Walens, Lin, Damrauer, et al. (2020) — Adaptation and Selection Shape Clonal Evolution of Tumors During Residual Disease and Recurrence

Bibliographic Reference

Walens, A., Lin, J., Damrauer, J. S., McKinney, B., Lupo, R., Newcomb, R., Fox, D. B., Mabe, N. W., Gresham, J., Sheng, Z., Sibley, A. B., De Buysscher, T., Kelkar, H., Mieczkowski, P. A., Owzar, K., & Alvarez, J. V. (2020). Adaptation and selection shape clonal evolution of tumors during residual disease and recurrence. Nature Communications, 11, 5017. https://doi.org/10.1038/s41467-020-18730-z

Core Argument

This study uses lentiviral-mediated cellular barcoding in an inducible Her2/neu (MMTV-rtTA;TetO-neu; MTB;TAN) mouse model of breast cancer to directly track clonal dynamics during primary tumor growth, oncogene-induced regression, residual disease (dormancy), and spontaneous recurrence. The central question is how the clonal composition of tumors changes during these phases — a process that is nearly impossible to study in human patients because residual disease is clinically undetectable and recurrent tumors are typically sampled without paired primary/residual tissue.

The study’s key finding is that clonal diversity decreases progressively — during tumor regression, through the residual disease phase, and during recurrence — but that recurrent tumors arise via at least two mechanistically distinct evolutionary routes. Approximately half of recurrent tumors exhibit clonal dominance driven by de novo Met gene amplification, producing oligoclonal tumors that are sensitive to the Met inhibitor crizotinib. The other half exhibit a polyclonal architecture similar to primary tumors, dependent upon an autocrine IL-6-Jak/Stat3 signaling pathway, and are sensitive to Jak inhibitors. A smaller subset shows extreme single-clone dominance (barcode 8240:8518) without Met amplification, suggesting yet another route. The authors conclude that both adaptive mechanisms (EMT as a coordinated response to oncogene inhibition) and selective mechanisms (expansion of Met-amplified clones) work together to shape tumor evolution during residual disease and recurrence.

Methods

Mouse model: The study uses the MMTV-rtTA;TetO-neu (MTB;TAN) inducible mouse model of breast cancer. Doxycycline (dox) administration induces Her2/neu expression, leading to invasive mammary tumor formation. Dox withdrawal induces Her2 downregulation and complete tumor regression, mimicking anti-Her2 targeted therapy. A population of residual cells survives Her2 downregulation in a dormant, non-proliferative state and eventually reinitiates proliferation to form recurrent tumors [lines 78-86].

Cellular barcoding: A primary MTB;TAN tumor was digested, cultured with dox, and ~200,000 cells were infected at MOI 0.1 with a lentiviral barcode library (~60 million unique barcodes). After selection and 12 population doublings, a population of ~100 million cells containing ~20,000 unique barcodes was generated (~5,000 cells per barcode). One million barcoded cells were injected bilaterally into the mammary glands of nu/nu mice [lines 131-170].

Experimental cohorts: Primary tumors (n=6), early residual tumors at 4 weeks post-dox withdrawal (n=6), late residual tumors at 8 weeks post-dox withdrawal (n=6), and recurrent tumors (n=12, median 75 days to recurrence) [lines 153-170].

Clonal analysis: Barcodes from ~100,000 cells per tumor were amplified and sequenced. Clonal complexity was quantified using the Shannon Diversity Index. Inter-tumor similarity was assessed using Jensen-Shannon divergence. A simulation experiment confirmed that reduced complexity in residual tumors was not due to failure to detect rare barcodes [lines 326-356, 726-740].

Secondary analyses: Whole-exome sequencing (WES) on 3 primary and 10 recurrent tumors; RNA-seq (bulk and single-cell) on primary and recurrent tumors; qPCR for Met copy number; qRT-PCR for EMT markers; drug sensitivity testing with crizotinib (Met inhibitor) and Jak Inhibitor I; western blotting for pStat3; GSEA for pathway enrichment [lines 1043-1112, 1115-1132, 1135-1140, 1702-1741].

Key Findings

1. Clonal diversity decreases progressively across all phases — regression, residual disease, and recurrence. The starting cell population had ~17,000 unique barcodes. Primary tumors averaged 4,544 barcodes (range 3,063–5,648), with reduced Shannon Diversity Index. Early residual tumors (4 weeks post-dox) showed further significant reduction (P=4.8x10^-4 vs. primary for Shannon Index), and late residual tumors (8 weeks post-dox) showed continued attrition (P=0.02 vs. early residual). “The decrease in clonal complexity between 4 and 8 weeks post-dox withdrawal was similar in magnitude to the decrease in complexity that accompanied tumor regression,” suggesting ongoing selective pressure during dormancy [lines 337-358, 670-740].

2. Approximately half of recurrent tumors exhibit (oligo)clonal dominance with Met amplification, arising from distinct independent clones. Six of 12 recurrent tumors had Met amplification by qPCR. “All Met-amplified tumors had low clonal diversity” and “the most abundant barcode(s) in each Met-amplified tumor was different,” with distinct amplicon boundaries by qPCR mapping, suggesting “these Met-amplified tumors arose from different Met-amplified clones” possibly arising de novo. The barcodes marking Met-amplified clones could be detected at low frequencies in all other tumors, including primary and residual tumors [lines 1043-1088].

3. The other ~50% of recurrent tumors have polyclonal architecture similar to primary tumors and are dependent on autocrine IL-6-Jak/Stat3 signaling. Three recurrent tumors contained thousands of barcodes with relatively even distributions and clonal architecture “very similar to primary tumors.” These tumors lacked Met amplification. RNA-seq revealed IL-6 was expressed “between 4.9-fold and 29-fold higher in non-amplified recurrent tumors compared to Met-amplified recurrent tumors” (P=0.006). pStat3 levels were elevated in non-amplified recurrent tumor cells, and treatment with a pan-Jak inhibitor “nearly completely inhibited the growth of recurrent tumor cells” [lines 779-810, 1702-1741].

4. All recurrent tumors undergo epithelial-to-mesenchymal transition (EMT), regardless of clonal architecture or Met amplification status. Both Met-amplified and non-amplified recurrent tumors showed “downregulated expression of the epithelial markers E-cadherin (Cdh1) and Epcam, and upregulated the mesenchymal marker Ddr2” compared to primary tumors. GSEA of cells shortly after Her2 downregulation showed EMT as the top enriched gene set, suggesting “EMT is an adaptive response to Her2 downregulation that occurs in the absence of selection.” The authors conclude that “adaptation and selection can work together to shape the evolution of tumors during residual disease and recurrence” — EMT as adaptation, Met amplification as selection [lines 1855-1878, 1135-1140, 1660-1699].

5. Recurrent tumors have distinct drug sensitivities based on their clonal architecture, which are therapeutically exploitable. Met-amplified recurrent tumor cells required HGF for proliferation and were sensitive to the Met inhibitor crizotinib in vitro. A Met-amplified cell line from an autochthonous recurrent tumor showed significantly impaired growth with crizotinib in vivo. Non-Met-amplified polyclonal recurrent tumor cells were insensitive to both HGF and crizotinib but were highly sensitive to Jak kinase inhibition, a finding that directly links clonal architecture to treatment vulnerability [lines 1092-1112, 1735-1741].

Concepts Introduced or Used

• Cellular barcoding: A method where random, inert DNA barcodes are introduced into a population of cells via lentiviral transduction. Each barcode marks all progeny of a single cell, enabling tracking of clonal dynamics across time, conditions, and tissue locations at single-clone resolution. This study uses the CellTracker library from Cellecta with ~50 million unique barcodes. subclonal-reconstruction

• Clonal dominance: A pattern where one or a few subclones comprise the vast majority of a tumor population, reflected in a low Shannon Diversity Index. In this study, ~50% of recurrent tumors showed this pattern, driven by Met amplification. clonal-sweep

• Residual disease / minimal residual disease: The population of tumor cells that survives initial therapy and persists in a clinically undetectable state. This study reveals that rather than being static, residual disease involves continued attrition of clones under ongoing selective pressure. therapy-resistance

• Tumor dormancy: A prolonged state of non-proliferative persistence of cancer cells following therapy. In the MTB;TAN model, residual cells survive Her2 downregulation in a dormant, non-proliferative state for 1-2 months before reactivation. This study finds that clonal diversity continues to decline during dormancy.

• Clonal recurrence vs. polyclonal recurrence: Two distinct evolutionary routes to tumor relapse identified in this study. Clonal recurrence proceeds through expansion of one or few clones (often Met-amplified). Polyclonal recurrence proceeds through tumor-wide reactivation of most clones present in the residual tumor, preserving the clonal architecture of the primary tumor. branching-evolution

• Met amplification: Copy-number amplification of the Met receptor tyrosine kinase gene, identified as the driver of clonal recurrence in ~50% of tumors. Each Met-amplified tumor had a unique barcode and distinct amplicon boundaries, suggesting independent de novo events. therapy-resistance

• Jak/Stat pathway: The Janus kinase/signal transducer and activator of transcription signaling pathway, activated by IL-6 in an autocrine loop in polyclonal recurrent tumors. This pathway drives the tumor-wide reactivation of residual cells without requiring genetic selection. therapy-resistance

• Shannon Diversity Index: A measure of clonal complexity incorporating both the number of clones (barcodes) and their evenness of distribution. Used throughout the study as the primary metric for comparing clonal architecture across tumors and time points.

• Jensen-Shannon divergence: A measure of dissimilarity between barcode abundance distributions, used to compare clonal composition between tumors. Primary tumors were highly similar to one another but progressively diverged from residual and recurrent tumors.

• EMT as adaptive response: The study presents evidence that EMT is an adaptive (non-selective) response to Her2 inhibition — a coordinated transcriptional change that occurs in all cells, not the result of selecting preexisting mesenchymal clones. dual-regime-evolution

Entities Referenced

Genes and pathways:

• Met (c-Met receptor tyrosine kinase) — Amplified in ~50% of recurrent tumors; drives clonal dominance recurrence; target of crizotinib
• Her2/neu (Erbb2) — The oncogene whose expression is dox-regulated in the MTB;TAN model; its downregulation induces tumor regression
• IL-6 (Interleukin-6) — Cytokine expressed 4.9 to 29-fold higher in polyclonal recurrent tumors; proposed mediator of autocrine/paracrine reactivation
• Jak/Stat3 (Janus kinase / Signal transducer and activator of transcription 3) — Signaling pathway activated downstream of IL-6; target of Jak Inhibitor I
• Cdh1 (E-cadherin) — Epithelial marker downregulated in all recurrent tumors
• Epcam — Epithelial marker downregulated in all recurrent tumors
• Ddr2 — Mesenchymal marker upregulated in all recurrent tumors
• Vimentin (Vim) — Mesenchymal gene upregulated in recurrent tumor cells
• Timp1, Twist1 — Mesenchymal genes enriched in recurrent tumor cell clusters by scRNA-seq
• S100A6 — Mesenchymal gene identified in scRNA-seq clusters
• Pik3ca — Mutated in one recurrent tumor cell line (the only canonical driver mutation found by WES)
• Tfrc — Reference gene used for copy-number normalization
• Mdfic, Cftr, Cav2 — Genes flanking Met used to map amplicon boundaries

Mouse model:

• MTB;TAN (MMTV-rtTA;TetO-neu) — Inducible Her2/neu mouse model of breast cancer
• nu/nu — Athymic nude mice used as orthotopic transplantation recipients

Methods and resources:

• CellTracker lentiviral barcode library (Cellecta) — ~50 million unique barcodes
• Agilent SureSelect XT Mouse All Exon Capture Kit — WES capture
• HiSeq 4000 / HiSeq 2500 — Sequencing platforms
• 10x Genomics Chromium — Single-cell RNA-seq platform
• Seurat — R package for scRNA-seq analysis
• DESeq2 — R package for differential expression
• Mutect2 — SNV caller for WES
• CODEX2 — CNA caller for WES
• STAR — RNA-seq aligner

Limitations (as stated by authors)

1. Mouse model vs. human disease. The authors acknowledge that “it remains unclear how the clonal composition of tumors changes during relapse” in human patients, and that this mouse model, while recapitulating key features, may not fully capture the complexity of human breast cancer recurrence [lines 15-17, 49-73].

2. Local vs. distant recurrence. The study examines clonal dynamics in local mammary gland recurrences, while “most deaths from breast cancer result from distant relapse.” The authors note “important differences between local and distant relapse, including the bottleneck of dissemination itself and stresses imposed by the foreign microenvironment” [lines 1907-1926].

3. Timing of Met amplification unresolved. The authors state “we cannot rule out the possibility that the donor primary tumor had many independent Met-amplified clones, each of which was labeled with a different barcode.” They note it is “currently working to implement” an approach to prospectively isolate barcoded clones and measure Met copy number to definitively determine whether amplification is preexisting or de novo [lines 1770-1781].

4. Residual tumor sampling constraints. Residual tumors were “too small to identify grossly” and required fluorescence-microscope microdissection for DNA isolation. This limited the ability to perform additional molecular analyses (e.g., RNA-seq, proteomics, drug sensitivity testing) directly on residual tissue [lines 322-327, 1333-1336].

5. Limited genomic characterization of primary tumors. WES revealed “very few non-synonymous SNVs” and no identifiable driver mutations in the top 20 most frequently mutated human breast cancer genes (except Pik3ca in one cell line). The authors note that “a number of copy-number alterations” were found in both Met-amplified and non-amplified recurrent tumors, but no focal high-level amplifications besides Met were identified, limiting mechanistic dissection of non-Met-amplified recurrences [lines 1124-1132].

6. Single donor tumor. The barcoding experiments were performed with tumor cells derived from a single MTB;TAN donor tumor (#1). A separate cohort using donor tumor #2 confirmed the pattern of progressive complexity decrease and Met amplification in half of recurrent tumors, but the generalizability across genetically distinct tumors is not fully established [lines 131-140, 1334-1348].

7. EMT as adaptation — causal evidence. While the data support EMT as an adaptive response to Her2 inhibition, the authors note that “experimental induction of EMT accelerates recurrence in this model,” suggesting that “the ability to undergo a full EMT is nonetheless rate-limiting for recurrence” — i.e., adaptation alone is not sufficient; the capacity to adapt is itself a substrate for selection [lines 1855-1867].

Relevance to Clonal Evolution

This study provides some of the strongest direct experimental evidence in the wiki corpus for bottleneck dynamics during therapy, because cellular barcoding enables clonal tracking at a resolution that is impossible in human studies (where only endpoint sequencing is available). The findings are relevant across multiple themes:

Therapy-induced bottleneck and the bottleneck paradox. The progressive decrease in clonal diversity from ~17,000 barcodes in the injected population to ~4,500 barcodes in primary tumors, to fewer in residual tumors, and ultimately to as few as 1-2 dominant barcodes in clonal recurrent tumors, is a direct observation of a therapy-induced population-bottleneck. This mirrors the bottleneck paradox documented in population-bottleneck and miething2019-clonal-evolution-myeloma — severe bottlenecks can either purge diversity (promoting cure) or select for resistant clones (enabling relapse). The finding that recurrent tumors emerge with distinct clonal architectures (clonal dominance vs. polyclonal) shows that the outcome of the bottleneck is not deterministic but depends on the adaptive strategy that prevails.

Compression-entrenchment and the dual-route finding. The Met-amplified clonal dominance route is a classic example of compression-progress-evolution: one clone finds a “compression” (Met amplification) that dramatically enhances fitness under the selective pressure of Her2 loss, entrenching itself and dominating the post-therapy landscape. The polyclonal Jak/Stat-dependent route represents a fundamentally different adaptive strategy — one that does not depend on a single genetic event but on coordinated signaling (autocrine IL-6) that reactivates the entire residual population. This parallels the plasticity-driven adaptation described in geng2016-genetic-diversity-phenotypic-plasticity, where a plastic response (here, EMT + IL-6 secretion) allows survival and regrowth without requiring a genetic selective sweep.

Dual-regime evolution. The two routes map onto the dual-regime-evolution framework. The Met amplification route is a Darwinian (genetic) adaptation: a specific gene amplification event is selected from heterogeneous clones, producing a clonal sweep. The Jak/Stat polyclonal route is closer to a non-Darwinian (epigenetic/transcriptional) adaptation: the reactivation of thousands of clones without genetic selection suggests a coordinated transcriptional program (EMT + cytokine signaling) that operates at the chromatin or signaling level. The authors’ explicit framing of “adaptation and selection” as complementary forces — EMT as an adaptive response (non-selective), Met amplification as selection-driven — directly supports the dual-regime hypothesis that both modes of evolution operate simultaneously in tumors.

Clonal evolution during dormancy is an active process. One of the most striking findings is that “the residual disease stage itself is also accompanied by a decrease in the number of clones and their diversity within tumors” and that “the reduction in diversity between early and late residual disease was as large as the reduction that accompanied tumor regression.” This challenges the view of dormancy as static quiescence and positions it as a dynamic period of ongoing selection, consistent with models in which residual cells continue to experience stress (microenvironmental, metabolic, immune) that shapes the eventual recurrent population.

Implications for therapy. The finding that clonal architecture predicts drug sensitivity — Met-amplified clonal tumors respond to crizotinib, while polyclonal non-amplified tumors respond to Jak inhibitors — connects clonal evolutionary analysis directly to treatment stratification. By identifying the mode of recurrence (genetic selection vs. adaptive reactivation), one may select the appropriate targeted intervention. This resonates with the broader theme of therapy-resistance and the clinical imperative to understand the evolutionary path of residual disease.

Relationship to persister cell biology. The authors note parallels to Hata et al. 2016 (EGFR-mutant lung cancer resistance via preexisting vs. de novo mutations) and Echeverria et al. 2019 (TNBC residual/recurrent tumors with similar clonal composition to untreated tumors). The polyclonal recurrence route — where “the majority of these residual cells can acquire the ability to resume proliferation in the absence of additional clonal selection” — extends the persister-cell paradigm by showing that some residual populations can reactivate en masse, not just through resistant subclones. The authors propose this “polyclonal, drug-tolerant intermediate stage seems to be common across different tumor types and therapies.”

Revision history

• 2026-07-05 — Source summary created. (clonal-evolution)