Rapamycin, the macrolide immunosuppressant and active pharmaceutic in drug-eluting stents (DES), has a well-recognized antiproliferative action that involves inhibition of the mTOR pathway after binding to the cytosolic protein FKBP12. TGF receptor-type I (TGFRI) spontaneous activation is inhibited by the association with FKBP12. We hypothesized that rapamycin, in addition to inhibition of mTOR signaling, activates TGFRI independent of TGFβ. Human umbilical vein endothelial cells (HUVECs) were treated with rapamycin (10 nmol/l) and/or TGFβ RI kinase inhibitor (TGFRIi, 100 nmol/l) for 24 h. Rapamycin induced SMAD phosphorylation (SMAD1, SMAD2, and SMAD5) and PAI-1 up-regulation, which was specifically abrogated by SMAD2 knockdown. TGFRIi efficiently blocked phosphorylation of SMAD2, but not SMAD1/5. Interestingly, the inhibitor did not alter cell proliferation arrest induced by rapamycin. Active TGFβ secretion was not affected by the treatment. Neutralizing TGFβ experiments did not influence SMAD2 phosphorylation or PAI-1 expression indicating that activation of this pathway is independent of the ligand. In addition, rapamycin induction of endothelial-to-mesenchymal transition (EndMT) was potentiated by IL-1β and efficiently blocked by TGFRIi. In vivo, the prothrombogenic effects of rapamycin and up-regulation of PAI-1 in murine carotid arteries were reduced by TGFRIi treatment. In conclusion, we provide evidence that rapamycin activates TGF receptor independent of its ligand TGFβ, in concert with promotion of PAI-1 expression and changes in endothelial phenotype. These undesirable effects, the prothrombogenic state, and activation of EndMT are SMAD2-dependent and independent of the therapeutic rapamycin-induced cell proliferation arrest.
Rapamycin, also known as Sirolimus, is a macrolide immunosuppressant that served as the active element in first generation of drug-eluting stents (DES). The success of this device in suppressing neointimal proliferation and preventing clinical restenosis prompted subsequent generations of stents to incorporate rapamycin analogs as their active agent and today DES include zotarolimus, everolimus, and ridaforolimus as well as sirolimus. The control of proliferative vascular events is classically ascribed to binding of drug to the FK506 binding protein 12 (FKBP12) [1,2]. Rapamycin–FKBP12 complex inhibits mammalian target of rapamycin (mTORC1) signaling promoting cell cycle arrest [1,2]. The DES effect using these pharmacological agents is substantial but also comes at the potential expense of thrombotic risk [3–5]. Though vascular smooth muscle cells are affected first and foremost, there are also effects on endothelial cells that could reduce the native antithrombotic protective elements of intact blood vessels.
Rapamycin, for example, induces endothelial tissue factor (TF) [6,7] and plasminogen activator inhibitor-1 (PAI-1) [8–10] expression. These effects are cited as evidence for the prothrombogenic potential of the drug—this though is debated  and, even if present, is not mechanistically defined. This is the issue we investigated.
PAI-1 is a member of serine protease inhibitor (serpin) family and a major inhibitor of fibrinolysis, the physiological breakdown of blood clots [12,13]. PAI-1 expression is induced by TGFβ receptor activation, specifically from receptor TGFRI/ALK5. FKBP12 interacts with TGF receptor-type I (TGFRI) and its spontaneous activation is kept in check by local complexation of FKBP12 [14–17]. Thus, we hypothesized that rapamycin, in addition to its inhibition of mTORC1, activates TGFRI in the absence of TGFβ, interfering with endothelial cell health. If true these data could explain the mode, nature and variability of rapamycin-induced effects on ECs and the vessel wall, and a possible means by which to deal with the thrombotic complications of DES.
Human umbilical vein endothelial cells (HUVECs) were obtained from Invitrogen and maintained in EGM-2 media (PromoCell, Heidelber, Germany) supplemented with 2% fetal bovine serum (FBS), 10 U/ml penicillin and 10 μg/ml streptomycin. Cells were used up to 8th passage. Before cell treatment, medium was changed to EGM-2 supplemented with 1% FBS for 24 h. HUVECs were treated with 10 nmol/l rapamycin (LC Laboratories, Woburn, MA, U.S.A.) and/or 100 μmol/l TGFβ RI kinase inhibitor (TGFRIi, ALK5 inhibitor, Calbiochem, Merck Millipore, Germany) for 24 h. Control cells were treated with 0.1% DMSO vehicle. For long treatment experiments, cells were treated with 10 nmol/l rapamycin and/or 10 ng/ml IL-1β (PeproTech Inc., Rocky Hill, NJ, U.S.A.) with or without TGFβ RI kinase inhibitor (100 μmol/l) for 4 days. Medium was refreshed every day.
SMAD1, SMAD2, and SMAD5 silencing were performed using small interfering RNA (siRNA) specific for each protein (sc-29483, sc-38374, and sc-38378 respectively). HUVECs were plated at 70% confluence and starved with EGM medium without serum for 16 h. Lipofectamine 2000 (Invitrogen) was used to transfected cells with 40 pmol of siRNA in Opti-MEM I medium (Invitrogen) for 6 h. After this period, cells were washed and EGM-2 fresh medium was replaced. Treatment with rapamycin and/ or TGFR inhibitor was performed 24 h after transfection.
TGFβ neutralizing antibody treatment
HUVECs were 1 h previously incubated with 7.5 μg/ml of TGFβ neutralizing antibody (R&D) before rapamycin (10 ng/ml) treatment. TGFβ treatment (5 ng/ml) was used as positive control. Cells were lysed after 30 min or 24 h for Western blot evaluation of SMAD phosphorylation or PAI-1 respectively.
HUVEC cultured medium was collected after 30 min or 24 h and TGFβ content was measured using TGF-beta 1 Emax ImmunoAssay System according to manufacturer’s instructions (Promega, Madison, WI, U.S.A.). All the samples were not submitted to acidification step in order to detect just active TGFβ1. Proliferation assay was performed by BrdU incorporation using ELISA colorimetric assay kit according to manufacturer’s instructions (Roche Applied Science, Germany). Expression of critical proteins was determined by immunocytochemistry and Western blot.
Cells were fixed with 4% paraformaldehyde for 30 min. After permeabilization (0.1% Nonidet P40 for 15 min), cells were incubated in blocking buffer (1% BSA) for 1 h. Primary antibodies for the following proteins were incubated overnight at 4°C: VE-cadherin (Cell Signaling Technology, Inc., Danvers, MA, U.S.A.), Phalloidin (Life Technologies, Molecular Probes, Inc., Eugene, OR), Calponin (EMD Millipore, Darmstadt, Germany), PECAM-1 (Abcam, Cambridge, MA, U.S.A.), and SM22α (Abcam). Secondary antibodies Alexa488 or Alexa555 (Life Technologies, Molecular Probes, Inc.) were incubated for 2 h and nucleus was stained with DAPI 10 µg/µl. Immunocytochemistry were analyzed on Laser Scanning Microscope 510 (Carl Zeiss).
Cells were lysed in SDS buffer (60 mM Tris/HCl pH 6.8, 5% glycerol, 2% SDS). Twenty-five to fifty micrograms of cell lysates were run on SDS/polyacrylamide gels, transferred to PVDF membranes (Millipore, Billerica, MA, U.S.A.), and incubated 1 h in a blocking buffer (5% bovine serum albumin (BSA), 10 mM Tris/HCl, pH 7.6, 150 mM NaCl, and 0.1%Tween 20). Each membrane was incubated overnight against a specific antibody: SMAD2 (Cell Signaling), pSMAD2 (Cell Signaling), SMAD1 (Cell Signaling), SMAD5 (Cell Signaling), pSMAD1/5 (Cell Signaling), PAI-1 (Santa Cruz Inc., Dallas, TX, U.S.A.), VE-cadherin (Cell Signaling), Calponin (Millipore), PECAM1 (Abcam), and SM22α (Abcam). After incubation with peroxidase conjugated secondary antibodies, detection was performed with enhanced chemiluminescence reagents (GE Healthcare, Sweden). Protein levels of GAPDH (R&D Systems, Minneapolis, MN, U.S.A.) were used to normalize the results.
Experimental animals and animal procedures
C57BL/6 mice (20–30 g) were treated with rapamycin (4 mg/Kg body weight/day) and/or TGFβ RI kinase inhibitor (25 mg/Kg body weight/day) by manually restraining the animals and injecting the drug solutions intraperitoneally using a 25- or 27-gauge syringe. After 3 days, the carotid artery was collected for immunohistochemistry or subjected to photochemical thrombosis experiment.
All animal procedures followed institutional guidelines for the care and use of laboratory animals. This study protocol was approved by the committee of animal care of Massachusetts Institute of Technology (# 0409-032-12).
The carotid segments were perfusion fixed in situ via cardiac puncture at 80 mmHg with an initial infusion of saline with 14.8 mM KCl, followed by 4% phosphate-buffered formalin. After 24–48 h, the artery segments were dehydrated in graded ethanol baths, immersed in citrisolv, and embedded in paraplast (Oxford, St Louis, MO, U.S.A.). Transverse sections (3 μm-thick) of the tissue were affixed onto slides positively charged. The slides were submitted to deparaffinization in xylenes and then rehydrated. Heat-induced antigen retrieval was done with Tris-EDTA buffer (pH 9.0) using pressure cooker. Slides were incubated in blocking buffer (Tris Buffer and 1% casein) for 1 h. PAI-1 (Santa Cruz) antibody was used at 1:50 dilution overnight at 4°C. After wash steps with Tris-buffer, slides were incubated with ALEXA 555 (Molecular Probes) for 2 h and nucleus stained with diamino-2-phenylindole (DAPI) at 10 µg/µl. Immunohistochemistry was analyzed on Laser Scanning Microscope 510 (Carl Zeiss). Images were quantified by IMAGEJ software.
Photochemical thrombosis experiment
Thrombosis in murine arteries followed by a protocol which we have previously described and validated . Mice were anesthetized with a ketamine–xylazine mixture and a Rose Bengal dye solution (10 mg/ml) was injected retro-orbitally into mice at a dosage of 50 mg/kg. The carotid artery of the mouse was then exposed by careful dissection and a flow probe (Transonic, Inc., Ithaca, New York) was placed around the artery, and a 2.0-mW laser with a wavelength of 543 nm (ThorLabs Inc., Newton, New Jersey) was applied to the artery. Blood flow was monitored until an occlusive thrombus was formed that stopped flow for at least 2 min.
All data are expressed as mean ± SEM. Comparison among groups were performed using two-way analysis of variance (ANOVA) followed by Tukey’s post-hoc test for comparison. Student t test was used for siRNA experiments, TGFβ release, and photochemical thrombosis experiment. Values of P<0.05 were considered statistically significant.
Rapamycin induces TGFRI activation and regulates PAI-1 expression via SMAD2
We evaluated SMAD signaling to determine TGFRI activation by rapamycin. In endothelial cells, TGFβ activates SMAD2/3 signaling through ALK5/TGFRI and SMAD1/5/8 pathway through ALK1/TGFRI. Rapamycin was able to activate both types of receptors as indicated by phosphorylation of SMAD2 (Figure 1A) and SMAD1/5 (Figure 1B). The well demonstrated TGF-induced PAI-1 expression was also observed in rapamycin-treated cells and abrogated by SMAD2 knockdown. siRNA to SMAD1 and SMAD5 did not interfere with PAI-1 expression indicating that only SMAD2 signaling is activated in rapamycin-induced PAI-1 expression (Figure 1C). The efficiency of SMADs silencing is represented in Supplementary Figure S1A.
SMAD activation and PAI-1 expression by rapamycin in endothelial cells
A commercially available TGFRI inhibitor, blocking specifically SMAD2 phosphorylation (Figure 1A and B) and PAI-1 expression (Figure 1D), was used on the following experiments to understand the TGFRI activation by rapamycin better. SMAD2 activation by rapamycin is observed for at least 6 h, and TGFRI inhibitor efficiently blocked its phosphorylation during this time (Supplementary Figure S1B). Indeed, induction of PAI-1 expression by rapamycin and its inhibition by TGFRI inhibitor can be observed even after 96 h treatment (Supplementary Figure S1C).
Classically, rapamycin is known to inhibit mTORC1 activity and the downstream signaling such as p70S6K [1,2]. Interestingly, cell proliferation arrest induced by rapamycin was not altered by the TGFRI inhibitor (Figure 2A). Similarly, the reduction in p70S6K phosphorylation was maintained in the presence of TGFRI inhibitor (Figure 2B) and SMADs silencing (Figure 2C), suggesting that rapamycin activates two independent pathways to control cell proliferation and cell phenotype.
Cell proliferation arrest induced by rapamycin is not altered by TGFRI inhibitor or SMAD knockdown
Rapamycin induces TGFRI pathway independent of TGFβ
To verify if TGFR activation occurs by increased release of active TGFβ induced by rapamycin, we performed an ELISA assay to measure TGFβ1 in the cell culture media. There was no difference in active TGFβ release in rapamycin-treated cells compared with control after 30 min (Figure 3A) and 24 h (Figure 3B). Additionally, the increase in SMAD2 phosphorylation and PAI-1 expression induced by rapamycin was not affected by neutralizing TGFβ antibody (Figure 3C and D). The neutralizing effect of TGFβ antibody was demonstrated by reduction in SMAD2 phosphorylation and PAI-1 expression in TGFβ treated cells (Figure 3C). These data indicate rapamycin activates TGF receptor independently of TGFβ.
Rapamycin induces TGFRI pathway independent of TGFβ
Rapamycin induces endothelial-to-mesenchymal transition
TGFβ has been implicated in endothelial-to-mesenchymal transition (EndMT) contributing to fibrosis [19–21] and remodeling after injury , therefore we examined EndMT in rapamycin activation of the TGF pathway. Because inflammatory costimulation has been described to efficiently induce EndMT , we also performed the study in the presence of IL-1β. EndMT is characterized by loss of endothelial markers and increased expression of mesenchymal markers [19,24]. Here, we evaluated VE-Cadherin and PECAM-1 as endothelial markers and SM22 and calponin as mesenchymal markers. Although rapamycin has shown a small decrease in VE-Cadherin and PECAM1, it increased the mesenchymal marker SM22 and changed actin fiber pattern from cortical to fiber stress characteristic of mesenchymal and smooth muscle cells (Figure 4). In the presence of IL-1β, the induction of EndMT by rapamycin was potentiated showing more pronounced decrease in VE-Cadherin and PECAM-1 than in rapamycin-only treated cells. Additionally, rapamycin and IL-1β treated cells showed induction of calponin expression (Figure 4B and F), which was not observed by the treatment of IL-1β alone (Supplementary Figure S1C). The TGFRIi reverted mesenchymal marker expression in rapamycin and rapamycin combined with IL-1β treated cells but did not prevent the fiber stress pattern of actin organization.
Rapamycin induces endothelial-to-mesenchymal transition, which is potentiated in the presence of IL-1β.
Blockage of TGFRI diminishes rapamycin-induced PAI-1 expression and prevents thrombogenicity in vivo
Mice treated with rapamycin exhibited increased carotid arterial PAI-1 expression, which was attenuated by TGFRI inhibitor (Figure 5A and B). Similarly, vessels expressing more PAI-1 were more thrombogenic than vessels with less PAI-1 (Figure 5C and D), demonstrating that TGFRI inhibition prevents the thrombogenicity induced by rapamycin through PAI-1 down-regulation.
Blockage of TGFRI activation decreases rapamycin-induced PAI-1 expression and thrombogenicity in vivo
Rapamycin effects are manifest as a chemotherapeutic agent and when presented in a local sustained and controlled manner as an inhibitor of local proliferative vascular disease. Use in the latter as the active element of DES is accompanied as well by increased thrombogenicity possibly in tune with inhibition of restoration of an intact endothelium. [8,25] Rapamycin increases intracellular calcium which activates protein kinase C-α, promoting disruption of p120 catenin–VE-cadherin and VE-cadherin degradation .
Loss of endothelial barrier function is real [26,27] but the mechanisms proposed are indirect and over long time-scales. As DES, for example, release a heavy drug load early after implantation, we wondered if the effects of rapamycin were even more direct. Rapamycin classically inhibits mTOR complexation through binding to a specific binding protein, FKBP12 [4,5]. mTOR is present in all cells and acts in two different protein complexes: mTORC1 and mTORC2. mTORC1 is involved in protein synthesis and cell cycle arrest, while mTORC2 participates in cell survival and actin cytoskeleton organization . While inhibition of mTORC1 occurs immediately after binding of rapamycin to FKBP12, blockage of mTORC2 needs prolonged exposure of cells to rapamycin and its effect is controversial [28,30,31]. FKBP12 is a ubiquitous protein and its interaction with TGFRI inhibits basal signaling of the growth factor receptor [14–17]. Fibroblasts from FKBP12 knockout mice showed enhanced activity of TGFRI and inhibition of cell cycle . As FKBP12 is a common component between TGF and rapamycin signaling, we asked if rapamycin could activate TGFRI by FKBP12 displacement from the receptor and by this means alter endothelial phenotype even without the TGFβ ligand.
The present work demonstrated that rapamycin phosphorylates SMAD1/5 and SMAD2 in a time-dependent manner in concert with the activation of TGFRI. In addition, rapamycin increases PAI-1 expression and induces EndMT similar to effects described for TGFβ. Although SMAD 1, 2, and 5 are activated by rapamycin, only SMAD2 seems to contribute to changes of endothelial phenotype. The implications of activation of SMAD1/5 by rapamycin should be further investigated but the balance of specific SMAD activation of TGF receptor may well explain disparate effects of DES on vascular healing.
Others have shown effects of rapamycin on PAI-1 [8–10]; however, the mechanism underlining this process remains unclear. To our knowledge, the present work is the first to demonstrate that rapamycin induces PAI-1 expression by TGFRI activation in endothelial cells independently of TGFβ. Some have reported that rapamycin activates the TGFβ pathway by increasing expression and/or activation of TGFβ [32,33]. In rat mesangial cells, rapamycin activates latent TGFβ and increases pSMAD2 and PAI-1 . The increased oxidative stress after 30 min of rapamycin treatment resulted in secretion of active TGFβ . The difference of action can be explained by differences in regulation between cell types: depletion of FKBP12 in mesangial cell reduces SMAD2 phosphorylation instead of increases as demonstrated in endothelial cell type.
We are suggesting that rapamycin can activate TGFRI by FKBP12 displacement from the receptor. Indirect action on TGF pathway can also occur, as regulation of accessories receptor like β glycan, endoglin, SARA (SMAD anchor for receptor activation), and SMAD7 [34–36]. Our in vivo experiments corroborated activation of TGF pathway by rapamycin, where the increase in PAI-1 expression after drug treatment correlates with increased thrombogenicity and TGFR inhibitor reduces this effect (Figure 5).
We also show that rapamycin can induce EndMT and affect locally the health of endothelial cells adding further to understand the endothelial toxicity of rapamycin. EndMT, a process where endothelial cells lose their specific markers and morphology and acquire mesenchymal cell-like characteristics, has been implicated in heart valve mesenchyme formation , cardiac and renal chronic fibrosis [19–21], and vascular remodeling . The most described regulator of EndMT is TGFβ and we demonstrate that rapamycin induces expression of mesenchymal markers in endothelial cells through activation of TGFRI.
EndMT induced by rapamycin showed some aspects distinct from the process induced by TGFβ. While both TGFβ and rapamycin induced the expression of mesenchymal markers, rapamycin treatment had a slight reduction in endothelial markers. Interestingly, TGFRI inhibitor reverted mesenchymal marker expression but did not prevent the fiber stress pattern of actin. This may be to an additional effect of rapamycin on mTORC2. In fact, it is demonstrated that rapamycin promotes actin stress fibers in HUVEC by activation of mTORC2  and this complex is involved in actin cytoskeleton modulation through regulation of F-actin and actin-related proteins . Though rapamycin does not recapitulate exactly the EndMT process, it is clear that rapamycin drives the endothelial cells to an activated state. Data from literature suggest that different dose of rapamycin results in distinct effect in cells [40–43]. High dose (∼100 nM) reduces EndMT induction [40,41], and low doses (10–20 nM) promote the mesenchymal transition . This can be in part explained by the existence of two distinct mTOR complexes with different sensitivity to rapamycin treatment.
Normal endothelial function is essential for retention of vascular health, inhibiting thrombus formation, inflammation, and smooth muscle cell proliferation. The incomplete re-endothelialization of stent strut surfaces is a recognized substrate for stent thrombosis, and DES are thought to delay endothelial recovery allowing for persistent fibrin build up around the stent sturts [44,45]. We provide evidence that endothelial cells in the presence of rapamycin are dysfunctional and thrombogenic, suggesting that the thrombogenicity of DES is not only from the loss of complete endothelial monolayer but also from a prothrombotic endothelial cell phenotype. We suggest that rapamycin may act by two distinct pathways—one that inhibits mTORC1 and leads to cell cycle arrest, and the other that activates SMAD2 and increases PAI-1 expression and EndMT (Supplementary Figure S2).
Overall, the present study shows that rapamycin activates the TGFβ pathway and modify endothelial phenotype to a more thrombotic state independent of TGFβ. Understanding the impact of DES on endothelial function is essential to find new approaches that prevent restenosis and maintain the vascular health.
Rapamycin has a well-recognized antiproliferative action that involves inhibition of the mTORC1 pathway after binding to the cytosolic protein FKBP12. TGFRI spontaneous activation is inhibited by the association with FKBP12. We hypothesized that rapamycin, when interacting with FKBP12, allows TGFRI activation independently of TGFβ.
Our findings have critical basic and applied implications. We demonstrated that rapamycin activates TGF receptor independent of its ligand TGFβ, promotes PAI-1 expression, and induces changes in endothelial phenotype. Such profound effects on signaling of ligand-independent binding are in and of itself fascinating. We further showed that these effects were associated with endothelial activation and a thrombogenic state in culture and in vivo, are SMAD 2-dependent and independent of the therapeutic rapamycin-induced cell proliferation arrest.
These findings may then explain the variability in disparate findings in the past, the mechanism of stent thrombosis and a means of obviating these complications in practice.
We thank Melissa St. Pierre for performing the surgical procedures of the photochemical thrombosis experiments.
The authors declare that there are no competing interests associated with the manuscript.
A.A.M. was supported by the J.P. Lemann Foundation as a Jorge Paulo Lemann Harvard Medical School Cardiovascular Fellow at Brigham & Women’s Hospital. T.G.-S. is recipient of fellowship from Sao Paulo Research Foundation (FAPESP 2014/06844-9). J.E.K. is supported in part by grants from Sao Paulo Research Foundation (FAPESP 2013/17368-0) and E.R.E. is supported in part by a grant from the US National Institutes of Health (R01 GM 49039).
Conception and design: A.A.M. and E.R.E.; Analysis and interpretation: A.A.M., T.G.-S., J.E.K., and E.R.E.; Data collection: A.A.M. and T.G.-S.; Writing the article: A.A.M., T.G.-S., and E.R.E.; Critical revision of the article: A.A.M., T.G.-S., J.E.K., and ERE; Statistical analysis: A.A.M. and T.G.-S.; Obtained funding: A.A.M., J.E.K. and E.R.E.; Overall responsibility: A.A.M. and ERE.
FK506 binding protein 12
human umbilical vein endothelial cells
mammalian target of rapamycin
plasminogen activator inhibitor-1
platelet/endothelial cell adhesion molecule 1
TGF receptor-type I
TGFβ RI kinase inhibitor
These authors contributed equally to this work.