Clonal Evolution

Mikutenaite, Karadoulama, Favero, et al. (2025) — Clonal Evolution and Transcriptional Plasticity Shape Metastatic Dissemination Routes in Prostate Cancer

Properties1
tagsprimary-research, selection, phylogenetics, clinical

12 min read

Bibliographic Reference

Mikutenaite, M., Karadoulama, E., Favero, F., Locallo, A., Rodriguez Gonzalez, F. G., Kiriy, D., Keshavarzian, T., Furlano, K., Graefen, M., Bristow, R. G., Gerhauser, C., Plass, C., Sauter, G., Simon, R., Lupien, M., Minner, S., Schlomm, T., & Weischenfeldt, J. (2025). Clonal evolution and transcriptional plasticity shape metastatic dissemination routes in prostate cancer. Nature Communications, 16, 11338. https://doi.org/10.1038/s41467-025-66704-w

Core Argument

This study integrates single-nuclei RNA sequencing (snRNA-seq) and whole-genome sequencing (WGS) from 43 spatially distinct tumour samples across 5 patients with locally advanced prostate cancer to reconstruct clonal evolution trajectories and transcriptional changes driving metastasis at single-cell resolution. The authors find extensive clonal heterogeneity, including both monophyletic and polyphyletic metastatic dissemination, and demonstrate ongoing clonal evolution in the primary tumour after metastatic spread has occurred (lines 25–39).

Metastatic seeding converges on disease trajectories involving both genomic and transcriptional changes, including androgen receptor (AR) independence and activation of estrogen, WNT, and JAK-STAT pathway activity in spatially distinct areas (lines 35–38). The key advance is the simultaneous analysis of both genetic (copy-number-based) and transcriptional evolution at single-cell resolution across multiple spatially mapped tumour regions, which the authors use to argue that an intricate interplay between clonal evolution and cellular plasticity drives metastatic seeding. This suggests that integrative prognostic markers — combining genomic and transcriptional features — may improve patient management over current pathology or single-biopsy genomics alone (lines 38–39, 1163–1177).

Methods

Study design and sample collection. Five patients with high-risk, locally advanced prostate adenocarcinoma were selected from the German ICGC cohort (lines 113–115, 1233–1241). Four patients were treatment-naive; one (PCAL25) received neoadjuvant androgen deprivation therapy six weeks before surgery. One patient (PCAL03) had synchronous bone metastasis at the time of surgery (lines 113–116). Following radical prostatectomy and pelvic lymph node dissection, whole prostates were cryopreserved and 10 regions with high cancer cell content (>80%) were selected per patient based on topographic pathology reports. Tissue punches from these areas, along with corresponding regional lymph node metastasis samples, underwent nuclei isolation followed by snRNA-seq and low-pass bulk WGS (lines 116–124). In total, 43 spatially distinct samples were analysed.

Single-nuclei RNA sequencing. After quality filtering (doublet removal, high mitochondrial content exclusion, ambient RNA removal via CellBender), more than 356,000 single nuclei were retained from the 43 samples (lines 128–129). Malignant cells were predicted using Numbat (line 132). Cell type assignment used consensus among scPred, SELINA, and literature-based references (lines 133–135). Pathway activity inference used decoupleR with PROGENy and CollecTRI (lines 1367–1378).

Phylogenetic reconstruction. Copy number alteration (CNA) profiles were inferred from area-specific snRNA-seq and bulk low-pass WGS data, integrating genomic segments from both platforms. Phylogenetic trees were built using MEDICC2 (lines 165–172, 1441–1463). For comparison, single-biopsy bulk-WGS phylogenetic trees were also constructed from the original ICGC cohort data (lines 170–171, 1479–1527). The branch length in phylogenetic trees is based on the number of CNAs, described by the authors as “only a simplistic estimate of the actual evolutionary timing” (lines 177–178).

Key Findings

  1. Both monophyletic and polyphyletic metastatic dissemination are observed. “We find extensive clonal heterogeneity, including both monophyletic and polyphyletic metastatic dissemination, and ongoing clonal evolution in the primary tumour after metastatic spread” (lines 33–35). In PCAL10, monophyletic seeding from clone PCAL10-C2 to the lymph node was identified (lines 577–578). In PCAL25, single-cell analysis “revealed that metastatic tumour clades originated from polyphyletic clonal lineages within the primary tumour” (lines 979–981). For PCAL03, the single-biopsy WGS tree “inferred a different seeding clone… suggesting polyphyletic metastatic dissemination, where several distinct clones in the primary tumour independently seed metastases” (lines 560–564).

  2. Ongoing clonal evolution continues in the primary tumour after metastatic spread. Both the abstract (lines 33–35) and the phylogenetic reconstructions show that clones in the primary tumour continued to diversify and diverge after the subclone(s) that seeded the metastases had already disseminated. In PCAL34, for example, “a distinct lineage, comprising clones PCAL34-C6 and PCAL34-C5, was restricted to area T06” (lines 507–508), demonstrating continued intra-prostatic clonal divergence post-dissemination.

  3. Metastatic seeding converges on both genomic and transcriptional changes, including androgen receptor independence and activation of estrogen, WNT, and JAK-STAT pathway activity. “Metastatic seeding converges on disease trajectories involving both genomic and transcriptional changes, including androgen receptor independence and activation of estrogen-, WNT- and JAK-STAT- pathway activity, in spatially distinct areas” (lines 35–38). Specifically, JAK-STAT activity was “exclusive to the PCAL10-C2 clone, regardless of the area” (lines 629–631), and the PCAL25-C3 metastatic lineage “displayed transcriptional plasticity, with exclusive WNT activity in the metastatic LNM1, while AR and JAK-STAT pathway activity was increased in the other metastatic area, LNM2” (lines 990–992). The authors further conclude: “These findings suggest convergent transcriptional programmes across different lineages, including AR suppression, JAK-STAT and WNT activity, even with distinct genotypes” (lines 1014–1016).

  4. Transcriptional plasticity enables metastatic adaptation without requiring new mutations. The finding that spatially distinct metastases from the same polyphyletic lineages exhibited different pathway activities (e.g., WNT-active vs. JAK-STAT-active in PCAL25’s two lymph node metastases) — despite deriving from the same ancestral clones — demonstrates that transcriptional reprogramming can drive metastatic adaptation independently of new genomic alterations. As the authors summarise: “substantial evidence for lineage plasticity as a key step in metastatic potential, including suppression of AR signalling and up-regulation of JAK-STAT, oestrogen-like signalling and the WNT pathway” (lines 1179–1182). The convergence on these transcriptional programmes across genetically distinct clones supports the operation of non-genetic mechanisms in metastasis.

  5. Single-biopsy bulk-WGS captures only a fraction of tumour heterogeneity, with implications for clinical management. “Our single-biopsy bulk-WGS tree structures were overall in agreement with the single-cell multi-region phylogenetic trees, but only captured a minor fraction of the clonal diversity” (lines 173–176). Furthermore, “single-area biopsies assessed by genome sequencing and/or pathology grading will be suboptimal in detecting locally advanced disease” (lines 1163–1165). The authors also found that “Gleason grade does not always correlate with metastatic potential” (line 1162), and that metastatic seeding areas “were localized preferentially toward the periphery of the prostate” (lines 1144–1145, 1093–1094).

Concepts Introduced or Used

  • Monophyletic dissemination — a pattern in which a single primary tumour subclone seeds all metastatic lesions. Observed in PCAL10, where clone PCAL10-C2 seeded the lymph node monophyletically (line 577–578).
  • Polyphyletic dissemination — a pattern in which multiple distinct subclones within the primary tumour independently seed metastases. Observed in PCAL03 and PCAL25 (lines 562–564, 979–981). Maps to branching evolutionary patterns in clonal-evolution.
  • Transcriptional plasticity — the ability of cancer cells to alter their gene expression programmes (pathway activity, transcription factor activity) without acquiring new genomic mutations. In this study, demonstrated by convergent pathway activation (WNT, JAK-STAT, oestrogen) across genetically distinct clones (lines 1014–1016, 1179–1182). Operates as a non-Darwinian evolutionary mechanism in the dual-regime-evolution framework.
  • Clonal evolution trajectories — the reconstructed phylogenetic paths of tumour subclones from the most recent common ancestor (MRCA) to metastatic lineages, inferred from CNA-based phylogenies using MEDICC2 (lines 165–172, 1441–1463). Related to clonal-evolution.
  • Single-nuclei RNA sequencing (snRNA-seq) — a single-cell transcriptomic method adapted for fresh-frozen tissue. Used here for cell-type classification, clonal reconstruction, and pathway activity inference from the same nuclei that provided WGS data (lines 108–124). Methodological significance: this study demonstrates the feasibility of integrated snRNA-seq + WGS from the same tissue sample for clonal and transcriptional analysis.
  • Phylogenetic reconstruction — inference of clonal relationships from copy number alteration profiles, built using MEDICC2. The authors note branch length “is based on the number of CNAs, and is, therefore, only a simplistic estimate of the actual evolutionary timing” (lines 177–178). The CNA-based phylogenetic approach is a simplification; branch lengths are approximate estimates of evolutionary timing.
  • Androgen receptor independence — loss of AR signalling activity, identified as a convergent feature of metastatic seeding clones across multiple patients. AR-negative clones showed enrichment of SHOX2 and MECP2 transcription factor activity (lines 636–638). Clinically relevant for ADT treatment decisions: AR-independent clones may not respond to androgen deprivation.
  • Weighted phylogenetic distance — a custom metric developed to assess clonal heterogeneity between spatially distinct samples, ranging from 0 (identical clonal composition) to 1 (no similarity in clonal composition) (lines 1403–1423).

Entities Referenced

  • Genes and pathways: Androgen receptor (AR), NCOA3, SHOX2, MECP2, EZH2, FOXJ1, DNAI1, DNAH11, MKi67, MSH2/MSH6, TP53, MYC, NKX3-1, CDKN1B, TCF7, SATB2, ONECUT2. Pathways: JAK-STAT, WNT, oestrogen, androgen, TRAIL.

  • Methods and tools: snRNA-seq, whole-genome sequencing (WGS), low-pass WGS, MEDICC2 (phylogenetic tree building), Numbat (CNA calling and malignant cell classification), decoupleR (pathway activity), PROGENy (pathway gene sets), CollecTRI (transcription factor regulons), scPred and SELINA (cell type prediction), CellBender (ambient RNA removal), STACAS (semi-supervised data integration), Sequenza (allele-specific copy number), ACE (absolute copy number from low-pass WGS), GATK, MuTect2, Strelka2, SomaticSeq, SigProfilerAssignment, MSIsensor2.

  • Cancer type: Prostate adenocarcinoma (locally advanced, high-grade).

  • Other entities: Tumour microenvironment (TME), lymph node metastasis (LNM), epithelial-mesenchymal transition (EMT), whole-genome doubling (WGD), microsatellite instability (MSI), chromothripsis, MMR (mismatch repair) signatures SBS15 and SBS21.

Limitations (as stated by authors)

The authors list the following limitations directly (lines 1200–1216):

  1. Cohort size and statistical power: “while our study investigated more than 350,000 cells from 43 tumour areas, larger and better-powered study cohorts will be needed for validation. We find evidence for transcriptional plasticities across the cohort, but note that our cohort is not powered to draw broader conclusions on the prevalences of e.g. JAK-STAT-, Oestrogen- and WNT signalling” (lines 1203–1206).

  2. Missing evolutionary programmes: “additional transcriptional programmes and clonal evolution patterns not present in our cohort are likely to contribute to driving metastatic dissemination, e.g. stem-like and neuroendocrine programmes” (lines 1207–1208).

  3. Resolution limited to CNAs: “While CNAs were sufficient to identify the majority of the clonal heterogeneity, using single-nucleotide variants would provide even greater resolution, in particular for copy-number-stable tumours” (lines 1208–1209).

  4. Metastatic site scope: “we used locoregional lymph nodes as metastatic sites to investigate dissemination patterns. Distant metastatic organs such as bone and liver are considered optimal proxies for aggressive metastatic disease” (lines 1209–1211). However, they note that “a recent study on genetic heterogeneity from multi-region primary and metastatic sites found that locoregional lymph nodes represented a good marker for aggressive disease” (lines 1211–1216).

Additionally, the authors note earlier: “we acknowledge that the spatial resolution of our sampling is, nevertheless, limited, and that the presence of metastatic clones in a given region does not exclude their existence elsewhere in the tumour” and that “spatial relationships within a prostate tumour can be convoluted, following non-Euclidean paths” (lines 1145–1149).

Relevance to Clonal Evolution

This paper provides the most direct empirical evidence in the current wiki corpus for the dual-regime-evolution model — the claim that transcriptional plasticity (a non-Darwinian/epigenetic evolutionary regime) operates alongside clonal evolution (a Darwinian/genetic regime) to drive metastasis. Several specific connections follow.

Dual-regime evidence. The finding that convergent transcriptional programmes (AR suppression, WNT activation, JAK-STAT activation) emerge across genetically distinct clones — with the same genotype producing different pathway activities in different metastases (PCAL25, lines 990–992) — demonstrates that transcriptional plasticity enables metastatic adaptation without requiring new mutations. This is the strongest evidence yet in the wiki corpus that the epigenetic/transcriptional level can drive adaptation independently of genetic change, which is the core empirical prediction of dual-regime-evolution.

Monophyletic vs. polyphyletic seeding and evolutionary modes. Monophyletic dissemination (PCAL10, lines 577–578) corresponds to a clonal-sweep pattern: one clone achieves sufficient fitness to dominate and seed all metastases. Polyphyletic dissemination (PCAL25, PCAL03, lines 560–564, 979–981) corresponds to branching evolutionary patterns with multiple clones seeding independently. These map directly onto the evolutionary modes described in clonal-evolution — the contrast between single-lineage dominance and multi-lineage branching. The fact that single-biopsy WGS detected only monophyletic seeding when multiple polyphyletic lineages were actually present (PCAL25, lines 981–983) underscores that sampling resolution qualitatively shapes which evolutionary models can be inferred.

Ongoing evolution after metastasis. The observation that clonal evolution continues in the primary tumour after metastatic spread (lines 33–35, 507–508) means the primary tumour is not evolutionarily “finished” at the point of dissemination. This is consistent with the compression-progress-evolution framework’s prediction of continuous clonal exploration — the primary tumour does not cease evolving once metastatic competency is achieved.

Convergent selective pressures. The convergence on specific pathways (AR independence, WNT, JAK-STAT) across spatially distinct metastases in different patients suggests shared selective pressures — possibly the lymph node microenvironment imposing consistent evolutionary demands. This convergence is what compression-progress-evolution would predict as progress toward a metastatic phenotype under directional selection, mediated by transcriptional plasticity rather than only genetic change.

Methodological significance for dual-regime testing. The integration of snRNA-seq + WGS from the same nuclei is methodologically significant because it allows separation of genetic from transcriptional adaptation within individual cells. This is exactly what dual-regime-evolution requires for empirical testing: the ability to observe the transcriptional regime operating on a fixed genetic background. The authors’ weighted phylogenetic distance metric (lines 1403–1423) and PCA integrating TME composition, cell states, and pathway activity (lines 1050–1084) provide analytical tools for disentangling the two regimes.

Links to bottleneck dynamics. The observation that seeding clones are localised in spatially confined peripheral regions (lines 1091–1094) and that the metastatic founder clone often had low cancer cell fraction and low Gleason grade (lines 578–581, 998–1001) connects to population-bottleneck dynamics: the clone that ultimately metastasises may not be the most abundant or histologically aggressive clone in the primary tumour, but rather a less conspicuous subpopulation with specific transcriptional and genomic properties.

Graph View

Backlinks