Single-nuclei transcriptomes from human adrenal gland reveal distinct cellular identities of low and high-risk neuroblastoma tumors

Authors

O. C. Bedoya-Reina

W. Li

M. Arceo

M. Plescher

P. Bullova

H. Pui

M. Kaucka

P. Kharchenko

T. Martinsson

J. Holmberg

I. Adameyko

Q. Deng

C. Larsson

C. C. Juhlin

P. Kogner

S. Schlisio

Doi

10.1038/s41467-021-24870-7

PMID: 34493726 · DOI: 10.1038/s41467-021-24870-7 · Journal: Nature Communications (2021)

TL;DR

Bedoya-Reina et al. sequenced single-nuclei/single-cell transcriptomes of 11 neuroblastoma tumors spanning different INRG risk groups and INSS stages, 3 postnatal human adrenal glands, and 5 mouse adrenal glands using Smart-Seq2. They identify a previously undescribed NTRK2+/CLDN11+ cholinergic progenitor population unique to postnatal human adrenal gland, and show that high-risk neuroblastoma tumors (typically in children >18 months) contain an undifferentiated cluster (nC3) that transcriptionally resembles this progenitor, while low-risk tumors resemble noradrenergic/chromaffin cells. The two clinical risk groups appear to represent two distinct cellular identities whose gene programs correlate with age-at-diagnosis and survival in a 498-patient SEQC bulk RNA-seq cohort.

Cohort & data

  • Neuroblastoma single-nuclei: 11 tumors, 0–79 months at diagnosis; 5 high-risk (MYCN-amplified and/or 11q-deleted), 1 intermediate, 5 low-risk; INSS stages spanning 1, 2/2B, 3, 4, and 4S. 4,224 single nuclei sequenced, 3,212 high-quality retained, average 709,676 reads/nucleus (Smart-Seq2 nuc-Seq).
  • Normal human adrenal gland (AG): 3 postnatal donors, 1,536 single nuclei sequenced, 1,322 high-quality, ~485,000 reads/nucleus (Smart-Seq2).
  • Normal mouse AG: 5 samples, 1,920 single whole cells, 1,763 high-quality, ~670,000 reads/cell (Smart-Seq2).
  • External comparators: GOSH-cohort 10X neuroblastoma data (n=8 tumors, Kildisiute et al. 2021), mouse E12–E13 adrenal anlagen (Furlan et al. 2017), human 8–14 PCW fetal adrenal gland (Dong et al. 2020), and the 498-sample SEQC bulk RNA-seq neuroblastoma cohort for survival/risk analyses (queried via the R2 Kaplan Scanner). Deconvolution was also applied to a 172-sample NB172 NCI TARGET bulk-sequenced cohort.
  • Validation: RNAscope multiplex fluorescent ISH on postnatal AG (ages 0 y, 4 y, adult) and on high-risk MYCN-amplified and low-risk 4S tumor sections.
  • Assay / methods: single-nuclei RNA-seq (snRNA-seq) and single-cell RNA-seq (scRNA-seq) with Smart-Seq2 full-length chemistry; RNAscope multiplex in situ hybridization for spatial validation.

Key findings

  • Clustering of 1,322 postnatal human AG nuclei (PAGODA) yielded 10 clusters including a novel progenitor cluster hC1 uniquely present in human (not mouse) postnatal gland, expressing SOX6, BCL11A, ERBB3, NTRK2 (encoding TRKB), RTTN, PTPRZ1, TP63, ASXL3, POU6F2, SEMA3E, LAMA3, DOCK7, and CLDN11 (all FDR <0.01, Welch’s t-test).
  • hC1 cells lack the canonical multipotent Schwann cell precursor (SCP) markers SOX10, FOXD3, S100B, and show no significant shared specific gene signature with mouse E12–E13 SCPs or human 8–14 PCW fetal adrenal SCPs (Fisher’s exact test, Supplementary Fig. 2c–d).
  • RNA velocity, entropy, and pseudotime analyses place hC1 as a proliferating progenitor (expresses MKI67, ASPM, BUB1) that gives rise to chromaffin cells (hC4), transitioning from an undifferentiated RTTN+ state through a noradrenergic DBH+ stage to an adrenergic PNMT+ state.
  • RNAscope ISH in postnatal AG (0 y, 4 y, adult) showed NTRK2+/CLDN11+ double-positive cells in adrenal capsule and medulla, mutually exclusive from TH+ chromaffin cells and from SOX10+ SCPs at all ages; most abundant at age 0, declining with age. Both hC1 progenitors and hC4 chromaffin cells express CHRNA7 (nAChR α7), supporting cholinergic identity.
  • Clustering of 3,212 neuroblastoma nuclei yielded 10 clusters: undifferentiated nC2/nC3, noradrenergic (NOR) nC5/nC7/nC8/nC9, mesenchymal stroma (MSC) nC1, endothelial nC4, macrophages nC6, T-cells nC10. High-risk cases had significantly higher proportions of stroma and undifferentiated cells; low-risk (including 4S) cases were dominated by NOR clusters (FDR <0.01, chi-square).
  • Undifferentiated nC3 cells significantly over-express MYCN, ALK, BRCA1, BRCA2, progenitor markers BCL11A, NTRK2, SOX5, SOX6, TP63, LGR5, USH2A, and migratory marker CLDN11; NOR clusters over-express NTRK1 (TRKA), TH, DBH, PHOX2A, PHOX2B, and ISL1 (all FDR <0.01, Welch’s t-test).
  • Mesenchymal markers PRRX1, YAP1, and PDGFRA are expressed in both nC3 and MSC nC1, whereas PDGFRB is significantly higher in MSC nC1.
  • Undifferentiated nC3 shares a significant gene signature with the postnatal human progenitor hC1 (including co-expression of NTRK2, collagens COL1A2/COL6A3/COL12A1, migratory genes CLDN11/DOCK7, and progenitor genes BCL11A/ERBB3/RTTN/TP63/ASXL3/POU6F2/SOX6), but not with SCP or sympathoblast populations from fetal adrenal gland. NOR clusters instead resemble human and mouse postnatal chromaffin cells and fetal (8–14 PCW) sympathoblast/chromaffin (FDR <0.01, Fisher’s exact).
  • Copy-number inference (inferCNV with immune cells as background, validated in 7 samples): undifferentiated nC3 harbored recurrent 17q gains, 1p losses, and 11q losses in samples K87, K10, 23, K55, and K3. Microarray confirmed 1p deletion (23, K10, K55), 11q loss (23, K87), and 17q gain (K10, K3, K87).
  • In a MYCN-amplified high-risk tumor (K10), RNAscope showed regions of TH+/MYCN+/NTRK1+/NTRK2−/ALK− cells with enlarged nuclei (possibly differentiating) surrounded by small-nucleus NTRK1−/NTRK2+ cells. A low-risk stage 4S tumor (K6) was homogeneously NTRK1+/TH+/NTRK2− with no PDGFRA or PRRX1 expression.
  • In the 498-sample SEQC cohort, the nC3 specific signature correlated directly with age-at-diagnosis (FDR = 8.12×10⁻³¹, Fisher’s exact) and with poor survival; NOR nC9 signature correlated inversely with age (FDR <1×10⁻⁵⁰). Kaplan-Meier log-rank tests (Bonferroni-corrected p <0.01) confirmed NOR nC7 and nC8 signatures associate with better outcome.
  • Gene ontology of nC3 genes correlated with poor survival enriched for DNA damage/repair and cell cycle; nC3 genes correlated with older age enriched for cell motility; NOR nC9 poor-survival genes enriched for RNA splicing/processing (FDR <0.01, Fisher’s exact).

Genes & alterations

  • MYCN — amplification; highly expressed in undifferentiated nC3 cluster of high-risk tumors. Five of the 11 tumors are high-risk with MYCN-amplified and/or 11q-deleted genotypes.
  • ALK — significantly over-expressed in undifferentiated nC3 cluster; co-expressed with MYCN in discrete tumor regions (RNAscope validation in K10).
  • NTRK1 (TRKA) — marker of noradrenergic NOR clusters (nC5/nC7/nC8/nC9) enriched in low-risk tumors; a known favorable-outcome biomarker.
  • NTRK2 (TRKB) — co-defining marker for the postnatal human progenitor hC1 and the high-risk undifferentiated nC3 cluster; associated with poor outcome.
  • PHOX2B, PHOX2A, TH, DBH, CHGA, CHGB, PNMT, ISL1 — noradrenergic/adrenergic panel marking NOR tumor clusters and postnatal chromaffin cells.
  • BRCA1, BRCA2 — significantly over-expressed in undifferentiated nC3 (high-risk-enriched) cluster.
  • Chromosomal alterations by inferCNV with microarray confirmation: recurrent 17q gain, 1p loss, 11q loss in nC3 cells of samples K87, K10, 23, K55, K3.
  • Progenitor/migratory gene program shared between hC1 (normal) and nC3 (high-risk tumor): BCL11A, ERBB3, RTTN, TP63, ASXL3, POU6F2, SOX6, CLDN11, DOCK7.
  • Mesenchymal markers in nC3/MSC nC1: PRRX1, YAP1, PDGFRA; PDGFRB specific to MSC nC1.
  • Cholinergic receptor CHRNA7 — expressed in both hC1 progenitors and hC4 chromaffin cells, supporting a cholinergic progenitor identity.

Clinical implications

  • Provides a transcriptional rationale for age-at-diagnosis as a neuroblastoma prognostic factor: low- and high-risk disease may arise from different progenitor pools (embryonic sympathoadrenal lineage vs a postnatal TRKB+ cholinergic progenitor unique to humans), suggesting two disease entities rather than a single disease spectrum.
  • Confirms NTRK1 vs NTRK2 as mutually exclusive outcome biomarkers at the single-cell level and validates this in situ by RNAscope.
  • nC3 signature stratifies 498 SEQC patients by survival (Kaplan-Meier Bonferroni-corrected p <0.01), supporting its potential as a prognostic transcriptomic biomarker, though the authors do not propose a specific assay.
  • The authors speculate that postnatally arising TRKB+/CLDN11+ cholinergic progenitors could be the cell-of-origin for high-risk neuroblastoma in older children; if validated, this would reframe target discovery away from SCP-based models.

Limitations & open questions

  • Small cohort (11 tumors, 3 human AG donors) — population-level claims rely on deconvolution into the 172-sample TARGET and 498-sample SEQC bulk cohorts, not on primary single-cell sampling across many patients.
  • Smart-Seq2 depth is high but total cell numbers are modest; rare subpopulations could be missed.
  • The identified hC1 progenitor is absent in mouse adrenal gland, so functional validation via genetic models is not straightforward.
  • The putative lineage link (hC1 → nC3) is based on transcriptional similarity and RNA velocity, not lineage tracing; the authors do not demonstrate that hC1 cells transform into tumor cells.
  • Cluster annotations for reference fetal data (Dong et al. 2020) are “currently debated” (authors note labels follow Kildisiute et al. and Bedoya-Reina & Schlisio revisions).
  • Whether CHRNA7-mediated cholinergic signaling is therapeutically actionable in high-risk NBL is not addressed.

Citations from this paper used in the wiki

  • “Here we deep-sequence full-length coverage RNA from single nuclei of tumors (n=11) across different risk groups” — abstract / introduction.
  • “We identify a cluster of TRKB+ cholinergic cells unique to human postnatal adrenal gland, that differ from previously described embryonic Schwann cell precursors (SCP). This TRKB+ population of cells shares a specific gene signature with a cluster of undifferentiated cells of mesenchymal nature enriched in high-risk neuroblastoma.”
  • “The gene signature of these mesenchymal cells is in turn coupled with lower patient survival probability and older age-at-diagnosis when tested in a larger cohort of 498 neuroblastoma patients [SEQC].”
  • “undifferentiated nC3 clusters presented a significantly high expression of MYCN, ALK, BRCA1, and BRCA2 genes and progenitor markers BCL11A, NTRK2, SOX5, SOX6, TP63, LGR5, and USH2A (FDR < 0.01, Welch’s t-test).”
  • “the noradrenergic clusters (NOR, nC5, nC7, nC8, and nC9) expressed significantly high NTRK1 (TRKA) and noradrenergic markers TH, DBH, PHOX2A, PHOX2B, and ISL1 (FDR < 0.01, Welch’s t-test).”
  • “For most of samples, a significantly higher number of cells presented gains in 17q, losses in 1p or in 11q, in the undifferentiated nC3 cluster (K87, K10, 23, K55, and K3)… 1p deletion was confirmed by microarrays for samples 23, K10, and K55, 11q loss for samples 23 and K87, and 17q gains for samples K10, K3, and K87.”
  • “Eleven neuroblastoma samples were collected from patients at different ages (0–79 months at diagnosis)… five patients were classified as high-risk, one as intermediate risk and five as low-risk” — Methods.

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