GPNMB–SMAD4 pathway links microglial inflammation to demyelination in diabetic cognitive dysfunction

Hyperglycemia appears to strip microglia in the hippocampus of a protective protein called GPNMB, and a new study reports that restoring the protein rescues memory deficits in diabetic mice by stabilizing SMAD4 and suppressing NF-κB signaling. The work, published in Cell & Bioscience, proposes the GPNMB/SMAD4/NF-κB cascade as a therapeutic target for diabetic cognitive dysfunction, a common hyperglycemia-induced complication of the central nervous system driven by hippocampal neuroinflammation.

The research team, led by BingXue Song, Jiahui Li and senior author Wei Shan, established two complementary models: streptozotocin-induced diabetic C57BL/6J mice to represent the disease in vivo, and primary hippocampal microglia stimulated with high glucose in vitro. To raise GPNMB levels in the brain, they used stereotactic intracerebroventricular injection of an adeno-associated virus carrying the GPNMB gene, and they manipulated SMAD4 activity with the agonist KGN and the inhibitor A83-01 in rescue experiments. Cognitive performance was assessed with the Morris water maze, while immunofluorescence and immunohistochemistry tracked microglial morphology, GPNMB colocalization and myelin integrity.

The central finding is a downregulation of GPNMB. Both messenger RNA and protein levels of GPNMB fell in the hippocampi of diabetic mice and in microglia exposed to high glucose. When the researchers overexpressed GPNMB via the viral vector, several measures improved: cognitive impairment in the water maze was rescued, excessive microglial M1 polarization was inhibited, inflammatory cytokine profiles normalized, and levels of the myelin proteins MBP and CNPase were restored both in vivo and in vitro.

A key mechanistic question was how GPNMB exerts these effects. The team identified SMAD4 as a direct binding partner. Bioinformatic prediction, molecular docking, co-immunoprecipitation and immunofluorescence all supported a physical GPNMB–SMAD4 interaction. SMAD4 is known to restrain NF-κB-mediated microglial inflammation, but the physical and functional relationship between the two proteins in diabetic brain injury had not been explored before.

The signaling logic that emerges from the results is a de-repression mechanism. Hyperglycemia depleted SMAD4 and triggered NF-κB activation, pushing microglia toward an inflammatory M1 state. When SMAD4 was stimulated, NF-κB-driven neuroinflammation was suppressed. Critically, when the researchers blocked SMAD4 with A83-01, the neuroprotective, anti-inflammatory and anti-demyelinating benefits of GPNMB overexpression were fully eliminated, placing SMAD4 downstream of GPNMB in the pathway.

According to the authors’ conclusion, hyperglycemia suppresses microglial GPNMB, disrupts GPNMB–SMAD4 binding, and thereby de-represses NF-κB signaling. This cascade drives M1 microglial polarization, neuroinflammation, hippocampal demyelination and cognitive deficits. In their framing, GPNMB overexpression mitigates diabetic cognitive dysfunction through SMAD4 stabilization and NF-κB suppression, making the three-step pathway a viable therapeutic target.

The study’s scope and limitations deserve emphasis. All findings come from a streptozotocin-induced mouse model and from high-glucose-stimulated primary hippocampal microglia, not from human tissue or clinical trials. The cognitive endpoint was measured with a single behavioral assay, the Morris water maze, and molecular readouts relied on western blot, quantitative PCR and immunostaining. The article is being shared early as a peer-reviewed, accepted manuscript that remains subject to further edits before the final version of record, and the authors note that statistical comparisons used Student’s t-test and one-way ANOVA with Tukey’s post-hoc testing.

Within those constraints, the work adds a specific molecular link to a growing picture of diabetes-related brain injury. GPNMB, a glycoprotein previously described as exerting anti-inflammatory neuroprotection in neurodegenerative diseases, had an undefined role in diabetic cognitive dysfunction until now. By showing that its downregulation reduces SMAD4 availability, thereby enhancing NF-κB signaling and inducing neuroinflammation, the study connects a metabolic insult to a defined transcriptional switch in microglia, and from there to observable myelin loss in the hippocampus.

Whether pharmacological manipulation of this axis can translate into therapy remains an open question. The rescue experiments used viral gene delivery and small-molecule SMAD4 modulators in mice, and the authors describe the GPNMB/SMAD4/NF-κB cascade as a target rather than a demonstrated treatment. Still, the finding that a single SMAD4 inhibitor abolished all of GPNMB’s observed benefits gives the pathway a clear testable hierarchy, and it suggests that preserving SMAD4 availability in inflamed microglia may be a productive direction for future research into diabetic cognitive dysfunction.

Subject of Research: The role of the GPNMB–SMAD4–NF-κB pathway in microglial neuroinflammation and demyelination in diabetic cognitive dysfunction

Article Title: GPNMB interacts with SMAD4 to mediate microglial neuroinflammation and induce demyelination in diabetic cognitive dysfunction

Article References: Song, B., Li, J., Zhao, Y., Liu, W., Wang, Y., Yu, S., Yu, H., Huang, T., Zuo, Z., & Shan, W. (2026). GPNMB interacts with SMAD4 to mediate microglial neuroinflammation and induce demyelination in diabetic cognitive dysfunction. Cell & Bioscience. https://doi.org/10.1186/s13578-026-01664-2

Image Credits: AI Generated

DOI: 10.1186/s13578-026-01664-2

Keywords: diabetic cognitive dysfunction, GPNMB, SMAD4, NF-κB, microglia, neuroinflammation, demyelination, hippocampus, hyperglycemia, AAV-GPNMB, Morris water maze, C57BL/6J mice