Dynamics of genomic clones in breast cancer patient xenografts at single cell resolution

PMID: 25470049 · DOI: 10.1038/nature13952 · Journal: Nature (2015)

TL;DR

Eirew, Steif, Khattra and colleagues generated 30 serially-passaged patient-derived breast cancer xenograft (PDX) lines from 55 patients implanted into NOD/SCID/IL2rγ⁻/⁻ (NSG) and NOD/RAG1⁻/⁻IL2rγ⁻/⁻ (NRG) mice, and applied whole-genome sequencing, deep targeted amplicon sequencing of 3,187 SNVs and 132 SVs, Affymetrix SNP6.0 CNA/LOH profiling, and single-cell targeted sequencing of 210 nuclei (from PDXs SA494 and SA501) to track clonal dynamics across up to 16 transplant generations over 3 years. In all 15 patient-xenograft series subjected to WGS, clonal selection occurred at engraftment, ranging from extreme expansion of clones representing <5% of the originating tumour to moderate polyclonal engraftment, and clonal dynamics frequently continued over serial passages. Single-cell phylogenetic inference confirmed that bulk-population PyClone mutation clusters correspond to real clonal genotypes, and replicate transplants of the same starting tumour into different mice produced concordant clonal expansion patterns (median Pearson r 0.91–0.94 across same-passage replicates in SA501, SA535, SA532, SA429), indicating that clonal fitness is largely deterministic and reproducible — a finding with direct consequences for the use of BRCA PDXs in drug-response and tumour-biology studies.

Cohort & data

• Patients: 55 BRCA patients contributed tissue (organoid suspensions); 15 patient series (10 primary tumour-derived, 5 pleural effusion / metastasis-derived) had matched tumour, normal, and xenograft whole-genome sequencing.
• Xenograft lines: 30 serially transplanted PDX lines, up to 16 generations over 3 years, in NSG and NRG mice; transplant sites included subcutaneous (SC), subrenal capsule (SR), and mammary fat pad (MFP). See subcutaneous-xenograft.
• Dataset: brca_bccrc_xenograft_2014 (EGA accession EGAS00001000952; processed data available at cbioportal.org).
• Sequencing volumes: WGSS on 47 tumour/xenograft/normal DNA samples (median depth 45.1×); deep targeted amplicon sequencing across 56 additional xenograft passages validating 3,187 somatic SNVs (100–300 per series) and 132 SVs; Affymetrix SNP6.0 array for CNA/LOH. Single-cell targeted re-sequencing of 210 nuclei across 40 SNV + 7 germline amplicons (SA494) and 45 SNV + 10 germline amplicons (SA501) using microfluidic isolation.
• Subtypes represented: Examples spanning ER+, HER2+, and triple-negative (TN) tumours, and both primary and metastatic origin.
• Methods: whole-genome-seq, targeted-dna-seq, single-cell-dna-seq, single-cell-genotyping, affymetrix-snp6, pyclone (Bayesian SNV clustering), titan-cna (CNA/LOH clonal inference), phylogenetic-tree-reconstruction (MrBayes 3.2).

Key findings

• Universal clonal selection on engraftment. All 15 tumour–xenograft series showed clonal selection: 6.5–32.1% of high-confidence SNVs were prevalent in the xenograft but at or below detection in the tumour, and 0.2–19.4% of SNVs went the opposite direction, while 53.0–92.9% of SNVs were shared across the tumour-xenograft pair (PMID:25470049).
• Extreme minor-clone expansion in several cases. PyClone identified mutation clusters with 75–100% prevalence in xenografts but 0–15% prevalence in tumours (e.g., clusters 3, 4, 3, 2, 8, 2, 2 of SA494, SA495, SA500, SA530, SA532, SA533, SA535) — implying initially minor (<5%) clones expanded to dominance in the PDX.
• Mutation load consistent with primary breast cancers. 4.3–27.7 × 10³ somatic SNVs genome-wide (57–1040 in coding regions); 34–67% of genome under CNA/LOH; SV burden consistent with prior breast cancer studies (PMID:22495314, PMID:22522925). Tumour-xenograft pairs showed comparable nucleotide-substitution patterns, indicating mutational processes are preserved post-engraftment.
• Polyclonal post-engraftment evolution. SA493, SA494, SA495, SA500, and SA531 showed polyclonal sub-structure specific to the xenograft after initial expansion, indicating that engraftment selection is permissive to additional clonal evolution. Polyclonal engraftment without strong initial bottleneck was evident in SA493, SA501, SA531, SA532.
• Two modes of clonal dynamics across serial passages. Cases with strong initial selection (SA500, SA530, SA494, SA535) stabilised over later passages, whereas cases with moderate initial dynamics (e.g., SA501 clusters 2, 3, 8) exhibited progressively more dramatic dynamics — including monotonic minor-to-dominant clonal conversion over multiple passages. Both modes occurred across ER+, HER2+ and triple-negative subtypes and across primary and metastatic origins.
• Concordant SNV and CNA clonal dynamics. TITAN-based CNA clonal inference recapitulated SNV-based PyClone dynamics: minor-subclone expansion (SA494, SA495, SA532, SA533) and polyclonal engraftment (SA493, SA501) appeared in both data modalities.
• Xenograft-specific TP53 deletion in SA500. A xenograft-specific deletion event encompassing TP53 was identified in SA500 and coincided with retention of a somatic SNV, demonstrating a structural event arising or selected post-engraftment.
• Single-cell validation of PyClone clusters. Bayesian phylogenetic inference on 62 SA494 tumour + 58 SA494 X4 nuclei placed tumour and xenograft cells in distinct clades; PyClone cluster 2 (dominant tumour clone) and cluster 3 (minor engrafting clone) SNVs were mutually exclusive in tumour vs xenograft nuclei, confirming a <5% originating clone expanded to dominance. Across 90 SA501 nuclei from passages X1, X2, X4, five major clonal genotypes (A–E) were resolved, with Genotype E confined exclusively to X4, mirroring bulk PyClone predictions of cascading subclonal evolution.
• Reproducible deterministic dynamics in replicate transplants. In 4 of 5 series tested, parallel clonal dynamics of the same mutation cluster occurred in independent replicate mice (SA501: 2/2 replicates at X3, 4/4 at X4; SA535 X1 3/3; SA532 X1 3/3, X2 3/7, X3 2/2; SA429 X2 3/5). Independent per-replicate PyClone analyses gave median Pearson correlations of 0.94 (SA501), 0.93 (SA535), 0.91 (SA532), 0.91 (SA429), and 0.46 (SA496) between same-passage replicates — strong evidence for shared, deterministic selective mechanism rather than stochastic drift.
• Reproducibility across mouse strains and sites. SA535 and SA532 showed concordant clonal expansion patterns in NSG vs NRG hosts; established-xenograft transplant-site changes (SA495 X3-4, SA499 X3-4, SA429 X1-2, SA496 X1-2) were not associated with unusually strong dynamics, though initial engraftment in MFP vs SR site was preferred for SA496 (4/4 MFP vs 0/4 SR) and SA429 (2/4 MFP vs 0/4 SR).

Genes & alterations

• TP53 — SA500 acquired (or selected for) a xenograft-specific deletion event encompassing TP53 that coincided with retention of a somatic SNV, demonstrating PDX-specific structural alteration at this locus (PMID:25470049).
• Note: The paper does not nominate individual gene drivers responsible for clonal fitness differences — the principal claim is that clonal genotypes defined by aggregate SNV/CNA clusters are reproducibly selected. No other specific gene-level alterations are highlighted in the main text beyond TP53 and the general SNV/CNA mutation load characterisation.

Clinical implications

• Functional PDX studies must control for clonal architecture. Because clonal selection at engraftment is universal and often extreme (<5% clones can dominate the xenograft), any drug-response or tumour-biology study using breast cancer PDXs without explicit clonal characterisation risks measuring behaviour of a non-representative subclone rather than the originating tumour.
• PDX clonal dynamics are reproducible enough to be experimentally useful. Because replicate transplants produce concordant clonal expansion patterns, PDX lines can support parallel-arm experiments (e.g., drug vs vehicle) on biologically matched clonal populations across mice — provided researchers ascertain clonal composition at the relevant passage.
• Histology and imaging do not reflect clonal change. The authors explicitly note that the genomic clonal dynamics observed here are not evident from histopathological or imaging characteristics, which remain broadly stable — undermining histology-only PDX QC.
• The paper does not claim any patient-treatment or prognostic biomarker.

Limitations & open questions

• Low tumour cellularity hindered baseline mutation discovery in SA429 and SA496, limiting comparability of tumour vs xenograft clonal architecture in those series.
• Specific driver aberrations underlying clonal fitness are not identified. The authors state determination of the precise aberrations giving rise to selective clonal fitness “still faces considerable challenges”; clonal genotypes are treated as composite markers, not as nominations of single causal events.
• Most-parsimonious interpretation is pre-existing variation, not PDX-induced mutation. The authors favour clonal selection of pre-existing subclones over de-novo somatic mutation in the mouse, but cannot rule out a contribution from in-mouse mutation accumulation.
• Mouse stromal/immune microenvironment vs human is not directly addressed — selection pressures driving the reproducible dynamics may be specific to immunodeficient murine biology, with unclear bearing on which clones would dominate in the autologous human host.
• No drug-response data in this paper. Whether the dominant PDX clones recapitulate the drug sensitivities of the originating tumour clones is left as future work — and is the central downstream question for clinical PDX use.
• SA496 was an outlier in replicate-transplant correlations (median Pearson r 0.46 vs ≥0.91 for the other four cases tested), indicating that not every tumour exhibits deterministic clonal selection — and the basis of that variability is unresolved.

Citations from this paper used in the wiki

• “In all 15 cases examined, clonal selection on engraftment was observed in both primary and metastatic breast tumours, varying in degree from extreme selective engraftment of minor (<5% of starting population) clones to moderate, polyclonal engraftment.” (Abstract)
• “all 15 samples also show clusters of SNVs prevalent in the xenograft while at or below the limit of detection in the tumour (range 6.5–32.1% of SNVs … and vice versa (range 0.2–19.4%) … implying clonal selection on initial engraftment.” (Results, p. 3)
• “These included a xenograft-specific deletion event containing TP53 (in SA500) that coincided with retention of a somatic SNV.” (Results, p. 4)
• “In 4/5 series examined, parallel clonal dynamics of the same mutation cluster(s) were observed … SA501 2/2 replicate mice at passage X3 and 4/4 at X4; SA535 X1 3/3; SA532 X1 3/3, X2 3/7 and X3 2/2; SA429 X2 3/5.” (Results, p. 5)
• “median Pearson correlations 0.94, 0.93, 0.91, 0.91, 0.46 for SA501, SA535, SA532, SA429, SA496.” (Results, p. 5)
• “the population dynamics of genomically-defined clones are replicated when transplants are carried out in multiple mice, implying that the basis of selection is non-random.” (Discussion, p. 6)
• “ascertainment of clonal dynamics will prove essential for fully informed future studies of drug response and tumour biology in xenografts of human breast cancers.” (Discussion, p. 6)

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