Exploiting nanopore sequencing for characterization and grading of IDH-mutant gliomas

Abstract

The 2021 WHO Classification of Central Nervous System Tumors recommended evaluation of cyclin-dependent kinase inhibitor 2A/B (CDKN2A/B) deletion in addition to codeletion of 1p/19q to characterize IDH-mutant gliomas. Here, we demonstrated the use of a nanopore-based copy-number variation sequencing (nCNV-seq) approach to simultaneously identify deletions of CDKN2A/B and 1p/19q. The nCNV-seq approach was initially evaluated on three distinct glioma cell lines and then applied to 19 IDH-mutant gliomas (8 astrocytomas and 11 oligodendrogliomas) from patients. The whole-arm 1p/19q codeletion was detected in all oligodendrogliomas with high concordance among nCNV-seq, FISH, DNA methylation profiling, and whole-genome sequencing. For the CDKN2A/B deletion, nCNV-seq detected the loss in both astrocytoma and oligodendroglioma, with strong correlation with the CNV profiles derived from whole-genome sequencing (Pearson correlation (r) = 0.95, P < 2.2 × 10^-16 to r = 0.99, P < 2.2 × 10^-16 ) and methylome profiling. Furthermore, nCNV-seq can differentiate between homozygous and hemizygous deletions of CDKN2A/B. Taken together, nCNV-seq holds promise as a new, alternative approach for a rapid and simultaneous detection of the molecular signatures of IDH-mutant gliomas without capital expenditure for a sequencer.

Keywords: 1p/19q codeletion; CDKN2A/B; CNV; IDH; astrocytoma; glioma; long-read technology; molecular marker; oligodendroglioma.

FIGURE 1

Performance of nCNV‐seq approach with 3 well‐characterized cell lines. (A) Cell lines and known molecular features. (B) CNV profile after 48 h of sequencing of BT88 DNA and with data down‐sampled to 10, 50, 100, and 500 Mb (bin ratios = 5 K). Left: nCNV‐seq. Right: nanopore‐based WGS. Deletion of 1p/19q was observed 25 min after sequencing start. (C) CNV profile of HOG and U87 genomes with data down‐sampled to 50 Mb. At 50 min, intact 1p/19q was observed in HOG, and CDKN2A/B loss was presented in U87. CNV, copy number variations; nCNV‐seq, nanopore‐based CNV sequencing approach; WGS, whole genome sequencing.

FIGURE 2

A total of 19 clinical samples used for nCNV‐seq. (A) Classification of samples from patients with the IDH‐mutant gliomas based on WHO Classification 2021. (B) Summary of demographic and molecular features of the 19 IDH‐mutant gliomas. Each row represents a single case. IDH1‐R132H was tested by immunostaining and/or WGS. pTERT was tested by WGS. ATRX, alpha‐thalassemia/mental retardation syndrome X‐linked; CDKN2A/B, cyclin‐dependent kinase inhibitor 2A/B; EPIC array, Infinium MethylationEPIC BeadChip DNA methylation arrays; FISH, fluorescence in situ hybridization; IDH, isocitrate dehydrogenase; NGS, next‐generation sequencing; WHO CNS5, 2021 WHO Classification of Tumors of the Central Nervous System, Fifth Edition.

FIGURE 3

Detection of chromosome gains and losses with three different methods: nCNV‐seq, Illumina WGS, EPIC array. (A) Representative patient with 1p/19q codeleted oligodendrogliomas. (B) Representative patient diagnosed with astrocytoma with 1p/19q intact and CDKN2A/B deleted. Pearson's correlation between nCNV‐seq and WGS is shown for both patients. chr, chromosome; EGFR, epidermal growth factor receptor; PTEN, phostphatase and tensin homolog deleted on chr 10.

FIGURE 4

Comparative results. (A) CNV profile of a normal diploid genome. (B) CNV profile of normal with in silico removed the bin of deletion markers, 50% deletion of chromosome 1p and 19q (hemizygous deletion) and 100% deletion of CDKN2A/B (homozygous deletion). (C) Scatter plots depicting the relationship between tumor fraction and predicted tumor fraction using IchorCNA. The diagonal line (blue line) reflected the perfect positive relationship between the variables (red dash). (D) Tumor fraction (ranging from 0 to 1). In silico mixing normal to cancer at different ratio from 0 or the absence of tumor cells (brown), 0.25 or relatively low tumor burden (green), 0.50 equal balance between tumor cells and non‐tumor cells (red), 0.75% or 75% of the tissue sample comprises tumor cells (purple), 1 represents the maximum tumor burden (yellow). This can be considered as the tumor fraction base lines. (E) CNV profiles of 3 patients with no CDKN2A/B deletion (top), with CDKN2A/B homozygous deletion and ~40% tumor (middle), and with CDKN2A/B homozygous deletion and ~80% tumor cells (bottom). The copy number prediction results were represented as horizontal lines, each colored according to its specific copy number classification. These include copy neutral (NEUT, 2 copies, in blue) hemizygous deletions (HETD, 1 copy, represented in green), and homozygous (HOMD, red). For enhanced visibility, we have color‐highlighted the regions associated with CDKN2A/B, 1p, and 19q. On the right panel of the figure, a line plot was implemented to indicate the homozygous deletion of CDKN2A/B. This plot delineates the levels of neutral (NEUT, blue), hemizygous (HETD, green), and homozygous (HOMD, red) deletions of CDKN2A/B based on the closest 0.80 tumor fraction (TF) from our tumor‐normal admixtures model (see in Method section), relative to the 0.84 TF prediction from ichorCNA.

FIGURE 5

Five CNV profiles were generated from EPIC array data, nCNV‐seq, and nCNV‐seq with tumor fraction. The homozygous deletion status of each method was evaluated by visual assessment of copy number profiles with the previous report cutoff, ≤−0.415 for EPIC array [33], and ≤−1 for nCNV‐seq (without considering tumor fraction). For nCNV‐seq with taking tumor fraction into account (right panel), green star was the predicted point of hemizygous deletion, red star was the expected point of homozygous deletion. The copy number prediction results were represented as horizontal lines, each colored according to its specific copy number classification. These include copy neutral (NEUT, 2 copies or diploid cell, in blue) hemizygous deletions (HETD, 1 copy, represented in green), and homozygous deletion (HOMD, red).

FIGURE 6

Schematic of nCNV‐seq workflow. Total sample‐to‐answer time is approximately 4 h. Time required for individual steps is indicated.