The starting point for the new work is a paradox that has haunted the field for years. More than ninety percent of pancreatic ductal adenocarcinomas — the most common and deadliest form of the disease — are driven by a mutation in the KRAS gene, an oncogene long considered untouchable. KRAS sits at the top of the MAPK signaling cascade, a molecular relay that carries growth instructions from receptors on the cell surface through RAS, RAF and MEK to ERK, the kinase that delivers the final commands to the nucleus. Because this pathway runs at full throttle in virtually every pancreatic tumor cell, drugs that block MEK or ERK should, in principle, strangle the cancer. In the clinic, they have not. Tumors adapt, rewire and rebound, and even the newest generation of allele-specific KRAS inhibitors, while producing genuine excitement, faces the same specter of resistance. The new study traces the root of the problem not to the pathway itself but to what the pathway controls: the identity of the tumor cell.

Cancer biologists have known for more than a decade that pancreatic tumors are mosaics rather than masses of identical cells. Single-cell analyses have repeatedly shown that, within a single patient, malignant cells can occupy sharply different states. Some resemble the progenitor cells that build the pancreas during embryonic development, expressing genes for duct formation, secretion and tissue architecture. Others adopt an aggressive, basal-like program reminiscent of squamous tissue, loaded with inflammatory and stress-response modules and associated, in clinical datasets, with poorer survival. These states are not fixed. Under the pressure of chemotherapy or targeted drugs, cells can switch between them, and that plasticity is widely believed to underlie the rapid resistance that makes pancreatic cancer so lethal. The missing piece has been a mechanistic account of who directs the switch. The prime suspects are transcription factors — proteins that recognize short DNA motifs in enhancers and promoters and switch entire gene programs on or off — because they sit at the apex of gene-regulatory hierarchies. The challenge has been identifying, from among the hundreds of factors active in a tumor, which ones actually hold the power.

To find the puppeteers, the researchers married two kinds of evidence that are rarely combined at this depth. First, they profiled pancreatic tumors and laboratory model systems at single-cell resolution, capturing both gene expression and chromatin accessibility — a measure of which regulatory regions of the genome are physically open and in use. Open chromatin is where active transcription factors bind, so reading the two layers together allowed the team to infer not merely which factors were present in a cell but which were functionally engaged with its DNA. Computational analyses then reconstructed the regulatory networks, or regulons, controlled by each factor, separating the drivers of malignant identity from the many genes that simply come along for the ride. Second, the investigators overlaid this map of identity onto functional data: the consequences of experimentally perturbing transcription factors, and the dependency profiles generated by CRISPR-based screens, in which genes are systematically disabled one at a time to reveal which ones a cell cannot survive without. The ambition was to connect, within a single framework, what a cancer cell is with what it needs.

The analysis revealed two fundamentally different regulatory modules governing the malignant cells. The first is built from developmental transcription factors — the same family of proteins that, in the embryo, coaxes the pancreas into existence from the foregut endoderm, specifying ductal and acinar lineages and orchestrating the organ’s branching architecture. Factors of the FOXA, GATA and SOX families, together with members of the KLF family, are largely silent in healthy adult tissue, yet pancreatic cancer reactivates them, effectively resurrecting a fetal program inside the tumor. In the new study, these developmental factors sustain a progenitor-like malignant state: they occupy the enhancers of lineage-specific genes and keep an entire identity program running. When the team suppressed their activity in experimental systems, that state collapsed, and the gene-expression profile of the cells shifted dramatically. The finding establishes that these embryonic factors are not inert relics of the tumor’s history but active, ongoing guardians of its identity — and therefore legitimate targets in their own right.

The second module answers to a very different master: the KRAS–MAPK pathway itself. When ERK, the terminal kinase of the cascade, is active, it phosphorylates a collection of short-lived transcription factors — among them members of the ETS family such as ELK1 and the ETV proteins, and components of the AP-1 complex assembled from FOS and JUN. These molecules carry the pathway’s signal from the cytoplasm into the chromatin within minutes, converting each pulse of growth-factor stimulation into a burst of gene expression. The study shows that, in pancreatic cancer, these MAPK-responsive factors do far more than relay instructions; they hold together a second, distinct malignant state. Because their activity is a direct readout of KRAS signaling, that state is structurally coupled to the oncogenic engine driving the disease. The coupling offers a new lens on why MAPK inhibitors produce such volatile responses in patients: shutting down the pathway does not simply remove a growth signal, it destabilizes the identity of an entire population of tumor cells, forcing them to adapt or die.

The most consequential discovery, however, concerns what those identities imply for treatment. Working from the premise that a cell’s gene-expression program dictates which genes it depends on, the team compared the transcription factor–defined states with genome-wide dependency data. The pattern that emerged was strikingly modular. Cells in the developmental, progenitor-like state relied on one set of essential genes; cells in the MAPK-driven state depended on a largely different set, involving distinct chromatin regulators, signaling molecules and stress-response machinery. Identity and vulnerability, in other words, travel together. A genetic weakness that is lethal to one state can be irrelevant to its neighbor, which offers a coherent explanation for one of clinical oncology’s most familiar frustrations: therapies that produce dramatic responses in some patients leave others untouched, and the responses that do occur are often brief. It also reframes tumor heterogeneity — long treated as a static complication of tumor architecture — as an active, mappable axis of drug response that can, in principle, be predicted from a tumor’s transcriptional state.

For therapy, the implications run in two directions. The first is diagnostic. If the dominant transcription factor program in a patient’s tumor can be identified — through single-cell profiling, or eventually through simpler biomarkers of state — that information could indicate which genetic dependencies are actually targetable in that patient, turning treatment selection from a gamble into an informed decision. The second is combinational. Because malignant states are plastic, attacking a dependency in one state may simply push the surviving cells into the other, where an entirely different dependency holds sway. The study’s framework therefore argues for drug combinations calibrated to a tumor’s state architecture — pairing, for example, a MAPK-pathway inhibitor with an agent aimed at the vulnerabilities unique to the state that MAPK signaling sustains, or coupling an attack on the developmental module with coverage of the state cells adopt when it fails. Rational combinations of this kind, the data suggest, are the logical counter to a cancer whose central survival strategy is changing costumes.

None of this will be easy. Transcription factors have a well-earned reputation for being undruggable: their DNA-binding surfaces are broad, featureless and lack the deep pockets that small-molecule drugs usually require. Yet the landscape is changing. Degraders designed to eliminate transcription factors rather than block their binding, molecules that disrupt the interactions between factors and their cofactors, and drugs aimed at the epigenetic machinery these factors recruit have all begun to bring the once-forbidden protein class within reach of medicinal chemistry. The new study strengthens the case for that effort by demonstrating that, in pancreatic cancer, transcription factors are not accessories to malignancy but the structural engineers of its heterogeneity. It also points to a practical research agenda: systematically perturbing state-defining factors across many models while mapping dependencies in parallel, to build a reference atlas that predicts, for any given tumor, which vulnerabilities accompany which identity.

The principle at stake extends well beyond the pancreas. Cell-state plasticity complicates treatment in many cancers, from melanoma and lung adenocarcinoma to ovarian and breast tumors, and the identity-to-dependency logic described here could in principle be mapped in each of them. For pancreatic cancer — a disease in which direct KRAS inhibitors are only now reaching patients after four decades of effort — the message of the new work is both sobering and energizing. Switching off the oncogene is necessary, but it may not be sufficient, because the malignant identities that the oncogene has organized do not dissolve overnight, and the cells that survive the transition may be the very ones that seed resistance. What the Nature Genetics study delivers is a search strategy rather than a finished drug: identify the transcription factor guarding each malignant state, define the dependency that state cannot live without, and combine the two attacks so that the tumor is left with nowhere to turn. It is a blueprint for converting one of cancer’s most formidable defenses — its versatility — into its most exploitable flaw.

The study describes how specialized regions known as macrophage niches can influence the biological state of non-small cell lung cancer stem cells. A niche is more than a physical location inside a tumor. It is a dynamic microenvironment made up of immune cells, connective-tissue cells, blood vessels, signaling molecules and extracellular matrix components. Together, these elements provide chemical and mechanical instructions that can alter how cancer cells grow, move, survive and respond to treatment. In this setting, macrophages appear to help push tumor cells into flexible states associated with dormancy, epithelial–mesenchymal transition and drug resistance.

Cancer stem cells are a small but important population within many tumors. Unlike the majority of cancer cells, they can self-renew and generate different tumor cell types, allowing a malignancy to rebuild itself after treatment. Their ability to remain alive in a slow-growing or non-dividing state is particularly significant. Dormant cells may escape therapies designed to kill rapidly dividing cells, then become active again months or years later. This biological “pause” can help explain why a patient may initially respond well to treatment but later experience relapse.

The research links macrophage-rich environments to this dormant behavior. Macrophages are part of the innate immune system and normally remove damaged cells, coordinate inflammation and support tissue repair. Tumors can reshape these cells through chemical signals, creating tumor-associated macrophages that often suppress immune attacks and promote tissue remodeling. Within lung tumors, these macrophages may release growth factors, cytokines and other signaling molecules that activate survival programs in cancer stem cells. Instead of simply stimulating proliferation, the signals can encourage cells to enter a protected, low-activity state.

A central concept in the study is tumor plasticity, the ability of cancer cells to change their characteristics without necessarily acquiring new genetic mutations. This flexibility allows a tumor cell to shift between states depending on environmental pressure. A cell may become highly invasive during one phase, dormant during another and actively dividing when conditions improve. Such changes are controlled through networks involving transcription factors, cell-surface receptors, metabolic pathways and epigenetic modifications, which alter gene activity without changing the underlying DNA sequence.

The researchers also connect macrophage niches with epithelial–mesenchymal transition, commonly known as EMT. During EMT, cells lose features associated with epithelial tissue, such as strong cell-to-cell adhesion, and acquire mesenchymal traits that increase mobility and invasiveness. In cancer, EMT can help tumor cells detach from the primary mass, migrate through surrounding tissue and enter the bloodstream. It is also associated with stem-like properties and resistance to several forms of therapy. The study suggests that macrophage-derived signals may help lung cancer stem cells move along this spectrum, linking dormancy and metastatic potential rather than treating them as separate phenomena.

Chemoresistance is another major consequence of this cellular flexibility. Many chemotherapy drugs are most effective against cells that are actively replicating or depend heavily on particular metabolic pathways. Dormant cancer stem cells may avoid these vulnerabilities by reducing proliferation and changing their metabolism. EMT-associated cells can also increase drug-export mechanisms, strengthen DNA-damage responses and activate anti-apoptotic pathways that prevent programmed cell death. If macrophage niches maintain these protective states, they could act as local shelters where cancer cells survive treatment and later repopulate the tumor.

These findings may help explain why targeting tumor cells alone is often insufficient. A therapy can eliminate a large fraction of malignant cells while leaving behind a smaller population protected by its microenvironment. The results point toward combination strategies that attack both cancer stem cells and the macrophage signals supporting them. Potential approaches could include therapies that reprogram tumor-associated macrophages, block specific communication pathways between macrophages and cancer cells, or force dormant cells into a vulnerable state before conventional treatment. Such strategies remain an area of investigation and would require careful testing because macrophages also perform essential functions in normal immunity and tissue repair.

The study does not suggest that every macrophage in every lung tumor behaves identically, nor that dormancy, EMT and drug resistance are controlled by a single mechanism. Tumors are highly diverse, and the behavior of a macrophage niche can depend on its location, molecular composition and interaction with other immune and stromal cells. Translating these observations into patient treatments will require determining which signaling pathways are most important in individual tumors and identifying reliable biomarkers that reveal when cancer stem cells are being protected by their surroundings. Even so, the work reinforces a growing principle in cancer biology: treatment resistance is not solely a property of tumor cells, but can emerge from an ongoing conversation between malignant cells and the ecosystem around them.

Subject of Research: The role of macrophage niches and tumor plasticity in regulating dormancy, epithelial–mesenchymal transition and chemotherapy resistance in non-small cell lung cancer stem cells.

Article Title: Tumor plasticity in macrophage niches promotes dormancy, EMT and chemoresistance of non-small cell lung cancer stem cells.

Article References: Zeuner, A., Sette, G., Rossi, S. et al. “Tumor plasticity in macrophage niches promotes dormancy, EMT and chemoresistance of non-small cell lung cancer stem cells.” Cell Death Discovery (2026). https://doi.org/10.1038/s41420-026-03261-1

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41420-026-03261-1

Keywords: non-small cell lung cancer, cancer stem cells, macrophages, tumor microenvironment, tumor plasticity, dormancy, epithelial–mesenchymal transition, EMT, chemoresistance, cancer relapse