Glioblastoma kills most of its victims within 18 months of diagnosis. There is no cure, and the therapies that exist — surgery, radiation, temozolomide chemotherapy — have changed little for years. That statistics-desperate patient population is why any credible new target in this cancer draws attention. The question is always the same: does the biology actually hold up?
A paper published this week in the Proceedings of the National Academy of Sciences (PNAS) makes a case that has the field paying attention. A team at the Centre for Cancer Biology at the University of Adelaide, in Australia, has identified a second, previously unrecognized mechanism by which the protein CD47 drives glioblastoma growth — one that operates entirely separately from the immune-evasion role that made CD47 a popular drug target in the first place. The paper, received by PNAS on January 29, 2026 and accepted February 13, was edited by Karen H. Vousden of the Francis Crick Institute in London.
CD47 is well-established as a "don't eat me" signal — a protein on the surface of cancer cells that tells macrophages to stand back, effectively hiding tumors from the immune system. Drug developers have been targeting this mechanism for years. What the Adelaide team found is that CD47 does considerably more than that. According to the paper, CD47 also sequesters a protein called ITCH, which would otherwise tag another protein, ROBO2, for destruction. When CD47 is present at high levels — as it is at the invasive edges of glioblastoma tumors — ITCH is occupied and ROBO2 accumulates unchecked. ROBO2, in turn, promotes tumor cell migration and invasion into healthy brain tissue.
The researchers demonstrated this in lab models and animal experiments. When they removed or blocked CD47, tumor cell proliferation, migration, and invasion dropped sharply. The effect persisted even in the absence of immune cells, confirming that the tumor-promoting mechanism operates independently of CD47's role in immune evasion. In some cases, tumors lacking CD47 grew more slowly and survival time in the models nearly doubled. Patients whose tumors had higher CD47 levels had significantly poorer survival outcomes, according to data cited in the paper from News-Medical.
The finding has a direct bearing on the pharmaceutical industry's rocky history with CD47 as a target. Gilead Sciences acquired Forty Seven, the primary developer of the anti-CD47 antibody magrolimab, in 2020 for $4.9 billion. The bet was that blocking CD47's "don't eat me" signal would unmask tumors for immune destruction. It didn't work out: Gilead removed the remaining solid tumor trials from its pipeline on April 29, 2024, per Fierce Biotech, effectively ending the magrolimab program. ALX Oncology's evorpacept, another CD47 inhibitor, also failed two Phase II trials as of 2025. Forty-seven CD47-targeting drugs remain in clinical development globally, per Fierce Biotech — a field that has been quietly regrouping after those high-profile setbacks.
The Adelaide research offers a possible explanation for why the trials struggled: anti-CD47 drugs were blocking CD47's immune-evasion function but appear to have left its direct tumor-promotion pathway intact. The dual mechanism means that achieving therapeutic benefit in glioblastoma may require more complete or differently targeted inhibition than previous approaches attempted.
"We've known for some time that CD47 acts as a kind of 'don't eat me' signal that helps cancer cells hide from the immune system," Dr. Nirmal Robinson, senior research fellow at the University of Adelaide and a senior author of the study, said in commentary cited by Drug Target Review. "What we've discovered is that CD47 is doing much more than that — it's actually driving the cancer's ability to spread and grow." Dr. Ruhi Polara, a research associate at the University of Adelaide who co-led the work, provided additional context to the same outlet: blocking CD47 in the tumor cells themselves, not just in the immune context, may be necessary for meaningful clinical effect in glioblastoma.
The research was conducted in collaboration with Professor Stuart Pitson and the Centre for Cancer Biology team at the University of Adelaide. The full author list spans 21 researchers across multiple institutions. The work is peer-reviewed and published in PNAS — a substantive publication venue — but the results are still preclinical. The jump from lab and animal models to human glioblastoma therapy is substantial and by no means guaranteed.
What happens next will depend on whether the findings replicate and whether drug developers see a practical path to targeting the CD47-ITCH-ROBO2 axis specifically. The 43 drugs currently in CD47 clinical trials were designed around the immune-evasion hypothesis. If the ROBO2 pathway proves central to glioblastoma progression, some of those programs may need rethinking — or may benefit from the biological clarity the Adelaide team has provided. Either way, a target that looked partially compromised by clinical failure is now showing a second front worth defending.
Sources: PNAS paper | Drug Target Review | News-Medical | Fierce Biotech | Newswise | Scientific Frontline
