The integrated diagnosis for infiltrating gliomas requires demonstration of IDH1 or IDH2 mutation and 1p/19q codeletion in the tumor to diagnose oligodendroglioma or anaplastic oligodendroglioma, IDH-mutant and 1p/19q-codeleted. IDH1 R132H, ATRX and p53 immunohistochemistry may help triaging cases for 1p/19q testing [3]. Recently, Filipski et al. [2] have reported that loss of H3K27me3 (trimethylation of lysine 27 of histone 3) stain could potentially stratify IDH-mutant diffuse gliomas based on cell type, recommending H3K27me3 stain as an additional marker in the immunohistochemical workup. A separate study reported differential loss of H3K27me3 staining in oligodendrogliomas and astrocytomas, which was only true using the monoclonal antibody [1]. We have therefore assembled a validation cohort of genetically well-characterized adult diffuse gliomas to assess the clinical utility of H3K27me3 immunohistochemistry in this setting.

We identified 117 adult gliomas initially diagnosed at UCSF and Izmir KCU Ataturk EAH, in which sufficient tissue to be included in tissue microarrays in duplicate or triplicate was available. Three neuropathologists (MP, FD and TT) performed morphologic classification based on H&E-stained sections without immunohistochemistry or molecular results. Immunohistochemistry was performed on five-micron sections using standard techniques, and all stains were independently scored by two neuropathologists (MP and TT) with subsequent consensus on discrepant cases. All cases with positive immunohistochemistry for IDH1 R132H mutant protein (H09, Dianova GmbH, Hamburg, Germany) were classified as IDH mutant. Immunonegative cases were classified as mutant or wild type based on Sanger sequencing of IDH1 and IDH2 genes covering exon 4 regions. Chromosome 1p/19q status was separately assessed by fluorescence in situ hybridization using the Vysis 1p36/19q13 Dual Color Probe Kit (Abbott Laboratories, Abbott Park, IL). Complete loss of ATRX (HPA001906, Sigma-Aldrich, St. Louis, MO) in neoplastic cells in the presence of internal positive control was considered as ATRX-altered. Immunohistochemistry for p53 (DO-7, Dako, Agilent, Santa Clara, CA) was considered positive when more than 50% of tumor nuclei showed strong staining. Only complete loss of H3K27me3 (clone C36B11, 1:50 dilution, Cell Signaling Technology, Danvers, MA) staining in the presence of positive internal control was considered significant. Patchy/mosaic staining was recorded, but was considered as retained expression.

The study cohort included 102 cases after exclusion of four IDH1R132H-negative cases without sequencing results, two IDH-mutant gliomas without 1p/19q codeletion status, two H3K27M-mutant diffuse midline gliomas and seven gliomas with failed H3K27me3 staining (one oligodendroglioma, three IDH-mutant and three IDH-wild-type astrocytomas). Among oligodendrogliomas, 21 (75%) showed complete loss of H3K27me3 staining, whereas 7 (25%) had retained expression. H3K27me3 loss was seen in nine (27%) IDH-mutant astrocytomas (grades II–IV) and in only one (2.5%) IDH-wild-type astrocytoma (Table 1). Examples of the staining patterns among oligodendrogliomas are shown in Fig. 1a–c. Analyzing only IDH-mutant tumors, 21/28 oligodendrogliomas and 9/34 astrocytomas have loss of H3K27me3 staining, showing a statistically significant association between 1p/19q codeletion and loss of H3K27me3 staining (χ2 test, p < 0.001), consistent with the previous reports [1, 2].

Table 1 H3K27me3 stain among diffuse gliomas
Full size table
Fig. 1
figure 1

a–c Immunohistochemistry for H3K27me3 on three different IDH-mutant and 1p/19q-codeleted oligodendrogliomas, 400 × magnification. a Loss of nuclear staining in the background of retained staining in endothelial cells. Note the dot-like staining in negative nuclei corresponding to the inactivated X chromosome in this female patient. b Patchy staining pattern with variable amounts of staining in tumor cells, considered as retained expression. c Retained nuclear expression in tumor cells. d Algorithmic immunohistochemical staining panel starting with H3K27me3. Blue bars correspond to “IDH-mutant and 1p/19q-codeleted” oligodendrogliomas, and red bars correspond to astrocytomas (not “IDH-mutant and 1p/19q-codeleted”). Note nine oligodendrogliomas with retained H3K27me3 expression (12.7%). e Algorithm starting with morphologic features followed by IDH1R132H, ATRX and H3K27me3 stains

Full size image

To assess the utility of the H3K27me3 stain in the decision of which cases would require 1p/19q testing, we have applied the algorithm suggested by Filipski et al. (Fig. 1d). H3K27me3 was retained in 71 of 102 cases. Seven (10%) of these were oligodendrogliomas, which would have been misclassified if classified by immunohistochemistry alone (Fig. 1c and Supplementary Fig. 1 [Online Resource 1]). This contradicts with the previous report suggesting that the retained nuclear expression of H3K27me3 is only seen in astrocytomas with a predicted probability score of 0.9995 [2]. Filipski et al. suggested that diffuse gliomas with H3K27me3 loss, retained or non-conclusive ATRX and positive IDH1R132H are oligodendrogliomas with a predicted probability of 0.9678. However, using the same sequential immunohistochemistry, we found that only 72% of such cases were oligodendrogliomas in our cohort. p53 immunohistochemistry did not provide additional information beyond ATRX.

We have also applied an alternative algorithm starting with IDH R132H and ATRX stains (Supplementary Fig. 2 [Online Resource 2]). Three of 43 IDH1 R132H-immunonegative gliomas were oligodendrogliomas, which carried non-canonical IDH1/2 mutations. Among 58 IDH1 R132H-positive gliomas, ATRX stain was retained in 34 and had failed in three. These 37 would require 1p/19q testing to identify 25 oligodendrogliomas in this group. Alternatively, adding H3K27me3 and classifying those with retained H3K27me3 as astrocytomas would cause misclassification of seven oligodendrogliomas.

As we have been moving toward an integrated diagnosis for the tumor classification, many of the recent studies focus solely on the molecular alterations, or surrogate markers, and no longer include the histologic criteria. In an algorithm starting with the histologic impression (Fig. 1e) as advocated by WHO and cIMPACT-NOW recommendations [3], all 59 cases with astrocytic morphology were found to be astrocytoma (1p/19q-intact). Thirty-seven cases exhibited classic oligodendroglial morphology, and six cases showed mixed features (formerly oligoastrocytoma). Among 28 genetically confirmed oligodendrogliomas, 27 showed oligodendroglial and one showed mixed morphology. All three oligodendrogliomas with non-canonical IDH1/2 mutations showed classic oligodendroglial morphology and retained ATRX expression, further highlighting the role of histology to guide ancillary testing. Twenty-three of 24 gliomas with classic oligodendroglial morphology, positive IDH1 R132H and retained ATRX exhibited 1p/19q codeletion with a positive predictive value of 96%. One case with intact 1p/19q showed rare p53 positive cells and H3K27me3 loss; therefore, these stains did not provide additional information (Supplementary Fig. 3 [Online Resource 3]). We believe this predictive value is sufficient to consider a tumor as oligodendroglioma, even in the absence of 1p/19q results, especially if the resources are limited.

In summary, while H3K27me3 stain may still be helpful in the clinical practice, retained expression should not be used as the sole evidence against 1p/19q codeletion when other morphologic and immunohistochemical features suggest the possibility of an oligodendroglioma.

Our study was limited to assess the clinical utility of H3K27me3 for the diagnosis and classification of diffuse gliomas. But as stated by Filipski et al., the loss of H3K27me3 nuclear staining among the majority of the IDH-mutant and 1p/19q-codeleted oligodendrogliomas is unexpected and studies to understand the mechanism of decreased H3K27me3 in diffuse gliomas are necessary.