Direct Targeting of the mTOR (Mammalian Target of Rapamycin) Kinase Improves Endothelial Permeability in Drug-Eluting Stents

Objective—Drug-eluting stents that release canonical mTOR inhibitors are commonly used to treat coronary artery disease. However, these stents accelerate the development of atherosclerosis within the stent, known as neoatherosclerosis, which is a major cause of late stent failure. Prior research indicated that canonical mTOR inhibitors bind to FKBP12.6, causing it to detach from calcium release channels. This leads to the activation of PKCα and the separation of p-120-catenin from VE-CAD, resulting in endothelial barrier dysfunction. The relevance of these findings to drug-eluting stents was previously unknown. Newer direct mTOR kinase inhibitors do not bind FKBP12.6 and could potentially improve endothelial barrier function while maintaining their ability to prevent artery re-narrowing after stenting. However, their actual effects were unknown. This study examined the effects of two different pharmacological targeting strategies: the canonical mTOR inhibitor everolimus and the mTOR kinase inhibitor Torin-2, on endothelial barrier dysfunction after stent placement.

Approach and Results—Using a rabbit model of stenting and techniques such as Evans blue dye staining, confocal microscopy, and scanning electron microscopy, it was found that stents releasing everolimus resulted in long-term endothelial barrier dysfunction compared to bare metal stents. This dysfunction was reduced when stents releasing the mTOR kinase inhibitor Torin-2 were used. Sixty days after stent placement, everolimus-eluting stents showed large areas of Evans blue dye staining and evidence of p120 VE-CAD separation, consistent with endothelial barrier dysfunction. These findings were not present in bare metal stents and were significantly less pronounced in Torin-2–eluting stents. To demonstrate the role of endothelial barrier dysfunction in neoatherosclerosis, animals were fed a high-cholesterol diet for an additional 30 days, starting 100 days after stenting. Everolimus-eluting stents showed significantly more macrophage infiltration, a sign of neoatherosclerosis, compared to both bare metal stents and Torin-2–eluting stents.

Conclusions—The results of this study indicate that the interaction between FKBP12.6 and canonical mTOR inhibitors is a major cause of increased vascular permeability and neoatherosclerosis. This issue can be overcome by using mTOR kinase inhibitors. This research suggests that further refinement of molecular targeting of the mTOR complex may be a promising strategy.

Introduction

Endothelial cells form a protective layer between the blood flow and the vessel wall. Endothelial barrier dysfunction is believed to be the initial step in atherosclerosis because it allows lipoproteins to enter the subendothelial space. Drug-eluting stents, which release canonical mTOR inhibitors like sirolimus or everolimus, are used to treat coronary artery disease because they are more effective at preventing the artery from re-narrowing compared to bare metal stents. The primary way these stents work is by inhibiting mTOR and its downstream targets, including eukaryotic initiation factor 4E–binding protein-1, S6 kinase, and Akt, which reduces the formation of new tissue within the artery wall. However, the use of these stents is linked to the rapid development of neoatherosclerosis, which can lead to blood clot formation within the stent due to plaque rupture and is the second leading cause of late stent failure. Neoatherosclerosis develops more quickly in drug-eluting stents compared to bare metal stents, occurring over months to a few years, whereas native vessel atherosclerosis develops over decades.

While previous data suggest that endothelial permeability plays a significant role in the development of atherosclerosis in native vessels, its role in neoatherosclerosis was previously unknown. Endothelial cells control the passage of leukocytes and lipoproteins into the subendothelial space through endothelial adherens junctions. Vascular endothelial cadherin mediates the integrity of cell-to-cell connections in endothelial cells through calcium-dependent interactions. VE-CAD is stabilized by a group of proteins, including β-catenin and p120, which are anchored to the actin fibers of the endothelial cytoskeleton. When catenin association is lost, the cell adhesion mediated by cadherin becomes unstable.

Recent research identified a mechanism by which sirolimus increases endothelial permeability that is independent of the mTOR signaling pathway. Sirolimus, and similar drugs like everolimus used in drug-eluting stents, inhibit mTOR by binding to FKBP12, a protein that stabilizes RyR2 intracellular calcium release channels. By displacing FKBP12.6, a vascular-specific form of FKBP12, sirolimus leads to a calcium-dependent activation of PKCα, phosphorylation of p120, and the separation of p120 from VE-CAD, which impairs endothelial barrier function. These findings suggest that endothelial barrier function is compromised by drug-eluting stents and that strategies to directly inhibit mTOR without involving FKBP12.6 might improve endothelial barrier function and subsequently reduce neoatherosclerosis, while still preventing artery re-narrowing.

Using a model of arterial stenting, this study examined the role of FKBP12.6 in increasing endothelial permeability and tested a strategy to overcome this. The study compared endothelial barrier function, assessed by Evans blue dye staining, confocal microscopy to observe the co-localization of p120 and VE-CAD, and scanning electron microscopy to examine the surface structure of the endothelium, in bare metal stents, everolimus-eluting stents, and Torin-2–eluting stents. Torin-2 is a newer generation selective inhibitor that directly blocks the ATP-binding site of mTOR kinase and does not bind to FKBP12.6.

To investigate the link between impaired endothelial barrier function and neoatherosclerosis, a proof-of-concept study was conducted where animals were fed a high-cholesterol diet, and the extent of macrophage infiltration in the stented arteries was examined.

Materials and Methods

The authors stated that all supporting data are available within the article and its online-only Data Supplement.

In Vitro Experiments

Western blotting and transendothelial electrical resistance measurements were performed on human aortic endothelial cells, which were maintained in a specific endothelial cell growth medium. Cells between passages 3 and 6 were used for the experiments. For all experiments, the cells were washed twice with a phosphate-buffered saline solution after each treatment for the specified durations. The cells were treated with either everolimus (500 nmol/L) or Torin-2 (100 nmol/L) dissolved in dimethyl sulfoxide. Control cells received only dimethyl sulfoxide. The concentration of everolimus was chosen to approximate average tissue levels observed 60 to 90 days after stenting and was based on preclinical data from a rabbit iliac artery model. The concentration of Torin-2 was based on its IC50 value for mTOR inhibition, which is lower than that of everolimus. The drugs were applied to the cells along with the full endothelial growth medium at 37°C with 5% carbon dioxide. Ryanodine (50 µmol/L), an alkaloid and stabilizer of the ryanodine release channel, was used to pretreat cells for 60 minutes before everolimus treatment. The relative effects of various signaling molecules in the mTOR pathway were evaluated by using small interfering RNA to reduce the levels of FKBP12.6, AKT, and p70S6K in the human aortic endothelial cells. The cells were plated at 50% confluence for transfection the following day. The small interfering RNA transfection and its effectiveness in reducing target protein levels had been previously validated.

Test Devices and Grouping

Three different devices were implanted in the rabbits. The test device was a thin-strut durable polymer metallic everolimus-eluting stent (Promus Element; Boston Scientific, Marlborough, MA), measuring 3.0×16 mm, with a platinum-chromium platform and an 81 μm strut thickness. This stent consisted of an inner layer of poly n-butyl methacrylate polymer surrounded by poly (vinylidene fluoride-co-hexafluoropropylene), which is composed of vinylidene fluoride and hexafluoropropylene monomers as the drug matrix layer containing 100 µg/cm2 of everolimus. This everolimus-eluting stent was compared with a thin-strut bare metal stent of the same backbone (OMEGA; Boston Scientific, Marlborough, MA), measuring 3.0×12 mm, used in 20 instances (platinum-chromium platform, 81 μm strut thickness) as a control. A custom-made Torin-2–eluting stent, measuring 3.0×16 mm (platinum-chromium platform, 81 μm strut thickness using the OMEGA backbone), was manufactured by Boston Scientific using the same techniques employed for the everolimus-eluting stent. This involved a poly n-butyl methacrylate primer layer with poly (vinylidene fluoride-co-hexafluoropropylene) plus 100 µg/cm2 Torin-2 (Selleck Chemicals, Houston, TX) in the external layer. It has been shown that using this polymer in combination with a drug releases the majority of the drug within 3 months.

Animal Model

The study protocol was approved by the Institutional Animal Care and Use Committee of the MedStar Health Research Institute and adhered to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health. A total of 35 healthy male New Zealand white rabbits (5–6 months old; Millbrook Laboratories, Amherst, MA) were included in the study. A model of iliofemoral stent implantation, as previously described, was used to study the endothelialization and inflammatory reaction after stenting. The animals received aspirin (40 mg) orally 12 hours before stent implantation and continued until they were euthanized. Immediately after the endothelial layer of both iliac arteries was removed using a balloon, the devices were deployed in the denuded sections of the left and right iliac arteries. Stents were assigned to the animals in a manner such that each animal received stents of 2 different types. Standard follow-up angiography using contrast media was performed to confirm that the stents were open before they were removed. Evans blue permeability, VE-CAD/p120 co-localization, and endothelialization were assessed in one group of animals, 60 days after implantation (17 animals, 34 stents: 13 bare metal stents, 12 everolimus-eluting stents, and 9 Torin-2–eluting stents). Two bare metal stents and 3 Torin-2–eluting stents were excluded from this analysis due to evidence of damage to the endothelium that occurred during the harvesting process. The inhibition of S6K and Akt, as well as neointimal formation, was assessed using Western blotting in another group of animals (5 animals, 10 stents: 3 bare metal stents, 4 everolimus-eluting stents, and 3 Torin-2–eluting stents) 28 days after implantation. The stents were cut lengthwise, and one half was used for histological measurement of neointima formation and strut coverage by plastic embedding and sectioning. An additional 4 animals (3 bare metal stents, 3 everolimus-eluting stents, and 2 Torin-2–eluting stents) were used to analyze endothelial mTOR inhibition. To understand the different atherosclerotic reactions within the stent, another assessment was performed at 130 days after the animals were fed a cholesterol-enriched diet (0.15% weight/weight; Bio Serv, Flemington, NJ) for 30 days, from day 100 until they were euthanized (6 animals, with 4 stents per group). One Torin-2–eluting stent was excluded from this analysis because of evidence of damage to the endothelium during harvesting.

Evans Blue Staining Analysis

Evans blue dye is commonly used to study the permeability of blood vessels and cellular membranes because it is non-toxic, can be administered as a dye within a living organism, and binds to serum albumin, using it as a carrier molecule. The complex of Evans blue and albumin can be visually identified by its distinct blue color within the tissue. Evans blue was injected intravenously 1 hour before the animals were euthanized to allow for circulation of the dye. After fixation, the stents were bisected lengthwise. Images of the inner surface of the stents were captured using a mounted camera. The quantification of Evans blue staining was performed using HALO software, version 2.0, with an area quantification module. The measurements are presented as a percentage of the total area measured.

Confocal Microscopy of Stents

The halved stents were treated with Triton-X 0.1% and then incubated with primary antibodies against VE-CAD (dilution 1:200), p120 (dilution 1:400), and RAM11 (rabbit macrophages, dilution 1:200), followed by incubation with Alexa Fluor-labeled secondary antibodies (donkey anti-mouse IgG, 488 nm, dilution 1:150; and donkey anti-goat IgG, 555 nm, dilution 1:150). DAPI was used to stain the cell nuclei. The specimens were mounted with the inner surface facing up on glass slides, and images were acquired using a 10× 0.45NA objective on a Carl Zeiss LSM 880 with Zen software, version 2.3. Three-dimensional projections were created using maximum intensity projection processing of Z-stacked images. Areas were measured using calibrated Zen software. The quantification of the RAM11-positive area was done using Zen Blue, version 2.3, with a fixed threshold. High-resolution images were taken with a 20× 1.0NA objective using a super resolution AiryScan module. As a negative control, secondary antibodies were used alone without incubation with primary antibodies. The cell shape index, defined as the ratio of the X and Y dimensions of endothelial cells, was assessed from 8 randomly selected cells from 20× images of the proximal, middle, and distal areas of each stent.

Scanning Electron Microscopy

After confocal microscopic analysis, the halves of the stented arteries designated for scanning electron microscopy were rinsed in 0.1 mol/L sodium phosphate buffer and then fixed again in 1% osmium tetroxide for approximately 30 minutes. The samples were then dehydrated using increasing concentrations of ethanol, dried using a critical point dryer, and coated with a thin layer of gold. The specimens were examined using a Hitachi Model S3400N or S3600N scanning electron microscope. Low-magnification images (×15) were taken of the inner surface of the arteries to estimate how well the newly formed tissue had incorporated the implant. From these images, the extent to which the struts of the stent were covered by new tissue was semi-quantified by visual estimation from the proximal to the distal end of the stent. High-magnification scanning electron microscopy images (400–600×) from randomly selected areas were taken (3 areas from 3 stents at each time point). From these high-magnification images, the number of monocytes present on the surface (per square millimeter) was counted. Corresponding areas from the confocal microscopy analysis were also identified, and the number of surface monocytes was categorized based on whether there was high or low co-localization of VE-CAD and p120 in those areas. The monocyte count is based on extrapolations using the relative ratios of high and low co-localization.

Statistical Analysis

Continuous variables that followed a normal distribution are presented as mean ± standard error, unless otherwise specified. The Shapiro-Wilk test was used to determine if the data followed a normal distribution. Generalized estimating equations with linear or Gamma regression models were used to compare the data from the stents using SPSS software, version 22. The correlation between VE-CAD/p120 co-localization and Evans blue staining was assessed using the non-parametric Spearman correlation test because the data did not follow a normal distribution (the reported P-value is two-tailed). In vitro cell culture comparisons of variables with a normal distribution were tested using ANOVA. Data graphs were created using GraphPad Prism, version 7.02, for Windows. A P-value of less than 0.05 was considered statistically significant.

Results

mTOR Inhibition by Everolimus Impairs Endothelial Barrier Function via Calcium-Dependent Activation of PKCα, Whereas Torin-2 Does Not

The study compared the relative effects of two different mTOR inhibitors: everolimus, which requires FKBP12.6 to inhibit mTOR, and Torin-2, a newer generation direct mTOR inhibitor that does not. A dose-response curve was generated by comparing the inhibition of mTOR targets Akt and p70S6K (S6K) in human aortic endothelial cells treated overnight with everolimus or Torin-2. Western blotting for phosphorylated Akt and S6K showed that both everolimus and Torin-2 inhibited S6K at doses of 10 nmol/L or higher, with higher doses of both agents being required to inhibit phosphorylated Akt. Only higher doses of everolimus increased phosphorylated PKCα, while Torin-2 had no effect. Endothelial permeability was measured using transendothelial electrical resistance. Higher doses of everolimus (100 and 500 nmol/L) significantly decreased transendothelial electrical resistance, whereas only 500 nmol/L of Torin-2 demonstrated a significant decrease, and this decrease was not as substantial as that caused by everolimus. To compare the effect of the two different agents in vitro on various aspects of endothelial cell behavior, 500 nmol/L everolimus was used, a dose chosen to approximate average tissue levels observed over 60 to 90 days based on preclinical data from the rabbit iliac model. For Torin-2, a dose of 100 nmol/L was used, based on its IC50 value for mTOR inhibition, which is lower than that of everolimus. The addition of everolimus (500 nmol/L) to a cultured layer of human aortic endothelial cells that had grown to cover the surface caused a significant decrease in transendothelial electrical resistance relative to the control culture media (dimethyl sulfoxide), indicating impaired endothelial barrier function. Pretreatment with ryanodine (50 μM) was used to stabilize RyR2 calcium release channels and significantly reduced the effect of everolimus on transendothelial electrical resistance, suggesting that the effect of everolimus on endothelial barrier dysfunction depends on RyR2 activation. Treatment with 100 nM Torin-2 caused a significantly smaller reduction in transendothelial electrical resistance compared with 500 nM everolimus, but still more than dimethyl sulfoxide. Reducing the levels of FKBP12.6 in Torin-2–treated human aortic endothelial cells using small interfering RNA caused an additional significant decrease in transendothelial electrical resistance compared with Torin-2–treated cells that received a scrambled small interfering RNA. To assess the relative contribution of downstream targets of mTOR versus FKBP12.6, human aortic endothelial cells were transfected with small interfering RNA directed against Akt, p70S6K, and FKBP12.6, and transendothelial electrical resistance was measured 24 hours later. All groups showed significant differences compared to cells treated with scrambled small interfering RNA, with FKBP12.6 small interfering RNA having the greatest impact on transendothelial electrical resistance. The relative inhibitory effect of everolimus (500 nmol/L) versus Torin-2 (100 nmol/L) on mTOR was confirmed by Western blotting in human aortic endothelial cells. Both drugs showed significantly less phosphorylated p70S6K and phosphorylated Akt compared with cells treated with dimethyl sulfoxide. The effect of everolimus versus Torin-2 on phosphorylated PKCα was then confirmed using Western blotting. In cultured human aortic endothelial cells, everolimus, but not Torin-2, resulted in a significant increase in phosphorylated PKCα compared to the control, consistent with the dose curve data. Consistent with this, a significant increase in intracellular free calcium concentration was observed as early as 10 minutes after everolimus treatment of human aortic endothelial cells, which was not seen after Torin-2 treatment. The interaction of p120 with VE-CAD in human aortic endothelial cells treated with everolimus versus Torin-2 was examined using immunoprecipitation for p120 and was significantly decreased (relative to Torin-2–treated cells) after 2 hours of drug exposure. To confirm the effect of everolimus versus Torin-2 on VE-CAD junction formation, the interaction of p120 and VE-CAD was examined by immunostaining in human aortic endothelial cells treated with dimethyl sulfoxide, Torin-2, or everolimus overnight. Everolimus treatment increased the movement of VE-CAD from the cell membrane into the cytoplasm (relative to dimethyl sulfoxide) with evidence of intracellular deposits and gaps between endothelial cells, whereas in cells treated with dimethyl sulfoxide, there was co-localization of p120 and VE-CAD at the cell borders with no evidence of gap formation. In Torin-2–treated cells, there was evidence of co-localization of p120 and VE-CAD at cell borders with minimal gap formation between endothelial cells.

To understand the effects of bare metal stents, everolimus-eluting stents, and Torin-2–eluting stents on the vascular endothelium, stents made of the same platinum-chromium backbone were constructed. To produce drug-eluting stents, commercially available Promus stents containing 100 µg/cm2 of everolimus were used. For Torin-2–eluting stents, stents with similar backbone and polymer structure were constructed containing 100 µg/cm2 Torin-2. Bare metal stents, everolimus-eluting stents, or Torin-2–eluting stents were implanted in rabbit iliac arteries. Compared with bare metal stents, both Torin-2–eluting stents and everolimus-eluting stents significantly reduced neointimal formation as measured by the thickness of the new tissue over the stent struts. The stented segment was also harvested for Western blotting. Western blotting of the newly formed tissue for phosphorylated forms of p70S6K and Akt, and of isolated endothelial cells from the stented arteries for mTOR, showed that the phosphorylated forms of these proteins were equally and significantly suppressed compared with bare metal stents.

Vascular Permeability Is Enhanced After Everolimus-Eluting Stents but Not After Bare Metal Stents or Torin-2–Eluting Stents and Correlates With p120 VE-CAD Dissociation

The permeability of the regenerated endothelium over bare metal stents, everolimus-eluting stents, and Torin-2–eluting stents implanted for 60 days was then examined. This time point was chosen because it was estimated that all stents would have complete endothelial coverage by then. Animals were randomly assigned to receive bare metal stents, everolimus-eluting stents, or Torin-2–eluting stents. All animals underwent Evans blue dye perfusion at the time of euthanasia; the stented arteries were analyzed using en face confocal immunofluorescence for VE-CAD and p120, and high-magnification scanning electron microscopy to examine the stents for endothelial coverage. The Evans blue perfusion results showed little to small areas of endothelium staining with Evans blue dye in bare metal stents and Torin-2–eluting stents. In contrast, large areas of permeability were observed in everolimus-eluting stents. In these highly permeable areas, p120 was located at the cell borders by confocal immunofluorescence, whereas VE-CAD was localized to the cytoplasmic portion of the cells. In areas that did not show Evans blue staining, there was complete co-localization of p120 and VE-CAD at cell borders throughout the stent, which was most evident in bare metal stents and Torin-2–eluting stents. Overall, a negative correlation was found between the area of Evans blue-positive dye staining and the amount of VE-CAD/p120 co-localization per stent, as calculated by Pearson correlation. In cells with p120/VE-CAD co-localization, the shape of the endothelial cells was quantified according to the type of stent. The ratio of the length (x-axis) to the width (y-axis) of each cell was measured in en face confocal images as a cell shape index. The results showed a significantly lower cell shape index in Torin-2–eluting stents and everolimus-eluting stents compared with bare metal stents. Cobblestone-shaped cells were predominant in everolimus-eluting stents, whereas spindle-shaped cells were primarily seen in bare metal stents. In Torin-2–eluting stents, the cell shape was intermediate. Further examination using matched high-magnification scanning electron microscopy and en face confocal imaging of areas with poor versus high co-localization of p120 and VE-CAD showed that the former areas were also associated with platelets and leukocytes adhering to the surface, whereas the latter areas were not. Overall, everolimus-eluting stents showed more monocyte adherence on the surface than Torin-2–eluting stents and bare metal stents.

Proof-of-Concept Neoatherosclerotic Change (Macrophage) Is Greater in Everolimus-Eluting Stents Relative to Bare Metal Stents and Torin-2–Eluting Stents

To understand the relevance of endothelial barrier dysfunction in the development of neoatherosclerosis within drug-eluting stents, a proof-of-concept study was performed in 6 rabbits implanted with bare metal stents, Torin-2–eluting stents, or everolimus-eluting stents (n=4 stents per group) and fed a normal diet until 100 days after stenting. At this point, their diet was switched to a cholesterol-enriched diet (0.15% weight/weight). This diet was continued until the animals were euthanized at 130 days. This experimental design was chosen to test the hypothesis that endothelial barrier dysfunction is an important factor in the development of neoatherosclerosis. At 130 days, endothelial barrier dysfunction persisted, with everolimus-eluting stents demonstrating significantly greater Evans blue permeability compared to bare metal stents and Torin-2–eluting stents, respectively. The areas of VE-CAD-p120 co-localization in bare metal stents and Torin-2–eluting stents were significantly greater than in everolimus-eluting stents (100% and 96.7±6.4% versus 69.8±22.1%, respectively). Using immunostaining against RAM11 to identify macrophages, there were greater areas of macrophage infiltration on the surface of everolimus-eluting stents, followed by Torin-2–eluting stents, with no macrophage infiltration observed in bare metal stents. Cross-sectional images were analyzed to quantify the composition of the newly formed tissue within the stent (both inflammatory cells and smooth muscle cells) and showed a similar result, with a greater number of inflammatory cells in the newly formed tissue of everolimus-eluting stents compared to Torin-2–eluting stents and bare metal stents.

Discussion

Drug-eluting stents are widely used to treat localized blockages in coronary arteries. However, this study demonstrates that their use is complicated by an increase in the permeability of the endothelial layer, which may contribute to the development of neoatherosclerosis within the stents. Although the exact mechanisms behind neoatherosclerosis are not fully understood, endothelial cells are a primary defense against the development of atherosclerosis, making them a key area of focus. This research reveals that endothelial barrier dysfunction is likely a significant factor in the development of neoatherosclerosis and appears to be a direct consequence of unintended effects of current generation mTOR inhibitors. Everolimus-eluting stents showed greater permeability compared to bare metal stents and Torin-2–eluting stents. Furthermore, in a model designed to mimic neoatherosclerosis, animals fed a high-cholesterol diet and receiving everolimus-eluting stents showed greater infiltration of macrophages. The data also suggest that using a stent that releases a newer generation selective inhibitor of mTOR that directly competes for ATP binding and does not bind FKBP12.6 may be a better option for future drug-eluting stents because it avoids the unwanted effects on permeability that are partly caused by the displacement of FKBP12.6. Taken together, these findings suggest that clinical studies using stents that release new-generation direct mTOR inhibitors, such as Torin-2–eluting stents, may successfully reduce the development of endothelial barrier dysfunction, which could lead to improved outcomes in patients requiring revascularization.

Direct mTOR inhibitors are currently being evaluated for use in humans primarily in cancer treatment, where targeting the mTOR kinase is thought to result in a more potent response. Concerns exist regarding the development of resistance to mTOR kinase inhibitors through feedback activation of PI3K. The strong inhibition of mTORC1 is thought to promote feedback activation of PI3K- and PDK1-driven phosphorylation of Akt at T308. This study only examined the ability of Torin-2 to inhibit phosphorylation of Akt at S473 and found complete inhibition at the tested dose. Incomplete inhibition of Akt phosphorylation at T308 remains a theoretical concern, given the roles that both S6K and Akt play in the proliferation and migration of smooth muscle cells.

There are also important limitations to this study. Although the animal models currently used have limitations in their ability to fully replicate human conditions, results obtained with the rabbit model have generally been representative of human responses, albeit with a different timeline for healing. It is likely that the time course of endothelial barrier dysfunction observed in these animal experiments would be even longer in humans. Pathology data indicate that the incidence of neoatherosclerosis progressively increases over time, with nearly half of all drug-eluting stents showing the presence of neoatherosclerosis by 2 years. Moreover, the proof-of-concept rabbit model of neoatherosclerosis used in this study did not account for the effect of the healing process because the cholesterol diet was not administered until 90 days after stent implantation. Lastly, the release of Torin-2 from the stent was not directly measured, although it is thought that the drug release and tissue drug levels would be similar for Torin-2–eluting stents and everolimus-eluting stents given that the drug load and polymer used were identical. Although similar mTOR inhibition by Torin-2–eluting stents and everolimus-eluting stents was shown in neointimal cells and endothelium using immunoblotting, the possibility of different degrees of endothelial mTOR inhibition by these two agents cannot be completely ruled out. Similarly, it is acknowledged that there was greater neointimal formation in Torin-2–eluting stents compared to everolimus-eluting stents, which illustrates the inverse relationship between healing and the suppression of new tissue formation. Additional experimental research would be needed before pursuing initial human applications.

In conclusion, the data from this study demonstrate that canonical mTOR inhibitors used in drug-eluting stents contribute to the induction of endothelial barrier dysfunction, and this dysfunction may be a driving factor in neoatherosclerosis. The off-target effect of these agents on displacing FKBP12.6 from RyR2 and inducing calcium-dependent activation of PKCα and disruption of the p120–VE-CAD interaction in the endothelium can be overcome by the use of direct mTOR inhibitors. The potential of newer mTOR inhibitors, such as the ATP-competitive mTOR inhibitor used in this study, as therapeutic options for local delivery in drug-eluting stents should be considered.

Supplementary On-line Methods

Study Design

Based on preliminary data, it was estimated that a sample size of four to six rabbits per group would provide 80-90% statistical power to detect an expected difference of roughly 20% in the endpoints (Evans Blue dye staining or p120 VE-cadherin co-localization) with a standard deviation of 15%, using a two-sided alpha of 0.05. Stents were excluded from the study if there was evidence of damage to the endothelial layer upon examination by scanning electron microscopy. Two bare metal stents and three Torin-2–eluting stents were excluded from the 60-day group for this reason. For the 130-day group, one Torin-2–eluting stent was excluded for a similar reason. Otherwise, there were no other exclusion criteria, and all other stents were included in the study results. The Animal Model section provides specific numbers for each experiment. All endpoints were prospectively selected and remained unchanged throughout the course of the study.

Based on previous work, it was hypothesized that major differences would occur in barrier function, as assessed by Evans Blue dye staining and p120/VE-cadherin immunostaining, between everolimus-eluting stents and bare metal stents, and between everolimus-eluting stents and Torin-2–eluting stents. Healthy male New Zealand White rabbits were used in all experiments. Each animal received two of the three different stent types (bare metal stents, everolimus-eluting stents, or Torin-2–eluting stents), which were chosen randomly based on the overall study design. Data were processed randomly without regard to the study group. Endpoints such as the area of Evans Blue staining and measurements of p120/VE-cadherin co-localization area were made in a blinded fashion. Animal caretakers were also blinded to the stent types.

The study protocol was approved by the Institutional Animal Care and Use Committee of the MedStar Health Research Institute and adhered to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health.

Transendothelial electrical resistance: Electrical resistance across the endothelial cell monolayer was measured using the Electric Cell-substrate Impedance Sensing instrument and its associated software, according to the manufacturer’s instructions. ECIS-8W10E + PET array chambers containing gold film electrodes delineated with an insulating film were used. The ECIS electrode array containing cells was placed in an array holder located in the CO2 incubator. Each of the 8 wells has two sets of 20 circular 250 μm diameter active electrodes located on inter-digitated fingers to provide measurements of cells upon a total of 40 electrodes. Cells were cultured for 5 to 10 days to reach full confluence. Before transendothelial electrical resistance measurements, electrodes were calibrated and stabilized according to the manufacturer’s recommendations. For transfected cells, measurement was begun 24 hours after transfection. Before each experiment, endothelial monolayers were washed with endothelial growth medium and used for measuring changes in transendothelial electrical resistance. Transendothelial electrical resistance was measured in real-time using ECIS software and is expressed as specific electrical resistance (Ω cm2). Data presented were normalized to baseline values for each treatment group, and statistical comparison was made between treatment groups to exclude variation in baseline monolayer characteristics (n = 3-5 wells for each treatment group).

Immunoblotting

Human aortic endothelial cells were lysed using RIPA buffer supplemented with a protease inhibitor cocktail, a phosphatase inhibitor, and PMSF. The cell lysate was separated on a 10% precast polyacrylamide gel using Tris/Glycine/SDS buffer and then transferred to a nitrocellulose membrane. The membranes were incubated with commercially available antibodies against phospho-PKCα (Ser657), PKCα, VE-Cadherin, p120-catenin, phospho-Akt (S473), Akt, phospho-p70S6K (Thr 389), p70S6K, and beta-actin. Reactive bands were detected using HRP-conjugated secondary antibodies and chemiluminescence. The signals were quantified by analyzing the area and density using Quantity One 1-D Analysis Software. For each experimental group, a ratio of phosphorylated protein to total protein or beta-actin was calculated and normalized to the control group. The experiments were repeated three to four times with human aortic endothelial cells (n = 3-4 per group), unless otherwise stated.

Live Cell Imaging and Intracellular Calcium Measurements

For live cell imaging to measure intracellular calcium levels, human aortic endothelial cells were treated according to the protocol described earlier after being plated on fibronectin-coated glass coverslips. The groups were then loaded with Fluo-3 AM and Fura Red in phenol-free media for 30 minutes prior to imaging and were maintained at 37°C with a stage heater during imaging. Images were acquired every 10 seconds for 20 minutes after loading, using a Carl Zeiss LSM 700 (λ excitation = 488 nm) with a Definite focus module. Regions of interest were selected, and the mean fluorescence intensity was obtained for each region using Zen (Black) software. After subtracting background fluorescence, the ratio of Fluo-3 to Fura Red (F/Fo) excitation ratio was calculated, and the values were normalized to the time point t=0. The experiment was repeated three times.

Immunoprecipitation

Human aortic endothelial cells were lysed in RIPA buffer (20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM Na2EDTA, 1 mM EGTA, 1% NP-40, 1% sodium deoxycholate, 2.5 mM sodium pyrophosphate, 1 mM beta-glycerophosphate, 1 mM Na3VO4, 1 μg/ml leupeptin) with protease and phosphatase inhibitors. A total of 500 μg of protein was isolated and incubated with 10 μg of p120 antibodies at 4°C for 48 hours. The resulting complexes were collected using protein A/G agarose beads. Reactive bands were detected and quantified as described in the immunoblotting section. For each experimental group, the amount of co-precipitated protein was calculated and normalized to the amount of p120. The experiments were repeated three times (n = 4 per group).

Confocal Imaging of Human Aortic Endothelial Cells

Human aortic endothelial cells were cultured as described above on fibronectin-coated coverslips in 6-well plates. Confluent cells were fixed in 4% paraformaldehyde and permeabilized with 0.05% Triton-X. The cells were incubated overnight with primary antibodies against VE-Cadherin (1:200 dilution) and p120 (1:400 dilution), and then with Alexa Fluor-labeled secondary antibodies (donkey anti-mouse IgG 488 nm and donkey anti-goat IgG 555 nm). DAPI was used to stain the cell nuclei. Images were acquired using a Carl Zeiss LSM 880 with Zen software (Black edition) version 2.3. High-resolution images were taken with a 20x 1.0NA objective using a super-resolution AiryScan module. For a negative control, secondary antibodies were used alone without primary antibody incubation.

Neointimal Thickness

Halves of the cut stents from the 28-day group were fixed with 10% neutral buffered formalin. The stented vessel segments were dehydrated in a graded series of ethanol and embedded in methylmethacrylate polymer. After polymerization, 2 to 3 millimeter segments were cut from the proximal, mid, and distal portions of each stent. Sections of 4 to 6 microns thick were cut from each of the segments using a rotary microtome equipped with a tungsten carbide blade, mounted on slides, and stained with hematoxylin and eosin, as well as Movat pentachrome stains (elastin stain). Slides were scanned on an Axio Scan Z1, and morphometric measurements were performed using calibrated Zen (Blue edition) software version 2.3. Neointima thickness was measured from the lumen to the internal elastic lamina.

Immunoblotting of Neointimal Tissue and Endothelium

Halves of the cut stents from the 28-day group were immediately flash frozen in liquid nitrogen. The neointimal tissue was later separated from the stent and sonicated using a MICROSONTM XL sonicator in supplemented RIPA buffer containing PMSF and a protease cocktail, as well as a phosphatase inhibitor. Protein extracts were separated and blotted as described for human aortic endothelial cells, Torin 2 using similar antibodies. For endothelial mTOR inhibition experiments, endothelial cells were isolated from freshly harvested stented arteries by luminal digestion in PBS containing collagenase/dispase. The endothelial cell suspension was labeled with an anti-VE-Cadherin antibody and analyzed using a flow cytometer, which showed that the majority of the cells were endothelial cells. Protein extraction and Western blotting were performed as described above.