The prognostic impact of subclonal IDH1 mutation in grade 2–4 astrocytomas

Abstract Background Isocitrate dehydrogenase (IDH) mutations are thought to represent an early oncogenic event in glioma evolution, found with high penetrance across tumor cells; however, in rare cases, IDH mutation may exist only in a small subset of the total tumor cells (subclonal IDH mutation). Methods We present 2 institutional cases with subclonal IDH1 R132H mutation. In addition, 2 large publicly available cohorts of IDH-mutant astrocytomas were mined for cases harboring subclonal IDH mutations (defined as tumor cell fraction with IDH mutation ≤0.67) and the clinical and molecular features of these subclonal cases were compared to clonal IDH-mutant astrocytomas. Results Immunohistochemistry (IHC) performed on 2 institutional World Health Organization grade 4 IDH-mutant astrocytomas revealed only a minority of tumor cells in each case with IDH1 R132H mutant protein, and next-generation sequencing (NGS) revealed remarkably low IDH1 variant allele frequencies compared to other pathogenic mutations, including TP53 and/or ATRX. DNA methylation classified the first tumor as high-grade IDH-mutant astrocytoma with high confidence (0.98 scores). In the publicly available datasets, subclonal IDH mutation was present in 3.9% of IDH-mutant astrocytomas (18/466 tumors). Compared to clonal IDH-mutant astrocytomas (n = 156), subclonal cases demonstrated worse overall survival in grades 3 (P = .0106) and 4 (P = .0184). Conclusions While rare, subclonal IDH1 mutations are present in a subset of IDH-mutant astrocytomas of all grades, which may lead to a mismatch between IHC results and genetic/epigenetic classification. These findings suggest a possible prognostic role of IDH mutation subclonality, and highlight the potential clinical utility of quantitative IDH1 mutation evaluation by IHC and NGS.

Diffusely infiltrating glial neoplasms are among the most common intracranial tumors, second only to meningioma. Historically, these tumors have been defined based on histologic features alone. However, in the past decade, the discovery of numerous molecular features which could be used to subdivide diffuse gliomas into improved prognostic groups has led to a revolution in the definition and grading of these tumors. Currently, adult-type diffusely infiltrating gliomas are subdivided into IDH-mutant astrocytoma, IDH-mutant and 1p/19q-codeleted oligodendroglioma, and IDH-wild-type glioblastoma. 1 With an incidence of 0.44 per 100 000 individuals and nearly 3000 cases annually in the United States, IDH-mutant astrocytoma comprises approximately 11% of all of these diffusely infiltrating gliomas. 2 Since the prognostic importance of IDH1/2 mutations was demonstrated in astrocytic neoplasms, [3][4][5] much work has been done to further refine this category and to identify additional prognostic factors to explain the remaining heterogeneity in clinical outcomes in these tumors. [6][7][8][9][10][11] This led to the inclusion of homozygous CDKN2A/B deletion as a codified molecular criterion for grade 4 status in IDH-mutant astrocytoma, equivalent to the histologic findings of microvascular proliferation and palisading necrosis. 1,12 Numerous other studies have suggested additional prognostic factors including CDK4 amplification, 8,9,11,13 mismatch-repair deficits, 14 and genome-wide copy number variation (CNV)/chromosomal instability (CIN). 10,[15][16][17][18] One additional factor that may have significant prognostic implications in gliomas is the presence of subclonal populations of tumor cells with mosaic expression of certain molecular alterations, including IDH mutations. [19][20][21][22][23][24][25][26][27] In this report, we discuss 2 institutional cases of WHO grade 4 IDH-mutant astrocytoma with mosaic IDH1 R132H expression by immunohistochemistry and very low IDH1 variant allele frequencies (VAF) compared to other identified pathogenic mutations, including TP53 and/or ATRX, corresponding to subclonal IDH1 mutation (ie, a low IDH1 mutation TCF). We also evaluate 2 large, independent, publicly available datasets to further investigate the frequency and clinical importance of subclonal IDH mutation.

Molecular Profiling
Clinically validated next-generation DNA sequencing (NGS) was performed on institutional cases at FoundationOne Laboratories (https://www. foundationmedicine.com/test/foundationone-cdx) (Foundation Medicine, Inc., Cambridge, MA) using targeted high throughput hybridization-based capture technology for detection of genomic alterations. For institutional case 1, whole genome DNA methylation profiling and classification was performed at the New York University Molecular Pathology laboratory using the Illumina EPIC Human Methylation array, which assesses 850 000 CpG sites, according to the manufacturer's protocol and is used for tumor classification (www.molecularneuropathology.org), as previously described. 15,[28][29][30][31][32]

Public Dataset Analysis
To identify similar cases in public datasets, we queried the cBioPortal For Cancer Genomics (http://www. cbioportal.org/) 33,34 to evaluate 2 large cohorts of diffuse gliomas totaling 2126 7,35 cases. The original histologic diagnoses reported included "diffuse glioma, "

Importance of the Study
IDH-mutant astrocytomas are biologically distinct from IDH-wild-type glioblastoma and have significantly better progression/recurrence-free and overall survival after initial resection. However, there remains significant heterogeneity of clinical outcomes in IDHmutant astrocytomas that is unexplained by the current glioma classification and grading system alone. Our study demonstrates that a small subset of IDHmutant astrocytomas has a low fraction of tumor cells with IDH mutations, and these subclonal IDH-mutant astrocytomas have shorter overall survival intervals compared to grade-matched IDH-mutant astrocytomas with high IDH mutation tumor cell fraction (TCF)s. These findings suggest a potentially overlooked source of variability in clinical outcome amongst IDH-mutant astrocytomas, and demonstrate the importance of quantitative IDH mutation evaluation by next-generation sequencing.
Vij et al.: Subclonal IDH1 R132H mutation "oligodendroglioma, " "anaplastic oligodendroglioma, " "oligoastrocytoma, " "anaplastic oligoastrocytoma, " "astrocytoma, " "anaplastic astrocytoma, " and "glioblastoma. " All cases represent the first known biopsy or resection specimen and were manually reclassified using IDH1/2, ATRX, TP53, and 1p/19q status according to the 2021 World Health Organization (WHO) Classification of Central Nervous System Tumors, leaving a combined total of 466 IDH-mutant astrocytomas, WHO grades 2-4. 1 Mutation, global and individual CNV, gene expression data, methylation profile status, tumor purity, and survival data were downloaded and analyzed as previously described in detail. 13,17,18,36,37 The variant allele frequency (VAF) for each pathogenic mutation was defined as the ratio of mutant alleles divided by total alleles and was corrected in instances of homozygous mutation or allelic loss in the case of TP53, as well as gender in the case of ATRX. 38 The TCF with IDH mutation was estimated using the ratio of IDH1 or IDH2 VAF divided by the corrected VAF of other pathogenic mutations that were found at clonal levels 23 (TP53 or ATRX if TP53 mutation was not present), and a conservative threshold of ≤0.67 was used as a working definition of subclonal IDH mutation. Eighteen cases were identified that met these criteria for subclonal IDH-mutant astrocytoma, and were compared to 156 cases with clonal IDH mutation (IDH TCF ≥ TP53 and/or ATRX). This simplified method of determining the TCF with IDH1 mutation had an 85% concordance rate for identifying subclonal mutations and a 95% concordance rate for identifying clonal mutations compared to previously proposed methods for determining mutation clonality. 24,27 Statistical Analysis The significance of differences between Kaplan-Meier survival curves were calculated using the Mantel-Cox test (Log-rank test). Differences in patient age and CNV were evaluated using student's t-test. Proportion of cases with specific molecular alterations as well as histologic grade and patient gender were calculated using Fisher's Exact test. All statistical calculations were performed with GraphPad Prism version 9 (GraphPad, La Jolla, CA).

Institutional Case Histories
Institutional case 1:.-A 68-year-old female with a 5.3 cm complex solid-cystic necrotic mass in the right frontal lobe convexity and an adjacent 5.0 cm complex, predominantly cystic mass in the right parasylvian cortex. The resected tissue demonstrated brisk mitotic activity, microvascular proliferation, and palisading necrosis. Immunohistochemical (IHC) studies demonstrated strong GFAP and OLIG2 staining, approximately 50% MIB-1/Ki-67 proliferation index, frequent nuclear p53 staining, loss of p16 staining, intact ATRX expression, and largely negative IDH1 R132H expression, although scattered cells demonstrated positive cytoplasmic expression (<1% of total tumor cells) ( Figure 1). In addition, there was partial loss of MSH2 and MSH6 with intact nuclear expression for MLH1 and PMS2. NGS results demonstrated low IDH1 R132H variant allele frequency (VAF) of 0.8% with significantly higher VAF for additional mutations, including TP53 (89% VAF) and PIK3R1 (48% VAF), and homozygous CDKN2A/B deletion, among other alterations (Table 1). In addition, in 87% of cells, there was an MSH2 mutation and a high overall tumor mutation burden (14 mutations/Mb). No additional alterations were suggestive of IDH-wild-type glioblastoma (EGFR amplification, TERT promoter mutation, or simultaneous gain of chromosome 7/loss of chromosome 10) were detected. Despite the extremely low IDH1 VAF, 39 DNA methylation profiling was consistent with high-grade IDHmutant astrocytoma with a high degree of certainty (confidence score = 0.98) (Figure 2).
Institutional case 2:.-A 42-year-old female with a past medical history of seizures and left fronto-insular glioma, not otherwise specified that was originally biopsied in 2012 and subsequently treated with radiotherapy and temozolomide, developed memory difficulties and right upper and lower extremity weakness in 2022. MRI demonstrated enhancing progression of her left frontotemporalinsular diffusely infiltrating glioma. Subsequent biopsy demonstrated a hypercellular infiltrating glioma with mitotic figures, insipient microvascular proliferation, and necrosis most consistent with treatment effect. IHC studies demonstrated GFAP and OLIG2 immunoreactivity, an 18% MIB-1/Ki-67 proliferation index, increased p53 staining, loss of p16 staining, loss of nuclear ATRX expression, and scattered IDH1 R132H reactivity ( Figure 3). NGS demonstrated an IDH1 R132H mutation (11% VAF), ATRX mutation (22% VAF), TP53 mutation (69% VAF), and homozygous CDKN2A/B loss (Table 1). No tissue was available from the 2012 biopsy specimen for additional molecular testing.
Frequency and effect of subclonal IDH mutation.-To further investigate the long-term impact of low IDH mutation VAF, we identified 18 subclonal IDH-mutant astrocytoma cases (3.9% of total IDH-mutant astrocytomas) in publicly available datasets 7,35 with IDH-mutant TCF ≤0.67 and an additional 156 cases with clonal IDH1 mutation. TCF was established as the ratio between the IDH1 mutation VAF and the VAF of additional clonal pathogenic mutations (TP53 and/or ATRX) after correction for heterozygous, homozygous, and hemizygous mutations, as this information was uniformly available in all institutional and publicly available cases, and allows for simple identification of cases with low IDH-mutant TCF (subclonal IDH-mutant cell population) using only the results from standard NGS reports. Subclonal IDH mutation appears to occur across all WHO grades, and no significant difference was identified in the gender, age, grade, O-6-methylguanine-DNA methyltransferase (MGMT) promoter methylation frequency, or level of overall CNV between subclonal and clonal IDH-mutant astrocytoma cohorts or in grade-forgrade analysis ( Table 2). One of the institutional cases had mutations in a mismatch-repair gene (MSH2); however, only 1 of the 18 subclonal IDH-mutant cases had such a mutation (PMS2), which was not significantly higher than the clonal IDH-mutant cohort. Global DNA methylation profiling and classification were available for 13 of the 18 subclonal IDH-mutant astrocytomas. All 13 of these matched with IDH-mutant, non-1p/19q-codeleted astrocytoma; 3 of these were consistent with glioma CpG island methylation phenotype (G-CIMP) "low" and the remaining 10 were G-CIMP "high. " 40,41 No significant differences were noted in recurrence/ progression-free survival between the subclonal and clonal IDH-mutant astrocytomas, although there were nonsignificant trends toward shorter progression-free survival in subclonal  Table 2). Compared to IDH-wildtype glioblastoma from the same publicly available cohorts (n = 983); however, subclonal IDH-mutant astrocytomas had

Discussion
Our results demonstrate that while rare, subclonal IDH mutation does occur and appears to have a significantly worse effect on survival compared to grade-matched IDH-mutant astrocytomas with clonal IDH mutation. Subclonality of IDH1 mutations was identified in the institutional cases included in this report via 2 distinct but complimentary methods: The relative percentage of tumor cells with IDH1 R132H immunoreactivity compared to the number of total tumor cells with other pathogenic mutations detectable by immunohistochemistry, including loss of nuclear ATRX reactivity and increased nuclear p53 immunostaining (Figures 1 and 3), and the confirmation of this finding with the ratio of VAF of IDH1 mutations compared to the VAF of other pathogenic mutations in the same tumor sample to establish an IDH1 mutation TCF consistent with subclonality. While the institutional cases were insufficient to determine the clinical impact of subclonal IDH mutations, we were able to leverage two large publicly available glioma datasets 7,35 to evaluate the prognostic impact of this finding. These data suggest that there is a significantly adverse impact on overall survival associated with subclonal IDH1 mutation compared to grade 3 and grade 4 clonal cohorts (Table 2 and Figure 4), a relationship that has been suggested previously. 24 Although many studies have demonstrated intercellular genomic heterogeneity as a fundamental feature of diffuse glioma and other solid cancer progressions, 19,20,22,25,35,42 IDH mutation is considered to be an early oncogenic event initiating glioma evolution, 4,40,[43][44][45] and as such is generally found with high penetrance across the tumor cell population of a given glioma. The identification of subclonal IDH1 mutation suggests that in a minority of cases, this mutation was either acquired in a subpopulation of an extant IDH-wild-type glioma or the mutation was present early in tumorigenesis and then lost in a subpopulation of tumor cells that had an apparent survival advantage so as to subsequently become the majority of the tumor volume. 23 Institutional case 1 had an extremely low IDH-mutant TCF, as well as a mutation in the mismatch-repair gene MSH2 and corresponding elevated tumor mutation burden (14 mutations/ Mb), which could provide a potential mechanism for the IDH1 mosaicism; the genomic instability resulting from this mutation 14 may have either resulted in IDH1 mutation in a subpopulation of tumor cells late in tumor progression or a revertant mutation in the IDH1 R132 codon. Alternatively, increasing CNV with tumor progression/ molecular evolution or as a result of radiation and che motherapies 13,17,31,46 could result in the loss of the IDH1 R132H mutated allele in a subset of cells, a possibility in institutional case 2. Given the lack of additional molecular features typically associated with IDH-wild-type glioblastoma in any of the institutional or public subclonal astrocytomas (EGFR amplification, simultaneous gain of chromosome 7/loss of chromosome 10, and TERT promoter mutation) as well as the DNA methylation profiling result consistent with IDH-mutant astrocytoma in institutional case 1 and all 13 of the publicly available subclonal cases for which DNA methylation profiling classification data were available, it is unlikely that these tumors represent IDH-wild-type glioblastomas with late IDH1 mutation. Additionally, the presence of a mutation in an MMR-related gene does not appear to be of increased frequency in cases with subclonal IDH1 mutation, seen in 2/20 total subclonal cases compared to 6/156 clonal cases (P = .2264) (Tables 1 and 2). Sublclonal IDH-mutant astrocytomas also have significantly better progressionfree survival and OS compared to IDH-wild-type glioblastoma ( Supplementary Figure 1), which suggests that they are not simply IDH-wild-type tumors with a late IDH1 mutation in a relatively small subclone. The presence of subclonal IDH1 mutation is, therefore, more likely due to IDH1 mutation as an early event, followed by loss of this mutation or loss of the allele and clonal expansion of the "IDH-wild-type subclone. " 21,40,47 The likelihood that the IDH mutation is lost in a large subset of cells is of particular interest in light of the methylation profiling results, 39 as these cases appear to maintain their IDHmutant methylation phenotype despite the partial loss of the underlying IDH mutation that originally drove this phenotype, suggesting that the epigenetic footprint of IDH mutation may be strong enough to persist even in the absence of the mutation itself. Previous studies have estimated tumor cell purity and determined the TCF containing specific mutations extrapolated from genomic data using more complicated formulas establishing the expected and observed VAF for each identified mutation. 24,27,[48][49][50] Here, we opted for a simplified method of estimating the TCF with IDH1 mutation, ie, comparison of the IDH1 VAF with the VAF of other known pathogenic mutations present in the same tumor sample (TP53 or ATRX). Given that these additional mutations would not be expected to be present in the non-neoplastic background brain parenchyma, the ratio of IDH1 VAF to these other mutations allows for estimation of maximum IDH1  In conclusion, while the mechanism of this subclonal mosaic IDH1 mutation remains uncertain, these data demonstrate that diffuse astrocytomas with subclonal IDH1 mutation may have worse clinical outcomes than their grade-matched clonal IDH1-mutant counterparts, although these patients have later recurrences and survive significantly longer than IDH-wild-type glioblastoma patients. This feature is important to recognize as the vast majority of pathology reports simply state a qualitative presence or absence of IDH1/2 mutation (as well as other pathogenic mutations), even when NGS studies report the relative VAF of each alteration. The possibility of subclonal IDH1 mutation is especially important to investigate in cases with discordant IHC and NGS findings or methylation results suggesting an IDH mutation in the presence of minimal IDH1 R132H IHC staining, as in institutional case 1. These findings suggest a potentially overlooked source of variability in clinical outcome in cohorts of patients with IDH-mutant astrocytomas, and demonstrate the importance of quantitative IDH mutation assessments. It is important to consider the subclonal tumor pattern and continuing molecular evolution in determining diagnostic criteria, personalizing treatment, and designing effective clinical trials.

H&E IDH1 R132H
GFAP p53 ATRX Ki-67 Figure 3. Immunohistochemical findings in institutional case 2, demonstrating mosaic IDH1 R132H in relation to p53 and ATRX staining. All panels were taken at 100x except insert in panel B, which was taken at 400x. Scale bar = 200 µm and applies to each panel.