Research ArticleVascular Biology

TGF-β Signaling Mediates Endothelial-to-Mesenchymal Transition (EndMT) During Vein Graft Remodeling

See allHide authors and affiliations

Science Translational Medicine  12 Mar 2014:
Vol. 6, Issue 227, pp. 227ra34
DOI: 10.1126/scitranslmed.3006927


Veins grafted into an arterial environment undergo a complex vascular remodeling process. Pathologic vascular remodeling often results in stenosed or occluded conduit grafts. Understanding this complex process is important for improving the outcome of patients with coronary and peripheral artery disease undergoing surgical revascularization. Using in vivo murine cell lineage–tracing models, we show that endothelial-derived cells contribute to neointimal formation through endothelial-to-mesenchymal transition (EndMT), which is dependent on early activation of the Smad2/3-Slug signaling pathway. Antagonism of transforming growth factor–β (TGF-β) signaling by TGF-β neutralizing antibody, short hairpin RNA–mediated Smad3 or Smad2 knockdown, Smad3 haploinsufficiency, or endothelial cell–specific Smad2 deletion resulted in decreased EndMT and less neointimal formation compared to controls. Histological examination of postmortem human vein graft tissue corroborated the changes observed in our mouse vein graft model, suggesting that EndMT is operative during human vein graft remodeling. These data establish that EndMT is an important mechanism underlying neointimal formation in interpositional vein grafts, and identifies the TGF-β–Smad2/3–Slug signaling pathway as a potential therapeutic target to prevent clinical vein graft stenosis.


The treatment of coronary atherosclerosis is a major global health care expenditure. A substantial component of this expenditure is attributable to patients undergoing percutaneous coronary intervention or coronary artery bypass graft (CABG) surgery, with >160,000 CABG procedures performed annually in the United States alone (1). Veins are common conduits used in these surgical procedures; however, about 40% of vein grafts will fail within 12 to 18 months of implantation (2). This high failure rate is attributable to adverse vascular remodeling, yet the disease mechanisms regulating the pathophysiological changes involved remain poorly understood (3). Vascular remodeling is the result of the complex, dynamic interactions between various cell populations, including endothelial cells, mesenchymal smooth muscle–type cells, activated thrombocytes, and infiltrating inflammatory cells (46). The cellular contributions to the remodeled vessels are generally considered to be lineage-restricted (7), although the origin of the cells forming the neointima is highly controversial (8).

Endothelial-to-mesenchymal transition (EndMT) is the process by which endothelial cells lose their cell-specific markers and morphology and acquire a mesenchymal cell-like phenotype. EndMT is regulated by a complex orchestration of several signaling pathways (9) and has been implicated in chronic fibrosis (10) and in heterotopic ossification owing to fibrodysplasia ossificans progressiva (11). Recent studies indicate that EndMT may be involved in murine transplant arteriopathy (12), in the rejection of human transplant lesions (12), and in vascular cerebral cavernous malformations (13). However, the role of EndMT and the mechanisms regulating cell phenotype adaptation during human vein graft remodeling are poorly understood.

Here, we investigated EndMT during the neointimal hyperplasic response that occurs after grafting veins into the arterial circulation in both mice and humans. We demonstrated that endothelial lineage–derived cells comprise most vascular neointimal cells that formed during murine vein graft remodeling. Early failed human grafts also displayed features indicative of EndMT. Furthermore, we show that the transforming growth factor–β (TGF-β)–Smad2/3–Slug signaling pathway plays a pivotal role in regulating vein graft EndMT, with in vivo reduction of TGF-β signaling decreasing both neointimal formation and the relative contribution of endothelial lineage–derived cells to the neointima. These findings enhance our understanding of the molecular mechanisms underlying vascular remodeling and reveal novel targets for potential therapeutic interventions aimed at improving clinical outcomes after interpositional vein grafting.


Endothelial lineage–derived cells contribute to neointimal formation during vascular remodeling in mice

To examine the contribution of endothelial-derived cells to the neointima during vascular remodeling, interpositional vein grafting was performed using two independent endothelial lineage tracing systems: the constitutive EndotrackYFP or EndotrackLacZ (1416) and the tamoxifen-inducible Ind.EndotrackYFP transgenic mouse models (17). In these models, endothelial-specific cre gene expression irrevocably activates the reporter gene, resulting in continuous YFP or LacZ gene expression regardless of subsequent changes in cellular phenotype (10, 1419). Before grafting, yellow fluorescent protein (YFP) expression within the jugular vein and perivascular region was observed in 98.5 ± 0.9% (SEM) of all EndotrackYFP and in 52.1 ± 6.1% (SEM) of all Ind.EndotrackYFP endothelial cells (fig. S1A). YFP expression in native, ungrafted jugular veins was only observed in endothelial cells, which also stained positive for the endothelial cell marker CD31 (fig. S1A).

We grafted a branch of the jugular vein from the endothelial lineage–tracing mouse models into the femoral artery of genetically matched wild-type recipients through end-to-end anastomosis (20). Graft placement with chronic exposure to arterial pressure induced a remodeling program, leading to neointimal formation. Endothelial-derived YFP-positive (YFP+) cells were recruited to the neointima, representing 51.7 ± 3.3% and 27.8 ± 2.9% (SEM) of the total neointimal cell population at day 35 in the EndotrackYFP and Ind.EndotrackYFP veins, respectively (Fig. 1A). These findings were confirmed in long-term fate-tracking experiments, where endothelial-derived cells persisted and contributed to the neointima, at least until 90 days after transplantation (fig. S1B). No reporter gene activation was observed in nonrecombined EndotrackYFP or Ind.EndotrackYFP grafted veins at day 35 (Fig. 1A). Similar results were obtained using veins isolated from the EndotrackLacZ lineage-tracing model (fig. S1C).

Fig. 1. Endothelial cell lineage tracing during vein graft remodeling.

(A) Endothelial YFP expression in ungrafted (uninjured) jugular veins from EndotrackYFP and Ind.EndotrackYFP mice, as well as in grafted veins from recombined and nonrecombined EndotrackYFP and Ind.EndotrackYFP mice grafted into wild-type recipients at 35 days. Scale bars, 30 μm (uninjured) and 100 μm. (B) The number of YFP+ cells per neointimal cells was quantified. Data are means ± SEM (n = 5). (C) Endothelial-derived (YFP+) and non–endothelial-derived (YFP) cells in the neointima over time after engraftment. Data are means ± SEM (n = 4). (D) Bromodeoxyuridine-positive (BrdU+) cells relative to the total number of neointimal cells over time after engraftment. Data are means ± SEM (n = 4). P values in (B) to (D) were determined by Student’s t test.

Consistent with other mouse models of neointimal formation (4, 5), we found a progressive increase in the total number of neointimal cells between 3 and 14 days after vein grafting (Fig. 1B). Whereas the number of nonendothelial lineage (YFP) neointimal cells plateaued after 14 days, the number of neointimal YFP+ cells continued to increase up to 35 days after grafting (Fig. 1B). Similar to a previous report (21), we detected cell cycle reentry of YFP+ cells (Fig. 1C and fig. S1, D and E).

There was a very low possibility that donor-derived YFP+ hematopoietic cells contribute to vascular remodeling. Nevertheless, we analyzed ungrafted jugular veins before transplantation by fluorescence-activated cell sorting (FACS) and did not detect any donor-derived hematopoietic stem cells or myeloid progenitor cells (fig. S2A). We also analyzed peripheral blood and bone marrow of recipient animals at 35 days after grafting and did not detect any YFP+ cells after vein grafting (fig. S2, B to D). We did not identify any YFP+ macrophages, T lymphocytes, or neutrophils within grafted vessels by immunohistochemistry (fig. S3, A to C), indicating that donor hematopoietic cells do not populate the vein graft or its neointima. We excluded possible non–endothelial-dependent Tie2 promoter activation using the transgenic Tie2GFP mouse line in which green fluorescent protein (GFP) is expressed transiently when the Tie2 promoter is active (fig. S3D) (22). These experiments demonstrate that endothelial lineage–derived cells contribute to neointimal formation during vein graft remodeling.

Vein graft vascular remodeling involves EndMT

We addressed the phenotypic fate of the vein graft endothelium with respect to neointimal formation. Staining for three endothelial-specific markers (CD31, VE-cadherin, and endoglin) revealed that neointimal, but not luminal, YFP+ cells progressively lost expression of these markers between 7 and 14 days after grafting (Fig. 2, A and B). With the exception of luminal cells, we did not detect expression of endothelial-specific markers on neointimal YFP+ cells at day 35.

Fig. 2. EndMT during vein graft remodeling.

EndotrackYFP veins were grafted into wild-type recipients (n = 5) and analyzed over the course of 35 days. (A) Immunofluorescence staining of the endothelial marker CD31 in grafted veins. L, lumen. (B) Proportion of YFP+ cells expressing CD31, VE-cadherin, and endoglin over time. (C) SMA expression by YFP+ cells. (D) Percentage of YFP+ cells expressing immature VSMC markers SMA and SM22α and mature VSMC markers smoothelin, calponin, and SM-MHC. (E) SM-MHC expression in non–endothelial-derived (YFP) neointimal cells. (F) Percentage of non–endothelial-derived (YFP) neointimal cells expressing both immature and mature VSMC markers. (G) Expression of CD31 and SMA in endothelial lineage–derived (YFP+) cells 7 days after grafting. Scale bars, 10 μm (A, C, E, and G). Data in (B), (D), and (F) are means ± SEM (n = 5). *P < 0.05, **P < 0.01, ***P < 0.001 versus day 0, Student’s t test.

We examined YFP+ cells for coexpression of early vascular smooth muscle cell (VSMC) markers, smooth muscle actin (SMA) and SM22α, and the mature markers calponin, smoothelin, and VSMC-specific myosin heavy chain (SM-MHC). At day 0, no YFP+ cells were detected that coexpressed any VSMC markers (Fig. 2, C to F). Thereafter, we observed a progressive increase in the number of YFP+ cells that expressed the early VSMC markers SMA and SM22α (Fig. 2, C and D). By day 35, nonluminal neointimal YFP+ cells lost the expression of all examined endothelial markers but uniformly expressed SMA and SM22α (Fig. 2, A to D). FACS analysis confirmed these histological findings: At day 0, most YFP+ cells were CD31+ [83.2 ± 2.7% (SEM)], at day 35, nearly 85% of YFP+ cells—representing the neointimal endothelial cell population—were SMA+, whereas only 15% of these YFP+ cells were CD31+ (fig. S4, A and B). Less than 10% of YFP+ cells expressed the more mature VSMC markers calponin or smoothelin (Fig. 2D). The fully differentiated VSMC marker SM-MHC was virtually undetectable in YFP+ cells (Fig. 2E). By contrast, most YFP cells in the neointima expressed all five VSMC-specific markers (Fig. 2, E and F), consistent with these cells being of resident VSMC origin. At 7 days, 93.9 ± 3.9% of cells costaining for CD31 and SMA were YFP+, indicating that these cells were of endothelial cell origin and in transition. At 7 days, 16.0 ± 0.1% (SEM) of total neointimal cells were triple-positive for CD31, SMA, and YFP (Fig. 2G).

To further confirm that a proportion of neointimal cells were derived from endothelial cells, we used a smooth muscle cell–specific lineage-tracing model Ind.SMCtrackYFP (23). The recombination efficiency of the Ind.SMCtrackYFP system was 50.2% (YFP+/SM22α+ cells) or 51.1% (YFP+/SM-MHC+ cells) (fig. S5A). After vein grafting, we found that only 8.3% of all neointimal cells were YFP+ (fig. S5B), which, on the basis of the recombination efficiency of this mouse model, suggests that 16% of the neointimal cells were of VSMC origin.

Adventitial fibroblasts have also been attributed to neointimal formation during vascular remodeling (24); however, we were unable to detect a significant number of YFP+ cells that coexpressed the fibroblast markers fibroblast-specific protein 1 (FSP-1) or fibroblast activation protein α (FAP) in the EndotrackYFP grafted veins (fig. S6A). Only 10.4 ± 3.0% (SEM) of neointimal cells coexpressed FSP-1 and SMA, representing the myofibroblast population and accounting for a minor population of neointimal cells (fig. S6B). The matrix markers fibronectin, tenascin, and collagen III and the mesenchymal markers N-cadherin and Thy-1 were all expressed within the neointima at 35 days after grafting (fig. S7).

TGF-β signaling is activated during vascular remodeling

TGF-β signaling in vascular remodeling is well documented (2527) and has been implicated in EndMT during development and fibrotic remodeling (18, 28). In endothelial cells, TGF-β is known to mediate the sequential phosphorylation of ALK5 and Smad2/3, but its impact on ALK1 and Smad1/5/8 signaling remains controversial (29, 30). Murine jugular vein endothelial cells expressed TGFβRI/Alk5, TGFβRII, and ALK1 (fig. S8A). In grafted veins, phosphorylated Smad2/3 (p-Smad2/3) was detected in endothelial cells at days 3 and 7 after grafting, which declined by day 14 and was undetectable by day 35 (Fig. 3A and fig. S8B). Individual p-Smad3 and p-Smad2 immunostaining showed similar activation as p-Smad2/3 (fig. S8C). In contrast, p-Smad1/5/8 was evident as early as 3 days after grafting but was present throughout the 35-day observation period in YFP+ cells (Fig. 3A and fig. S8, C and D).

Fig. 3. TGF-β signaling during EndMT in vivo.

EndotrackYFP veins were grafted into wild-type recipients (n = 3). (A) Immunofluorescence staining of p-Smad1/5/8 and Smad2/3 after grafting in both endothelial-derived (YFP+) and non–endothelial-derived (YFP) neointimal cells. Scale bars, 5 μm. (B) Immunostaining of the TGF-β–regulated transcription factors Slug and Twist in endothelial-derived cells after grafting. Scale bars, 10 μm. (C) Western blot analysis of p-Smad3, p-Smad2, p-Smad1/5/8, and Slug from day 0 to day 35 after grafting. β-Actin served as a loading control.

Next, we quantified the expression of established TGF-β–regulated transcription factors implicated in developmental EndMT: Slug, Twist, and Snail (31). Slug and Twist were activated predominantly in YFP+ cells from 3 to 14 days after vein grafting, but were absent by day 35 (Fig. 3B and fig. S9, A and B). Conversely, in the early phase of remodeling (days 0 to 7), activation of Snail was barely detectable but was evident at day 35 after grafting, where it was primarily restricted to YFP cells (fig. S9C). Immunoblotting using individual p-Smad and Slug antibodies corroborated the immunostaining data (Fig. 3C). The transient activation of Smad2/3 is consistent with the up-regulation of Twist and Slug, and supports our hypothesis that TGF-β signaling triggers endothelial cell phenotypic fate changes during vein graft remodeling.

TGF-β–Smad2/3 signaling activates EndMT in murine endothelial cells

FACS analysis confirmed that TGFβRII, TGFβR1/Alk5, Alk1, and CD31 are expressed in murine endothelial hemangioendothelioma (mEOMA) cells and murine aortic endothelial cells (mAECs) (fig. S10A). Treatment of mEOMA cells and mAECs in vitro with TGF-β1 decreased the expression of the endothelial markers VE-cadherin and CD31, with a concurrent morphological change and up-regulation of SM22α and F-actin (fig. S10B), indicative of in vitro EndMT. In mEOMA cells, knockdown of Smad3 or Smad2 (fig. S11A) prevented TGF-β–induced EndMT (fig. S11B). Immunoblotting confirmed these immunostaining results (fig. S11C).

TGF-β–Smad2/3 signaling modulates EndMT during vein graft remodeling

The role of TGF-β in vein graft remodeling and the initiation of EndMT were investigated in vivo. Recipient mice were pretreated with pan–TGF-β neutralizing antibody before grafting EndotrackYFP veins that were also treated by immersion in pan–TGF-β neutralizing antibody or control immunoglobulin G (IgG) ex vivo before grafting. Inhibition of TGF-β1 was verified by measuring plasma levels of total TGF-β1 (Fig. 4A) and by confirming diminished activation of Smad2/3 at the 7-day time point (fig. S12A). Consistent with our hypothesis, blocking TGF-β signaling inhibited EndMT by reducing neointimal area and YFP+ cell contribution by day 35 compared to IgG-treated controls (Fig. 4, B to D). TGF-β inhibition did not substantially alter the extracellular matrix composition (fig. S12B).

Fig. 4. A TGF-β neutralizing antibody reduces TGF-β signaling and EndMT in vivo.

EndotrackYFP veins were immersed in TGF-β neutralizing antibody or control IgG for 4 hours before grafting. TGF-β signaling was blocked in recipient wild-type animals by treatment with a pan–TGF-β neutralizing antibody before grafting, and every 14 days thereafter (n = 5 grafts per group). (A) Plasma levels of TGF-β1 after vein graft. Data are means ± SEM (n = 5). **P < 0.01, ***P < 0.001, ****P < 0.0001, analysis of variance (ANOVA) with Newman-Keuls multiple comparison test. (B) YFP expression and corresponding hematoxylin and eosin (H&E) histology of vein grafts from treated mice at day 35. Scale bars, 15 μm (YFP images) and 100 μm (H&E images). (C) Percentage of endothelial-derived (YFP+) cells in neointima of vein grafts treated with control or TGF-β neutralizing antibody. (D) Quantification of neointimal area in treated vein grafts harvested at day 35 after grafting. (E) Percentage of SM22α+ and SM-MHC+ cells in the neointima at day 35. Data are means ± SEM (n = 5). P values in (C) to (E) were determined by Student’s t test.

TGF-β–Smad2/3–Slug signaling axis regulates EndMT during vein graft remodeling

We have shown thus far that TGF-β activates Smad2/3 and mediates EndMT both in vitro and in vivo in mice. We performed in vivo studies using EndotrackYFP mouse veins incubated ex vivo with lentiviral short hairpin RNA (shRNA) targeting Smad3 or Smad2 before grafting. We observed a significant decrease in Smad3 or Smad2 mRNA after corresponding shRNA treatment (fig. S13A). Both Smad2 and Smad3 shRNA–treated grafts displayed significantly reduced YFP+ cells and EndMT, and decreased neointimal formation at day 35 (Fig. 5).

Fig. 5. Knockdown of Smad3 or Smad2 attenuates EndMT in mouse vein grafts.

EndotrackYFP veins transduced with Smad3 or Smad2 shRNA were grafted into wild-type recipients (n = 5). (A) Immunofluorescence staining and H&E histology of the vein grafts harvested at day 35. Scale bars, 10 μm (YFP images) and 100 μm (H&E images). (B) Quantification of the neointimal area and endothelial-derived (YFP+) cells after knockdown of Smad3 or Smad2. Data are means ± SEM (n = 5). P values were determined by Student’s t test.

We next tested whether the inhibition of Smad2 or Smad3 signaling in vein grafts altered the expression of Snail, Slug, and Twist, as well as Smad-dependent activators of the mesenchymal program, myocardin, and myocardin-related factor a (MRFa) and MRFb. These experiments revealed that Smad3 knockdown significantly attenuated the expression of Slug and MRFb at the 3- and 7-day early time points (fig. S13B). Smad2 knockdown attenuated only the expression of Slug, whereas changes in MRFb expression did not reach statistical significance (fig. S13B). We used chromatin immunoprecipitation to demonstrate the direct binding of Smad3 to the Slug promoter both in vitro (fig. S13C) and in vivo (fig. S13D), indicating that TGF-β–Smad2/3 signaling in vein graft remodeling transcriptionally regulates Slug directly.

We further investigated the role of Smad3 in EndMT during vein graft adaptation in a Smad3 genetic knockout model, Smad3;EndotrackYFP mice. Homozygous Smad3 gene deletion results in severe growth deficiency and thus renders these mice unsuitable for vein graft experiments. Jugular veins from heterozygous Smad3;EndotrackYFP mice were grafted into genetically matched wild-type controls. At 35 days after grafting, Smad3 haploinsufficiency attenuated neointimal formation and EndMT (Fig. 6, A and B). Activation of Smad3, but not Smad1/5/8, was significantly decreased in these grafts (fig. S14). Lentiviral-mediated overexpression of Slug rescued the effect of Smad3 haploinsufficiency on neointimal formation and EndMT after vein grafting (Fig. 6C). The role of endothelial Smad2 was examined in the Smad2Fl/Fl;Ind.EndotrackYFP line. Depletion of endothelial Smad2 resulted in more favorable vascular remodeling and reduced EndMT (Fig. 6D). These findings are consistent with our Smad3 and Smad2 shRNA knockdown data (Figs. 4 and 5) and support the important role of Smad2/3 in TGF-β–dependent EndMT during vein graft remodeling.

Fig. 6. Neointimal formation and EndMT in vein grafts are regulated via Smad3 and Slug signaling.

All data were obtained at day 35 after grafting. (A) YFP and H&E staining of Smad3+/- EndotrackYFP veins grafted into wild-type recipients. Scale bars, 10 μm (YFP images) and 100 μm (H&E images). (B) Quantification of neointima area and endothelial-derived (YFP+) cells in the neointima. (C) H&E staining and quantification of neointima in Smad3+/- mouse veins overexpressing Slug that had been grafted into wild-type recipients (n = 5). (D) Vein grafts from Ind.EndotrackYFP and Smad2Fl/Fl;Ind.EndotrackYFP mice that had been transplanted into wild-type recipients (n = 5) were stained for YFP or H&E. Scale bars, 15 μm (YFP images) and 100 μm (H&E images). Quantification of YFP+ endothelial-derived cells and neointimal area. Data in (B) to (D) are means ± SEM (n = 5). P values were determined by Student’s t test.

EndMT is observed in human endothelial cells and in human vein graft remodeling

We confirmed by FACS analysis that TGFβRII, TGFβR1/Alk5, and Alk1 are expressed in human umbilical vein endothelial cells (HUVECs) (fig. S15A). HUVECs treated with TGF-β1 showed increased expression of SM22α and Slug, and decreased expression of CD31 and VE-cadherin, compared with no treatment (Fig. 7A), which is indicative of EndMT. Similar to murine vein grafts, TGF-β1–treated HUVECs show increased gene expression of Slug, Snail, and Twist (fig. S15B), suggesting that the TGF-β–Smad2/3–regulated transcriptional program is shared between species and plays a central role in orchestrating EndMT in human cells. Data from gene expression profiles comparing HUVECs, TGF-β1–treated HUVECs, and VSMCs show that TGF-β1–treated HUVECs acquired a gene expression profile that was distinct from both untreated HUVECs and VSMCs (Fig. 7B). TGF-β1–treated HUVECs lost expression of markers present in native HUVECs [for example, CD31, vWF (von Willebrand factor), and endothelin 1] and gained expression of markers present in primary VSMCs [for example, αSMA, PDGF-β (platelet-derived growth factor-β), and LOX] indicative of EndMT in human cells.

Fig. 7. EndMT in human endothelial cells and in human vein graft remodeling.

(A) HUVECs were treated with TGF-β1 (10 ng/ml) and stained for CD31, VE-cadherin, SM22α, Slug, and SM-MHC. Scale bars, 10 μm. Images are representative of n = 3 experiments. (B) Human Endothelial Cell Biology and Fibrosis RT2 PCR Arrays were used to assess gene expression in baseline HUVECs ± TGF-β1 and human aortic VSMCs. Data represent significant (P ≤ 0.05, Student’s t test) twofold differentially regulated gene expression in HUVECs compared to HUVECs with TGF-β1. (C) EndMT observed in tissue samples during human vein graft remodeling. Immunofluorescence costaining of early (<1 year, n = 5) and late (>6 years, n = 5) human vein grafts with endothelial (VE-cadherin and vWF) and immature (SMA and SM22α) or mature (SM-MHC and calponin) VSMC markers. Scale bars, 100 μm. Graph depicts quantification of EndMT (double-positive staining) in human vein grafts (n = 5 per group). P values were determined by Student’s t test.

To better understand the role of EndMT in human vein graft remodeling, we evaluated histological sections from newly acquired early-phase failed (<1 year, n = 5) and long-term patent vein grafts (>6 years, n = 5) stained with multiple endothelial cell markers (VE-cadherin, vWF, and CD31) and immature (SMA and SM22α) or mature VSMC markers (SM-MHC and calponin). A subset of cells in the early vein graft neointimas costained for both endothelial and immature SMC markers, but not for mature VSMC markers (Fig. 7C, top panels), indicating that endothelial cells transition and acquire a VSMC-like phenotype during human vein graft remodeling. In the long-term patent grafts, only the luminal endothelial cells stained for endothelial cell markers, and no double-positive cells were observed in the neointima (Fig. 7C, bottom panels). The human data corroborate our murine studies and underscore a role for EndMT in both human and mouse vein graft remodeling.


Here, we present evidence that EndMT is a central component of the vascular remodeling that occurs during vein adaptation to the arterial circulation in a murine model. At 35 days after vein grafting, half of the cells populating the neointima were of endothelial origin. These endothelial-derived cells lost their endothelial phenotype and acquired smooth muscle cell–like properties, yet failed to develop a fully “mature” smooth muscle cell phenotype. We provide a mechanistic model demonstrating that TGF-β–Smad2/3–Slug signaling is a principal pathway regulating EndMT during vein graft remodeling. Examination of postmortem human vein graft tissue confirmed that the process of EndMT is operative in human vein graft remodeling. Together, these data identify new pathways regulating vascular remodeling that may be targeted for clinical interventions to improve outcomes after interpositional vein grafting in humans.

Supporting our in vitro findings, Medici et al. recently showed that TGF-β signaling promotes Snail-mediated EndMT via both Smad-dependent and Smad-independent signaling (32), and studies using murine models have shown that endothelial cells participate in cardiac fibrosis via EndMT (11, 18). It is well known that during development, TGF-β–mediated signaling contributes to emerging mesenchymal cell populations. Therefore, this study and others collectively support the hypothesis that TGF-β–mediated developmental programs are reactivated in response to injury or disease, and we propose that adult vein graft neointimal formation represents yet another biological setting where EndMT contributes to cardiovascular pathogenesis. Treatment strategies to attenuate TGF-β signaling in experimental cardiovascular disease models may include administration of a TGF-β neutralizing antibody or a small-molecule inhibitor to prevent vascular remodeling (33), as we have demonstrated in mouse models to stop EndMT. The role of other TGF-β family members and downstream signaling pathways activated upon vein graft surgery and subsequent remodeling remains unclear. Specifically, neither the biological effects of BMP9/10-induced Smad1/5/8 signaling nor the TGF-β–induced Smad1/5/8 activation was examined in this study. The role of BMP9/10-Smad1/5/8 in vein graft remodeling is currently unknown, and its relationship to Smad2/3 requires further investigation.

The use of genetically engineered mouse models enabled us to identify the cellular origins and molecular mechanisms contributing to the underlying pathology of vein graft remodeling, which is currently not possible with larger animal models or with humans. The cellular origins of the neointima during vascular remodeling have, in part, been attributed to differences in surgical technique (34). The vascular remodeling that occurs in end-to-end anastomoses differs from end-to-side anastomoses in that the neointima in the former is composed primarily of donor-derived cells, whereas that in the latter is composed primarily of recipient-derived cells (35, 36). Our study used end-to-end anastomoses to graft a branch of the jugular vein into the femoral artery. This technique resulted in substantial donor endothelial cell preservation, despite the increase in graft diameter and initial disruption of the venous endothelial layer because of increased arterial pressure. Although it may be argued that the combination of pressure-induced disruption of the endothelial cell layer with the preservation of select endothelial cell groups may facilitate EndMT, other dissimilarities between these vein graft models remain to be defined to account for the differing neointimal cellular contributions. Among the established murine vein graft models, our surgical model most closely replicates clinical vein-to-arterial grafting (34).

However, there are inherent differences between mouse and human anatomy and physiology, including differences in size, heart rate, and hemodynamic output, limiting the degree to which one can extrapolate these findings to human vein graft remodeling. For example, unlike human veins, mouse veins consist of a considerably thinner smooth muscle cell layer; the distending forces of acute-onset arterial pressure and flow might induce different degrees of vascular damage in comparison to human vein grafts. Further, because of size disparity, different surgical techniques are used to transplant mouse versus human veins, which may affect the remodeling process. Furthermore, endothelial cell lineage–tracing studies are not feasible in humans; therefore, we examined postmortem failed and long-term patent vein grafts. Corroborating our in vitro and murine studies, early-phase human vein grafts displayed molecular features that were strongly suggestive of EndMT.

In summary, our findings contribute to the greater understanding of the molecular mechanisms underlying human vascular remodeling and reveal novel targets against which potential therapeutic interventions can be developed for improving clinical outcomes after interpositional vein grafting. Toward translation, in a clinical study, treatment with the angiotensin II receptor blocker losartan significantly slowed the rate of pathologic aortic root dilation in a small cohort of pediatric patients with Marfan syndrome (37). From the perspective of attenuating vein graft neointimal formation, given that TGF-β signaling is up-regulated within hours after vein graft surgery (38) and that neointimal formation plateaus and slows over the ensuing weeks and months, we speculate that the optimal window for clinical therapeutic intervention with anti–TGF-β therapies may be perioperative and continued in the subsequent 2 to 6 weeks after graft surgery. Translation to the clinic will first require additional proof-of-concept studies in larger animal models to confirm the potential clinical benefit of modulating the TGF-β signaling pathway after vein graft surgery. These findings contribute to the greater understanding of the molecular mechanisms underlying human vascular remodeling and reveal novel targets that may be translated into potential therapeutic interventions to improve clinical outcomes after interpositional vein grafting.


Study design

The study objective was to use in vivo endothelial cell lineage–tracing murine models to track and quantify donor vein endothelial cells grafted into the arterial circulation, characterize the molecular mechanisms regulating EndMT, and identify potential therapeutic targets. For all surgeries, numerical identifiers linking genotype/treatment of the donor sample were blinded to the scientist assessing/quantifying the results (for example, counting costained cells). All experiments were replicated as indicated. The 3-, 7-, 14-, and 35-day time points were selected on the basis of (20), and our preliminary studies showed that there is no additional neointimal formation after the 35-day time point. A minimum of four patent mouse vein grafts were used in data analysis. Postmortem human samples were selected on the basis of available samples that most closely approximated the mouse time points. Anonymized unstained sections of paraffin-embedded human saphenous vein grafts were available from the CVPath Institute Sudden Coronary Death registry (CVPI-SCDRr) from hearts submitted in consultation (39).

Generation of animals

The National Heart, Lung, and Blood Institute (NHLBI) Animal Care and Use Committee approved all animal experiments. Primers used for genotyping are listed in table S1. EndotrackYFP and EndotrackLacZ mice were created by crossing the Tie2Cre (Jackson Laboratories) and R26RstopYFP and R26RstopLacZ lines, respectively (F. Constantini, Columbia University) (1416). The Ind.EndotrackYFP line was created by breeding tamoxifen-inducible endSclCreERT (C.-M. Hao, Vanderbilt University) (17) and R26RstopYFP lines. The indSM22αCreT line was created by crossing SM-CreERT2 (ki) (R. Feil, University Tübingen, Germany) (23) and R26RstopYFP lines. The Smad3;EndotrackYFP line was created by crossing heterozygous 129-Smad3<tm1Par>/J (Jackson Laboratories) and EndotrackYFP lines. Smad2Fl/Fl;Ind.EndotrackYFP was created by crossing the Smad2Fl/Fl (M. Weinstein, Ohio State University) (40) and Ind.EndotrackYFP lines. C57Bl/6J, 129S2/SvPasCrl, and transgenic (Tie2GFP)287Sato (22) mice were purchased from Jackson Laboratories.

Induction of cre gene was achieved in 4-week-old mice by intraperitoneal injection of 1 mg of tamoxifen dissolved in 100 μl of sterile peanut oil once daily for 1 week, followed by 1 week off, then a second course of injections daily for 1 week.

Vein grafting

Vein graft surgeries were performed as described in (20), where a vein graft was interpositioned via end-to-end anastomoses, as detailed in Supplementary Materials and Methods.


For immunostaining on fresh-frozen tissues, thawed slides were washed in phosphate-buffered saline (PBS) and blocked using 5% donkey serum in PBS with 0.1% Triton X-100. For staining on formalin-fixed, paraffin-embedded human samples, slides were deparaffinized and underwent antigen retrieval. After blocking, primary antibodies were applied in 3% bovine serum albumin overnight at 4°C, followed by the corresponding secondary antibody for 1 hour at room temperature. Primary antibodies are shown in table S2. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI). Control slides were routinely stained in parallel by substituting IgG or the specific IgG isotype. Protocol for BrdU immunolabeling is detailed in Supplementary Materials and Methods. All images were acquired using either a Zeiss Axioscope with AxioVision v4.3 software, a Zeiss LSM 510 UV laser scanning confocal microscope (Carl Zeiss GmbH), or Nikon Eclipse TE300 inverted fluorescence microscope.

Counting of costained cells

Slides were scanned with a confocal microscope, as described, at the stated magnification. Images were analyzed in a blinded manner. Specimen areas were divided into numbered grids, and every cell pertaining to the neointimal region within the grid was counted. Total cell number within each grid was assessed by DAPI staining. Antibody-specific staining was assessed with filters specific for the corresponding secondary antibody fluorophore. YFP+ cells were counted from images manually and expressed as the number of YFP+ cells relative to total neointimal cells.

Morphology of neointima and cell counting

For morphological analysis of neointimal area, we first determined the ideal luminal area by calculating the area within the perimeter of the internal elastic lamina. Next, we measured the luminal area within the neointima (residual lumen). The difference between these values represents the neointimal area.

Cell lines and culture conditions

HUVECs (Lonza) and mEOMA cells (American Type Culture Collection) were maintained according to the manufacturer’s instructions, detailed in Supplementary Materials and Methods.

shRNA transfection

shRNA constructs for Smad3, Smad2, and scrambled controls were purchased from Open Biosystems, prepared according to the manufacturer’s protocol, and used as described in Supplementary Materials and Methods.

Neutralizing anti–TGF-β antibody treatment

Recipient mice were injected intraperitoneally with neutralizing anti–pan–TGF-β (10 mg/kg) (Sigma) or control IgG 1 day before surgery, and every 2 weeks thereafter. Donor veins were pretreated with neutralizing antibody (1 μg/ml) or control IgG for 4 hours at room temperature. Total TGF-β1 from plasma at specified time points was measured by enzyme-linked immunosorbent assay according to the manufacturer’s instructions (R&D Systems).

RNA extraction, reverse transcription, and real-time polymerase chain reaction

RNA was isolated from cultured cells with the RNeasy kit (Qiagen) according to the manufacturer’s instructions. Details regarding extraction from tissue samples are outlined in Supplementary Materials and Methods. Primers are listed in table S1.

Western blot

Samples were lysed in radioimmunoprecipitation assay buffer containing protease and phosphatase inhibitors (Sigma). For each sample, 20 μg of total protein was analyzed by Western blot (Bio-Rad 4–20% Mini-PROTEAN TGX Gel) and transferred onto nitrocellulose membranes. Antibodies used are listed on table S2.

Fluorescence-activated cell sorting

Jugular veins and vein grafts were digested with Liberase TM (0.13 U/ml) (Roche Applied Science) at 37°C for 20 to 25 min, and digestion was stopped with 20% fetal bovine serum. Bone marrow and peripheral blood cells were incubated in Gey’s solution and stained in FACS buffer, as detailed in Supplementary Materials and Methods.

Statistical analyses

For cell culture, each experiment was repeated at least three times. Data are means ± SEM. After testing the normality of data by Shapiro-Wilk test, we applied Student’s t test for two-condition comparison or one-way ANOVA test for time course comparison. If the condition levels are more than two, the Newman-Keuls multiple comparison correction was performed after ANOVA. Differences were deemed significant when P < 0.05. Statistical analyses were performed with Prism version 4.00 (GraphPad Software) and verified by Y. Yang (NHLBI).


Materials and Methods

Fig. S1. Endothelial cell lineage tracing and kinetics during vein graft remodeling.

Fig. S2. FACS analyses for YFP+ KSL (cKit+, Sca1+, Lin) hematopoietic stem cells.

Fig. S3. YFP+ cells in the neointima of EndotrackYFP grafted into wild-type recipients are not bone marrow–derived.

Fig. S4. FACS analysis of EndMT during vein graft remodeling.

Fig. S5. Vascular smooth muscle–specific recombination efficiency in Ind.SMCtrackYFP veins.

Fig. S6. Immunostaining using fibroblast and myofibroblast markers.

Fig. S7. Extracellular matrix proteins and mesenchymal markers in grafted jugular veins.

Fig. S8. Evidence of TGF-β signaling in EndMT at various time points in EndotrackYFP grafted veins.

Fig. S9. Localization of TGF-β–regulated transcription factors during vein graft remodeling.

Fig. S10. In vitro EndMT in mouse cells.

Fig. S11. TGF-β–Smad2/3 signaling activates EndMT in murine endothelial cells.

Fig. S12. TGF-β neutralizing antibody attenuates Smad2/3 activation.

Fig. S13. Knockdown of Smad3 or Smad2 attenuates neointimal formation.

Fig. S14. Smad3 and Smad1/5/8 presence 7 days after vein grafting.

Fig. S15. Expression of TGF-βR and TGF-β–induced gene targets in HUVECs.

Table S1. Primers used.

Table S2. Antibodies used.


  1. Acknowledgments: We thank F. Constantini (Columbia University) for providing R26RstopYFP and R26RstopLacZ mice, C.-M. Hao (Vanderbilt University) for the tamoxifen-inducible endSCLCreERT mice, and M. Weinstein (Ohio State University) for Smad2Fl/Fl mice. We thank the staff of the Laboratory of Animal Medicine and Surgery facility and L. McKennett for assistance with the transgenic mice. We acknowledge the professional skills and advice of C. A. Combs and D. Malide of the NHLBI Light Microscopy Core Facility and Z.-X. Yu and colleagues at the NHLBI Pathology Core Facility. Statistical advice was provided by Y. Yang of the NHLBI DNA Sequencing and Computational Biology Core. Funding: This project was funded by the Intramural Research Program of the NHLBI. J.N. was supported by the Commission on Higher Education and the Philippine Council on Advanced Science and Technology Research and Development, Department of Science and Technology (PCASTRD-DOST) in the Philippines. J.M. was supported by an NIH Marshall Scholarship. J.C.K. is supported by NIH grant K08HL111330. J.D.L. was supported by a Boehringer Ingelheim Fonds MD Fellowship. Author contributions: B.C.C. and H.S. performed surgeries; J.N., J.M., D.Y., C.S.H., A.N., F.F., G.C., A.D.W., C.B., A.E., and M.B. designed and/or performed immunostaining, Western blotting, FACS, histological staining, and cell culture experiments; D.Y., C.S.H., and M.B. worked with human samples; J.N., D.Y., C.S.H., A.N., A.D.W., J.D.L., and M.B. analyzed the data; R.L.S., B.T., and L.E.B. performed mouse colony management, coordinated animal surgery experiments, and conducted specimen collection. M.B. provided funding; M.B., D.Y., C.S.H., A.N., and J.C.K. wrote the paper; and R.F., R.V., and E.L. provided specific reagents. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All murine models are available with materials transfer agreement. Postmortem human samples were obtained from CVPath Institute Inc.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article