Loss of H3K27 trimethylation is frequent in IDH1-R132H but not in non-canonical IDH1/2 mutated and 1p/19q codeleted oligodendroglioma: a Japanese cohort study

Oligodendrogliomas are defined by mutation in isocitrate dehydrogenase (NADP(+)) (IDH)1/2 genes and chromosome 1p/19q codeletion. World Health Organisation diagnosis endorses testing for 1p/19q codeletion to distinguish IDH mutant (Mut) oligodendrogliomas from astrocytomas because these gliomas require different treatments and they have different outcomes. Several methods have been used to identify 1p/19q status; however, these techniques are not routinely available and require substantial infrastructure investment. Two recent studies reported reduced immunostaining for trimethylation at lysine 27 on histone H3 (H3K27me3) in IDH Mut 1p/19q codeleted oligodendroglioma. However, the specificity of H3K27me3 immunostaining in this setting is controversial. Therefore, we developed an easy-to-implement immunohistochemical surrogate for IDH Mut glioma subclassification and evaluated a validated adult glioma cohort. We screened 145 adult glioma cases, consisting of 45 IDH Mut and 1p/19q codeleted oligodendrogliomas, 30 IDH Mut astrocytomas, 16 IDH wild-type (Wt) astrocytomas, and 54 IDH Wt glioblastomas (GBMs). We compared immunostaining with DNA sequencing and fluorescent in situ hybridization analysis and assessed differences in H3K27me3 staining between oligodendroglial and astrocytic lineages and between IDH1-R132H and non-canonical (non-R132H) IDH1/2 Mut oligodendroglioma. A loss of H3K27me3 was observed in 36/40 (90%) of IDH1-R132H Mut oligodendroglioma. In contrast, loss of H3K27me3 was never seen in IDH1-R132L or IDH2-mutated 1p/19q codeleted oligodendrogliomas. IDH Mut astrocytoma, IDH Wt astrocytoma and GBM showed preserved nuclear staining in 87%, 94%, and 91% of cases, respectively. A high recursive partitioning model predicted probability score (0.9835) indicated that the loss of H3K27me3 is frequent to IDH1-R132H Mut oligodendroglioma. Our results demonstrate H3K27me3 immunohistochemical evaluation to be a cost-effective and reliable method for defining 1p/19q codeletion along with IDH1-R132H and ATRX immunostaining, even in the absence of 1p/19q testing.


Introduction
The current World Health Organisation classification for CNS tumors recommends integrated diagnosis based on combined phenotypic and genotypic findings [1] Although originating from common progenitor cells harboring Isocitrate Dehydrogenase (NADP(+)) (IDH) mutations, oligodendrogliomas differ from diffuse astrocytomas by combined whole-arm losses of chromosome 1p and 19q (1p/19q codeletion) and frequent Telomerase Reverse Transcriptase (TERT) promoter mutations. In contrast, astrocytoma typically exhibits Tumor Protein P53 (TP53) and ATRX Chromatin Remodeler (ATRX) mutations [2][3][4][5][6][7]. From a clinical perspective, these gliomas require different treatments and have different outcomes; therefore, the distinction of oligodendroglioma and astrocytoma is crucial. Subclassification of IDH mutant (Mut) glioma into astrocytomas and oligodendrogliomas requires testing for 1p/19q codeletion. Several different methods have been used to identify 1p/19q status, but clear consensus guidelines or standard protocols for practical use have not been established [8]. Fluorescent in situ hybridization (FISH) is a commonly used method for detecting 1p/19q codeletion. PCR-based loss of heterozygosity analysis, multiplex ligation-dependent probe amplification, and array comparative genomic hybridization can also test 1p/19q status with high reliability [9][10][11][12]. However, these techniques are labor-intensive and require substantial infrastructure investment, making their global application difficult in countries with less developed healthcare systems.
Although H3K27me3 has been reported to be involved in several brain tumor entities, comprehensive data about H3K27me3 in IDH Mut gliomas are controversial. Recently, Filipski et al. [23] reported that loss of H3K27me3 staining can potentially discriminate between oligodendroglial and astrocytic tumor lineages. Similarly, Feller et al. [24] and Kitahama et al. [25] reported lower H3K27me3 in oligodendroglioma by data-independent acquisition (DIA)-based mass spectrometry and immunostaining, respectively. However, using sequential IHC, Pekmezci et al. [26] did not consider H3K27me3 to be a specific marker for the classification of diffuse gliomas. Therefore, we have assembled a cohort of adult diffuse gliomas to determine whether simple H3K27me3 immunostaining can be a reliable method to triage cases for 1p/19q testing.
All immuno-positive cases for IDH1-R132H were classified as IDH1 Mut. Negative immunostaining of ATRX in neoplastic cells in the presence of an internal positive control was considered to indicate a loss of ATRX expression. Immunohistochemistry for p53 was positive when more than 50% of tumor nuclei showed intense staining.
Scoring of H3K27me3. Human colonic mucosa was used as a positive control according to the antibody datasheet. Preserved H3K27me3 in endothelial cells and immune cells served as an internal positive control. H3K27me3 immunostaining was assessed as H3K27me3positive (nuclear retention) or -negative (nuclear loss) in a blinded manner. Complete nuclear loss or dot-like H3K27me3 staining in neoplastic cells was regarded as nuclear loss, as described previously [23]. Each slide was scanned using a Nanozoomer XR Scanner (Hamamatsu, Japan) and viewed using NDP. Scan version 3.2.4 software. JPEG images for each case were captured from three randomly selected areas at 20× magnification using NDP.view 2 software. At first, using the PatholoCount software Ver 1.0 (Mitani Corporation, Tokyo, Japan), we scored H3K27me3 immunostaining positive when more than 25% of cells show diffuse staining and negative when more than 75% of cells show loss of staining. Later, an automated, blinded quantification was performed based on the previously described methodology [22]. Quantification of immunostaining in each JPEG was conducted using Matlab's image processing toolbox. The algorithm used background-foreground separation with a global threshold set using Otsu's method. We recorded the average intensity of extracted pixels of each area. A case's final score was calculated by averaging three random areas chosen from a section. H3K27me3 staining patterns in different glioma subtypes are illustrated in Fig. 1a-l. To assess the variability between PatholoCount scoring and automated quantification, we evaluated the scoring results obtained from the same section. This comparison showed that the results were identical for all cases in terms of positive or negative. We used the scoring value of automated quantification for the analysis of our data.

DNA sequencing
DNA sequencing of IDH1 codon 132 and IDH2 codon 172 was performed in IDH1-R132H immuno-negative cases using an Applied Biosystems 3130 Genetic Analyzer and Sequencing Analysis Finch TV 1.4.0 software. DNA was extracted from FFPE tumor tissue using a DNA tissue extraction kit (Qiagen; Cat: 56404). The extracted DNA was quantified using a NanoDrop 1000 (Thermo Scientific). A fragment of 129 bp spanning the R132 codon of IDH1 was amplified using forward primer 5′-CGG TCT TCA GAG AAG CCA TT-3′ and reverse primer 5′-GCA AAA TCA CAT TAT TGC CAAC-3′. Likewise, a fragment of 293 bp spanning the R172 codon of IDH2 was amplified using forward primer 5′-GCT GCA GTG GGA CCA CTA TT-3′ and reverse primer 5′-TGT GGC CTT GTA CTG CAG AG-3′.

Fluorescence in situ hybridization
Fluorescence in situ hybridization (FISH) was performed on 3-µm thick FFPE tissue sections to assess the chromosome 1p/19q status using the Vysis 1p36/19q13 Dual Color Probe Kit as described previously (Abbott Laboratories, Abbott Park, IL, USA) [8]. Briefly, paraffin sections were deparaffinized, permeabilized, and hybridized using a probe kit. Changes in the 1p and 19q probe signals compared with controls were used to determine the presence of 1p/19q codeletion. For each sample, approximately 100 well-defined nuclei were scored for signals from the probes 1p36 (red)/1q25 (green) and 19q13 (red)/19p13 (green) under fluorescence microscopy at 1000× magnification. FISH results are expressed as a percentage of tumor cells with a deleted signal. Established criteria for deletion (1)(p36)/deletion (19)(q13) were considered when 50% of nuclei or more displayed only one red (n × red signal) and two green signals (2n × green signal).

Statistical analysis
Statistical analysis was performed using JMP ® Pro 15.2.0 (SAS) software (Cary, North Carolina, USA). The associations among 1p/19q deletion with H3K27me3 and ATRX staining, IDH1/2 mutation, and histopathological parameters were determined using the chi-squared test/Fisher's  (a, b). c Dot-like H3K27me3 staining in negative tumor cell nuclei was considered loss of H3K27me3 expression in IDH1 Mut 1p/19q codeleted oligodendroglioma. Arrows point to retained nuclear staining in endothelial cells and infiltrating lymphocytes (a-c). Retained nuclear H3K27me3 staining was observed in IDH1 Mut astrocytoma (d-f), IDH Wt Astrocytoma (g-i), and IDH Wt GBM (j-l). a-c 40× magnification (Scale bar = 20 μm); d-l 20× magnification (Scale bar = 50 μm). Mosaic plot analysis comparing the correlation between H3K27me3 (m, n) and ATRX immunoreactivity (o) among glioma subclasses. m IDH Mut 1p/19q codeleted oligodendrogliomas showed significantly lower H3K27me3 staining compared with other glioma subtypes. n Significant differential expression of H3K27me3 was seen between IDH1 and IDH2 Mut 1p/19q codeleted oligodendrogliomas. o Retained ATRX staining showing a statistically significant difference between the two IDH Mut glioma lineages. P ≤ 0.05 was considered significant. p-u Mutational analysis patterns among glioma subtypes. IDH1-R132H Mut oligodendroglioma cases showing a single amino acid transition from p arginine to histidine (R132H), q arginine to serine (R132S), and r arginine to leucine (R132L). IDH2-R172 Mut oligodendroglioma cases showing a single amino acid transition from s arginine to lysine (R172K), t arginine to serine (R172S), and u arginine to tryptophan (R172W). v-y Representative FISH images of IDH Mut glioma subtypes. A case of IDH Mut oligodendroglioma showing both 1p (v) and 19q (w) deletion. A case of IDH Mut astrocytoma showing intact 1p (x) and 19q (y). 1p/19q deleted cases show one red signal (target) and two green signals (control). NR nuclear retention, NL nuclear loss, Mut mutated, Wt wild type, GBM glioblastoma (See figure on next page.) exact test. Association with age and gender for IDH Mut gliomas and IDH Wt gliomas was determined using the chi-squared test. A partitioning model was deployed to predict H3K27me3 expression in IDH Mut 1p/19q codeleted gliomas. Hierarchical clustering based on the average intensity score was performed in R 3.6.3 (https:// cran.r-proje ct. org/) to visualize the relationship between IDH Mut 1p/19q codeleted gliomas and non-oligo gliomas (IDH Mut and Wt) with H3K27me3 staining.

Assessment of the predictive value of H3K27me3 in diffuse gliomas
To explore possible implications for clinical practice, we employed a recursive partitioning model to assess the value of H3K27me3 expression in diffuse gliomas to predict IDH Mut, and 1p/19q codeleted oligodendroglioma. Immunohistochemical analysis for H3K27me3, ATRX, and IDH1-R132H revealed that diffuse gliomas with a loss of nuclear H3K27me3 staining, retained ATRX staining, and IDH1-R132H positivity can be predicted as 1p/19q codeleted oligodendrogliomas with a probability score of 0.9835. In addition, glioma with retained nuclear H3K27me3, loss of ATRX staining, and IDH1-R132H positivity can be predicted as 1p/19q non-codeleted glioma with a probability score of 0.9823. Five of nine gliomas with preserved H3K27me3 were oligodendrogliomas that harbor non-canonical IDH1-R132L or IDH2-R172 mutations. Among 20 cases of 1p/19q, non-codeleted gliomas with preserved H3K27me3 staining, IDH1-R132H immunostaining did not provide additional information beyond that of ATRX (Fig. 3).

Discussion
Here we present an approach for H3K27me3 immunostaining for adult diffuse glioma and demonstrate its application in a routine diagnostic procedure. We show differences in H3K27me3 staining between oligodendroglial and astrocytic lineages and between IDH1-R132H and non-canonical IDH1/2 Mut oligodendrogliomas. While the loss of nuclear H3K27me3 was predominantly seen in IDH1-R132H Mut oligodendrogliomas, retained nuclear staining was mostly observed in IDH1 Mut astrocytoma regardless of the mutation type. However, H3K27me3 staining was always present in non-canonical IDH1/IDH2 Mut oligodendrogliomas ( Fig. 1n; Additional file 1: Table S1). Unsupervised hierarchical clustering showed two primary clusters, H3K27me3 nuclear loss (NL) and nuclear retention (NR), for both IDH Mut 1p/19q codeleted oligodendroglioma and non-oligo gliomas. Although complete differentiation was observed between NL and NR in IDH1 Mut 1p/19q codeleted oligodendroglioma, the cluster patterns show no difference in non-oligo gliomas between the groups (Fig. 2a, b). Therefore, H3K27me3 staining in non-oligo gliomas did not provide additional information between subgroups.
We applied a recursive partitioning model to assess the clinical utility of H3K27me3 immunostaining to predict IDH Mut 1p/19q codeleted oligodendroglioma. Our prediction models indicate the clinical utility of H3K27me3 IHC for the prediction of IDH1-R132H Mut 1p/19q codeleted oligodendroglioma along with IDH1-R132H and ATRX IHC. Consistent with a previous report [23], the high predicted probability score (0.9835) indicated that the loss of H3K27me3 with ATRX positivity is frequent to IDH1-R132H Mut 1p/19q codeleted oligodendroglioma (Fig. 3). However, our data contradict Fig. 3 Decision tree of recursive partitioning model providing the best split of the immunostaining. Blue bars correspond to IDH Mut 1p/19q codeleted oligodendrogliomas, and orange bars correspond to not IDH Mut 1p/19q codeleted gliomas. We considered IDH Mut 1p/19q codeleted oligodendroglioma as a dependent variable, and immunostaining (H3K27me3, ATRX, and IDH1-R132H) as predictors. NR nuclear retention, NL nuclear loss, Mut mutated the previous report, which suggested that the retained nuclear expression of H3K27me3 is only seen in astrocytoma [23]. Moreover, using the same sequential immunostaining, we found a discrepancy over the sensitivity of H3K27me3 immunostaining, as reported previously [26]. This discrepancy may be because our cut-off point to decide the loss of H3K27me3 staining was 20% lower than Pekmezci et al. Moreover, Pekmezi considered only complete loss as significant and patchy/mosaic staining as a retained expression [26]. However, following Filipski et al., [23], complete nuclear loss or dot-like nuclear retention was combined as the nuclear loss in our study. Filipski and Kitahama et al. [23,25] mentioned that dot-like nuclear staining corresponds to the inactivated X chromosome, which presumes to label the Barr body in the female subgroup of oligodendrogliomas. We also observed dot-like staining in eleven female cases of oligodendrogliomas.
When the integrated diagnosis approach was used to assess previous histological diagnoses, among 45 oligodendrogliomas, 39 showed oligodendrogliomas, and six showed mixed morphology. Thirty-five out of 39 oligodendroglioma cases that were positive for IDH1-R132H and ATRX, and reduced H3K27me3, exhibited 1p/19q codeletion. However, oligodendrogliomas with IDH2 mutations, retained ATRX, and preserved H3K27me3 expression showed classical oligodendroglial morphology and did not provide additional information about 1p/19q codeletion (Fig. 4). Thirty-three out of 46 cases showing astrocytic morphology were astrocytomas, and nine cases showed mixed features (formerly oligoastrocytoma). Four IDH Mut astrocytoma cases exhibited classic oligodendroglial morphology, and the integrated diagnosis was confirmed by intact 1p/19q chromosome status by FISH staining (Fig. 5).
Three IDH mutations (IDH1-R132x, IDH2-R172K, and IDH2-R140Q) occur predominantly in subsets of cancers and regulate central circuitry metabolism by producing the oncometabolite, 2-hydroxyglutarate (2-HG) [31]. Lu et al. [32] reported that 2-HG in IDH Mut tumors prevents the demethylation of repressive histone marks, such as H3K9me3 and H3K27me3, resulting in increased histone methylation. While IDH1 mutation causes a marked increase in hypermethylation at many genes, a small group of hypomethylated genes was also reported [33]. Papaemmanuil et al. [34] reported that IDH2-R172K-mutated acute myeloid leukemia (AML) showed severe disruption to central metabolism and was associated with different gene expression and DNA methylation compared with other IDH1 or IDH2 mutated AML. Although IDH1-R132H is the most frequent IDH mutation, other IDH mutations found in oligodendrogliomas have received less attention. Moreover, it is unknown whether IDH1-R132H and non-canonical IDH1/2mutated oligodendrogliomas have different prognostic and therapeutic characteristics. Genome-wide analyses would help to determine the underlying mechanism.
Immunohistochemistry is a cost-effective and accessible technique that can be readily adapted for detecting molecular surrogates [17]. Immunohistochemistry for the mutant specific IDH1-R132H is routine for diffuse adult glioma [35]. Moreover, H3K27me3 immunohistochemistry is used as a molecular surrogate to identify pediatric midline gliomas [1], malignant peripheral nerve sheath tumors [20], and H3K27M mutant gliomas [22]. Therefore, H3K27me3 immunostaining can be regarded as a sensitive and specific molecular surrogate for defining IDH1-R132H Mut 1p/19q codeleted oligodendroglioma in the absence of molecular testing.

Limitation of this study
The number of non-canonical IDH1/2 mutated 1p/19q codeleted oligodendrogliomas is small (n = 5). Thus, further investigations of the differential expression of H3K27me3 between IDH1-R132H and non-canonical IDH1/2 mutant oligodendrogliomas are required for prognostic and therapeutic application.

Conclusion
Our study revealed that loss of H3K27me3 nuclear staining among 1p/19q codeleted oligodendrogliomas is frequent in cases harboring IDH1-R132H mutation. We consider that H3K27me3 immunoreactivity could predict the 1p/19q codeletion status along with IDH1-R132H and ATRX immunostaining.