NEWS
Pancreatic Cancer Escapes Treatment by Switching Clone Identity
NYU mapping of 443,451 cells shows pancreatic tumors switch clone identity and drop PD-1, a clue to failed immunotherapy.
NYU scientists mapped 443,451 pancreatic cancer cells and found each tumor holds rival clones that switch identity and dodge immune attack. The work, published October 13, 2026, in Nature Genetics, tracks pancreatic ductal adenocarcinoma cell by cell instead of as a blended average.
That average is what most tests still report. Aggressive clones gained extra copies of MYC, turned down PD1, and raised TIGIT, a different immune brake. PD-1 drugs look for a signal those clones have already dimmed.
They Read 443,451 Cancer Cells, One by One
Despoina Kalfakakou, PhD, a postdoctoral fellow at NYU Grossman School of Medicine and Perlmutter Cancer Center, led the NYU Langone Health team with co-senior investigators Aristotelis Tsirigos, PhD, a professor of medicine and pathology at NYU, and Diane Simeone, MD, at the University of California San Diego. Oregon Health and Science University, including Rosalie Sears, PhD, also contributed.
The group profiled 62 primary tumors from 56 men and women. A subset of 11 tumors also got spatial maps, 53,236 spots that show where each clone sits in the tissue. Six tumors were sampled before and after chemotherapy, a small window on what treatment leaves behind.
Single-cell RNA sequencing measured gene activity in each cell. Copy-number changes, extra or missing stretches of DNA, then tied those activity patterns to distinct clones. Individual tumors contained two to 11 clones, mixing less aggressive classical cells with basal cells that invade and spread.
THE NYU CELL CENSUS
| What was counted | Number |
|---|---|
| Single cells | 443,451 |
| Primary tumors | 62 |
| Patients | 56 |
| Tumors with spatial maps | 11 |
| Spatial spots | 53,236 |
| Tumors sampled before and after chemo | 6 |
| Clones per tumor | 2 to 11 |
Kalfakakou’s paper on clonal heterogeneity in pancreatic tumors also rebuilt clone family trees inside each mass. Amplifications of a ligand or receptor enriched the matching cell-to-cell contacts. Deletions suppressed them. The genetic mix was not background noise. It wrote the way each clone talked to the tissue around it.
Two Labels Never Explained the Switch
For more than a decade, bulk tests have sorted pancreatic ductal adenocarcinoma into two transcriptional types. Classical cells look more like normal duct tissue. Basal cells look more injured and track with worse spread. Those labels came from studies that mashed thousands of cells into one score, so they could not show a cell leaving one state for the other.
Doctors still meet the disease as if it were one tumor, or at most those two types. The NYU maps say a single mass can hold both at once, in genetically distinct groups, and that a lineage can move from classical toward basal as it picks up new DNA changes. Tsirigos said even a small region can contain more than one genetically distinct population, and that a routine biopsy may capture that mix without the usual tests being able to tell the groups apart.
That is a practical problem, not a naming dispute. A test that reports the average gene activity of a tumor can miss the clone that is about to spread, or the clone that will survive the next round of chemo. Kalfakakou put it in one line: understanding a tumor’s average molecular profile may not be enough.
Extra MYC Copies Flip Cells Into Attack Mode
When the team lined those family trees against subtype, one genetic event kept showing up at the flip. Focal extra copies of 8q24, the chromosome stretch that holds MYC, were tied to classical-to-basal transitions more often than chance. Those flips came with a coordinated drop in immune-response programs. The clone that turned more aggressive also became harder for immune cells to see.
Lab tests then asked whether MYC was doing the work. In mouse models and in primary human pancreatic cancer cells, knocking MYC down did three things at once.
WHAT MYC KNOCKDOWN DID IN THE LAB
- Basal programs: Aggressive basal-like gene activity fell in the treated cells.
- Classical features: The same cells picked up the milder classical program again.
- Immune pathways: Immune-response signaling returned, and the cells sat more exposed to immune attack.
The switch, in other words, was not a one-way door in these models. Blocking MYC pushed basal cells back toward the classical state. Kalfakakou announced the paper on October 14, 2026, and walked through that MYC result in her own thread.
Our study provides the most granular view to date of pancreatic ductal adenocarcinoma, revealing how ongoing genetic reshaping drives tumor evolution over the course of disease. These findings suggest that targeting the mechanisms underlying tumor evolution, including MYC, may help restrain the cancer’s rapid progression.
Despoina Kalfakakou, PhD, postdoctoral research fellow, NYU Langone Health
Thrilled to share our study, now published in @NatureGenet! 🎉
Using single-cell & spatial transcriptomics, we investigated how clonal heterogeneity shapes pancreatic cancer progression and what this could mean for precision medicine.
🧵👇
— Despina Kalfakakou (@DespinaKalf) October 14, 2026
MYC is not a new suspect. It is a growth-control gene that cancer after cancer turns up, and the protein has no convenient pocket for a classic pill, which is why drugmakers spent years calling it undruggable. The new work does not invent that biology. It pins extra MYC copies to a specific identity change inside human pancreatic tumors and shows, in models, that turning MYC down can reverse the change.
Why PD-1 Drugs Keep Missing Pancreatic Tumors
Checkpoint drugs that help in melanoma and some lung cancers have done little for most people with pancreatic cancer. Dense scar-like tissue and a shortage of T cells inside the tumor are the usual explanations. The NYU clones add a cleaner mechanical reason.
In some clones the TIGIT gene was more active, making proteins that help cancer hide from immune cells. In those same cells, activity of the PD1 gene was suppressed. PD-1 drugs are built to block a pathway those clones have already turned down. TIGIT is the brake they still use.
That pattern is the piece most write-ups of the paper bury under the MYC headline. If the clone that is growing is the clone that dropped PD1 and raised TIGIT, a PD-1 antibody is aimed at the wrong handle. Tsirigos said that mismatch may help explain why some PD-1 immunotherapies are rarely effective against this disease, and that trials of MYC inhibitors or TIGIT inhibitors are already underway.
TIGIT blockade is still unproven in pancreatic cancer. Early randomized signals for the class came from lung cancer, not from the pancreas, and the broader solid-tumor record has been uneven. The NYU data do not show that a TIGIT drug will work. They show why a PD-1 drug was a long shot against the clones that matter.
Doctors Are Already Testing Drugs That Block MYC
No MYC inhibitor is on the market. A mini-protein called OMO-103, derived from Omomyc, is the furthest along in people. In its first solid-tumor study, the European MYCureX project reported 49% tumor shrinkage in one patient with pancreatic ductal adenocarcinoma and stable disease in 8 of 12 people who could be evaluated.
A later combination study is testing OMO-103 with standard chemotherapy, gemcitabine plus nab-paclitaxel, in people with previously untreated metastatic pancreatic cancer. That phase 1b trial, run by Peptomyc, started August 31, 2023, and listed an estimated primary completion date of May 31, 2026.
A separate early phase 1 study at OHSU Knight Cancer Institute opened January 27, 2026. The OMO-103 tumor biopsy trial is built to measure how the drug changes pancreatic tumor cells, with 12 participants planned and an estimated completion date of December 31, 2027. Sears, a co-investigator on the NYU paper, is based at Oregon Health and Science University, the same campus running that biopsy study. The paper does not claim to have designed the trial. It does supply the exact lab readout the trial is trying to see in patients: a shift away from basal programs and a return of immune-response signals.
MYC DRUGS ALREADY IN PEOPLE
| Study | What it tests | Recorded dates |
|---|---|---|
| NCT06059001 | OMO-103 plus gemcitabine and nab-paclitaxel in untreated metastatic disease | Started August 31, 2023; estimated primary completion May 31, 2026 |
| NCT07089940 | How OMO-103 changes tumor cells in biopsies | Opened January 27, 2026; 12 participants planned; estimated completion December 31, 2027 |
WHAT WE KNOW
- Clone mix: Each of the 62 tumors held two to 11 genetically distinct clones, not a single uniform mass.
- The switch: Extra 8q24 MYC copies tracked with classical-to-basal flips and a drop in immune programs.
- Lab reversal: MYC knockdown in mouse and human models restored classical features and immune-response pathways.
WHAT IS UNCONFIRMED
- Patient effect: Whether OMO-103 or any MYC drug will reverse basal clones inside human tumors is still an open trial question.
- TIGIT drugs: Whether blocking TIGIT will help in pancreatic cancer is unproven, even though the clones point at that brake.
- Chemo editing: Only 6 tumors were sampled before and after chemotherapy, too few to set a treatment rule.
Those limits belong next to the hope. The six paired tumors held 24 clones before chemotherapy and 31 after. Some clones were wiped out, some were kept, and new ones appeared. Chemo did not freeze the crowd. It remodeled it.
The Dangerous Clones Were the Ones That Spread Out
Spatial maps of the 11 tumors picked up a second pattern that bulk tests cannot see. Clones that were dispersed through the tissue, rather than sitting in a tight nest, were rich in programs cells use to detach and migrate. They showed immune suppression, tracked with metastatic potential and lymph-node colonization, and themselves drifted toward the basal form.
WHAT THE SPATIAL MAPS SHOWED
- Dispersed clones: Scattered groups carried epithelial-to-mesenchymal programs tied to movement and invasion.
- Immune cover: Those same groups showed immune suppression, matching the MYC-linked hide-from-T-cells pattern.
- Lymph nodes: The dispersed, basal-shifting clones correlated with colonization of lymph-node niches.
The National Cancer Institute’s SEER program puts 13.7% five-year relative survival for pancreatic cancer diagnosed in 2016 through 2022. For 2026 it projects 67,530 new cases and 52,740 deaths, 3.2% of new cancers and 8.4% of cancer deaths. Localized disease, 15.4% of diagnoses, has 43.6% five-year relative survival. Regional disease sits at 17.0%. Distant disease sits at 3.4%. In 2023, 113,931 people in the United States were living with the cancer.
Those figures are why a clone map is not an academic toy. Most people are diagnosed after the tumor has already left the pancreas. The NYU work says the cells that leave are not a random draw from the average tumor. They are a subset that rewrote its DNA, flipped toward basal, and dimmed the immune pathway PD-1 drugs were built to hit.
Kalfakakou said looking at tumors at clonal resolution could reveal aggressive populations that conventional tests miss, and that this could open paths for diagnostics and patient stratification. Tsirigos said single-cell clonal analysis offers a better way to watch a complex cancer as it develops, by mapping how it adapts at each stage. The next hard measurement is already on a protocol at OHSU: biopsies from people receiving OMO-103, asking whether the identity switch that reversed in a dish will reverse in a patient.
Frequently Asked Questions
What Is Pancreatic Ductal Adenocarcinoma?
It is the main exocrine cancer of the pancreas, starting in the duct cells that carry digestive enzymes. Exocrine tumors account for about 95% of pancreatic cancers, and they are the group the NYU maps describe. Neuroendocrine tumors of the pancreas are a separate, often slower disease and were not the subject of this study.
What Is the Difference Between Classical and Basal Pancreatic Cancer Cells?
Classical cells resemble normal duct tissue more closely and have been linked in older genomic studies to somewhat better responses to standard chemotherapy. Basal cells look less differentiated and track with worse spread. Those two labels date to gene-activity studies from 2011 and 2015. They are still not a routine test at every hospital, and the NYU work shows they are states a clone can move between, not fixed tribes.
What Does the MYC Gene Do in This Cancer?
MYC sits on chromosome 8q24 and acts as a master growth switch. The protein has no convenient pocket for a classic small-molecule drug, which is why no MYC inhibitor is marketed. In this study, extra copies of that 8q24 stretch were the genetic event most tied to a clone flipping into the basal state and dropping immune-response programs.
Why Do PD-1 Drugs Often Fail in Pancreatic Cancer?
Most of these tumors have few T cells inside them and a dense, fibrotic shell. The NYU clones add a second, clone-level problem: aggressive groups turned up TIGIT, another immune brake, while turning down PD1. A PD-1 antibody then aims at a pathway those clones have already dimmed. Outside rare mismatch-repair-deficient cases, checkpoint drugs have shown little benefit for unselected pancreatic cancer.
Can Patients Get a MYC Inhibitor Now?
No MYC inhibitor is approved. OMO-103 is in early-phase trials, including a chemotherapy combination study and a 12-person biopsy study designed to see how the drug changes tumor cells. Eligibility depends on the trial’s rules, stage of disease, and prior treatment. A pancreatic cancer center can check current listings; this paper is a map of clone behavior, not a prescription.
Disclaimer: This article is news reporting on a published laboratory and tissue-mapping study. It is for information only and is not medical advice, a diagnosis, or a treatment recommendation. It does not advise anyone to start, stop, or change cancer therapy, join a trial, or use an unapproved MYC or TIGIT drug. Readers should talk with a qualified oncologist who knows their case before making any treatment decision. Cell counts, survival figures, and trial dates reflect the Nature Genetics paper, SEER statistics, and trial records cited here and may change as new data are posted.
-
NEWS3 months agoMeta’s Smart Glasses Sell a Light That Needed Two Patches
-
NEWS3 months agoApple’s New Mac Studio Clusters Local AI on the Desk
-
NEWS3 months agoGoogle’s Free AI Pro Year Is Three Different Deals
-
NEWS3 months agoSong Yadong Stops Umar Nurmagomedov the Only Way Left
-
ENTERTAINMENT3 months agoLanterns Puts Sinestro in a Cell and Lets Him Talk
-
NEWS3 months agoThe Mahape Fake Apple Raid Repeats a Dual-Shift Pattern
-
ENTERTAINMENT3 months agoLUN8’s First London Concert Filled a 620-Cap Club
-
NEWS3 months agoOpenAI Agents Hit Hugging Face to Cheat a Test
