Four of the children were supposed to be on the better side of the ledger.

there is an overarching cell state of aggressive disease that defies conventional classification systems
Sam Behjati

Each had rhabdomyosarcoma, a childhood cancer that grows from cells resembling developing muscle. None of their tumors carried the gene fusion that oncologists use to pick out the most dangerous form of the disease, and the Wellcome Sanger Institute describes their cancers as non-high-risk tumors that behaved aggressively. All four died. A team from the Sanger, the University of Cambridge, Great Ormond Street Hospital and University College London went back into those tumors one cell at a time and found populations of cancer cells whose gene activity looked like the high-risk disease the label had ruled out.

The paper, “High-Risk Rhabdomyosarcomas Feature a Convergent Cell State”, ran in Cancer Research this summer under first author Holly J. Whitfield, with Karin Straathof of UCL and Great Ormond Street and Sam Behjati of Cambridge as senior authors, and is circulating again this week as “Why some ‘lower-risk’ childhood cancers turn deadly”.

The label and what it decides

The disease is rare, with about 55 children diagnosed in the United Kingdom each year, according to the Sanger. With so few patients, the system that sorts them into treatment groups carries enormous weight, because it decides who gets heavier chemotherapy, for how long, and who is spared it.

That system gives one molecular feature a great deal of authority. Some tumors fuse two normally separate genes, PAX3 or PAX7 with FOXO1, and fusion status has become a key prognostic marker; it sits alongside stage, site and surgical outcome rather than replacing them, but a fusion-negative result is generally the reassuring one. The distance between groups is wide. A 2022 Children’s Oncology Group review in Pediatric Blood & Cancer put four-year event-free survival for low-risk disease at approximately 90%, and five-year event-free survival for intermediate-risk disease at 50–75%. For patients with alveolar rhabdomyosarcoma on the group’s high-risk trials, three-year event-free survival ran as low as 6–16%. A child’s column on that table shapes the next year of their life.

Some fusion-negative children die anyway, and the classification has never explained why.

Sixteen tumors, read cell by cell

The team worked from tumor samples taken from 16 children: 4 with fusion-negative disease who died, 4 with fusion-positive disease, and 8 with fusion-negative disease who are still alive, per the Sanger release. Single-cell RNA sequencing measured gene activity in individual cancer cells, and spatial transcriptomics mapped where that activity sat inside each tumor. According to the paper’s abstract, the researchers then tested the result with nuclear transcriptomics, chromatin-accessibility profiling, perturbation experiments and published datasets, and the Sanger says the outside validation drew on data from research centers in the United States and the Netherlands.

THE DISCOVERY COHORT
4
fusion-negative, died
4
fusion-positive
8
fusion-negative, alive
The 16 children whose tumors were read cell by cell. The central comparison rests on the first group. Source: Wellcome Sanger Institute news release, August 2026

What they found is in the title. Cancer cells from fatal cases converged on a common cell state with a shared transcriptional landscape, whether or not the tumor carried the FOXO1 fusion. That state only partly overlaps with what the fusion itself does to a cell, and it carries features of nerve tissue that the authors did not expect. In the fusion-negative children whose disease behaved aggressively, the team also found rare genetic changes affecting the same cellular pathway as the high-risk tumors, so different mutations appear to have arrived at the same dangerous destination.

The label said these children were not high-risk. Their cells disagreed.

“By studying tumours one cell at a time, we uncovered hidden populations of aggressive cancer cells that conventional methods miss,” Straathof said in the Sanger release. Behjati went further, saying that there is an overarching cell state of aggressive disease that defies conventional classification systems.

A rulebook patched one gene at a time

The field has known for years that the fusion test misses some dangerous tumors, and its answer has been to write exceptions into the rulebook, one mutation at a time.

In 2021, Jack Shern of the National Cancer Institute and colleagues analyzed 641 patients from American and British trials in the Journal of Clinical Oncology. Among fusion-negative tumors, MYOD1 mutations turned up in only 3% of cases but carried a hazard ratio for an event of 6.84 in the Children’s Oncology Group cohort, while TP53 mutations, found in 13%, carried a hazard ratio of 2.07. The group’s next low-risk trial, ARST2032, was designed as the first to incorporate real-time molecular risk stratification: under that design, fusion-negative children whose tumors carried MYOD1 or TP53 mutations would no longer count as low-risk and would instead receive 42 weeks of VAC chemotherapy.

FUSION-NEGATIVE MUTATIONS AND RISK OF AN EVENT
6.84hazard ratio
MYOD1 mutation
2.07hazard ratio
TP53 mutation
Hazard ratios for an event in the Children's Oncology Group cohort, among fusion-negative tumors. Source: ASCO Post summary of Shern et al., Journal of Clinical Oncology, 2021

That was progress, and it was also an admission. Every newly identified bad mutation becomes one more exception, each covering a thin slice of the children it was written to protect, and the list only grows after enough of them have done badly to be counted. The Cambridge and London work proposes turning the question around: stop cataloguing the routes a tumor can take to lethality and ask whether its cells have arrived.

Four children carry the argument

That proposal currently rests on 4 fusion-negative children who died, set against 8 who lived. It is a sample for generating a hypothesis, measured with research-grade sequencing rather than a clinical test with a defined cutoff, and nothing yet shows that reading this cell state at diagnosis would change what happens to a child. ScienceDaily’s write-up says plainly that further research is needed before the findings can change routine care.

The publicity also has its own priorities. The Sanger release spends its hopeful paragraphs on finding markers unique to the high-risk cells that an engineered CAR-T immunotherapy might target, a therapy with no approved use in rhabdomyosarcoma and the most expensive branch of the idea. The cheaper payoff gets a passing phrase about avoiding unnecessary toxicity for other lower-risk patients. A marker that separates the fusion-negative children who are in danger from the ones who are not would let oncologists escalate treatment for the first group early and hold back for the second with more confidence than a single fusion test allows. The work was paid for by charities, a Cancer Research UK and Children with Cancer UK Innovation Award co-funded by Alice’s Arc, with support from Wellcome, and the release names no industry funder.

Straathof works between UCL and Great Ormond Street, where the sorting happens at diagnosis and the first question asked of a tumor is still, in effect, which gene broke. Her paper’s evidence, four children deep, is that the question which tracked these deaths was a different one: what the cells had become.

Sources

  1. Whitfield et al., Cancer Research, 2026
  2. Wellcome Sanger Institute news release, August 2026
  3. ScienceDaily, October 2026
  4. Prinses Máxima Centrum repository: abstract and authors, Whitfield et al., 2026
  5. ASCO Post summary of Shern et al., Journal of Clinical Oncology, 2021
  6. Haduong et al., Pediatric Blood & Cancer, 2022