An AI model trained on single-cell data has helped researchers identify PBX1 as one regulator of how blood stem cells change with age. In mice, increasing Pbx1 in young stem cells reproduced part of an older pattern: weaker red-blood-cell production and a relative shift toward platelets. The work is an important mechanistic study, but it does not establish PBX1 as an anti-aging treatment or show that targeting it will benefit people.
The findings were published in Science Advances.
The stem cells that keep blood production running
Hematopoietic stem cells are rare, long-lived cells found mainly in bone marrow. They can renew themselves and produce every major blood-cell lineage, including oxygen-carrying red blood cells, infection-fighting white blood cells and platelets that help stop bleeding.
That balance shifts with age. Older hematopoietic stem cells often become less effective at rebuilding the blood system. They tend to produce fewer red cells and lymphocytes while leaning more toward myeloid cells and the megakaryocyte lineage, which produces platelets.
These changes may contribute to anemia and impaired immune function. They may also intersect with clonal hematopoiesis, in which one altered blood stem-cell clone expands disproportionately, and with risks including thrombosis and blood cancer. Those conditions are complex, however, and the new study did not show that PBX1 causes them in people.
How Geneformer helped narrow the search
The researchers first used single-cell RNA sequencing to examine gene activity in young and old mouse blood stem cells. Old cells carried two programs at the same time. One resembled a very primitive, undifferentiated stem-cell state. The other showed preparation for megakaryocyte and platelet production.
To identify regulators linking those programs, the team fine-tuned Geneformer, a transformer-based AI model originally trained on roughly 30 million single-cell profiles. It treats genes somewhat like words in a sentence and learns how their relationships change with cell state.
The model compared young and old hematopoietic stem and progenitor cells and simulated the effects of increasing or deleting individual genes. It predicted that raising Pbx1 activity could move young cells toward an older transcriptional state. The researchers then combined that result with conventional gene-expression analysis, chromatin profiling and a laboratory screen of 143 blood-related transcriptional regulators.
That distinction matters. Geneformer generated testable priorities; it did not discover a treatment by itself. This reflects a broader lesson from AI-guided aging research: computational screening can make experiments more efficient, but biological validation remains essential.
What PBX1 is
PBX1 is a transcription factor, a protein that binds DNA and helps control which genes a cell uses. It works with proteins including HOX and MEIS and has established roles in development, cell differentiation and the quiet, self-renewing state of blood stem cells.
The study found higher Pbx1 expression in old mouse stem cells. Network analysis placed it near the center of age-associated gene relationships. A separate laboratory screen also found that Pbx1 delayed a marker of stem-cell differentiation, and reducing Pbx1 in old cells pushed that marker in the opposite direction.
What happened when researchers changed Pbx1
Increasing Pbx1 in young mouse blood stem cells reproduced part, but not all, of the aged state. Up to 73.3 percent of the genes activated by Pbx1 were also increased in old cells. The overall shift was still modest, and several other age-related programs were not explained by Pbx1.
After transplantation into irradiated mice, young stem cells engineered to overexpress Pbx1 produced fewer donor-derived red blood cells while preserving platelet output. That created a relative platelet bias similar to an early feature of transplanted old stem cells.
The mechanism appeared to involve Gata1, a transcription factor that must rise at the right time for red-cell development. Pbx1 overexpression suppressed Gata1 induction and reduced access to part of the Gata1 gene region. Knocking down Pbx1 increased Gata1, supporting a regulatory connection.
The reverse experiment was less straightforward. Knocking down Pbx1 in old stem cells reduced the platelet bias, but it also reduced red-cell production. In old mice recovering from bone-marrow injury caused by 5-fluorouracil, a research compound called T417 produced only modest changes in blood recovery. Two treated mice died during follow-up, and the reported delay in platelet recovery was not statistically significant.
Why the result could matter
A regulator that helps lock older blood stem cells into a primitive, platelet-leaning state could eventually improve understanding of age-related anemia, abnormal blood-cell production and thrombotic risk. PBX1-centered programs might also help researchers study why certain stem-cell states become more competitive during clonal hematopoiesis or are co-opted during leukemia.
Those are research directions, not demonstrated clinical effects. The experiments used mouse cells and mice. The study did not test human hematopoietic stem cells, did not show that changing PBX1 prevents anemia or thrombosis, and did not demonstrate prevention or treatment of clonal hematopoiesis or blood cancer.
There is no PBX1 anti-aging therapy
No approved therapy targets PBX1 to treat blood aging. T417 was developed as an experimental PBX1 transcriptional-signaling inhibitor in cancer research. Its limited use in this mouse study does not establish safety, dosing or effectiveness for people.
PBX1 also performs normal functions in stem-cell maintenance and development. The mixed knockdown results suggest that both timing and dose may matter. Blocking the pathway too broadly could disrupt blood production instead of improving it.
What researchers need to show next
- Confirm the PBX1 and GATA1 relationship in human blood stem and progenitor cells from donors across different ages.
- Determine whether PBX1 predicts clinically relevant blood abnormalities rather than only molecular signatures.
- Separate short-term changes in lineage output from long-term effects on stem-cell renewal and cancer risk.
- Develop more selective ways to adjust the pathway and test toxicity, dose and timing.
- Show reproducible benefits in multiple animal models before considering human trials.
Bottom line: Geneformer helped identify PBX1 as one hub in a larger network associated with mouse blood stem-cell aging. Increasing Pbx1 reproduced reduced red-cell production and relative platelet bias, but the study does not establish a human target or treatment. The next test is whether the same mechanism operates in human cells and whether it can be modified safely.