Research ArticleTissue Engineering

A completely biological “off-the-shelf” arteriovenous graft that recellularizes in baboons

See allHide authors and affiliations

Science Translational Medicine  01 Nov 2017:
Vol. 9, Issue 414, eaan4209
DOI: 10.1126/scitranslmed.aan4209

Growing grafts for hemodialysis

Patients undergoing hemodialysis for renal failure often receive an arteriovenous fistula, a connection between a vein and an artery. These surgical connections fail or cannot be attempted in some patients with compromised vasculature, who instead require vein grafts. As an alternative to autologous or synthetic grafts, Syedain et al. used a tissue engineering approach to generate vascular grafts from sacrificial fibrin scaffolds and human fibroblasts. Decellularized grafts were implanted into baboons and tested as hemodialysis access points. Over the course of 6 months, the grafts were recellularized with host cells and maintained sufficient burst pressure without evidence of immune rejection. Pending additional testing, these grafts represent an additional surgical option for hemodialysis access.


Prosthetic arteriovenous grafts (AVGs) conventionally used for hemodialysis are associated with inferior primary patency rates and increased risk of infection compared with autogenous vein grafts. We tissue-engineered an AVG grown from neonatal human dermal fibroblasts entrapped in bovine fibrin gel that is then decellularized. This graft is both “off-the-shelf” (nonliving) and completely biological. Grafts that are 6 mm in diameter and about 15 cm in length were evaluated in a baboon model of hemodialysis access in an axillary-cephalic or axillary-brachial upper arm AVG construction procedure. Daily antiplatelet therapy was given. Grafts underwent both ultrasound assessment and cannulation at 1, 2, 3, and 6 months and were then explanted for analysis. Excluding grafts with cephalic vein outflow that rapidly clotted during development of the model, 3- and 6-month primary patency rates were 83% (5 of 6) and 60% (3 of 5), respectively. At explant, patent grafts were found to be extensively recellularized (including smoothelin-positive smooth muscle cells with a developing endothelium on the luminal surface). We observed no calcifications, loss of burst strength, or outflow stenosis, which are common failure modes of other graft materials. There was no overt immune response. We thus demonstrate the efficacy of an off-the-shelf AVG that is both acellular and completely biological.


The annual incidence of hemodialysis-treated end-stage renal disease patients in the United States is estimated to be 100,000 to 105,000 with a prevalent population of 415,000 to 425,000 (1). Arteriovenous fistula (AVF), where a vein is surgically connected directly to an artery, is currently the preferred mode of vascular access for hemodialysis, because it has the highest primary patency rate combined with the lowest associated risk of morbidity and mortality compared to the alternatives, arteriovenous graft (AVG) or central venous catheter. However, 30 to 50% of AVF thrombose, fail to mature sufficiently for use, or develop issues refractory to intervention (2). In addition, an AVF requires between 2 and 3 months to fully mature before cannulation for hemodialysis.

Currently, expanded polytetrafluoroethylene (ePTFE) is the most common material used in construction of prosthetic AVGs. However, ePTFE tubes are more prone to clotting and infection, have worse primary and secondary patency rates, and thus require a greater degree of clinical care to maintain patency compared to AVF. ePTFE tubes have a primary patency rate of 50% at 12 months, which is often due to the development of intimal hyperplasia and associated stenosis of the outflow vein (3). The AVG has one advantage over a successful AVF: It is typically ready for use within several weeks, often eliminating the need for a period of catheter-based dialysis until the AVF is suitable for use. If an alternative material to ePTFE could be identified for AVG construction that improved hemocompatibility and durability, it has the potential to be widely adopted.

Tissue engineering (TE) has led to some promising alternative materials. The quest for a tissue-engineered vascular graft began more than 30 years ago (4) and has finally approached clinical reality in a few cases, including the original conceptions of implanting a living tissue tube grown using autologous cells (5) or a synthetic polymer tube seeded with autologous cells (6). In the most recent clinical success pioneered by Dahl et al. (7), TE—growing a tissue tube using donor cells—is combined with regenerative medicine (RM)—implanting an acellular scaffold conducive to recellularization by host cells via decellularization of the tissue tube. Many of these “off-the-shelf” grafts can be fabricated from a single biopsy to facilitate commercialization and clinical utility, because it is a nonimmunogenic allograft by virtue of decellularization. This new TE/RM approach has yielded promising results in a phase 2 clinical trial as an AVG (8). In their study, human allogeneic smooth muscle cells were grown on a tubular polyglycolic acid (PGA) scaffold, subjected to pulsatile distention in a bioreactor for 8 weeks during which most of the PGA hydrolyzes (9), and then the tubular construct was decellularized using detergents (7). In an alternative approach to an off-the-shelf graft, Wystrychowski et al. (10, 11) implanted a graft fabricated from their pioneering sheet-based TE approach that had been devitalized via dehydration.

In contrast to the use of a synthetic polymer scaffold, we previously developed a completely biological vascular graft grown in vitro from human dermal fibroblasts (hDFs) entrapped in sacrificial fibrin gel tubes (12). Over a period of about 8 weeks, the fibroblasts convert the fibrin gel tube into a tube of circumferentially aligned dense collagenous matrix, which is then detergent-decellularized to create an off-the-shelf allograft that can be stored refrigerated in phosphate-buffered saline for at least 6 months without loss of mechanical properties (13). It has physiological compliance and a burst pressure that meets or exceeds that of native arteries (14). We have shown in preclinical studies in sheep (up to 6 months’ duration) that these grafts, grown from ovine dermal fibroblasts, become populated by appropriate host cells including endothelium formation without a sustained inflammatory response or an immune response as arterial grafts and tubular heart valves (1317) and even grow with young lambs into adulthood (17). The complete absence of residual synthetic polymer fragments from these grafts at the end of 6 months ensures the absence of any associated persistent inflammatory response to such residual polymer reported in other studies (1820).

On the basis of this TE/RM approach, we now report the results of implanting our completely biological vascular grafts as AVG into the baboon, which has been used as a critical preclinical model given the expected minimal immunogenicity of human extracellular matrix in an old-world nonhuman primate (7). Over the course of 6 months, implanted AVGs were periodically monitored for patency and diameter with ultrasound while being cannulated with a dialysis needle. Explanted grafts were characterized for mechanical properties and composition via biochemical assay, histology, and immunohistochemistry. These were performed to answer key questions about the graft including immunogenicity, propensity for calcification, capacity for recellularization, matrix remodeling in relation to maintenance of stiffness and strength, and tissue repair after cannulation.


AVG pre-implant properties

Before implantation, the grafts were 6 mm in diameter and about 13 cm in length after cutting off the ends that included Dacron cuffs for handling (Fig. 1, A and B). They appeared highly collagenous and aligned circumferentially based on trichrome and picrosirius red staining and were devoid of cell nuclei (Fig. 1, C and D). Residual DNA and SDS were quantified as 14.5 ± 8.0 μg/cm3 and 54 ± 43 μg/cm3, respectively. The graft wall thickness was 0.48 ± 0.05 mm. In terms of mechanical properties, UTS was 3.8 ± 0.7 MPa and modulus was 10.2 ± 2.2 MPa in the circumferential direction (Fig. 1E). Graft burst pressure was 3164 ± 342 mmHg, or 99.8% of the value reported for the human internal mammary artery (3196 ± 1264 mmHg) (Fig. 1F) (21). Graft suture retention was 199 ± 56 gram-force, or 144% of the value reported for the native internal mammary artery (138 ± 50 gram-force) (Fig. 1G) (21).

Fig. 1. Pre-implant AVG appearance and physical properties.

(A) Side view and (B) end-on view image of a 6-mm-diameter decellularized tissue-engineered vascular graft. (C) Trichrome-stained cross section of the graft showing circumferentially aligned collagen fibers (blue). (D) Picrosirius red–stained cross section under polarized light confirming collagen fiber orientation primarily in the circumferential direction. (E) Ultimate tensile strength (UTS) and modulus of the graft. (F) Burst pressure and (G) suture retention force with comparison to human internal mammary artery (IMA) (21). Scale bars, 200 μm.

AVG patency

After surgical exposure of the target vein (initially the brachial vein), we observed profound venospasm and a very thin vein that prompted us to alter our venous outflow target to the larger and thicker cephalic vein. The first graft remained patent for 6 months (BAVG1, fig. S1). We subsequently experienced continued occurrence of spasm and unexplained rapid, occlusive thrombosis in the next five cases, during which time we instituted small iterative changes in anticoagulation (added clopidogrel for the first 14 days with or without 7-day preloading), and applied topical lidocaine and papaverine reactively and then proactively for spasm before abandoning the cephalic vein for outflow. Once we returned to the brachial vein and minimized surgical exposure time, we no longer experienced early thrombotic events. A summary of implants in the five animals with early thrombosis (all in the cephalic vein) includes two animals (BAVG5 and BAVG6) that were euthanized at about 2 weeks postoperatively to confirm and study the thrombotic event and the remaining three animals (BAVG2, BAVG3, and BAVG4) that were subsequently reimplanted with a second graft in the contralateral arm using the brachial vein as outflow target (fig. S1). The regraft surgery was performed within 6 to 8 months of the first implant in all cases.

For the subsequent five grafts implanted using brachial vein outflow (two primary and three regrafts), the 3-month primary patency rate was 80% (4 of 5) (fig. S1). One patent graft was explanted by design at 3 months (BAVG7). Two of the remaining four grafts (50%) were patent at 6 months (BAVG2 and BAVG8); one regraft (BAVG3) failed within 3 weeks as a result of spontaneous rupture, and one regraft (BAVG4) clotted at an unknown time point after the 3-month assessment. See fig. S1 for a summary of the implantation chronology, conditions, and outcomes.

Considering the first graft redirected to the brachial vein in addition to the subsequent five grafts described previously but excluding grafts that rapidly clotted during development of the model, the overall 3-month and 6-month primary patency rates were 83% (5 of 6) and 60% (3 of 5), respectively. Patency survival for this subset of grafts is plotted in Fig. 2. Patency survival curves for grafts using cephalic versus brachial outflow, which are identical to early versus late cases during model development, are shown in fig. S2.

Fig. 2. Kaplan-Meier diagram for all grafts that did not rapidly generate an occlusive clot.

Percent graft patency is plotted as a function of time after implantation in baboons (n = 6).

AVG cannulation

All five grafts that were patent for at least 3 weeks were cannulated multiple times with only one instance of repeated pressure being required for hemostasis after needle puncture. Only one such instance occurred despite the relative lack of subdermal tissue to facilitate venipuncture closure in this model compared to typical human anatomy (Fig. 3A).

Fig. 3. AVG implanted into a baboon model.

(A) Graft image at implantation showing proximal anastomosis to axillary artery and distal anastomosis to brachial vein. (B) Ultrasound-based measurement of midgraft and distal graft diameter over the course of implantation. Color Doppler image of (C) distal graft-vein junction at 3 months, (D) midgraft at 6 months, and (E) distal graft-vein junction at 6 months (heat map scale in meters per second). (F) Angiogram image of the graft with contrast dye at 6 months. White arrows indicate distal anastomosis. Paired symbols (* and #) in the plot indicate P < 0.05.

Ultrasonography and angiography measurements

Midgraft diameter measured with ultrasound at 1 month after implantation was 5.7 ± 1.4 mm (n = 5), which was not different from the pre-implant diameter of 6 mm (Fig. 3B). A trend of increasing diameter attained statistical significance (P < 0.05) at 6 months (9.5 ± 0.9 mm, n = 3), which was also true for the graft diameter near the distal anastomosis (Fig. 3B). However, fig. S3 shows that the three patent 6-month explants had stable diameter over the final 3 months of implantation. The diameter at 2 months (7.4 ± 1.6 mm, n = 5) had increased but did not change during further follow-up (Fig. 3B). A similar trend was evident for the graft diameter near the distal anastomosis. Qualitatively, blood flow in the mid and distal regions of the graft was generally turbulent (Fig. 3, C and D) and not accurately measured. No graft stenosis was observed at the distal anastomosis (Fig. 3, C to E). It was also observed that native vein diameter distal to graft increased from 3.2 ± 0.7 mm to 6.4 ± 1.9 mm over the first month for the five patent grafts, consistent with outflow vein accommodation.

At the final cannulation during explant, contrast dye was injected through the needle and visualized using a C-arm. The patent grafts (one graft explanted at 3 months and three of four grafts explanted at 6 months, as detailed above) showed no stenosis (Fig. 3F). Angiography indicated that graft diameter increased toward the distal half of the graft compared with the proximal half in some cases (fig. S4).

AVG mechanical properties at explantation

Grossly, there was connective tissue adhesion to the abluminal surface of the grafts explanted at both 3 and 6 months (Fig. 4, A and B), and the luminal surface of the grafts appeared glossy (Fig. 4, C and D). Grafts were sectioned for histology and immunohistochemistry in the proximal, mid, and distal graft regions. Proximal and distal graft segments were analyzed for burst strength and wall thickness. The wall thickness had increased from the pre-implant value of 0.48 ± 0.05 mm to 0.86 mm at 3 months and 0.79 ± 0.10 mm at 6 months (Fig. 4E), consistent with extensive recellularization at both time points as seen in H&E-stained sections (Fig. 4, F and G). At 3 months (n = 1), the graft burst pressure was 4953 mmHg in the proximal region and 4735 mmHg in the distal region. The distal region tested was the region accessed with a dialysis needle at 1 and 2 months for this animal. At 6 months (n = 3), the average burst pressure in the proximal region was 4918 ± 530 mmHg, whereas the distal region cannulated at 1, 2, and 3 months was 3714 ± 195 mmHg (Fig. 4H). Averaging all measured burst pressures, at 3 months, the average burst pressure was 136% of the pre-implant values, and at 6 months, it was 156% of the pre-implant values.

Fig. 4. Explanted AVG appearance and physical properties.

Side view at (A) 3-month and (B) 6-month explantation showing loose connective tissue grown onto abluminal surface. End-on view at (C) 3-month and (D) 6-month explantation. (E) Graft thickness of the pre-implant grafts and explanted grafts in distal regions that were cannulated (shaded) and proximal region (solid). Hematoxylin and eosin (H&E) stain of midgraft section at (F) 3-month and (G) 6-month explantation. (H) Graft burst pressure measurement comparison of the pre-implant grafts with proximal segments (solid) and distal cannulated segments (shaded). Paired symbols (* and #) in plots indicate P < 0.05. In H&E images, luminal surface is marked with an asterisk (*). Scale bars, 100 μm.

AVG matrix remodeling, cellular invasion, and endothelialization

As noted above, host cells had extensively populated the one patent graft explanted at 3 months and showed similar distribution in the three patent explants at 6 months (Fig. 4, F and G). These H&E-stained sections also revealed small aggregates of inflammatory/immune cells including eosinophils, polymorphonuclear leukocytes, lymphocytes, and macrophages along the abluminal aspects of the grafts, but these were rated as absent to minimal/diffuse in all explants except one, the 3-month explant (BAVG7), where they were rated as marked/focal in some sections. Most trichrome-stained sections showed elongated cells with circumferential orientation in both 3-month and 6-month explants (Fig. 5, A, D, and G). Fibrin immunohistochemical staining confirmed a thin thrombus on the luminal surface of the 3-month explant (Fig. 5B), which was also evident in the trichrome staining. Minimal fibrin staining was evident in one 6-month explant but was evident near the luminal surface of another (Fig. 5, E and H). The grafts had abundant collagen as seen in the trichrome-stained sections, but lacked mature elastic fibers based on Verhoeff–van Gieson stain at both time points except near the proximal anastomosis in two 6-month explanted grafts (fig. S5). All grafts at 3 and 6 months stained negative for calcification with von Kossa stain (Fig. 5, C, F, and I).

Fig. 5. Histology of explanted AVG.

Sections from midgraft region with trichrome stain, fibrin immunostaining, and calcification indicator von Kossa stain at (A to C) 3-month (n = 1) and (D to I) 6-month (n = 3) explantation. Luminal surface is marked with an asterisk (*). Scale bars, 200 μm.

Many cells populating the interstitium of the grafts stained positive for vimentin in all locations at 3 and 6 months (Fig. 6, A, D, and G). Most elongated cells in a distinct region adjacent to the luminal surface of many sections also stained positive for α-SMA at both time points (Fig. 6, B, E, and H). The thickness of this region, when clearly identifiable, was highly variable but commonly between 100 and 300 μm. There was sporadic smoothelin staining of the elongated cells along the base of this region (Fig. 6, E and H insets). At 3 months, CD31 staining was limited to the distal graft region and absent midgraft (Fig. 6C). After 6 months, CD31 staining indicated a complete endothelium in the proximal and distal regions in two of three patent grafts, as well as in the midgraft region in one case (BAVG1) (Fig. 6F) but not completely in the other two subjects (Fig. 6I). Consistent with the ultrasonography, no stenosis was evident at the distal anastomosis based on histological assessment, with minimal to mild (generally <250 μm in thickness) intimal thickening and fibrosis in the vein at the junction with the graft (fig. S6).

Fig. 6. Cellularity of explanted AVG.

Sections from midgraft region stained for the interstitial cell marker vimentin, the myofibroblast/smooth muscle cell marker α–smooth muscle actin (α-SMA), and the endothelial marker CD31 at (A to C) 3-month (n = 1) and (D to I) 6-month (n = 3) explantation. Insets show staining for the mature smooth muscle cell marker smoothelin. Luminal surface is marked with an asterisk (*). Scale bars in black, 200 μm; scale bars in red, 50 μm.

AVG immunocompatibility

The criterion for a positive intradermal sensitivity test was defined as measurable induration or erythema (>2 mm in diameter) at the injection site of the homogenized graft observed at 24, 48, or 72 hours after injection, with the saline injection control site scored in the same way for reaction comparison. The test was negative in all cases, including the animals that received a repeat graft. Likewise, immunoglobulin G, E, and M (IgG, IgE, and IgM) blood concentrations were not elevated from baseline values at any time point, including the animals after a regraft, except for a transient increase in the IgE concentration of BAVG2r (fig. S7). Proliferation of peripheral blood mononuclear cells was not different for cells cultured with the graft ex vivo for 7 days (fig. S8). Values for both assays averaged over all animals and all time points convey absence of an overt immune response (Fig. 7).

Fig. 7. Assessment of systemic immune response to implanted AVG.

Plots of peripheral blood mononuclear cell (PBMC; T cell) proliferation (A) (n = 19, including 3 regrafts) and IgE, IgG, and IgM blood concentrations (B to D) (n = 49, including 3 regrafts) averaged over all animals at all time points. T cell proliferation values are normalized to paired untreated control wells. Wells treated with the positive control concanavalin A (ConA) show a response, but no response is evident with cells cultured with replicate graft material or PTFE. IgE blood baseline concentrations showed substantial variance, but no increase in IgE, IgG, or IgM is indicated. Paired symbols (*, $, and #) in the plot indicate P < 0.05. See figs. S7 and S8 for values for each animal at all time points.


The results from this study indicated reasonable patency rates and favorable matrix remodeling with extensive recellularization and endothelialization of patent grafts after refinement of the surgical approach. There was no evidence of calcification or aneurysm, and only a minimal inflammatory response was noted among the eight explanted grafts. The one graft that did not clot among the first six implanted (before measures taken to mitigate venospasm) was the first graft implanted, and this animal (BAVG1) only received daily aspirin therapy. However, the target vein (brachial) in this animal was ligated for technical reasons, and implantation was redirected to the cephalic vein. We hypothesize that ligation of the brachial vein and the resultant “delayed” exposure of the cephalic vein may have afforded decreased venospasm as well as potentially more venous flow from the forearm.

One of the three animals that received a regraft (BAVG3r) experienced graft rupture; however, it may not represent an intrinsic graft failure. Although there were abundant eosinophils in this graft near the graft rupture location, there was no evidence of an immune response based on IgG, IgM, and IgE concentrations and peripheral blood mononuclear cell proliferation, increased concentrations of which would have been expected if sensitization to the initial graft occurred. The lack of an immune response causing degradation of the graft matrix in a graft failure only a few weeks after implantation points to a mechanical defect in the graft before implantation or damage to the graft associated with the implantation. The relatively small variance in the mechanical properties of replicate grafts before implantation (Fig. 1, E and F) suggests damage to the graft associated with the implantation as the cause of rupture (for example, this graft may not be as robust to repeated clamping as ePTFE).

The eosinophils presumably localized in the graft subsequent to failure resulting from suspected damage to the collagen and other matrix proteins during the ensuing inflammation (22). The response of the graft to trauma is a potential concern that could affect its suitability for clinical use; however, we found no evidence to suggest any alteration in graft integrity directly or indirectly attributable to episodic (surgical) or repeat (cannulation) graft traumas. This potential concern will only be resolved in a larger study.

An important finding summarized in Fig. 7 is the absence of a systemic immune response in any animal and at any time point (noting that BAVG2r exhibited a transient increase in IgE blood concentration; fig. S7). An immune response could have occurred because of incomplete removal of antigens from the hDFs used to grow the tissue tubes before decellularization, remnants of the bovine fibrin used as the sacrificial tissue growth scaffold, or residual bovine serum proteins from the in vitro culture process. In previous studies, we have shown that this decellularization method using sequential immersion in SDS, Triton X-100, and deoxyribonuclease (DNase) reduces the DNA content by 93%, with β-actin and β2-microglobulin being undetectable by Western blot (14). In addition, although not part of the original study design, the general lack of a systemic or local immune response in the three regrafts suggests that no sensitization occurred in response to the initial grafts. The apparent IgE response in BAVG2r is consistent with a type I reaction (immediate hypersensitivity) to an allergen associated with IgE-mediated release of histamine and other inflammatory mediators from mast cells and basophils. These results showing absence of an overt immune response with adequate decellularization, although not complete decellularization, are consistent with the seminal studies reported by Niklason and colleagues (7, 8), although the scaffold, cells, and detergents used in those studies differed. Moreover, the grafts that clotted occlusively and were left in place for the three regraft subjects appeared to be largely resorbed by hemosiderin-laden macrophages in the graft remnant after its extreme constriction without adverse reaction of proximal tissue by the time of the 6-month regraft explant (nearly a year after the initial graft implantation), indicating that a clotted graft left in place would not necessarily pose a safety risk.

There was an increase in graft diameter over the course of the study, but there was no thinning of the graft wall indicative of aneurysm formation, nor was there reduction in burst pressure (including the burst pressure of the graft segment that had been cannulated multiple times). Rather, an increase in graft thickness and diameter occurred uniformly across the graft (or large regions thereof), suggesting an adaptation of the graft to the hemodynamic conditions. Although no statistical change in graft diameter occurred until 6 months after implantation, how the preceding trend of increasing diameter-related to recellularization (whether a passive mechanical response of the matrix or an active response of the tissue upon recellularization) could not be conclusively determined, because only a single graft was explanted at the 3-month time point. However, extensive recellularization had occurred by 3 months (Figs. 4F and 5A), suggesting that the diameter increase was a cell-mediated tissue response. The 33% increase in graft diameter (Fig. 3B) and 65% increase in graft thickness (Fig. 4E) at 6 months are consistent with many of the cells present staining for markers indicative of matrix-producing fibroblasts/myofibroblasts/smooth muscle cells (Fig. 6, D, E, G, and H) adapting to the local mechanical and chemical signals in the graft after implantation.

Although the vimentin and/or α-SMA–positive interstitial cells were apparently responsible for remodeling of the graft, there was no evidence of intimal hyperplasia at the distal graft anastomoses, which commonly occurs at the distal end of ePTFE grafts (23). In the landmark study by Dahl et al. (7), their tissue-engineered graft also showed reduced intimal hyperplasia when compared to ePTFE grafts, potentially due to more physiological compliance (23).

To limit the number of anesthetic events for each animal, cannulation with a dialysis needle was performed at defined monthly intervals. It remains to be seen whether the graft can tolerate the triweekly cannulation required for hemodialysis. The maintenance of pre-implant burst pressure of the graft segment that had been cannulated three times before the 6-month explantation indicates that cannulation sites had not created focal defects but had been “healed” by the vimentin/α-SMA–positive cells repopulating the graft interstitium.

A concerning outcome in this study was the occlusive clot that occurred in one animal that received a regraft and was maintained only on aspirin therapy for the duration (BAVG4r). Thrombosis occurred sometime after the 3-month cannulation, although the graft was patent the following day based on manual assessment. Another animal (BAVG2r) that received a regraft from the same production batch was maintained only on aspirin therapy for the duration and had a patent graft at the 6-month explantation. In the single animal explanted at 3 months (BAVG7), there was substantial endothelialization of the graft lumen toward the anastomoses. Endothelialization appeared to progress over the first 6 months; it was complete throughout the graft in one animal (BAVG1), but was incomplete in the mid region in the other two animals with patent grafts (BAVG2r and BAVG8). Thus, it may be that BAVG4 was hypercoaguable relative to the other animals or it yielded slower or lesser endothelialization. Thrombus and endothelial cell coverage of the luminal surface were inversely correlated, suggesting that the spontaneously forming endothelium was in a non-thrombogenic state. There is evidence based on the fibrin staining that thrombus deposited onto the luminal surface became organized into a distinct layer with subsequent endothelialization; however, this layer was not needed for maintenance of burst strength or graft recellularization, based on the 3-month explant. Both layers had circumferentially aligned cells, indicating that the circumferential alignment of the implanted matrix provided an effective template for remodeling of the graft into a circumferentially aligned tissue.

In a directly comparable tissue-engineered AVG study in baboons reported by Dahl et al. (7), the 3- and 6-month primary patency rates were 83% (5 of 6) and 100% (3 of 3), respectively, with no occurrences of rapid occlusive clots reported. We removed such occurrences in determining our 3- and 6-month primary patency rates of (5 of 6) 83% and (3 of 5) 60%, respectively, because they were considered to be associated with model development. A notable difference is that there was more robust α-SMA staining in the midgraft region in this study, as well as smoothelin staining indicative of mature smooth muscle cells (24, 25) along the base of the distinct luminal layer noted above. The grafts used in these two studies, although both grown in vitro from cells on a scaffold and decellularized before implantation as allografts, differed in many ways: scaffold (biopolymer fibrin versus synthetic degradable polymer PGA), cells (dermal fibroblasts versus vascular smooth muscle cells), detergents (SDS/Triton X-100 versus CHAPS or SDS), resulting matrix structure (circumferentially aligned versus isotropic), among others. It would thus be expected that differences in the recellularization and graft remodeling would occur. These differences could also explain why we experienced one relatively late (between 3 and 6 months) occlusive clot, whereas Dahl et al. did not (7). However, both these studies had a small number of subjects, and thrombosis was the major cause for intervention and loss of primary and primary-assisted patency in the follow-on phase 2 study of their graft at a rate of 1.89 per patient-year (8).

Our 6-month primary (unassisted) patency rate of 60% compares favorably to the rates of 70 and 46% for the two regional cohorts reported in the cited phase 2 study (8) and to the rates of 46.5 and 36.3% reported for groups that did and did not receive dipyridamole/aspirin therapy in a previous study (26). Antiplatelet therapy is not the standard of care for ePTFE AVG. It was used in this study because it was used in the directly comparable study by Dahl et al. (7) and the follow-on phase 2 study for almost all patients (8). We used it in our study for the same reason Dahl et al. stated (8): Our AVG is also an exposed collagenous matrix that is inherently thrombogenic until a normal endothelium spontaneously forms or passivation occurs. Adoption of antiplatelet therapy for this AVG, at least transiently, would be prerequisite for its clinical use, as is the case for the bovine products currently on the market in the United States.

All grafts fabricated from this study were made from a single isolation of neonatal hDFs. From a single aliquot of passage one (p1) cells obtained from the vendor, more than 10,000 of these grafts can be created from the resulting 0.6 trillion p7 cells. Other neonatal hDF isolations have been used in our previous studies, where high burst strength was also achieved (12); thus, although we have not assessed whether all the results reported for this study are unique to the cell isolation used, the ability to achieve high burst strength in a small-diameter tube of this extracellular matrix is not unique to this isolation.

In conclusion, this preclinical study of a completely biological, acellular off-the-shelf AVG exhibited encouraging patency rates once the model was sufficiently developed to mitigate rapid thrombosis. The grafts exhibited favorable biological response, including extensive recellularization by tissue cells and a developing endothelium on the luminal surface, and no calcifications, loss of burst strength, outflow stenosis, or overt immune response. These findings will help facilitate clinical testing of these grafts.


Cell culture

hDFs (Lonza) were maintained in a 50/50 mixture of Dulbecco’s modified Eagle’s medium and Ham’s F12 cell culture medium (DMEM/F12; Cellgro) supplemented with 15% fetal bovine serum (FBS; Thermo Fisher Scientific), penicillin (100 U/ml), and streptomycin (100 U/ml) (Gibco) at 37°C. Cells were expanded to p7 and used in the fabrication of the vascular grafts.

AVG fabrication

hDF-seeded fibrin gel was formed by adding bovine thrombin (Sigma-Aldrich) and calcium chloride in 20 mM Hepes-buffered saline to a suspension of hDF in bovine fibrinogen (Sigma-Aldrich). The final component concentrations of the cell suspension were as follows: fibrinogen (4 mg/ml), thrombin (0.8 U/ml), 5.0 mM Ca++, and 1 million cells/ml. Cell suspensions were mixed and injected into a tubular mold containing a glass mandrel pretreated with a 5% Pluronic F-127 solution. Subsequently, grafts were cultured in DMEM supplemented with 10% FBS (HyClone), penicillin (100 U/ml), streptomycin (100 μg/ml) (Gibco), insulin (2 μg/ml) (Sigma-Aldrich), and ascorbic acid (50 μg/ml) (Sigma-Aldrich). After an initial 2-week maturation period, the grafts were transferred to a custom pulsed flow-stretch bioreactor for an additional 3-week maturation period as previously described (12). The final dimensions of each graft were as follows: length, 12 to 15 cm; inner diameter, 4 mm; and thickness, ~500 μm.

AVG decellularization

After bioreactor culture, grafts were rinsed in PBS and incubated on an orbital shaker for 6 hours with 1% SDS (Sigma-Aldrich) in distilled water. The SDS solution was changed three times. The grafts were rinsed in PBS and incubated with 1% Triton X-100 (Sigma-Aldrich) in distilled water for 30 min, extensively washed with PBS for 7 days, and then incubated in DNase enzyme (Worthington Biochemical) in DMEM supplemented with 10% FBS overnight. Decellularized grafts in this study were stored for up to 12 months at 4°C in PBS until implantation. Four batches of grafts (3 to 5 grafts per batch) were used for the initial and repeat implants. After decellularization, all grafts were analyzed for tensile mechanical and compositional properties from an end piece of 10 mm. In addition, one graft from each batch was used for burst pressure and suture retention testing along with mechanical and compositional properties.

AVG burst pressure, suture retention, and tensile mechanical properties

For burst pressure testing, AVGs (n = 6) were mounted in a system designed for pressurizing individual grafts to failure. Each graft was cannulated and burst strength measured as described previously and per ISO 7198 (12). In addition, 2 cm length of graft (n = 6) was used for suture retention testing by placing a 7-0 prolene suture 2 mm from the end of the graft and pulled to failure on an Instron tensile testing system (Instron Testing Systems) with 5 N load cell attached. Each graft (n = 6) was measured at three locations equally spaced around the circumference. In addition, a circumferential strip from all grafts (n = 17) was pulled to failure with UTS and modulus calculated from stress-stain data (12).

AVG compositional properties

The DNA and SDS content of decellularized graft (n = 17) were quantitated. The sample volume was calculated using the measured length, width, and thickness of the tubular section determined with a caliper. The cell content was quantified with a modified Hoechst assay for DNA (27), and residual SDS was quantified using methylene blue (28). The DNA and SDS concentrations are reported per unit volume of matrix (wet basis).


All protocols and procedures were approved by the University of Minnesota Institutional Animal Care and Use Committee, conducted in compliance with the Animal Welfare Act, and animals were housed and cared for according to the standards detailed in the Guide for the Care and Use of Laboratory Animals. The cohort included eight purpose-bred male olive baboons (Papio anubis) aged between 8.8 and 12.6 years (median, 10.5 years) and weighed between 27.8 and 34.1 kg (median, 29.8 kg). They were singly housed in close proximity with compatible conspecifics, maintaining visual, auditory, and olfactory contact at all times. Baboons had free access to water and were fed biscuits (Harlan Primate Diet 2055C, Harlan Teklad) based on body weight. Their diet was enriched daily with fresh fruits, vegetables, grains, beans, nuts, and a multivitamin preparation. Semiannual veterinary physical examinations were performed in all animals. Animals participated in an environmental enrichment program designed to encourage sensory engagement, enhance foraging behavior and novelty seeking, promote mental stimulation, increase exploration and play and activity levels, and strengthen social behaviors, together providing opportunities for animals to increase time budget spent on species-typical behaviors. Baboons were trained to cooperate in medical procedures including hand feeding and drinking, shifting into transport cages for sedation and limb presentation for graft examination (29).

Implant procedures

Grafts were implanted between the axillary artery and cephalic or brachial vein, with repeat AVG construction in contralateral arms in three animals, as summarized in fig. S1. Briefly, the skin over the axillary artery and brachial or cephalic vein in the upper arm was incised and the artery and vein were dissected free from the surrounding tissues. The axillary artery was clamped proximally and distally and an arteriotomy was created. The graft was anastomosed to the artery in an end-to-side manner using a running 7-0 prolene suture. After completion of the anastomosis, the graft was back-bled, retrograde-flushed with heparinized saline (10 U/ml), and clamped distal to the anastomosis. Antegrade flow was reestablished into the brachial artery. The cephalic or brachial vein at the lower upper arm was chosen for outflow. The vein was exposed, and a subcutaneous tunnel was created through which the graft was placed. The inflow clamp was moved to allow for graft expansion and marked for length, and then the graft was backflushed with heparinized saline and reclamped just distal to the inflow anastomosis. The vein was controlled proximally and distally, a venotomy was created, the graft was trimmed, and an anastomosis was created in an end-to-side manner using running 7-0 prolene suture. Flow was established by sequentially removing the vein clamps and the inflow clamp, after which the venous sutures were tied. Vasospasm was treated with topical lidocaine or papaverine and warm saline. After assuring a thrill in the outflow vein, a palpable radial pulse, and hemostasis, the wound was closed in layers using 5-0 Monocryl (Ethicon) and topical skin adhesive (Ethicon). Analgesia [buprenorphine, 0.01 to 0.03 mg/kg intramuscularly (im) and ketoprofen, 1 mg/kg im] was administered just before case initiation and continued postoperatively for 24 hours (buprenorphine) and 72 hours (ketoprofen). Figure S1 illustrates the anticoagulation regimen used for each animal, with 324 mg/day aspirin starting on day 7 and continued for the duration and (in some subjects) 150 or 300 mg/day (preoperatively) and 75 mg/day (postoperatively) clopidogrel administered as shown.

Graft cannulation

The AVG was accessed with a standard hemodialysis needle (16 gauge × 1″) using a rope ladder technique at 1, 2, 3, and 6 months. Clinical monitoring of thrill was performed before needle placement and after access. The skin at the intended insertion site was prepped with ChloraPrep before insertion, and needle placement was verified with blood flow and flush patency. Immediately after graft access, hemostasis was established with direct pressure for 15 min. Hemostasis was then evaluated, and if residual bleeding was observed, additional direct pressure was applied.

Ultrasound and angiography evaluation of graft

Grafts were evaluated postoperatively for flow using ultrasound at months 1, 2, 3 and 6. At each time point, a Siemens ultrasound instrument (ACUSON Sequoia C512) with a 14-MHz transducer was used to measure diameter and flow in the midgraft and distal graft in addition to venous flow. At the time of euthanasia, angiography was also performed and images were acquired using a Philips C-arm (BV Pulsera) with contrast Omnipaque (iohexol) injected through the dialysis cannula inserted midway into the graft.

Explant analysis

Grafts were examined and imaged during explant and then cut into five segments. The proximal anastomosis (~2 cm of graft), distal anastomosis (~2 cm of graft), and midsection (~2 cm) were placed in formalin for histological evaluation. In addition, proximal (not cannulated, 3 to 5 cm) and distal (cannulated, 3 to 5 cm) sections of grafts were evaluated for burst pressure.

Explant histology, histochemistry, and immunohistochemistry

Samples were placed in 10% neutral buffered formalin solution and fixed at room temperature for 12 hours. After fixation, samples were transferred to 70% ethanol solution until processed for routine paraffin embedding. Samples were then sectioned 4 μm thick and stained with H&E, Masson’s trichrome, Verhoef–van Gieson, and von Kossa stains with standard protocols. For immunohistochemical staining, sections were cut at 4 μm, deparaffinized, and rehydrated. For SMA, CD31, and vimentin, sections were incubated with 3% hydrogen peroxide to quench endogenous peroxidase activity followed by 15 min in serum-free protein block (Dako). Sections were then subjected to appropriate antigen retrieval methods and incubated with the primary antibody at room temperature for 60 min (table S1). Detection was performed with EnVision+ Kits (Dako). For fibrin and smoothelin, sections were subjected to antigen retrieval followed by endogenous peroxidase quench using glucose oxidase (30) (fibrin) or 3% hydrogen peroxide (smoothelin). Sections were then blocked with normal serum supplied with ImmPRESS HRP Reagent Kit (Vector Labs) followed by primary antibody incubation, both for 60 min at room temperature. ImmPRESS HRP detection polymer was used according to Vector Labs instructions followed by color development with DAB 2 Component with Stabilizer (Biolegend).

Blood immunoglobulin concentration

Ig blood concentrations were measured using enzyme-linked immunosorbent assay (ELISA) kits according to the supplier (Life Diagnostics). Briefly, duplicate 100-μl aliquots of standards from the supplier, previously prepared test samples, and the diluted samples were dispensed into microtiter wells and incubated on an orbital microplate shaker at 100 to 150 rpm at room temperature for 45 min. The contents of the wells were aspirated and plates were blotted onto paper towels. The wells were quickly washed five times with 300 μl of 1× wash solution, striking the wells sharply onto absorbent paper in between each wash to remove all residual wash buffer. Enzyme conjugate reagent (100 μl) was added into each well, and plates were incubated on an orbital microplate shaker at 100 to 150 rpm at room temperature for 45 min. After repeating the washing step described above, 100 μl of 3,3′,5,5′-tetramethylbenzidine reagent was dispensed into each well and gently mixed on an orbital microplate shaker at 100 to 150 rpm at room temperature for 20 min. The reaction was stopped by adding 100 μl of stop solution to each well and by gentle mixing until all the blue color changed to yellow. Optical density at 450 nm was read for each well within 5 min. A standard curve was created from the standards, and sample values were determined from the standard curve.

T cell proliferation

PBMCs were isolated using Ficoll gradient density centrifugation and then cryopreserved. Cryopreserved cells were thawed and plated overnight. Cells were counted and suspended at 2.0 × 106 cells/ml in AIM V culture medium (Gibco). ConA medium was made by diluting ConA (5 mg/ml) (Sigma-Aldrich) into AIMV to a concentration of 0.01 mg/ml. Cells (2 × 105) were incubated with and without 5 mm × 5 mm sections of sterile-stored replicates of implant graft and Gore-Tex graft (ePTFE) in individual wells of a 96-well plate (in triplicate). ConA solution (100 μl) was added on day 3 to positive control samples and 5-bromo-2′-deoxyuridine (BrdU) was added on day 6 of the cell incubation to all wells. BrdU ELISA analysis on day 7 began by gently pipetting PBMCs off the test materials using medium in the well. After the graft and PTFE samples were removed, the plates were centrifuged, and a BrdU ELISA (Roche) procedure was performed with absorbance at 450 nm measured. A stimulation index was calculated for each well as Embedded Image, where the denominator was obtained for identically treated cells as used in the test sample wells but without being cultured with a test sample. All groups for each animal were prepared at the same time and run in the same plate.

Intradermal graft injection sensitivity

Intradermal injection was performed at day 0 and day 28 (±4 days). Graft material was homogenized in phosphate buffer solution at a concentration of protein (0.25 mg/ml) and injected intradermally. A 0.1-ml solution of the sterile homogenized graft and control phosphate buffer solution was injected at adjacent sites and monitored over 72 hours for erythema.

Statistical analysis

For all experiments, statistical significance of differences between groups was determined using Student’s t test for two treatments and one-way analysis of variance (ANOVA) for more than two treatments with the Tukey post hoc test in GraphPad Prism software for Windows. Any reference to a difference in Results and Discussion implies statistical significance at P < 0.05. In all cases, where the difference was significant, paired symbols are used to indicate the difference and these are explained in the figure legends. Error bars in the plots indicate standard deviations.


Fig. S1. Study design of AVG implantation.

Fig. S2. Kaplan-Meier diagram comparing graft patency for cephalic (n = 6) versus brachial (n = 5) outflow veins.

Fig. S3. Plot of graft diameters measured by ultrasound for all grafts contributing to the patency rates (n = 6).

Fig. S4. Full-size angiogram of all 6-month explants (n = 3).

Fig. S5. Verhoeff–van Gieson staining of explants.

Fig. S6. Histological section of graft-vein junction at distal anastomosis of BAVG1.

Fig. S7. IgE, IgG, and IgM blood concentrations measured for all animals and all time points.

Fig. S8. T cell proliferation measured for all animals and all time points.

Table S1. Immunohistochemistry antibodies and conditions.


  1. Acknowledgments: We acknowledge L. Mutch for coordinating scientific conduct of the animal study, N. Ferguson and B. Norris for technical assistance, and the Preclinical Research team and Research Animal Resources veterinarians for providing outstanding animal care. We also acknowledge the Mayo Clinic for their generosity in gifting both animals and customized enclosures. Xeno Diagnostics LLC (Indianapolis, IN) optimized and performed the Ig blood concentration and PBMC proliferative assays as described. Funding: Funding was provided by the John and Nancy Lindahl Children’s Heart Research Innovators Fund and the University of Minnesota Center for Translational Medicine (to R.T.T.). Author contributions: Z.H.S. designed and performed all experiments, M.L.G. designed and performed all animal experiments, T.B.D. designed and performed all animal surgeries, T.O. performed all pathology/histopathology, S.L.J. developed and performed fibrin/smoothelin immunostaining and DNA/SDS assays, R.J.S. designed the animal experiments, and R.T.T. designed all experiments. All authors contributed to writing the paper. Competing interests: Z.H.S. and R.T.T. are co-inventors on pending U.S. patent 13/771,676, “Decellularized biologically-engineered tubular grafts,” submitted by the University of Minnesota. All other authors declare that they have no competing interests. Data and materials availability: All relevant data are available in the manuscript. Requests for materials should be addressed to R.T.T.

Stay Connected to Science Translational Medicine

Navigate This Article