Hidden Gatekeeper Protein Decides Which Vessel Signals Endothelial Cells Hear

Every blood vessel in the human body is constantly listening. Endothelial cells, the thin cellular lining of our arteries, veins and capillaries, must interpret a crowded chorus of molecular signals to decide whether to sprout new branches, mature into stable tubes, or simply hold still and keep the vessel wall quiet. A new study published in the open-access journal iScience reveals that a small, long-known protein called FKBP12 acts as a hidden gatekeeper in this process, determining which of those signals each individual endothelial cell actually hears. The work, led by Zixin Huang, Sonja Lenhardt, Petra Snyder, Ahmed Bulldan, Felix Hausch and Alexander Loewer, combines engineered fluorescent reporters, live-cell microscopy and targeted protein degradation to expose a layer of control in the SMAD signaling pathway that had previously been invisible to bulk measurements.

The SMAD pathway is one of the central information conduits in animal cells. It is switched on by more than thirty ligands of the TGF-β superfamily, which bind to serine/threonine kinase receptors at the cell surface. When a ligand latches on, type II receptors phosphorylate and activate type I receptors, which in turn phosphorylate receptor-associated SMAD proteins. These activated R-SMADs partner with the common mediator SMAD4, travel into the nucleus and steer the transcription of target genes. The pathway splits into two major branches: the TGF-β branch, carried by ligands such as TGF-β, Activins and GDFs through the type I receptors ALK4 and ALK5 into SMAD2 and SMAD3, and the BMP branch, carried by BMP ligands through ALK1, ALK2 and ALK3 into SMAD1, SMAD5 and SMAD9. In endothelial cells this division of labor is beautifully choreographed during angiogenesis. Tip cells at the leading edge of a sprouting vessel rely on SMAD1/5/9 activation by BMP2, BMP6 and BMP7 to navigate, while stalk cells behind them depend on SMAD2/3 signaling through TGF-β to proliferate and form a proper lumen. In mature vessels, BMP9 and BMP10 keep the endothelium quiescent through SMAD1/5/9.

When this balance breaks, human disease follows. Mutations in ALK1, the BMP9 receptor, or its co-receptor endoglin weaken SMAD1/5/9 signaling and cause hereditary hemorrhagic telangiectasia, a disorder marked by fragile arteriovenous malformations and recurrent bleeding. Conversely, mutations in BMPR2 that heighten SMAD2/3 activity contribute to pulmonary arterial hypertension, in which vessels narrow and remodel abnormally. Understanding how individual cells weigh pro- and antiangiogenic inputs is therefore not just a question of basic biology but a prerequisite for new therapies. Yet most of what we know about SMAD signaling comes from population averages, which can mask the striking variability that individual cells display.

To see signaling cell by cell, the team built stable reporter lines in EA.hy926 cells, a widely used model of human vascular endothelium derived from fusing primary umbilical vein cells with a lung carcinoma line. They fused the cDNA of SMAD1 to the yellow fluorescent protein mVenus under a constitutive ubiquitin C promoter, and co-expressed a histone H2B-CFP nuclear marker so that automated image analysis could segment and track every nucleus over time. Parallel reporters were generated for SMAD5 and SMAD9. Western blotting confirmed that the tagged proteins were phosphorylated with the same kinetics as their endogenous counterparts after BMP9 stimulation, peaking around 1.5 hours, and RT-qPCR showed that key target genes such as Endothelin 1 and ID2 were induced normally, indicating that the reporters did not distort the pathway they were meant to observe.

Time-lapse microscopy at ten-minute intervals over sixteen hours then delivered the first surprise: enormous cell-to-cell heterogeneity. After BMP9 addition, essentially all cells pushed SMAD1 into the nucleus within about an hour, but what happened next differed dramatically. Some cells settled into stable signaling, while others repeatedly shuttled SMAD1 back and forth in pulsatile waves. The population median showed a sharp peak at roughly 1.5 hours followed by a gradual return to baseline, and even a medium-only control, containing no added ligand, elicited a small translocation peak, revealing exquisite sensitivity to trace amounts of BMP-family ligands left in the serum. Dose-response experiments showed that peak SMAD1 translocation saturated at about 0.375 nanograms per milliliter of BMP9, with a half-maximal effective concentration of just 0.06 nanograms per milliliter. Higher doses produced sustained nuclear occupancy and delayed adaptation, consistent with the idea that signaling duration is governed largely by ligand depletion as receptors internalize and degrade their cargo.

The three R-SMADs of the BMP branch, often assumed to be interchangeable, turned out to differ quantitatively in revealing ways. SMAD5 required the highest ligand concentrations to respond, showing the largest EC50 of the three, and failed to react even to the residual ligand in the medium-only control, suggesting it cannot compensate at low signal doses. SMAD9, by contrast, returned to the cytoplasm more slowly than SMAD1 or SMAD5 once the ALK1 inhibitor K02288 was applied, hinting that its prolonged nuclear retention may serve processes requiring sustained gene expression, such as pulmonary vascular remodeling. Pharmacological validation confirmed that BMP9 signaling runs through ALK1, with an IC50 of 192 nanomolar for SMAD1, and lower values of 47 and 107 nanomolar for SMAD5 and SMAD9 respectively.

The pivotal discovery came when the team screened a panel of TGF-β superfamily ligands. Despite the presence of receptors for BMP2, BMP6 and BMP7, only BMP9 and BMP10 drove robust nuclear translocation of SMAD1, SMAD5 and SMAD9 in these endothelial cells, even at very high ligand concentrations. Suspecting an intracellular brake rather than an extracellular one, the researchers focused on FKBP12, the FK506-binding protein known to dock onto the glycine-serine-rich region of type I receptors and physically block their phosphorylation by type II receptors. FKBP12 was highly expressed in the cells. To remove it cleanly, the team used a proteolysis-targeting chimera, a bifunctional molecule that recruits the von Hippel-Lindau ubiquitin ligase to FKBP12 and triggers its degradation. Western blots confirmed dose- and time-dependent FKBP12 loss, and the consequences were striking: with FKBP12 gone, BMP7 and Activin A, ligands that had previously been mute, suddenly provoked strong SMAD1 activation through ALK2, while the response to BMP9 and BMP10 remained almost unchanged.

FKBP12 did more than simply unlock a second receptor. Using a dual reporter line expressing YFP-SMAD1 and mCherry-SMAD2 simultaneously, the researchers showed that BMP7 normally produces a weak SMAD1 response and a robust SMAD2 response. Degrading FKBP12 shifted the balance: SMAD1 translocation rose, particularly at high BMP7 concentrations, while the SMAD2 response fell. Co-stimulation experiments with BMP9 and TGF-β, which engage distinct receptor complexes, produced no mutual attenuation, ruling out feedback loops or pathway competition as the explanation. Instead, the data point to limited shared receptor availability and differential receptor complex composition. The authors propose that BMP7-induced SMAD2 activation depends on heterodimeric ALK2/ALK4 complexes, whereas SMAD1 activation after FKBP12 removal relies on ALK2 homodimers, and that Activin A can additionally engage ALK4 homodimers. In this model, FKBP12 does not merely silence receptors; it actively determines which receptor assemblies form and therefore which branch of the pathway a given ligand will engage.

These signaling shifts translated into real cellular behavior. The SMAD1/5/9 target gene DLL4 dropped sharply after BMP9 treatment and stayed flat when only SMAD2/3 was engaged, but FKBP12 degradation made BMP7 a potent DLL4 repressor as well. The gene ID2, regulated by both branches, showed additive induction when both pathways fired. Over 48 hours of single-cell tracking, BMP9 significantly boosted endothelial motility, an effect abolished by ALK1 inhibition, while BMP7 increased motility more modestly, and combining BMP7 with FKBP12 degradation produced an additive enhancement that ALK4/5 inhibition could curtail. The findings suggest that SMAD1/5/9 and SMAD2/3 cooperate rather than compete in shaping endothelial phenotypes, with FKBP12 acting as a fine-tuning dial. Therapeutically, the implications are tantalizing. The immunosuppressant FK506, known clinically as tacrolimus, displaces FKBP12 from receptor complexes and is already being piloted to reduce bleeding in hereditary hemorrhagic telangiectasia, while FKBP12-specific PROTACs could offer a more selective route that spares the immune system. The authors caution that EA.hy926 cells are a hybrid model and that the receptor-complex model rests on pharmacological inference, so validation in primary endothelial cells and in vivo disease models remains the essential next step. Even so, the study reframes FKBP12 from a passive receptor cap into an active architect of ligand-specific signaling, one that cells may exploit, and clinicians may one day target, to restore the delicate balance their vessels depend on.

Subject of Research: FKBP12-dependent regulation of ligand-specific SMAD signaling in living endothelial cells

Article Title: FKBP12 determines ligand-specific SMAD signaling in individual living endothelial cells

Article References: Huang, Z., Lenhardt, S., Snyder, P., Bulldan, A., Hausch, F., & Loewer, A. (2026). FKBP12 determines ligand-specific SMAD signaling in individual living endothelial cells. iScience, 29(10), Article 117728. https://doi.org/10.1016/j.isci.2026.117728

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117728

Keywords: SMAD signaling, FKBP12, endothelial cells, BMP9, TGF-β, ALK1, ALK2, PROTAC, live-cell microscopy, angiogenesis, hereditary hemorrhagic telangiectasia, pulmonary arterial hypertension