Hypermutagenesis in untreated adult gliomas due to inherited mismatch mutations.

Hypermutagenesis refers to marked increase in the number of mutations due to continuous mutagenic process. Hypermutated tumors, have being found in several tumor types, are associated with inherited or acquired alterations in the DNA repair pathways. Hypermutation has been observed in a subset of adult glioma patients as a direct result of temozolomide(TMZ)-induced mutagenesis. In our study, we have identified a rare subset of treatment-naïve adult gliomas with de novo hypermutator phenotype and explored the evolution of spontaneous and treatment-induced hypermutagenesis. We conducted Whole-Exome Sequencing (WES), Whole-Transcriptome Sequencing (WTS), and Single-Cell Sequencing (SCS) of TMZ-naïve and post-TMZ-treated hypermutated tumors to identify distinct clinical or genomic manifestations that contribute to the development of hypermutation in untreated adult gliomas. TMZ-naïve hypermutated tumors were marked by absence of IDH1 somatic mutation and MGMT promoter (pMGMT) methylation, two genomic traits that were significantly associated with the TMZ-induced hypermutagenic event in glioblastoma, and harbored inherited alterations in the mismatch repair (MMR) machinery. The immediate family members of the TMZ-naive hypermutated glioma patients were also previous diagnosed with cancer development history, suggesting that germline dysfunction of the MMR pathway could potentially pose hereditary risk to genetic predisposition of carcinogenesis in gliomas. Lastly, both TMZ-naïve and post-TMZ-treated hypermutated tumors exhibited a significant accumulation of neoantigen loads, suggesting immunotherapeutic alternatives. Our results present new and unique understanding of hypermutagenic process in adult gliomas and an important step towards clinical implication of immunotherapy in glioma treatment.

Introduction

DNA repair mechanism is part of the essential cell homeostasis, ensuring genome integrity.1 Alterations in the DNA repair pathways, both hereditary and somatic, largely contribute to carcinogenesis and hypermutagenesis,2, 3, 4 manifesting specific genomic profiles and sensitivities to diverse therapies, including immunotherapy.5, 6, 7, 8, 9, 10, 11 As the number of systematic sequencing efforts have accumulated exponentially in recent years,12, 13, 14 prevalence of hypermutator genotype have become evident across multiple cancer types, including gliomas.4, 15, 16 Glioma is the most common primary malignant brain tumor in adults.17 The current standard regimen for glioma patients consists of surgical resection followed by radiotherapy and/or concomitant and adjuvant temozolomide (TMZ) treatment.18, 19 TMZ has been associated with improved prognosis in overall survival of the patients, specifically those with transcriptional silencing of the O‐6‐Methylguanine‐DNA Methyltransferase (MGMT) gene, mediated by promoter methylation. However, previous literatures have shown that a fraction of TMZ‐treated patients acquired a hypermutator phenotype,15, 16, 20 a direct result of TMZ‐induced mutagenesis due to inactivation of the DNA mismatch repair (MMR) pathway, and they subsequently acquired treatment‐induced therapeutic resistance.21, 22, 23, 24 In our study, we report a rare subset of adult glioma patients with de novo hypermutator phenotype with no prior therapeutic intervention. We investigate whether TMZ‐naïve hypermutagenesis manifests distinct genomic or clinical features that present secondary mutagenic trajectory, suggesting potential alternative avenue for malignant transformation.

Materials and Methods

Clinical manifestations and glioma specimens

With appropriate approval from the institutional review board, all glioma specimens were obtained from patients undergoing surgery at the Samsung Medical Center. The study protocol was approved by our institution's ethical committees, and written informed consents were received from all patients. After thorough analysis by pathologists, tumor specimens were snap‐frozen and preserved in liquid nitrogen for genomic analysis.

Whole‐exome sequencing

An Agilent SureSelect kit was used to capture exonic DNA fragments. An Illumina HiSeq 2000 instrument was used for sequencing and generating 2 x 101‐bp paired‐end reads.

Somatic mutation

The sequenced reads from the FASTQ files were aligned to the human genome assembly (hg19) using the Burrows‐Wheeler Aligner (BWA) version 0.6.2. The initial alignment BAM files were subjected to conventional preprocessing before mutation calling: sorting, removing duplicated reads, locally realigning reads around potential small indels and recalibrating base quality scores using SAMtools, Picard version 1.73 and Genome Analysis ToolKit (GATK) version 2.5.2. We used Mutect (version 1.1.4) and Somatic IndelDetector (GATK version 2.2) to distinguish high‐confidence predictions on somatic mutations between neoplastic and non‐neoplastic tissue pairs. Variant Effect Predictor (VEP) was used to annotate the acquired somatic mutations.

Copy number alteration

ngCGH python package was used to generate estimated copy number alterations. The patient‐matched normal WES data were used as reference to distinguish copy number changes in tumor specimens. The gene was labeled as “amplified” when the copy number was 3 or higher and “deleted” when it was 1.5 or lower.

Bulk RNA sequencing

RNA‐seq libraries were prepared using the Illumina TruSeq RNA Sample Prep kit. Sequenced reads were subjected to quality trimming and mapped onto hg19 using GSNAP, not allowing mismatch, indels, or splicing. The resulting alignments were sorted and summarized into BED files using SAMtools and bedTools. The BED files were used to calculate values of RPKM (reads per kilobase of transcript per million reads) for each gene, using DEGseq package.

Isolation of single cells for RNA sequencing and single‐cell analysis

We used the C1TM Single‐Cell Auto Prep System (Fluidigm) with the SMARTer kit (Clontech) to generate cDNAs. Cells were captured as a single isolate on a C1 chip, determined from bright‐field images under 100× magnification using the Axiovert200 inverted microscope (Carl Zeiss). Libraries were generated using the NexteraXT DNA Sample Prep Kit (Illumina) and sequenced on the HiSeq 2,500 using 100‐bp paired‐end mode of the TruSeq Rapid PE Cluster kit and the TrueSeq Rapid SBS kit. Subsequent RNA reads were filtered at Q33 using Trimmomatic‐0.30 and mapped to reference. For single‐cell analysis, we filtered out non‐neoplastic cells based on transcriptome profiles. We employed the expression signatures of normal astrocytes, oligodendrocytes, neurons, microglia and other immune cells to classify individual cells according to their expression profiles. We further validated our annotation of malignant cells from non‐malignant cells using the CONICSmat package to infer large‐scale copy number variations (CNVs) from single‐cell RNA‐seq data.

Glioma‐intrinsic transcriptional subtype analysis

To estimate glioma‐intrinsic transcriptional subtype, we evaluated single sample Gene Set Enrichment Analysis (ssGSEA) for each subtype markers. Each corresponding subtype scores were normalized and the subtype with the highest z‐score within each sample was assigned as its dominant subtype identity.

HLA typing

HLA typing for each sample was performed using POLYSOLVER algorithm.25

Somatic mutation immunogenicity predictions

We used the pVAC‐Seq pipeline with the NetMHCcons binding strength predictor to identify neoantigens.26 NetMHCcons integrates three state‐of‐the‐art methods NetMHC, NetMHCpan and PickPocket to give the most accurate predictions.26 As required, we used the variant effect predictor from Ensembl to annotate variants for downstream processing by pVAC‐Seq.27 For each single‐residue missense alteration, MHC binding affinity was predicted for all the wild‐type and mutant peptides of 8, 9, 10 and 11 amino acids in length. Based on widely accepted standards of the field, mutant peptides with binding affinity of <500 nM and corresponding wild‐type peptides with binding affinity >500 nM were considered as a predicted binder. The mutant peptide with the strongest binding affinity was selected for neoantigen counts.

MGMT promoter methylation

DNA was extracted from tumor tissue specimens using the QIAamp DNA Mini Kit (Qiagen) and the resulting DNA was modified with sodium bisulfite treatment using the EpiTect Bisulfite Kit (Qiagen). Afterward, the CpG island region of the MGMT was amplified and assessed using methylation/unmethylation specific primers. Methylation; TTTCGACGTTCGTAGGTTTTCGC (forward) and GCACTCTTCCGAAAACGAAACG (reverse). Unmethylation; TTTGTGTTTTGATGTTTGTAGGTTTTTGT (forward) and AACTCCACACTCTTCCAAAAACAAAACA (reverse).28, 29 Additionally, the reactions were validated by electrophoresis to verify the methylation status of the MGMT promoter. PCR products were 81 bp (methylation) and 93 bp (unmethylation), respectively.

Results

Genomic characterization of TMZ‐naive and TMZ‐treated hypermutated tumors

To assess distinct genomic characteristics of TMZ‐naïve hypermutated gliomas, we identified 14 patients (7 TMZ‐naïve and 7 TMZ‐treated) with a hypermutator phenotype among 243 patients whose Whole‐Exome Sequencing (WES) of both tumor specimens and matched normal blood were available (Table 1 and Fig. 1 a). All hypermutated tumors exhibited extremely large numbers of non‐synonymous somatic mutations with average of 98.2 mutations/Mb, significantly exceeding the average number of mutation rates found in non‐hypermutated tumors (Fig. 1 a). Patients with post‐TMZ‐treated hypermutated tumors have undergone 6 to 18 cycles of temozolomide treatment. Both TMZ‐naïve and post‐TMZ‐treated cases showed somatic alterations in the core oncogenic pathways that are frequently dysregulated in glioblastoma including p53, Rb and receptor tyrosine kinase (RTK)/Ras/Phosphoinositide 3‐kinase (PI3K) signaling pathways. Furthermore, glioma‐intrinsic transcriptional subtype analysis revealed high prevalence of proneural cellular state in both groups (Fig. 1 b). Consistent with prior notions that IDH1‐mutated tumors are more susceptible to acquiring a hypermutator phenotype,15, 20 majority of the TMZ‐treated tumors were marked by somatic mutations in IDH1, while all TMZ‐naïve tumors demonstrated absence of IDH1 mutation (P = 0.02, Fisher's exact test) (Fib. 1b).

Table 1

Clinical information of patients who were diagnosed with either treatment‐naïve or treatment‐received hypermutated gliomas

Patient Clinical Information
Patient	Sex	Age	Pathology	Prior Tx
BT1011	Male	68	GBM	None
BT1021	Female	40	GBM	None
BT103	Male	45	AA	None
BT1041	Female	21	GBM	None
BT105	Female	34	GBM	None
BT1061	Female	72	GBM	None
BT1071	Female	63	GBM	None
BT108	Female	56	AODG	RT + TMZ# 12
BT109	Female	56	GBM	RT + TMZ#6
BT110	Male	57	AA	RT + TMZ# 18
BT111	Female	53	AODG	RT + TMZ#9
BT112	Male	45	GBM	RT + TMZ# 17
BT113	Female	59	GBM	CCRT+TMZ#6
BT114	Male	24	GBM	CCRT+TMZ#6

AODG: Anaplastic Oligodendroglioma; GBM: Glioblastoma; AA: Anaplastic Astrycoma; RT: radiotherapy; CCRT: concurrent chemoraditionary therapy; TMZ: temozolomide.

Patients had family history of malignant tumors.

Figure 1

Genomic landscape of TMZ‐naïve and TMZ‐treated hypermutated gliomas. (a) Somatic mutational frequencies in TMZ‐naïve and TMZ‐treated hypermutated high‐grade gliomas compared to non‐hypermutated high‐grade gliomas. Data on the y axis are log‐transformed. (b) Genomic landscape of TMZ‐naïve and TMZ‐treated hypermutated gliomas. Number of TMZ cycle, MGMT promoter methylation status, glioma‐intrinsic transcriptional subtype, somatic alterations including single nucleotide variation (SNV) and small insertion/deletion (Indel), and copy number alterations are shown for each corresponding patient.

Previous studies have identified a subset of GBM tumors with characteristic DNA methylation profile, commonly known as glioma CpG island methylator phenotype (G‐CIMP).30, 31 The global methylator phenotype is often induced by the gain‐of‐function IDH1 mutation at arginine 132 and the resulting tumors frequently demonstrate molecular and clinical features that are distinct from Non‐G‐CIMP tumors. The hypermethylation of the genome in G‐CIMP tumors often encompass the promoter region of the O6‐methylguanine‐DNA methyltransferase (MGMT) gene. Subsequently, pMGMT methylation attributes to impairment of DNA repair mechanism32 and development of a hypermutator phenotype. Consistent with the previous reports, majority of the post‐TMZ‐treated hypermutated tumors demonstrated aberrant DNA methylation of the pMGMT (Fig. 1 b) followed by attenuated mRNA expression level (Supporting Information Fig. S1). Conversely, TMZ‐naïve tumors mainly exhibited unmethylated pMGMT status. Collectively, our results suggest that TMZ‐naïve hypermutated tumors follow an alternative evolutionary path in acquiring a hypermutator phenotype.

Dysregulation of mismatch repair (MMR) machinery

DNA mismatch repair mechanisms are composed of various essentials, necessary for maintaining genome integrity and faithful replications. Mismatch repair (MMR) safeguards genome integrity through correcting improper nucleotide pairings that arise from replication errors. Dysfunction of the MMR largely contributes to accumulation of spontaneous mutations during tumor progression. Both TMZ‐naïve and TMZ‐treated tumors acquired somatic mutation of the MMR encoding genes including MSH (MSH2, MSH3, MSH5, MSH6; 12/14 tumors), MLH (MLH1, MLH3; 7/14 tumors) and/or PMS family (PMS1, PMS2; 5/14 tumors) (Fig. 2 a). We also identified nonsynonymous germline mutations of the MMR associated genes in the TMZ‐naïve tumors, including MSH2, MSH3, MSH6 and MLH3. Lynch syndrome, an autosomal dominant genetic condition that is characterized by germline alterations in the MMR genes, is associated with increased risk of carcinogenesis including colon and endometrial cancers.33, 34, 35 As these genetic conditions are often hereditary, we expanded clinical assessment of the patients with germline MMR variations to their immediate family members. As suspected, several members of the patients’ families were previously diagnosed with cancer development history including brain tumor, lung, breast and prostate carcinomas. Consistently, several germline mutations have been formerly associated with cancer predisposing syndromes including rs63749919, (MSH6, Y969C), rs55740729, (MSH6, K1358D) and rs1114167705 (MSH6, F1088S). We postulated that combined inherent and somatic mutations of the mismatch repair genes were integral to the rapid progression of a hypermutator phenotype in TMZ‐naïve tumors, demonstrating a secondary pathway that drive hypermutagenesis in gliomas. Furthermore, we observed that the mismatch repair mutations were clonally represented in the tumors (high allelic fraction >35%), suggesting an explosive accumulation of somatic mutations after acquiring MMR pathway disruption (Fig. 2 b).

Figure 2

Dysregulation of mismatch repair (MMR) encoding genes. (a) Germline and somatic mutation of the MMR encoding genes in TMZ‐naïve and TMZ‐treated hypermutated tumors. The panel includes patients for whom whole‐exome sequencing data of both tumor and matched normal specimens were available. (b) Variant allele fractions of both somatic and germline MMR mutations are portrayed with respect to all other mutations in the genome. (c) Subtype expression phenotype and mutational profiles of individual tumor cells from patient, BT104 are shown. For each tumor cell, the subtype with the highest enrichment score was determined as the corresponding identity; Proneural (Purple) and non‐Proneural (White). PDGFRA, PMS1, MSH3, MLH3 and TP53 genomic alterations could be identified in the single‐cell expression data (green), despite the abundance of missing data (gray). The p value for co‐occurence of the Proneural subtype and PDGFRA mutation were obtained using the hypergeometric test.

To further interrogate the clonality of TMZ‐naïve hypermutated tumors at a single cell level, we curated single‐cell sequencing data from multi‐region tumor samples (main mass and resection margin) from an individual patient (BT104). To distinguish non‐malignant from malignant cells, we first employed the expression signatures of stromal and immune cells to classify individual cells based on their transcriptome profiles. Afterward, we estimated inferred copy‐number variations for individual cells using gene expression profiles over large chromosomal regions within each cell. Each tumor cell demonstrated large‐scale chromosomal aberrations, including amplification of chromosome 8 and deletions of 5,10 and 13 (Supporting Information Fig. S2). On the contrary, cells that were classified as non‐malignant cells lacked any detectable CNVs. Consistent with the bulk WES results, mutations of the MMR encoding genes including, PMS1, MLH3 and MSH3, were detected across all single cells with enough coverage at the corresponding genomic regions (Fig. 2 c). Additionally, we also discovered a high frequency of PDGFRA mutations (60%, 83/138), which significantly associated with a proneural expression phenotype. These observations were consistent with the bulk WTS results, where proneural transcriptional subtype was predominant among hypermutated tumors.

Distinct mutational signatures and elevated neoantigen loads in hypermutated tumors

Next, we evaluated mutational signatures of TMZ‐naïve and TMZ‐treated hypermutated tumors.36, 37 Of the six classes of base substitution in the mutation type, both groups exhibited robust presence of C‐to‐T transition (Fig. 3 a). TMZ‐naïve tumors were characterized by prominence of C‐to‐T substitutions at NpCpG trinucleotides (where N represents any nucleotide), which has been speculated to be a direct result of elevated rate in spontaneous deamination of 5‐methyl‐cytosine38 (Fig. 3 b). On the other hand, post‐TMZ treatment tumors were enriched with excess C‐to‐T substitutions at NpCpC and NpCpT trinucleotides, which is an evident indication of TMZ‐induced mutagenesis36, 37 (Fig. 3 b). Notably, majority of the driver mutations that were observed in the post‐TMZ treatment tumors consist of C‐to‐T substitutions at NpCpC and NpCpT contexts, suggesting that these tumors have undergone TMZ‐induced malignant progressions.

Figure 3

Mutational transition of TMZ‐naïve and TMZ‐treated hypermutated tumors. (a) Top panel exhibits mutational type, indicating mutational spectrum of each hypermutated tumors. Bottom panel notes mutational context, demonstrating base substitution mutation spectra of each somatic mutation. All 96 mutated trinucleotides are represented in a heatmap. The base corresponding to 5′ is shown on the vertical axis and the 3’s base is on the horizontal axis. (b) Representative nucleotide context of the C‐to‐T transitions in TMZ‐naïve (upper panel) and TMZ‐treated (bottom panel) hypermutated tumors. The height of the letters represents the occurrence frequency of the corresponding nucleotide at each position. (c) Log‐scaled Neoantigen load counts in non‐hypermutated, TMZ‐naïve hypermutated, and TMZ‐treated hypermutated tumors. P‐values were calculated using the Wilcoxon rank sum test. ***P < 0.001.

High mutational burden has been shown as an integral indication of neoantigen production and presents potential immunotherapeutic intervention across multiple cancer types.6, 8, 9, 39 Therefore, we have assessed neoantigenicity levels in both TMZ‐naïve and post‐TMZ‐treated hypermutated tumors compared to non‐hypermutated tumors. Notably, both hypermutated tumors exhibited a significant accumulation of neoantigen load (Fig. 3 c), suggesting potential immunotherapeutic strategy for these patients.

Discussion

In our study, we report for the first time a group treatment‐naïve adult glioma patients with a hypermutator phenotype and explore their spontaneous genomic profiles that constitute an alternative mutagenic process. Notably, TMZ‐naïve hypermutated tumors were primarily marked by absence of IDH1 somatic mutation and pMGMT methylation, two genomic features that were significantly associated with development of hypermutation in GBM.16 Transcriptional silencing of MGMT gene via promoter methylation, followed by dysfunction of the MMR encoding genes were presumably the underlying mechanism behind TMZ‐induced hypermutator phenotype.12, 40, 41 Furthermore, hypermutagenesis in glioma has been exclusive to recurrent tumors as it has been perceived as a direct result of TMZ intervention that subsequently lead to therapeutic resistance to standard treatment regimen. Therefore, it remained controversial whether inactivation of the MMR encoding genes or hypermutation was directly associated with acquisition of TMZ‐induced resistance. Our results have shown that spontaneous de novo hypermutated tumors follow an alternative evolutionary trajectory in acquiring continuous mutagenic process via combination of germline and somatic mutations of the mismatch repair genes. Despite having such features, patients with de novo hypermutation exhibited favorable clinical responses to the standard TMZ chemotherapy (Supporting Information Fig. S3), in contrast to the previous notion that inactivation of MSH6 could potentially confer TMZ resistance in glioblastoma.24 Although recent literatures have identified multiple brain tumor types to be associated with Lynch syndrome MMR mutational carriers, including MLH1 and MSH2,42, 43, 44 TMZ‐naïve patients primarily demonstrated high prevalence MSH3 and MSH6 germline variations. Our results suggest that inherited dysfunction of the MMR pathway through MSH3 or MSH6 mutation could potentially pose hereditary risk to genetic predisposition of carcinogenesis. Our speculations were further corroborated via expanded evaluation of the patients’ clinical history. Notably, patients with TMZ‐naïve hypermutator phenotype demonstrated a significantly higher incidence of cancer‐related occurrences in their immediate family members (Supporting Information Table S1). These observations suggest that inherited mutations in the MMR encoding genes could potentially promote tumor initiation and further impairment of the DNA repair mechanism via newly acquired somatic variation could result in hypermutation development.

A substantial body of evidence has highlighted significant association between patients with elevated neoantigen load and improved clinical response to checkpoint‐blockade immunotherapy across various cancer types including glioblastoma.6, 7, 8, 9, 39 Interestingly, our results showed increased neoantigen counts in both TMZ‐naïve and post‐TMZ‐treated hypermutated tumors, suggesting potential implication of checkpoint inhibitors in these patients. Even though hypermutagenesis in a newly‐diagnosed glioma is a rare event, it is important that all patients should to be evaluated for a hypermutated genotype as early identification of glioma patients prior to temozolomide treatment could potentially lead to new clinical trials in evaluating replacement of alkylating agents with checkpoint inhibitors. However, further studies are warranted in implementing immunotherapeutic approach in glioma patients as CNS malignancies are often void of cytotoxic immune cells such as T cells, which are critical components in immune checkpoint inhibition. Therefore, only a subset of glioma patients could be appropriate targets for immunotherapeutic approach in defined scenarios, such as presence of a hypermutator phenotype. As such, our results provide a conceptual groundwork for clinical practice against glioma patients with de novo hypermutation.

Although complete understanding of the underlying mechanism behind spontaneous hypermutagenesis in glioma requires an in‐depth experimental validation, we demonstrate that combined inherent and somatic mutations of the mismatch repair encoding genes could largely contribute to development of de novo hypermutator phenotype in gliomas. Furthermore, our results present an important step towards potential implementation of immunotherapy in glioma treatment.

Figure S1 MGMT mRNA expression level between TMZ‐naïve and TMZ‐treated hypermutated tumors. Relative mRNA expression of MGMT was calculated via RNA Sequencing.

Click here for additional data file.

Figure S2 Copy number variation inferred from single‐cell RNA sequencing. Blue represents copy number loss while red represents copy number gain.

Click here for additional data file.

Figure S3 Representative MRIs of BT102 and BT104 patients with pre‐ and post‐temozolomide treatments.

Click here for additional data file.

Table S1 Clinical features between patients with TMZ‐naïve non‐hypermutated and hypermutated tumors.

Click here for additional data file.