A team at Biberach University of Applied Sciences and Boehringer Ingelheim systematically compared two CRISPR/Cas9 workflows for generating CHO cell lines lacking nine polysorbate-degrading hydrolases. Polysorbate is a surfactant added to therapeutic protein formulations to prevent aggregation, but host cell-derived hydrolases can enzymatically chop it up, generating free fatty acids that form visible and subvisible particles and compromise drug stability. Removing the offending enzymes has therefore become a promising strategy for stabilizing biologics during manufacturing and storage.

The researchers compared a sequential editing strategy, in which genes were knocked out one or a few at a time over repeated rounds of transfection and single-cell cloning, with a multiplex strategy that removed several targets simultaneously. The sequential lineage required 383 days of cultivation, seven rounds of single-cell cloning, and seven electroporations to reach the final genotype. The multiplex route took only 250 days and four cloning rounds, though it needed eleven electroporations because some guide RNAs worked inefficiently, forcing repeated transfections. Both approaches deleted the same nine hydrolase genes, including lipoprotein lipase, phospholipases, and two large carboxylesterase gene clusters excised with paired guide RNAs spanning 506 and 1,629 kilobases of genomic DNA.

Functionally, both strategies succeeded. A fluorescence-based micelle assay showed that polysorbate retention climbed from roughly 30 to 45 percent in parental controls to more than 75 percent in final knockout clones from either workflow. RNA sequencing confirmed transcript depletion at the targeted loci, and genomic analysis verified editing at every intended site. On paper, the two cell lineages were equivalent. In practice, they were strikingly different organisms.

In fed-batch bioreactor cultures, the sequentially edited clones deteriorated in a stepwise fashion as knockouts accumulated. Peak viable cell density fell from about 25 million cells per milliliter in the parental line to 15 to 17 million in final clones, while harvest viability collapsed from above 90 percent to between 50 and 60 percent. The multiplex clones also declined, but more gently, retaining peak densities of 17 to 21 million cells per milliliter and viabilities between 55 and 80 percent, albeit with greater clone-to-clone variability. Metabolic behavior diverged too: sequential clones showed escalating glucose consumption, reaching 42 to 53 grams per liter by day ten, whereas multiplex clones stayed closer to parental levels at 19 to 27 grams per liter.

Perhaps the most striking physiological change emerged in the sequential lineage after knockout of the Pla2g15 and Lipa genes: cells became larger and carried more DNA, with increased cell diameter correlating with elevated DNA content measured by flow cytometry and with reduced proliferation. This shift was never reversed by subsequent editing steps. Intriguingly, one multiplex clone, M-KO4, independently developed the same enlarged, high-DNA, slow-growing phenotype, suggesting that clone-specific disruptions during repeated manipulation, rather than the knockout targets themselves, drive these undesirable traits.

The transcriptomic data told an even sharper story. Final sequential knockout clones showed 425 upregulated and 415 downregulated genes relative to the parental line, while final multiplex clones displayed a far smaller footprint of 162 up and 109 down. Direct comparison between the two endpoints revealed 894 genes higher and 1,003 genes lower in the sequential clones. On a two-dimensional UMAP projection of global gene expression, sequential intermediates drifted progressively away from the parental state, while the multiplex lineage stayed within a confined transcriptional space. The two workflows ended in largely non-overlapping transcriptomic states despite engineering identical genotypes.

Stage-resolved analysis of when differentially expressed genes first appeared pointed to clonal selection, not the editing events themselves, as the dominant driver. Early sequential intermediates changed only modestly, but the Pla2g15 and Lipa knockout step coincided with a jump of 87 up and 83 down genes, matching the inherited increase in cell size and DNA content. The largest sequential shift, 346 up and 250 down genes, followed deletion of the second carboxylesterase cluster, yet analysis of genes flanking the deleted regions showed no broad local expression changes, arguing against simple position effects. Meanwhile, the multiplex lineage accumulated changes gradually, and even its final step involving seven electroporations produced a shift no larger than several sequential stages. The authors conclude that repeated single-cell isolation and clonal adaptation, consistent with the known genomic plasticity of CHO cells, primarily shape the final cellular state.

The team also probed whether eliminating nine hydrolases triggers compensatory upregulation of alternative polysorbate-degrading enzymes, a key concern for product stability engineering. Across a curated panel of candidate serine hydrolases defined by the conserved catalytic triad associated with active CHO-derived enzymes, most genes showed minimal change. Only Pla1a and Abhd14b rose above threshold in sequential clones, and only Liph in multiplex clones, indicating workflow-specific rather than universal compensation. Notably, Ppt2, a lysosomal thioesterase paralog of the knocked-out Ppt1, was consistently downregulated in both lineages, the opposite of what a simple compensation model would predict.

The practical implications for cell line development are substantial. For non-essential targets such as problematic host cell proteins, the data favor multiplex editing, which reached the same functional endpoint in less time, with fewer cloning rounds, and with less accumulated drift in growth, DNA content, and gene expression. Sequential editing retains value when target biology is uncertain or when individual knockouts might impair essential functions, since intermediate phenotypes can be monitored before proceeding. Either way, the study delivers a sobering message: the engineering process itself, particularly the gauntlet of electroporation, cloning, and selection, can sculpt cellular phenotype and transcriptome as powerfully as the intended genetic modifications, and unfavorable traits acquired early tend to persist through the entire lineage. Choosing an editing strategy, the authors argue, is a critical design parameter in its own right.

Subject of Research: Comparison of sequential versus multiplex CRISPR/Cas9 editing workflows for hydrolase knockout in CHO cells

Article Title: Same hydrolase knockouts, different outcomes: CRISPR/Cas9 editing workflow shapes resulting cellular phenotype and transcriptome in CHO

Article References: Weiß, L., Schneider, S., Otte, K., & Fischer, S. (2026). Same hydrolase knockouts, different outcomes: CRISPR/Cas9 editing workflow shapes resulting cellular phenotype and transcriptome in CHO. Frontiers in Bioengineering and Biotechnology, 14, 1923843. https://doi.org/10.3389/fbioe.2026.1923843

Image Credits: AI Generated

DOI: 10.3389/fbioe.2026.1923843

Keywords: CHO cells, CRISPR/Cas9, gene knockout, polysorbate degradation, hydrolases, cell line development, transcriptomics, multiplex editing, clonal selection, biopharmaceutical manufacturing, RNA-seq, host cell proteins

The core of the argument is that metabolic dysregulation acts as a selective pressure on premalignant tissue, not merely as a passive risk factor in the conventional epidemiologic sense. Chronic hyperinsulinemia increases bioavailable insulin-like growth factor 1 (IGF-1) and sustains activation of the PI3K–AKT–mTOR and RAS–MAPK signaling cascades, the two master growth pathways that drive proliferation and survival. Paradoxically, in tissues where insulin resistance has developed, mitogenic sensitivity is preserved even as glucose disposal falters, creating an environment highly permissive to clonal selection. Elevated inflammatory mediators, including interleukin-6, leptin, and other adipokines, further amplify proliferative signaling while dampening apoptotic responses. A systematic review using the hyperinsulinemic–euglycemic clamp confirmed that severe insulin resistance is present in patients with cancer relative to matched controls, with a mean glucose disposal rate of 4.7 versus 7.5 mg/kg/min, and limited evidence suggests this metabolic adaptation may precede clinical diagnosis rather than follow it.

The mechanism by which metabolic state becomes inscribed into cellular behavior runs through the epigenome. Metabolic intermediates—S-adenosylmethionine (SAM), acetyl-CoA, α-ketoglutarate, and NAD+—serve as obligate substrates for chromatin-modifying enzymes, continuously coupling a cell’s metabolic condition to its epigenetic landscape. Shifts in one-carbon metabolism can directly alter histone methylation and downstream transcription across every cell division that occurs during the years of subclinical dysregulation preceding diagnosis. Among the relevant modifications, the authors focus on histone H3 lysine 9 (H3K9) methylation, catalyzed by the methyltransferases EHMT1 and EHMT2—known as GLP and G9a—which deposit mono- and dimethyl marks using SAM as the methyl donor. H3K9me2 at apoptotic and differentiation loci has been linked to drug resistance in prior studies, making this mark a candidate bridge between metabolism and treatment failure.

Two distinct inputs shape this chromatin landscape, and the authors are careful to separate them. Hypoxia is the driver of H3K9me2 gain: low oxygen stabilizes the G9a protein and increases H3K9me2 at tumor-suppressor and apoptotic loci, effectively silencing the genes that would otherwise command damaged or abnormal cells to die. SAM availability works separately. When SAM is limiting, EHMT1/2 preferentially preserve H3K9me1 while higher-order di- and trimethylation decline, so SAM depletion does not itself add to H3K9me2. One-carbon and SAM status is therefore best understood as setting the available H3K9me1 substrate pool, whereas hypoxia-stabilized G9a generates the resistance-associated H3K9me2. The consequence of both processes is an elevated apoptotic threshold and an enrichment of epigenetically primed, apoptosis-resistant cells before a tumor is histologically apparent or clinically detected. The authors note an unresolved gap: whether G9a regulation in premalignant epithelial compartments mirrors the dynamics seen in metabolic tissues such as liver and adipose, where G9a expression drops under high-fat conditions, has not been systematically characterized.

A compelling precedent for the idea that metabolic intervention can undo epigenetic hardening comes from pancreatic cancer, where pharmacologic inhibition of the metabolic pathway driving epigenomic reprogramming reversed both the chromatin state and the malignant phenotype. Against this mechanistic backdrop, the clinical prediction converges: tumors arising in metabolically dysregulated tissue should harbor larger fractions of pre-existing resistant subclones. Parallel mechanisms, not competing ones, produce a cell population more likely to progress, survive dormancy, and resist treatment. In this framing, resistance is less a response to therapy than the clinical unveiling of a cellular architecture established years earlier.

The population-scale data that speak to this hypothesis have arrived, perhaps unexpectedly, through the GLP-1 receptor agonists, the drug class behind the current obesity-treatment revolution. These agents reduce hyperinsulinemia, attenuate IGF-1 bioavailability, and suppress systemic inflammatory markers, making them the most informative pharmacologic probe of the metabolic hypothesis—though the authors stress they are not its therapeutic endpoint. The emerging observational signal is provocative. A 2026 meta-analysis of 24 observational studies enrolling nearly 4 million patients found a 30 percent reduction in obesity-related cancer risk, and a large retrospective cohort confirmed reduced overall cancer incidence in adults with obesity, with a hazard ratio of 0.83. Body mass index–independent reductions in acute myeloid leukemia and breast cancer incidence, persisting after adjustment for metabolic covariates, hint at a mechanism operating through metabolic signaling rather than adiposity reduction alone, though these data remain preliminary conference abstracts. In a preclinical colitis-associated carcinogenesis model, the dual agonist tirzepatide reduced tumor burden by suppressing metabolic pathways without inducing weight loss.

The most striking clinical evidence comes from a retrospective cohort of 841,831 patients with breast cancer drawn from the TriNetX network, in which propensity-matched GLP-1RA users demonstrated approximately 60 percent lower all-cause mortality at both 5 and 10 years compared with non-users, alongside a large relative improvement in recurrence-free survival. A finding of this magnitude in a dataset of this size commands attention and scrutiny, all the more so given its retrospective design. Intriguingly, when GLP-1RA users were compared not with untreated controls but with users of sodium-glucose cotransporter 2 inhibitors—another drug class that corrects insulin resistance through a distinct mechanism—the survival difference was attenuated, leaving only modest residual benefit after covariate adjustment. The implication is subtle but potentially transformative: if two mechanistically distinct agents that share the common property of metabolic normalization produce convergent oncologic signals, then the operative variable may not be either drug, but the metabolic environment that both drugs correct.

The counterweight is substantial, and the authors confront it directly. A meta-analysis of 48 randomized controlled trials enrolling 94,245 participants found no effect of GLP-1RAs on obesity-related cancer incidence, and a separate RCT-based meta-analysis reported increased thyroid and colorectal cancer risk, with Mantel–Haenszel odds ratios of 1.55 and 1.27 respectively. The authors attribute these discrepancies to trial design rather than failure of the mechanism: the trials were not designed with cancer endpoints, had inadequate follow-up durations for solid tumor latency, and lacked metabolically enriched patient selection. But they are careful not to treat these as exculpations of the null result. The metformin precedent is instructive: the MA.32 trial and a subsequent meta-analysis of 22 randomized trials confirmed no benefit on disease-free or overall survival, despite a compelling mechanistic rationale. The intervention was simply not potent enough. By that logic, meaningfully reversing the metabolic selective pressure may require a more potent, broader-acting intervention deployed in a metabolically high-risk population.

What elevates the hypothesis above speculation is its alignment with the somatic evolution literature, which indicates that resistance predominantly pre-exists. Multiple orthogonal approaches—cell-free DNA analysis, high-complexity barcode tracing, and single-cell sequencing—have collectively demonstrated that acquired resistance frequently arises from pre-existing subclones selected by therapeutic pressure. The same single-cell analyses that identify pre-existing resistant genotypes also document treatment-associated transcriptional reprogramming, indicating that selection and adaptation operate together. What has been missing is a specific account of how the resistant architecture becomes established in the first place, and the metabolic framework supplies one: chronic metabolic stress creates conditions in which EHMT1/2 gain the upper hand in premalignant tissue, concentrating repressive marks at genes governing cell death and differentiation, yielding a subpopulation already primed to resist apoptosis before any treatment is given. These clones, the authors argue, might not be generated by treatment but revealed by it. The framework explicitly does not claim to account for reversible drug-tolerant persister states or to exclude resistance genuinely acquired during treatment.

The authors specify the experiment that would directly test the hypothesis, and it has not been performed: correlating pretreatment fasting insulin, HOMA-IR, and IGF-1/IGF-binding protein 3 ratio with baseline subclonal architecture, H3K9me2 state, circulating tumor DNA heterogeneity, and treatment outcome in metabolically unselected patients initiating systemic therapy. Reverse causation predicts metabolic abnormality follows neoplastic change; this model requires the opposite ordering. Treatment-induced plasticity predicts resistant chromatin states emerge only after exposure; this model predicts they are demonstrable in treatment-naive tissue. Each prediction is testable, and none has yet been tested directly. The urgency is sharpened by young adults with early-onset colorectal cancer who are neither obese nor diabetic, a cohort for whom conventional risk factors do not suffice as explanation, and by the CHALLENGE trial, which provided the first level I evidence in stage III and high-risk stage II colon cancer that structured post-chemotherapy exercise—an intervention known to reduce fasting insulin, IGF-1 bioavailability, and inflammatory markers—improves disease-free survival. A cautionary note comes from the neoadjuvant setting, where concurrent GLP-1RA use during chemoimmunotherapy for early-stage triple-negative breast cancer was associated with reduced pathologic complete response in one retrospective cohort, though a subsequent multi-institution study found diabetes mellitus, not GLP-1RA use, independently associated with reduced response. The authors are explicit that the evidence does not yet establish that GLP-1 receptor agonists prevent recurrence or resistance, and that patients with established metabolic indications should be treated for those indications on their own merits. But the central untested prediction—that pretreatment metabolic state correlates with baseline subclonal architecture, chromatin state, and treatment outcome—is tractable, fundable, and immediately actionable, and the answers will matter whatever they turn out to be.

Subject of Research: How chronic metabolic dysregulation epigenetically primes pre-existing cancer treatment resistance before therapy

Article Title: Metabolic oncology: clonal selection and the pre-existing origins of cancer resistance

Article References: Perez, C. A., & Khaled, A. R. (2026). Metabolic oncology: clonal selection and the pre-existing origins of cancer resistance. Frontiers in Oncology, 16, Article 1951576. https://doi.org/10.3389/fonc.2026.1951576

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1951576

Keywords: cancer metabolism, treatment resistance, clonal selection, epigenetics, H3K9 methylation, EHMT1/EHMT2, GLP-1 receptor agonists, hyperinsulinemia, IGF-1, tumor evolution, metformin, early-onset colorectal cancer

The story of Tribidrag is itself a tale of scientific detective work. For decades, the identity of Zinfandel was one of the great mysteries of viticulture, with competing claims about whether the variety originated in Italy, where it thrives as Primitivo, or elsewhere in the Mediterranean. Genetic fingerprinting eventually traced the cultivar to Croatia, where it survives under its oldest recorded name, Tribidrag, and where old vines still cling to terraced slopes along the Adriatic coast. Because grapevines are propagated vegetatively, by cuttings rather than seeds, every plant of a given variety is in principle a genetic copy of its parent. Yet over centuries of cloning, somatic mutations accumulate, and growers have selected vines that differ in yield, berry size, sugar accumulation, and acidity. These selected lines, called clones, are the raw material of clonal selection programs that supply nurseries and vineyards worldwide.

The research team, led by Goran Zdunić of the Institute for Adriatic Crops and Karst Reclamation in Split and including collaborators at the University of Zagreb and Yunnan Agricultural University in Kunming, set out to bridge a persistent gap in clonal research. Although previous studies had documented variation among Tribidrag clones both in the vineyard and at the molecular level, no one had systematically connected the two. The question was deceptively simple: if two clones of the same variety look and behave differently, can we find the genetic differences responsible? To answer it, the researchers evaluated productive traits in eight clones over three consecutive growing seasons, measuring yield, yield components, and basic must chemistry, the sugar and acid parameters that determine the character of the resulting wine.

The field data told a clear story of phenotypic divergence. The eight clones varied significantly in productivity and showed significant phenotypic plasticity, meaning their performance shifted from year to year depending on growing conditions. Statistical analysis revealed a significant clone-by-year interaction, confirming that no single clone dominates in every season. Two clones maintained under the Foundation Plant Services collection in California, designated FPS03 and FPS06 and known as Primitivo selections, consistently showed lower productivity, smaller berries, higher sugar content, and lower total acidity in the 2021 and 2022 seasons compared with the Zinfandel 1–24 clone. That latter clone displayed greater productivity in all three years of the study, making it the most reliable performer in the trial. For winemakers, these differences matter enormously: smaller berries with concentrated sugars tend to produce more structured, intense wines, while higher-yielding clones favor volume and freshness.

With the phenotypic landscape mapped, the team turned to the genome. Using short-read whole-genome resequencing, they generated high-coverage sequence data for each of the eight clones and called variants across the entire reference genome of grapevine. The scale of the variation was remarkable. Each sample carried approximately 3.8 million single nucleotide polymorphisms, or SNPs, which are positions where a single DNA letter differs from the reference, along with roughly 300,000 insertions and deletions, known as InDels, where short stretches of sequence are present or absent. That level of variation within a single vegetatively propagated cultivar underscores how much somatic mutation and historical mixing of plant material can accumulate over centuries of clonal transmission. Most of the structural variants were located in intergenic regions, the stretches of DNA between genes, which is consistent with the expectation that purifying selection keeps functional gene sequences relatively stable while neutral variation piles up elsewhere in the genome.

To hunt for signatures of clonal selection, the researchers took a windowed approach, dividing the genome into 100-kilobase segments and counting pairwise SNP differences among the eight clones within each window. This method highlights genomic neighborhoods where particular clones stand out from the rest, suggesting that a mutation or set of mutations arose on a specific vine lineage and was then propagated along with it. The analysis revealed candidate regions that distinguish individual clones, providing a catalog of genomic intervals that future studies can mine for genes influencing berry size, sugar transport, acid metabolism, and other traits of viticultural interest. Because the windows are small enough to pinpoint candidate loci but large enough to be robust to sequencing noise, the approach offers a practical middle ground between whole-genome association studies and targeted candidate-gene sequencing.

Then came the twist. When the researchers performed a principal component analysis on the genome-wide SNP data, the eight clones resolved into three principal genetic clusters. If clonal selection had driven the phenotypic differences, one might expect the low-yielding, sugar-rich Primitivo clones to group together genetically, apart from the productive Zinfandel selections. Instead, the genetic groupings showed no clear geographic association among the samples and, more strikingly, no direct association with the phenotypic or productive traits of individual clones. The clones that look most different in the vineyard are not necessarily the ones that are most different in their DNA, at least not at the resolution of genome-wide SNP patterns. Measurable genetic differentiation is real, but it is not a simple mirror of measurable phenotypic divergence.

This disconnect is not a failure of the study; it is arguably its most valuable finding, and it echoes a pattern seen in other clonally propagated crops. Phenotypic differences among clones can arise from a small number of mutations with large effects, from epigenetic changes such as DNA methylation that whole-genome resequencing of the kind performed here does not capture, from structural rearrangements too large or too complex for short-read sequencing to resolve, or from interactions between genotype and environment that three field seasons, however carefully measured, only begin to sample. The significant clone-by-year interaction observed in the trial reinforces this last point: the same clone can rank differently for yield and must composition depending on the season, so any attempt to link genes to traits must account for environmental noise. The candidate SNP signatures identified in 100-kilobase windows now provide a starting list of genomic regions where deeper investigation, including long-read sequencing, epigenomic profiling, and functional validation, is most likely to pay off.

For the wine industry, the implications are practical as well as scientific. Clonal selection programs rely on the assumption that the clones they distribute are genetically distinct and predictably different in performance. The new study confirms that Tribidrag clones carry abundant genomic variation and that some of it marks individual clonal lineages, which strengthens the case for DNA-based tools to authenticate and track clonal material. At the same time, the weak link between genome-wide differentiation and field performance is a caution against assuming that a genetic profile alone can predict how a clone will behave in a given vineyard. Growers choosing between the consistently productive Zinfandel 1–24 and the smaller-berry, higher-sugar Primitivo selections FPS03 and FPS06 will still need multi-season field data, but they now have a far richer genomic map to guide the search for the mutations that ultimately matter in the glass.

The work also carries a conservation message. Tribidrag survives in Croatia as a heritage cultivar of limited distribution, and the study was supported by the Croatian Science Foundation through the RemainGrape project, which is dedicated to the prospection and genomic analysis of endangered Croatian grapevine cultivars and wild grapevine. By documenting both the phenotypic richness and the genomic depth of variation within a single ancient variety, the research makes the case that preserving old vineyards and clonal collections preserves an irreplaceable archive of somatic evolution. Every century-old vine carries mutations accumulated over generations of vegetative transmission, and each clone is a living record of both natural mutation and human selection. Sequencing eight clones has revealed how much of that record remains to be read, and how much of the connection between the grapevine genome and the wine in the bottle is still waiting to be decoded.

Subject of Research: Whole-genome sequencing and phenotypic analysis of clonal variation in the grapevine cultivar Tribidrag (Zinfandel)

Article Title: Whole-genome sequencing of eight grapevine clones of cv. Tribidrag (Vitis vinifera L.) reveals measurable differentiation not linked directly to phenotypic divergence

Article References: Zdunić, G., Duan, S., Wang, C., Lukšić, K., Marinov, L., Mucalo, A., Ozretić Zoković, M., Maletić, E., Pejić, I., & Chen, W. (2026). Whole-genome sequencing of eight grapevine clones of cv. Tribidrag (Vitis vinifera L.) reveals measurable differentiation not linked directly to phenotypic divergence. BMC Plant Biology. https://doi.org/10.1186/s12870-026-10110-x

Image Credits: AI Generated

DOI: 10.1186/s12870-026-10110-x

Keywords: Tribidrag, Zinfandel, grapevine, clonal variation, whole-genome sequencing, somatic mutation, Vitis vinifera, viticulture, SNP, phenotypic plasticity, plant genetics, clonal selection

The review, authored by Giulio Caravagna of the University of Trieste and Area Science Park, Trevor A. Graham of the Centre for Evolution and Cancer at The Institute of Cancer Research in London, and Andrea Sottoriva of Human Technopole in Milan, makes a deceptively simple but profound point. Sequencing is a snapshot, whereas evolution is a dynamic process. A single tumour biopsy tells us which mutations are present and at what frequencies, but it does not, by itself, reveal the underlying dynamics that produced them. Bridging this gap requires models. Population genetics, the discipline developed over the past century to understand how allele frequencies change in natural populations under the influence of mutation, selection and drift, provides exactly the framework needed to transform static measurements of variant allele frequencies into quantitative estimates of clonal fitness and evolutionary timings.

At the heart of the framework lies the concept of the site frequency spectrum, the distribution of mutations across different variant allele frequencies within a tumour sample. In a neutrally evolving tumour, one in which no clone enjoys a fitness advantage over its neighbours, mathematical theory predicts a characteristic power-law tail: a predictable excess of mutations at progressively lower frequencies, each arising in expanding lineages as passengers hitchhiking along with the growing clone. This ‘neutral tail’ has become a signature of neutral tumour evolution, first identified across cancer types in work led by the same research groups. Deviations from this expected distribution are the fingerprints of selection. When subclones carrying driver mutations expand faster than neutral expectations, the frequency spectrum distorts in characteristic and quantifiable ways, allowing researchers to estimate the strength of selection rather than merely guess at it.

The authors show how modern computational tools exploit these principles. Subclonal deconvolution, the process of resolving a bulk sequencing sample into its constituent clonal and subclonal populations, has traditionally been treated as a clustering problem. But clustering alone cannot distinguish between a tumour shaped by strong selection and one dominated by spatially constrained growth, where genetic diversity accumulates by neutral drift in separate geographic compartments. By embedding population genetic models directly into inference algorithms, for example approaches combining machine learning with branching process theory, researchers can distinguish genuine selection from the illusions created by tumour architecture and sampling bias. Simulations of mutation, drift and selection can be tuned until the synthetic genomic data they produce match patient samples, converting descriptive allele frequencies into estimates of evolutionary parameters such as selection coefficients and the timing of clonal expansions.

Timing is one of the most clinically valuable outputs of this framework. Clock-like mutational processes, such as the spontaneous deamination of methylated cytosines, accumulate at approximately constant rates, providing a molecular clock against which key events can be dated. Studies of clear cell renal cell cancer have used these principles to time landmark events in tumour evolution, revealing that many chromosomal catastrophes occur astonishingly early, sometimes decades before diagnosis. More recent theoretical work has shown that patient age itself can help distinguish selection from causation in cancer genomes, since a mutation that arises early and confers a growth advantage leaves a different statistical imprint than one that simply accumulates with time. Dating driver events, genome doublings and the origins of metastatic seeds transforms the tumour genome from a parts list into a chronicle.

The review also confronts the limitations and confounders that complicate evolutionary inference from real data. Bulk sequencing averages across millions of cells, obscuring rare subclones and entangling spatial structure with temporal dynamics. Multi-region sequencing and single-cell approaches help, but each introduces its own biases: sampling depth, copy number alterations that distort allele frequencies, and the fundamental fact that a biopsy represents only a fragment of a spatially extended population. Copy number changes in particular must be carefully modelled, since amplifications and deletions shift variant allele frequencies in ways that can mimic or mask selection. The authors emphasise that assumptions embedded in population genetic models, such as well-mixed populations or constant growth rates, must be tested rather than taken for granted, because spatially constrained tumour growth can generate patterns that superficially resemble selection in the absence of any fitness advantage.

Beyond genetics, the framework extends to epigenetic inheritance and phenotypic plasticity, two dimensions of cancer evolution that the standard genetic models handle poorly. Epigenetic states such as DNA methylation patterns are heritable across cell divisions and can be under selection, yet they are reversible and can switch stochastically, creating a one-to-many relationship between genotype and phenotype. Quantitative models adapted from evolutionary theory, including those describing phenotypic plasticity and stochastic switching in fluctuating environments, offer a way to measure the heritability, transition rates and fitness consequences of non-genetic states. This matters enormously for therapy, because drug-tolerant persister cells frequently arise through epigenetic reprogramming rather than genetic mutation, and their dynamics determine whether resistance emerges in weeks or years.

The ecological dimension of tumour evolution receives similar treatment. Cancers are not just populations of competing clones; they are ecosystems in which cells cooperate, cheat and interact with stromal and immune cells. Game theory and eco-evolutionary models capture frequency-dependent selection, in which the fitness of a clone depends on the composition of its neighbourhood, something classical population genetics assumes away. These models have practical consequences. Adaptive therapy strategies, which aim to maintain sensitive clones that suppress resistant ones rather than eradicate the tumour wholesale, draw directly on ecological and game-theoretic reasoning. Similarly, immune selection on neoantigens shapes both tumour antigenicity and response to checkpoint inhibitors, and can be quantified using selection metrics adapted from population genetics.

Ultimately, the review’s central message is one of reframing. Cancer genomes should be read not as catalogues of mutations but as quantitative records of evolutionary processes, in which every allele frequency, every frequency-spectrum distortion and every signature of mutational timing encodes information about the dynamics that produced them. The authors argue that population genetics provides the foundation for understanding and, ultimately, predicting the trajectories of cancer evolution. If the framework fulfils its promise, the implications for precision oncology are substantial: forecasts of relapse timing grounded in measured evolutionary parameters, treatment strategies designed around the predictable dynamics of resistance, and clinical decisions informed not merely by which mutations a tumour carries today, but by the evolutionary forces that will shape what it becomes tomorrow.

The intellectual roots of this framework stretch well beyond oncology. Population genetics matured through the study of natural populations, and its migration into cancer research has been gradual, beginning with early attempts to reconstruct individual tumour histories from genetic data and gaining momentum as sequencing costs fell. A parallel body of work on clonal haematopoiesis and on pre-malignant lesions in normal tissues has reinforced the point that the evolutionary processes described in the review operate long before a tumour is diagnosed. Studies of normal breast tissue, for example, have revealed rare aneuploid epithelial populations and copy number alterations shared with frank cancers, suggesting that the same population genetic tools can illuminate the earliest steps of carcinogenesis across a continuous spectrum from healthy tissue to invasive disease.

Selection in tumours is not exclusively positive. Adapted metrics such as the ratio of synonymous to non-synonymous mutations, borrowed directly from classical genetics, have revealed universal patterns of selection in cancer and somatic tissues, while complementary analyses indicate that negative selection acts on essential cellular functions and on the immunopeptidome, pruning mutations that would compromise basic biology or expose cells to immune attack. Copy number amplifications of wild-type regions may further allow tumours to tolerate otherwise deleterious coding mutations, a reminder that the interplay between different classes of genomic alteration can obscure simple fitness calculations. At population scale, large studies of somatic mutation across many individuals now provide the statistical power to measure these forces with unprecedented precision.

Practical questions of study design also fall within the framework’s remit. How many biopsies are needed to confidently identify truly clonal mutations in a heterogeneous tumour, and how sampling schemes affect the inferred frequency spectrum, are problems that can themselves be solved with evolutionary models rather than ad hoc rules. The review’s figures trace this progression, from conceptual overviews of genomic readouts through the dynamics of variant allele frequencies, the confounding effects of spatial structure, and the distinct evolutionary routes by which tumours respond to therapy and relapse. Together they map a research programme in which the reliability of every evolutionary claim is tied explicitly to the sampling strategy and the model assumptions behind it.

Subject of Research: Application of population genetics models to infer tumour evolutionary dynamics from cancer genome sequencing data.

Article Title: A guide to understanding tumour evolution through the lens of population genetics

Article References: Caravagna, G., Graham, T. A., & Sottoriva, A. (2026). A guide to understanding tumour evolution through the lens of population genetics. Nature Reviews Cancer. https://doi.org/10.1038/s41568-026-00973-5

Image Credits: AI Generated

DOI: 10.1038/s41568-026-00973-5

Keywords: tumour evolution, population genetics, cancer genomics, clonal selection, variant allele frequency, subclonal deconvolution, genetic drift, mutational signatures, phenotypic plasticity, adaptive therapy, tumour heterogeneity, molecular clock