitively recognized and confirmed as molecular hallmarks in oligodendrogliomas, but rarely reported in LEATs, representing an additional diagnostic feature in cases in which the small size of the biopsy raises differential diagnosis of oligodendrogliomas. We also strongly recommend the evaluation of IDH1 and IDH2 gene mutations as useful tool for the differential diagnosis between WHO I° LEAT and WHO II° gliomas. IDH1 mutations can be recognized using a mutation-specific antibody directed against the most common mutation site in WHO II° gliomas, but generally absent in LEAT.

18The combined molecular analysis of IDH1/IDH2 and BRAF mutations may represent the most powerful diagnostic tool in the evaluation of LEAT. A mutation of the BRAF oncogene (V600E mutation) has been recently shown in different entities within the LEAT spectrum, including “GG”, “desmoplastic infantile gangliogliomas” and “DNT”. The presence of BRAF V600E mutations can be evaluated by direct DNA sequencing and by BRAF V600E immunohistochemical detection. However, the low cellularity of these tumors and the frequent admixture of normal cell components may produce false-negative findings. Accordingly, immunohistochemical studies emphasize the importance of performing DNA sequencing from macrodissected tissue, or even to apply single-cell laser-capture microdissection. The identification of mechanisms underlying tumor development in BRAF wild-type LEAT cases represents a major challenge for future studies, which also requires a reliable terminology and histopathological classification of these tumor entities. Interestingly, evaluation of BRAF mutation in a large cohort of both pediatric and adult LEAT reveals a positive correlation with the oncofetal marker protein CD34. These studies also support the pathogenic role of BRAF in low-grade glial tumors arising in young age groups and including entities, such as pilocytic astrocytomas and pleomorphic xanthoastrocytoma. A remaining issue is represented by the possible prognostic value of the BRAF status on recurrence-free survival and postoperative seizure outcome, which will also require further evaluation.

19The search for additional mechanisms involved in LEAT development and epileptogenicity has identified the mTOR signaling pathway as promising candidate. The mTOR pathway acts as key regulator of cell size and growth control, proliferation, differentiation and survival during brain development, and increasing evidence supports the role of the TOR pathway in a wide variety of neurological disorders including both MCD and brain tumors. Enhanced mTOR pathway activation has been reported in LEAT, such as GG and DNT, suggesting a pathogenic link between these tumor entities and focal MCD, such as FCD ILAE Type IIb and cortical tubers in tuberous sclerosis complex (TSC),. Mutational analysis of TSC1 and TSC2 has been performed in LEAT, but failed to identify mutations. A somatic mutation was reported in intron 32 of the TSC2 gene in glial cells of one LEAT patient, but not in dysplastic neurons. Thus, additional efforts are necessary to clarify the mechanism underlying the deregulation of mTOR in LEAT.

20Understanding the mechanisms that underlie epileptogenesis in LEAT is essential to develop effective treatment in young patients in which drug-resistant epilepsy critically affects their daily life. Over the last decades, several hypotheses have been put forward to explain epileptogenesis in LEAT patients. These studies suggested the involvement of both tumor-related factors (i.e., tumor size, tumor location, cellular composition), as well as of peritumoral changes (i.e., hypoxia and acidosis, ionic and enzymatic changes, deposition of hemosiderin;). A better definition of the molecular biology of LEAT may not only aid the development of a more targeted treatment involving specific pathogenic pathways, but may also contribute to explain the epileptogenic nature of specific LEAT entities, such as GG (i.e., BNET, GNET) and DNT.

21Clinical management of patients with epilepsy due to brain tumors exemplifies an important area of conflict between medical specialities, i.e., neurooncology and epileptology. As a matter of fact, many brain tumors manifest with seizures and will not require attention by specialized epileptologists. Patients showing clinical symptoms for increased intracranial pressure with/without seizures will have brain imaging to localize a suspected brain tumor. Most centers will then move forward to neurosurgical resection to reduce mass effects and to obtain a written histopathological report guiding further treatment options. This applies for most adult or elderly patients suffering from rapid tumor growth in cortical or subcortical areas. But we need to consider the red flags, which require a different approach and careful neurophysiological attention. As prominent example, children or adolescents with a brain tumor manifesting in the temporal lobe should be carefully examined, as these neoplasms usually belong to the spectrum of LEAT and not to semi-malignant or malignant gliomas. The benign nature of most LEATs and their slow cell growth will not require surgical treatment or any adjuvant radiation or chemotherapy. Instead, control of spontaneous seizures should be the primary treatment target to reduce the risk for progressive cognitive impairment or adverse effects of medication. This can be exemplified by a mean disease onset of LEATs at 16.5 years (Table 1) and temporal lobe location in 80 % of our LEAT series. These patients should be consulted by a comprehensive and multi-disciplinary epilepsy center. Specialized expertise in reading neuroradiology images may already help to clarify the underlying nature of a given LEAT (Figs. 1, 2). If medical treatment does not achieve sustained seizure control, surgical resection should be recommended as primary and in many instances also curative treatment option, although a controlled and randomized clinical trial was performed so far only for surgical treatment of drug-resistant temporal lobe epilepsy. If epilepsy surgery is considered in patients suffering from brain tumors, we have to anticipate two distinct clinical scenarios: (1) Tumor is located in non-dominant and non-eloquent cortical and subcortical areas. We do not need invasive monitoring in this situation. A generous resection involving the tumoral area plus adjacent cortex can be performed, guided by intraoperative electrocorticography (ECoG) if available. (2) Tumor is located in or in close proximity to eloquent cortical and subcortical areas. As discussed above, the MRI-identified tumor may or may not be intrinsically epileptogenic and perilesional cortex may also contribute to the generation and early propagation of the epileptic seizures. In this clinical scenario where there is an anatomical proximity with functional areas in the brain, such as a dominant hippocampus, Brocca’s speech or prefrontal motor areas, the anatomical boundaries between the epileptogenic zone, the tumoral area and functional cortical and subcortical areas can be very poorly defined if based only on non-invasive data. In this particular situation, invasive monitoring is strongly recommended.

22There is, however, no agreement regarding optimal surgical strategies, and studies comparing lesionectomy with more extensive resection of the presumed peritumoral epileptogenic zone are lacking. Using not only a tumoral, but also an epilepsy surgery-oriented strategy, LEAT-associated epilepsies usually have excellent seizure outcome following appropriate surgical resection. Timing of surgical treatment is debatable in patients whose seizures are well controlled. When perceived benefits of surgical intervention are thought to overweigh any surgical or neurological risks, surgery can be offered early. In the decision making, one should also take into account the rare risk of tumor growth and malignant transformation, the real risk of later development of pharmaco-resistance, the potential side effects of long-term treatment with antiepileptic medications and the favorable outcome of surgical intervention, which of course depends on the location of the LEAT and its proximity to essential eloquent cortex.

23The predilection of LEATs to involve the temporal lobe (and in particular the anterior, basal and mesial compartments of the temporal lobe) has direct implications in the presurgical evaluation and tailored surgical therapy, which may or may not include the hippocampus in the area of resection. LEATs involving or abutting the mesial temporal structures are usually associated with a more widespread epileptogenic network and should be regarded as a distinct electroclinical group. Moreover, LEATs can be associated with focal cortical dysplasia ILAE Type IIIb and, therefore, a “curative” lesionectomy has to be extended beyond the MRI-identifiable lesion. Tumor recurrence and malignant transformation are rather unusual, but should never be excluded, especially when dealing with certain histopathological subtypes. In these instances, the focus of the neurology and oncology team is to not only limit tumor growth but also eliminate seizures, which can significantly impair quality of life. One cannot overemphasize the importance of appropriate surveillance plans following successful resection of such tumors. Lastly, controlled studies are needed to examine when it is appropriate to discontinue AEDs in patients with sustained seizure-freedom following surgical resection.

24In conclusion, this review and our proposal of the A–B–C terminology are meant to initiate an interdisciplinary consensus discussion between neuropathologists, neurooncologists, epileptologists and any other medical and research discipline toward a better understanding of each tumors biologic behavior and successful medical treatment in patients suffering from drug-resistant focal epilepsy.