Divergent clonal selection dominates medulloblastoma at recurrence

Authors

A. Sorana Morrissy

Livia Garzia

David J. H. Shih

Scott Zuyderduyn

Xi Huang

Patryk Skowron

Marc Remke

Florence M. G. Cavalli

Vijay Ramaswamy

Patricia E. Lindsay

Salomeh Jelveh

Michael D. Taylor

Marco A. Marra

Doi

10.1038/nature16478

PMID: 26760213 · DOI: 10.1038/nature16478 · Journal: Nature (2016)

TL;DR

Morrissy et al. combined a Sleeping Beauty transposon-driven, surgery-plus-craniospinal-irradiation mouse model of Shh medulloblastoma with whole-genome and whole-exome sequencing of 33 patient-matched diagnostic/recurrence pairs (46 total human samples) to ask whether targets identified at diagnosis are still present at relapse. In both species, the dominant clone at recurrence shares <12% of clonal genetic events with the diagnostic sample, instead arising via clonal selection of pre-existing minor subclones plus therapy-induced de novo mutations. Recurrent Shh medulloblastoma converges on TP53-pathway loss and DYNC1H1 / chr14q deficits in 79% of cases, while drug-actionable targets at recurrence differ from those at diagnosis in the majority of patients — arguing that targeted-therapy trials at relapse must re-biopsy to confirm target presence.

Cohort & data

  • Patient cohort: 33 patient-matched primary + recurrent MBL pairs — 15 pairs with germline controls profiled by WGS, 18 pairs without germline by WGS, 10 additional recurrence-only samples with germline by WGS, plus 3 FFPE pairs by WES (46 total samples). Patient samples are catalogued under cBioPortal study mbl_sickkids_2016. Raw sequencing was deposited at EGA (EGAD00001000946).
  • Subgroup coverage: all four molecular subgroups of medulloblastoma (Shh, Wnt, Group 3, Group 4) represented; Shh and Group 4 are the most analysed.
  • Mouse model: 38 Ptch+/−/Math1-SB11/T2Onc(2) mice underwent subtotal tumour resection and 18 × 2 Gy fractions of CT-guided craniospinal irradiation over 4 weeks; 11/18 treated mice (61%) developed local/metastatic relapse. Median medulloblastoma-free survival was 118 days for treated vs 5 days for untreated control mice.
  • Assays / methods: whole-genome-seq (BWA / GRCh37-lite alignment), whole-exome-seq (Nextera Rapid Capture for FFPE), targeted deep amplicon sequencing of 192 patient-specific SNVs across 20 patients, transposon-insertion sequencing in mice analysed via gCIS, strelka and MutationSeq for SNV calling, varscan for indels, EMu for mutation spectra, titan-cna for allele-specific CN, pyclone and EXPANDS for clonal-prevalence inference. RNA-seq supported subgroup-stability and pathway-enrichment analyses (PMID:26760213).

Key findings

  • Massive genetic divergence at recurrence. Across 14 patients with germline data, on average only 11.8% of somatic SNVs/indels were shared between matched diagnostic and recurrent tumours; in the Sleeping Beauty mouse model the overlap of clonal driver gCISs was <5% (PMID:26760213).
  • Mutational burden rises ~5-fold post-therapy. In 13/15 patient pairs the somatic mutational burden increased by an average of 5× at recurrence (P = 2.7 × 10⁻⁴). Structural-variant burden also rose.
  • Clonal restructuring. Clonal mutations grew 1.9-fold as a proportion of all mutations post-therapy (P = 8.7 × 10⁻³). 60.5% of damaging clonal primary mutations decreased in abundance post-therapy — 25.9% became subclonal and 34.6% disappeared. Only 25% of patients retained the full clonal SNV set. Damaging clonal mutations post-relapse outnumbered retained clonal events 5:1.
  • Recurrence arises from pre-existing minor clones. Ultra-deep targeted resequencing across 20 patients confirmed expansion of clones initially present at <5% in the therapy-naive tumour in 16/20 patients; assay sensitivity was demonstrated down to 2/10 000 (subclonal frequency).
  • Trunk-driver switching. Patient MB-REC-12 had a homozygous PTCH1 driver clonally dominant in the primary tumour that was completely eradicated by therapy; the recurrence was driven by a sister clone with CDKN2A / CDKN2B homozygous loss on wild-type chr9q, while still retaining Shh transcriptional identity.
  • Structural-variant turnover. MB-REC-14 lost a diagnostic TERT amplification at recurrence; MB-REC-09 acquired MYC-locus chromothripsis only at recurrence (P = 3.97 × 10⁻⁷).
  • Therapeutic-target instability. Of 9/15 patients with at least one drug-actionable alteration, only 4/9 (44.4%) retained the diagnostic target set at recurrence; 2/9 had a complete actionable-target switch, and 3/9 gained new actionable targets post-therapy. One case showed convergent evolution with distinct SMO mutations at the two time points.
  • TP53-pathway convergence at recurrence (Shh). TP53-pathway alterations (KEGG04115) appeared in 12/23 (52.2%) human recurrences and TP53-gene mutations in 6/23 (26.1%), predominantly in Shh tumours. In the mouse model, Trp53 insertions were subclonal at diagnosis but clonal at recurrence, with reduced p21 expression confirming loss-of-function. Drosophila brain tumours expressing dominant-negative p53 (p53^R159N) lost the radiation-induced apoptotic response.
  • DYNC1H1 / chr14q loss. Damaging DYNC1H1 mutations occurred in 16% of post-therapy Shh tumours and were mutually exclusive with TP53 mutations. Among therapy-naive Shh tumours, chr14q loss was prognostic (worse outcome; recurrence rate 50%, 7/14 with loss vs 30%, 7/23 balanced). Cumulatively, 6/15 (40%) Shh recurrences functionally lost one DYNC1H1 allele. TP53, DYNC1H1 or chr14q deficits affected 79% of recurrent Shh medulloblastoma.
  • Mutational signatures shift. Four SNV signatures were detected; recurrences had significantly elevated transversion rates, consistent with therapy-induced DNA damage.
  • Branched clonal evolution. EXPANDS modelling of 14 germline-matched cases showed branched evolution in all 14; in 8/14 all recurrent clones derived from a single primary lineage, while 6/14 retained higher phylogenetic similarity to the primary. 71.4% of patients had increased clonal-population counts post-therapy; Group 4 tumours were significantly more heterogeneous (Shannon Index, P = 0.029).
  • Transcriptional convergence. Relapsed Shh tumours showed enriched P53-signalling and apoptosis gene sets; relapsed Group 4 tumours enriched for extracellular-matrix and cell-surface receptor signalling.

Genes & alterations

  • TP53 — somatic mutations enriched at recurrence (26.1% of human recurrences); TP53-pathway alterations in 52.2%. Mouse model: clonal Trp53 transposon insertions emerge at recurrence and predict worse prognosis. Drosophila dominant-negative p53 abrogates radiation-induced apoptosis (PMID:26760213).
  • PTCH1 — homozygous driver in primary MB-REC-12 (heterozygous mutation + chr9q LOH) was eradicated at recurrence; the relapse derived from an ancestral chr9q-wild-type sister clone.
  • CDKN2A / CDKN2B — homozygous loss drove the recurrent clone in MB-REC-12 after PTCH1-driven primary clone was wiped out.
  • DYNC1H1 — damaging mutations in 16% of post-therapy Shh tumours; mutually exclusive with TP53 mutations; chr14q loss (containing the gene) is prognostic in Shh medulloblastoma only.
  • SMO — different mutations seen pre- and post-therapy in one patient, evidence of convergent evolution.
  • MYC — locus chromothripsis acquired only at recurrence in MB-REC-09 (P = 3.97 × 10⁻⁷).
  • TERT — amplification present at diagnosis but lost at recurrence in MB-REC-14, illustrating loss of a potential target.
  • CREBBP (CBP) — only mouse-model gCIS observed across all compartments (primary + local + metastatic recurrence).
  • ARID1B, TCF4 — additional mouse driver gCISs shared between local and metastatic recurrences but not diagnostic samples.

Clinical implications

  • Re-biopsy at recurrence should be mandatory for any targeted-therapy trial in medulloblastoma — putative drug targets identified in therapy-naive tumours are unreliable predictors of targets present in the dominant relapse clone (PMID:26760213).
  • Subclonal targetable events in primary tumours may anticipate the targets that will dominate at relapse; combination therapies that pre-empt the resistant subclone (e.g. add a TP53-pathway-aware agent up front) are a logical extension.
  • Molecular subgroup affiliation is stable even when trunk drivers switch — subgroup-level therapeutic strategies (e.g. targeting Shh-subgroup biology broadly) remain rational at recurrence, while gene-level targeted strategies are not.
  • TP53-pathway intactness predicts radiation response. Mouse and Drosophila data link p53 dysfunction to radiation-resistance and recurrence in Shh tumours, consistent with known fatal prognosis of TP53-mutant Shh medulloblastoma in humans.
  • chr14q loss / DYNC1H1 deficit is a prognostic marker specific to therapy-naive Shh medulloblastoma (50% recurrence with loss vs 30% balanced).

Limitations & open questions

  • Cohort size is modest (15 germline-matched WGS pairs) — recurrent-driver identification is statistically underpowered for individual genes; convergence is best seen at the pathway level.
  • Sample-quality heterogeneity: 18 pairs lack germline DNA (somatic calls enriched indirectly), and 3 pairs are FFPE WES; cross-cohort statistical comparisons are limited.
  • The Sleeping Beauty mouse model captures only Shh-subgroup medulloblastoma and the Ptch+/− lineage; clonal-selection dynamics in Group 3/Group 4 tumours rely on the smaller human transcriptomic comparison (n = 3 Group 4).
  • Whether ‘anticipatory therapy’ (treating the predicted resistance pathway up front) actually reduces recurrence in patients is untested.
  • The mechanistic basis of DYNC1H1 / chr14q loss in Shh medulloblastoma is not resolved.
  • Whether the same diagnosis-to-recurrence divergence holds for chemotherapy-only regimens (no radiation) is not addressed; cohort received radiation ± chemotherapy.

Citations from this paper used in the wiki

  • “Genetic events in recurrent murine medulloblastoma exhibit a very poor overlap with those in matched murine diagnostic samples (<5%). Whole-genome sequencing of 33 pairs of human diagnostic and post-therapy medulloblastomas demonstrated substantial genetic divergence of the dominant clone after therapy (<12% diagnostic events were retained at recurrence).” (Abstract)
  • “in 13/15 patients the somatic mutational burden increased by an average of fivefold at recurrence (P value = 2.7 × 10⁻⁴).” (Genetic divergence in recurrent human medulloblastoma)
  • “Two-thirds of patients have deleterious events in at least one gene for which an anti-neoplastic drug interaction has been defined (n = 9 of 15) … The current assumption that all putative drug targets present at diagnosis are retained post-therapy is only valid in 4 of these patients (44.4%).” (Genetic divergence in recurrent human medulloblastoma)
  • “we observed TP53 gene and pathway mutations, DYNC1H1 mutations, or chr14q losses in 79% of recurrent Shh medulloblastoma.” (Convergent biological pathways at relapse)
  • “Patient MB-REC-12 harbours a clinically compelling example of a homozygous PTCH1 driver mutation that is clonally dominant in the primary tumour and completely eradicated by therapy … a sub-lineage of these cells driven by CDKN2A and CDKN2B−/− loss successfully reconstituted the tumour post-therapy.” (Clonal selection drives recurrence)
  • “we find evidence for significant expansion of clones initially present at <5% in the therapy-naive tumour in 16/20 patients” (Clonal selection drives recurrence)

This page was processed by crosslinker on 2026-05-14.