Single-cell RNA-seq supports a developmental hierarchy in human oligodendroglioma

Authors

Itay Tirosh

Andrew S. Venteicher

Christine Hebert

Leah E. Escalante

Anoop P. Patel

Keren Yizhak

Aviv Regev

Mario L. Suvà

Doi

10.1038/nature20123

PMID: 27806376 · DOI: 10.1038/nature20123 · Journal: Nature (2016)

TL;DR

Tirosh, Venteicher, Suvà and Regev profiled 4,347 single cells from six untreated grade II human oligodendrogliomas (all IDH1 or IDH2 mutant and 1p/19q co-deleted) by Smart-seq2 single-cell RNA-seq and reconstructed their developmental programs from genome-wide expression signatures PMID:27806376. Most cancer cells were differentiated along two glial lineages (oligodendrocyte-like and astrocyte-like), while a rare subpopulation expressed a neural stem/progenitor program (SOX2/SOX4/SOX11 high) and accounted for nearly all of the proliferating cells. CNV sub-clones and subclonal point mutations (including a CIC mutation in MGH53) spanned all three compartments, arguing that oligodendroglioma architecture is primarily dictated by non-genetic developmental programs and supporting a cancer stem cell model with a neural-progenitor-like rather than oligodendrocyte-progenitor-like apex.

Cohort & data

  • 6 untreated grade II oligodendrogliomas (MGH36, MGH53, MGH54, MGH60, MGH93, MGH97), all with confirmed IDH1 or IDH2 mutation and 1p/19q co-deletion, collected at Massachusetts General Hospital under IRB 1999P008145 PMID:27806376.
  • 4,347 single cells passing QC (≥3,000 detected genes and housekeeping expression threshold); three tumours sequenced deeply (791–1,229 cells/tumour), three at medium depth (430–598 cells/tumour) PMID:27806376.
  • Assay: plate-based Smart-seq2 scRNA-seq on FACS-sorted live single cells (calcein AM⁺/TO-PRO-3⁻); 96 cells from MGH60 also profiled with a UMI-based protocol with concordant results PMID:27806376.
  • Orthogonal assays: bulk whole-exome sequencing for subclonal mutation calling (ABSOLUTE), FISH on 1p36/1q25 and 19q13/19p13 to confirm co-deletion, RNA in situ hybridization (ApoE, OMG, SOX4, CCND2, Ki-67), immunohistochemistry (GFAP, Ki-67), and mutation-sensitive qPCR for CIC genotyping of single cells PMID:27806376.
  • External reference cohort: 69 bulk oligodendroglioma samples from TCGA LGG used to correlate cell-cycle and stem/progenitor signatures PMID:27806376.

Key findings

  • Across 4,347 cells, combined PCA revealed two dominant glial lineage programs (oligodendrocytic, marked by high OLIG1, OLIG2, OMG; astrocytic, marked by high APOE, ALDOC, SOX9) — 137 “oligo-high” and 128 “astro-high” signature genes (Supplementary Table 1) PMID:27806376.
  • A third program (PC2/PC3, 63 genes) defines a rare “stem/progenitor” compartment enriched for neurodevelopmental TFs SOX4, SOX11, SOX2 and for NFIB, ASCL1, CHD7, CD24, BOC, TCF4, and CCND2 — overlaps glioblastoma CSC signatures (hypergeometric P = 1.5×10⁻⁴) and is preferentially expressed in prenatal human brain (Allen Brain Atlas, t-test P = 8×10⁻¹⁸) and in fetal human brain single cells (hypergeometric P = 0.006) PMID:27806376.
  • Proliferation is confined to the stem/progenitor compartment. 1.5–8 % of cells per tumour score as cycling (G1/S or G2/M signatures); almost all cycling cancer cells are stem/progenitor or undifferentiated, confirmed by MKI67 (Ki-67) staining and ISH across the discovery cohort and a 10-tumour validation cohort PMID:27806376.
  • The PC2/PC3 stemness program matches human NPCs (profiled by scRNA-seq in this study, t-test P = 2×10⁻³⁵) and activated mouse NSCs (t-test P = 3×10⁻⁶) more closely than OPCs, suggesting a tri-potent neural-progenitor-like cell of origin rather than an OPC PMID:27806376.
  • CCND2 is preferentially expressed in stem/progenitor tumour cells (both cycling and non-cycling), while differentiated cells express CCND1 and CCND3 — mirroring normal neural development and consistent with CCND2’s role priming cells for cell cycle entry PMID:27806376.
  • In bulk TCGA LGG (n = 69 oligodendrogliomas), the stem/progenitor signature correlates with the cell-cycle signature, supporting the single-cell findings in a larger independent cohort PMID:27806376.
  • Genetic subclones recapitulate the same hierarchy. Inferred-CNV subclones in MGH36 and MGH97, a subclonal CIC mutation (~30 % VAF) in MGH53, 22 subclonal point mutations, and a subclonal LOH event all span all three expression compartments — architecture is not driven by any single genetic clone PMID:27806376.
  • Mutation-sensitive qPCR on single cells (28 CIC-mutant vs 27 CIC-wild-type cells from MGH53) identified a CIC-specific expression program including upregulated ETV1 and ETV5 in CIC-mutant cells, consistent with loss of CIC repression; yet mutant and WT cells both populated all three states PMID:27806376.
  • CNV sub-clone composition differed quantitatively: MGH36 clone 1 had more stem/progenitors than clone 2 (Fisher’s exact P < 10⁻⁹); MGH97 clone 2 was enriched for stem/progenitors (P < 10⁻¹⁶), with matching differences in cycling-cell frequency — suggesting genetic evolution can modulate self-renewal/differentiation ratios even if it does not dictate the hierarchy itself PMID:27806376.

Genes & alterations

  • IDH1 / IDH2 — defining hotspot mutations (confirmed by Sanger/WES); single-base Bowtie alignment used to detect IDH1 point mutations in scRNA-seq reads PMID:27806376.
  • CIC — subclonal loss-of-function mutation in MGH53 (~30 % VAF); mutant cells show increased ETV1 and ETV5 expression consistent with derepression of CIC targets; mutation spans all three developmental compartments PMID:27806376.
  • 1p/19q co-deletion — confirmed in every tumour by CNV inference from scRNA-seq and by FISH (1p36/1q25 and 19q13/19p13 dual-colour probes) PMID:27806376.
  • Oligodendrocytic lineage markers OLIG1, OLIG2, OMG high in differentiated oligo-like tumour cells PMID:27806376.
  • Astrocytic lineage markers APOE, ALDOC, SOX9, GFAP high in differentiated astro-like tumour cells PMID:27806376.
  • Stem/progenitor program: SOX2, SOX4, SOX11, NFIB, ASCL1, CHD7, CD24, BOC, TCF4, CCND2 PMID:27806376.
  • Cyclin switching: CCND2 in stem/progenitors vs CCND1 / CCND3 in differentiated cells PMID:27806376.
  • MKI67 — proliferation marker, used for IHC confirmation of cycling-cell rarity (1.5–8 %) and stem/progenitor-biased distribution PMID:27806376.

Clinical implications

  • The stem/progenitor transcriptional program defines a small, proliferating subpopulation that is a plausible therapeutic target; differentiated oligo-like and astro-like cells are largely non-cycling and therefore unlikely to fuel long-term tumour growth PMID:27806376.
  • Because grade II oligodendroglioma stem/progenitor cells more closely resemble a tri-potent neural progenitor than an OPC, therapies targeting OPC biology may miss the proliferative apex of these tumours PMID:27806376.
  • The finding that CNV and point-mutation subclones each contain cells from all three compartments means that bulk-sequencing-based subclone targeting will not, on its own, eliminate the cycling CSC pool PMID:27806376.

Limitations & open questions

  • Grade II oligodendroglioma cells do not engraft in xenotransplantation, so the stem/ progenitor program was not functionally validated — its “stemness” is inferred from inverse association with differentiation, enriched proliferation, and similarity to normal NSC/NPC programs PMID:27806376.
  • Subclonal genetic information from scRNA-seq (inferred CNVs + ABSOLUTE-derived point mutations retrieved from RNA reads) is partial; a full phylogenetic tree cannot be reconstructed and a residual genetic contribution to the hierarchy cannot be categorically excluded PMID:27806376.
  • Generality beyond grade II IDH-mutant 1p/19q-codeleted oligodendroglioma (e.g. to astrocytomas, grade III lesions, IDH-wild-type glioma) is not established here and is flagged as a direction for future work PMID:27806376.
  • Opportunities for clinical translation (pharmacological targeting of the stem/progenitor program) are raised but not explored experimentally PMID:27806376.

Citations from this paper used in the wiki

  • “We profile 4,347 single cells from six IDH1 or IDH2 mutant human oligodendrogliomas by RNA sequencing (RNA-seq) and reconstruct their developmental programs from genome-wide expression signatures.” (Abstract)
  • “We focused on oligodendroglioma, an incurable glioma characterized by mutations in IDH1 or IDH2 and co-deletion of chromosome arms 1p and 19q.” (Introduction)
  • “The 20 highest-ranking genes include SOX4, SOX11 and SOX2, neurodevelopmental transcription factors critical to neural stem cells and glioma CSCs.”
  • “We found a small proportion of cells in each tumour (1.5–8%) that were proliferating … Almost all proliferating cancer cells were confined to the stem/progenitor and undifferentiated subpopulation of the tumour.”
  • “Strong correlation between our cell-cycle and stem/progenitor signatures across 69 bulk oligodendroglioma samples in The Cancer Genome Atlas.”
  • “We identified a subclonal mutation of CIC (~30% frequency in MGH53) … we identified a signature of expression changes between the mutant CIC and wild-type cells, including increased expression of ETV1 and ETV5 in mutant CIC cells.”
  • “Our single-cell profiles suggest that oligodendroglioma stem/progenitor cells more closely resemble a primitive tri-potent neural cell type, such as NSC or NPC than a more committed glial progenitor like an OPC.”

This page was processed by crosslinker on 2026-05-04.