According to WHO 2007 classification, main histological variants include classic, cellular, clear cell, papillary, tanycytic, and myxopapillary ependymomas [16]. As for meningioma, some histological subvariants are correlated with the WHO grade, for instance, myxopapillary and subependymoma are grade I. Of note, in the pediatric population, the tanycytic and myxopapillary ependymoma variants are exceptional and classic and clear cell ependymoma (grade II or III) are the most frequent (Figs. 23.1, 23.2, and 23.3). Clear cell ependymoma is more often observed in supratentorial location.

The current WHO 2007 distinguishes grade II from grade III ependymomas (anaplastic ependymoma) (Fig. 23.4), but important uncertainties about the prognostic value of this grading still persist, especially among children. Tihan et al. have performed a meta-analysis on 24 pediatric intracranial ependymoma studies using the WHO grading system between 1990 and 2005 [24]. Twelve studies found no impact of grading criteria or grade on survival, whereas five found a statistical significant negative impact of high grade on OS. The remaining studies found a variable prognostic impact in terms of EFS or PFS. The ratio of grade II versus III is strikingly heterogeneous in the different cohorts (ranging from 7 % to 89 %), highlighting the numerous potential selection biases in age, location, tissue sampling, and interobserver discrepancies in the application of the WHO grading scheme [10]. The literature on the issue remains controversial due to multiple factors: retrospective monocentric studies extending over a long period of time, mixture of hybrid adult and pediatric cohorts, inclusion of ependymoma from different locations, and inclusion of grade I myxopapillary and subependymoma variants. Furthermore, all these confounding factors are not always appropriately considered in multivariate analyses, particularly in this disease in which clinical characteristics as age and extent of surgery have major impacts on survival. From a histopathological standpoint, major difficulties stem from the absence of precise WHO guidelines, a variable definition of grading criteria (particularly extensive or focal anaplastic areas, cellularity, necrosis) and different cutoff values for the mitotic rate.

Currently, WHO grading criteria used to define an anaplastic ependymoma are “high mitotic activity often accompanied by microvascular proliferation and pseudopalisading necrosis” [16]. The exact definitions and thresholds are not yet clearly established. In general, most studies have identified the parameters of cellular proliferation as significant for prognosis [8, 13, 17, 27]. Ellison et al. have proposed a cutoff value for mitoses of more than 5 mitoses per 10 HPF, associated or not with microvascular proliferation and taking into account the presence of hypercellular nodules [7]. Tihan et al. used a mitotic rate greater than 10 per HPF to identify grade III lesions but failed to find a prognostic correlation between grade III and OS [24].

The development of limited, well-defined, reproducible histological criteria associated with simultaneous biomarkers might be a useful approach to optimize the WHO grading scheme and to identify patients who could benefit from additional treatment.

Immunohistochemical stains in ependymoma generally show strong cytoplasmic positivity for glial fibrillary acidic protein (GFAP), vimentin, S-100, and CD56. Perivascular pseudorosettes are GFAP positive and typically express S-100 and vimentin, whereas true rosettes are largely GFAP negative. EMA immunostains are variable from case to case, usually highlighting dot-like punctuate perinuclear staining, ringlike intracytoplasmic microluminal positivity, or less frequently a linear staining along the luminal surface of ependymal rosettes. Dot-like EMA positivity is neither necessary nor specific to the diagnostic of ependymoma but remains a strong positive argument. Olig2 is not expressed by ependymal cells or is limited to scattered positive nuclei. Negativity for Olig2 represents a good criterion to eliminate the diagnostic hypothesis of other gliomas mimicking ependymoma such as an astrocytic tumor or oligodendroglioma [30].

Architectural features such as clear delineation from surrounding nonneoplastic tissue and perivascular pseudorosettes are robust diagnostic criteria, but the frequent heterogeneity of classic ependymoma (cellular and paucicellular areas) as well as the existence of clear cell and papillary variants sometimes lead to challenging diagnostic problems.

A common diagnostic issue in classic or cellular ependymoma is the distinction from another malignant glioma, glioblastoma (GBM), since this entity also presents a perivascular tropism. Features identifying GBM are extensive pseudopalisading necrosis, absence of dot-like EMA positivity, extended Olig2 positivity, and more importantly peritumoral diffuse infiltration. Ependymoma particularly in the spinal cord or in the fourth ventricle must be distinguished from pilocytic astrocytoma (PA) since this tumor also displays perivascular formations and a frequent clear cell component. Features identifying PA are intratumoral Rosenthal fibers, extensive Olig2 positivity, and the absence of dot-like punctuate EMA positivity.

Clear cell ependymoma can be confused with oligodendroglioma, central neurocytoma, and metastatic renal carcinoma since all these entities comprise neoplastic uniform round cells with encircling clear cytoplasm. Features identifying clear cell ependymoma are perivascular pseudorosettes even if these present only focally and GFAP and vimentin positivity associated with negativity for Olig2 (or with some rare positive nuclei), NeuN, and synaptophysin.

Papillary ependymoma could be confused with choroid plexus tumors and metastatic papillary carcinoma, as this ependymoma variant is characterized by epithelial-appearing cells supported by fibrovascular stalks. Ependymoma, as opposed to choroid plexus papilloma, lacks the PAS-positive basement membrane and is widely positive for GFAP. Ependymoma exhibits much weaker staining for keratins, such as CAM5.2, CK7, and CK18 [26].

It is also important to precise that the presence of foci of true ependymal differentiation in a CNS tumor is not synonymous of an ependymoma. For example, ependymal features may be encountered in mixed neuronal-glial tumors such as dysembryoblastic neuroepithelial tumor, ganglioglioma, or angiocentric glioma [5, 14, 15, 21, 25] and also in embryonal tumors such as atypical teratoid/rhabdoid tumor, PNET, and embryonal tumor with abundant neuropil and true rosettes (ETANTR) [6]. Exceptional childhood neuroepithelial tumors can display ependymal differentiation such as chordoid glioma of the third ventricle or papillary tumor of the pineal region [14].

Cooperative studies – necessary to achieve an adequate cohort, sufficient for a robust statistical analysis in this relatively rare disease, and associated with the recent technical progress in genetics and transcriptional profiling – have permitted the identification of ependymoma molecular subtypes. Location stands out as a major driver for the biology of ependymoma [2, 23]. Studies on biomarkers have consistently confirmed the major differences between ependymomas in children and adults, well described also on a clinical basis, as besides different preferential location (supratentorial and spinal in adults, posterior fossa in children); ependymomas in children have been largely shown to have a worse prognosis [10].

In comparison to adult counterparts, pediatric ependymomas display a higher frequency of 1q gain (20 % vs. 8 %, p = 0.0040) which has been linked to an adverse prognosis in several retrospective cohorts and more recently in two infant and young children trial cohorts [11].

A microarray study on a mixed adult and pediatric population has shown two different prognostic groups in posterior fossa ependymoma, according to their gene expression profiles [29]. Group A showed worse outcome, more frequent location in the cerebellopontine angle and cerebellar invasion, younger age, and overexpression of laminin a2 (LAMA2) or tenascin C (TN-C); group B was composed mainly of adolescent/adult patients with intraventricular and intramedullary tumors overexpressing NELL2 and genes involved in ciliogenesis or microtubule assembly. The 5-year progression-free and overall survival was significantly worse for group A (44 vs. 75 %; p = 0.017 and 65 vs. 95 %; p = 0.0048). These data have been partially confirmed by an independent group [28] but need further validation.