Therapeutic Angiogenesis by Implantation of a Capillary Structure Constituted of Human Adipose Tissue Microvascular Endothelial Cells
Objective— We previously reported a novel technology for the engineering of a capillary network using an optical lithographic technique. To apply this technology to the therapy of ischemic diseases, we tested human omental microvascular endothelial cells (HOMECs) as an autologous cell source and decellularized human amniotic membranes (DC-AMs) as a pathogen-free and low immunogenic transplantation scaffold.
Methods and Results— Human umbilical vein endothelial cells were aligned on a patterned glass substrate and formed a capillary structure when transferred onto an amniotic membrane (AM). In contrast, HOMECs were scattered and did not form a capillary structure on AMs. Treatment of HOMECs with sphingosine 1-phosphate (S1P) inhibited HOMEC migration and enabled HOMEC formation of a capillary structure on AMs. Using quantitative RT-PCR and Western blot analyses, we demonstrated that the main S1P receptor in HOMECs is S1P2, which is lacking in human umbilical vein endothelial cells, and that inhibition of cell migration by S1P is mediated through an S1P2–Rho–Rho-associated kinase signaling pathway. Implantation of capillaries engineered on DC-AMs into a hindlimb ischemic nude mouse model significantly increased blood perfusion compared with controls.
Conclusion— A capillary network consisting of HOMECs on DC-AMs can be engineered ex vivo using printing technology and S1P treatment. This method for regeneration of a capillary network may have therapeutic potential for ischemic diseases.
Ischemic diseases, including peripheral arterial diseases (PADs), myocardial infarction, and cerebral infarction, threaten our health and our lives. When medication becomes ineffective, patients need to undergo percutaneous transluminal angioplasty or artery bypass grafting. Recently, therapeutic angiogenesis has been developed for no-option patients. The target of percutaneous transluminal angioplasty and artery bypass grafting is blood vessels with a diameter larger than 1 mm, whereas that of angiogenic therapy is capillaries and microvessels with a diameter on the micrometer order.
See accompanying article on page 1277
Therapeutic angiogenesis consists of 2 strategies: (1) growth factor therapy and (2) cell therapy. It has been applied to ischemic diseases, including PADs, such as arteriosclerosis obliterans and Burger’s disease, as well as myocardial infarction and cerebral infarction.1,2 The aim of growth factor therapy is to induce angiogenesis by gene therapy or by protein administration of angiogenic factors. Although early nonblinded and uncontrolled studies showed promising outcomes, some later double-blinded, randomized, and controlled trials have been less effective than anticipated.3 More recently, cell therapy has been developed as a second angiogenic therapy. Initially, peripheral blood-derived or bone marrow–derived mononuclear cells (MNCs) were transplanted as endothelial progenitor cells, which should be incorporated into newly formed vessels. However, an increasing number of reports indicate that transplanted MNCs augment angiogenesis by functioning as angiogenic factor–producing cells,4,5 progenitors of mural cells,6,7 or cells that stimulate muscle cells to produce angiogenic factors.8 Unselected MNCs, rather than MNCs enriched for CD34, CD133, or vascular endothelial growth factor receptors, have been preferentially used for cell therapy. The common reasons for their selection are convenience and the absence of a consensus on the most effective subpopulation. Moreover, the MNC population is a highly heterogeneous group containing multipotent adult progenitor cells, which could differentiate into hematopoietic cells as well as into nonvascular cells such as epithelial cells, muscle, and bone/cartilage.9 Furthermore, a case of human gastric cancer originating from bone marrow–derived cells has been reported.10
We therefore chose endothelial cells (ECs), which are mature and genetically stable, as the substrate cells for an engineered vascular endothelium. Moreover, we implanted the capillary networks engineered ex vivo by the use of printing technology, instead of by single-cell injection.11 We previously demonstrated that such implanted capillary networks functioned in vivo11 and improved blood perfusion in a mouse model of hindlimb ischemia.12 For clinical application, we tested human omental microvascular ECs (HOMECs) as an antilogous cell source and decellularized amniotic membrane as a pathogen-free and low immunogenic implantation scaffold.
An expanded Materials and Methods section is described in the supplemental materials, available online at http://atvb.ahajournals.org. All patients gave written informed consent according to the Declaration of Helsinki, and all protocols were approved by the Tokyo Medical and Dental University Bioethical and Animal Care Ethics Committees.
HOMECs were isolated from omental tissue using a modification of Kern’s method and were cultured as described elsewhere.13,14 Human omental tissue was obtained from patients undergoing abdominal surgery. HOMECs were cultured in MCDB131 supplemented with 10% fetal bovine serum (10%-MCDB131). Human umbilical vein ECs (HUVECs) were purchased from Cambrex (Walkersville, Md) and cultured on type I collagen–coated dishes (Iwaki, Tokyo, Japan) in endothelial basal medium-2 supplemented with 5% fetal bovine serum and EGM SingleQuots (Cambrex).
Immunocytochemistry of CD31, endothelial nitric oxide synthase, and VE-cadherin in HOMECs was performed using mouse monoclonal antibodies against CD31, endothelial nitric oxide synthase, and VE-cadherin or control mouse IgG as primary antibodies and Alexa-488 conjugated goat anti-mouse IgG as a secondary antibody. Nuclei were counterstained with TO-PRO-3. Images were photographed by a confocal microscope system (LSM 510, Zeiss, Oberkochen, Germany). Positive cells were counted in a low-power field (278 to ≈352 cells).
Preparation of Amniotic Membrane
Human amniotic membrane (AM) was prepared from human placentas as described elsewhere.15 Human placentas were obtained from Caesarean sections of women who were seronegative for human immunodeficiency virus, human hepatitis virus types B and C, and syphilis. Decellularization of the AM was conducted as described elsewhere.16
Ex Vivo Formation of Capillary Networks by ECs
The patterned glass substrate was prepared by coating glass with polyethylene glycol to create a nonadhesive area and by photomasking to create an adhesive area as described previously.11,12 Cells fluorescently labeled with PKH26GL (Sigma, St. Louis, Mo) were seeded on the patterned glass substrate at a density of 4×104 cells/cm2 and were cultured for 16 hours in 5%-MCDB131. The patterned cells were then transferred onto the AM or decellularized human AM (DC-AM) and incubated for a further 24 hours. PKH26GL-labeled cells on the substrate or on the AM or DC-AM were observed under a fluorescent microscope (Biozero BZ-8000, Keyence, Osaka, Japan).
The cell migration assay was performed using a modified Boyden chamber.17 HOMECs were placed in the upper insert (5×104 cells/well), and 100 nmol/L sphingosine 1-phosphate (S1P) or 10 μmol/L of the Rho-associated kinase (ROCK) inhibitor Y-27632 (Biomol, Plymouth Meeting, Penn) in 0.1%-MCDB131 was added into the lower chamber of 24-well transwell chambers with an 8-μm pore size filter (Corning, Acton, Mass). Before the migration assay, the Y-27632-treated cell groups were incubated in 0.1%-MCDB131 for 2 hours and then treated with or without 10 μmol/L Y-27632 for a further 30 minutes. After 7 hours of incubation, the cells that had migrated were counted in randomly selected high-power fields (HPFs). In some experiments, cells transfected with anti-S1P2 small interfering RNA (siRNA) or scrambled RNA were used.
Quantitative Reverse Transcription Polymerase Chain Reaction
The mRNA of S1P1, S1P2, S1P3, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was measured by quantitative RT-PCR using a First-strand cDNA synthesis kit for RT-PCR, the SYBR Green I Premix Ex Taq, and a Light Cycler System (Roche, Mannheim, Germany). The level of each mRNA was measured as the intensity of a fluorescent signal and normalized to the signal for GAPDH mRNA.
Western blotting of S1P1, S1P2, S1P3 and β-actin in HOMECs and HUVECs was performed using the respective polyclonal antibody and horseradish peroxidase–conjugated protein G antibody. The signals were detected using ECL Plus Western blotting detection reagents and were visualized using LAS1000 (Fuji, Tokyo, Japan). Equal loading was confirmed using an anti-β-actin antibody blot.
Transfection of siRNA for S1P2
Oligonucleotide duplexes, designed to block S1P2 expression (siRNA), or as scrambled RNA, were synthesized as previously described.18 Transfection of siRNA or scrambled RNA into HOMECs was performed using the Lipofectamine RNAiMAX reagent according to the manufacturer’s instructions. The cells were incubated for a further 22 to 24 hours and were then analyzed by quantitative RT-PCR and Western blotting and in cell migration assays.
Inhibition of Endogenous S1P2 Signaling by a ROCK Inhibitor
HOMECs were starved for 2 hours, pretreated with or without 10 μmol/L Y-27632 for 30 minutes, and then treated with or without 500 nmol/L S1P for a further 30 minutes. Immunocytochemical analysis of vinculin and F-actin was performed.
Implantation of the Capillary Structure Into Nude Mice
Ex vivo formation of capillary networks by PKH26GL-labeled HOMECs was performed as described above except that 500 nmol/L S1P was added during incubations on the glass substrate as well as on an AM. The capillaries engineered on AMs were implanted subcutaneously into the abdomen of 8-week-old BALB/c nude mice. A soft tissue wound was created by making a U-shaped incision in the skin. Host blood vessels under the dermis were scraped off, and the AM with the capillary structures were sutured there. Tissue sections from the abdominal tissue were analyzed by hematoxylin staining and by fluorescent microscopy at postoperative day 6.
Mouse Hindlimb Ischemia Model
Ischemia was induced by proximal femoral artery ligation, including the superficial and deep branch as well as the distal saphenous artery.12 Thereafter, DC-AM with engineered capillaries or DC-AM alone was placed onto the site of the excised arteries. Blood flow was measured using a laser Doppler blood flowmeter (Laser Doppler Imager; Moor Instruments Ltd, Devon, United Kingdom), and the blood flow ratio was expressed as ischemic limb/nonischemic limb. Tissue sections implanted with PKH26GL-labeled HOMECs on DC-AM were treated with Hoechst 33342 for nuclear staining and analyzed by confocal microscopy and fluorescent microscopy at postoperative day 6.
The statistical significance of differences was evaluated by ANOVA, with probability values calculated by the Student t test for in vitro assays and the Mann-Whitney test for an in vivo assay. A value of P<0.05 was accepted as statistically significant. Data are presented as the mean±1 SD or SE (*P<0.05, **P<0.01).
Characterization of HOMECs
HOMECs isolated from human omentum showed a cobblestone appearance typical of ECs and were identified as ECs by immunocytochemistry of CD31, endothelial nitric oxide synthase, and VE-cadherin (Figure 1). Positive cell ratios of these markers were 90%, 91%, and 94%, respectively.
Failure of HOMECs to Form a Capillary Structure on AMs
HOMECs were tested as an autologous cell source for the clinical application of engineered capillaries to ischemic diseases. We previously reported that HUVECs were aligned on a patterned glass substrate manufactured by the same optical lithographic techniques used in this study and that they formed a capillary structure when transferred onto AMs.11 Although HOMECs could be cultured on a patterned substrate and transferred onto AMs, they were scattered on the AMs and failed to form capillaries (Figure 2A).
Effect of S1P on Capillary Formation by HOMECs on AMs
To maintain the alignment of HOMECs on AMs, we tested whether the addition of S1P reinforces cell-cell adhesion, because S1P is known to enhance the localization of VE-cadherin at endothelial junctions in HUVECs.19 As expected, S1P treatment prevented cell scattering and resulted in alignment of the HOMECs on the AMs. However, this effect was not due to enhancement of VE-cadherin expression at cell-cell junctions or to the induction of VE-cadherin (data not shown). In the presence of S1P, HOMECs formed a tube-like structure on AM as early as 2 hours, whereas they started to migrate without S1P (Figure 2B). We therefore speculated that S1P might inhibit HOMEC migration, thereby inhibiting cell scattering and aligning the cells on the AM. Although S1P is known to induce EC migration,19–23 S1P significantly inhibited HOMEC migration, even though it did induce the migration of HUVECs as previously reported20–22 (Figure 2C).
Different Expression of S1P Receptors in HOMECs and HUVECs
As shown in Figure 2, the response of HOMECs and HUVECs to S1P differed, which prompted us to analyze the expression of S1P receptors (S1PRs) in both cell types at the level of both mRNA (Figure 3A through 3C) and protein (Figure 3D). We found that HUVECs expressed predominantly S1P1 and, at a lower level, S1P3, but they did not express S1P2, which is consistent with previous reports.19,21 Interestingly, S1P2 was the predominant receptor in HOMECs. The expression of S1P1 and S1P3 in HOMECs was lower than that of S1P2. The expression of S1P3 in HOMECs was similar to that in HUVECs.
A Role for S1P2 in the Inhibitory Effect of S1P on HOMEC Migration
To clarify whether S1P2 is involved in the inhibition of HOMEC migration by S1P, we performed a migration assay using HOMECs transfected with anti-S1P2 siRNA (siS1P2). As previously reported,18 this siRNA efficiently knocked down S1P2 expression at the level of both mRNA (Figure 4A) and protein (Figure 4B). Transfection of siS1P2 into HOMECs abrogated the decrease in migration induced by S1P (Figure 4C). These results indicate that the inhibition of HOMEC migration by S1P might be mediated through a HOMEC-specific S1PR, S1P2.
The Downstream Signaling Pathway of S1P2 in HOMECs
S1P2 has been shown to activate Rho and to inhibit Rac activity, resulting in increased contraction, stress fiber formation, and cell rounding and decreased cell motility (see reviews 24, 25). ROCK is a downstream kinase in the Rho pathway that regulates the actin cytoskeleton.24,25 Treatment with Y-27632, a ROCK inhibitor, abrogated the inhibitory activity of S1P on HOMEC migration (Figure 5A), indicating that ROCK is required for S1P-induced inhibition of HOMEC migration. To investigate the effect of S1P on cytoskeletal and adhesion components that function in migration, we performed immunostaining of F-actin and vinculin. S1P induced cortical actin ring formation and hindered focal adhesion and lamellipodia formation (Figure 5B versus 5C), whereas Y-27632 abrogated the effects of SIP (Figure 5C versus 5D and Figure 5E). These data suggest that the inhibitory effect of S1P on HOMEC migration may be mediated through an S1P2-Rho-ROCK signaling pathway.
Implanted Capillary Structures Function as Blood Vessels In Vivo
We next applied the capillaries engineered using S1P-treated HOMECs as described above to in vivo. S1P made HOMECs form a tube-like structure on AM as early as 2 hours, resulting in curtailment of the operation period. However, the effect of S1P did not last for 24 hours (Figure 2B). Therefore, we treated HOMECs on AM for 2 hours and implanted them into nude mice. To evaluate implanted capillaries, we performed histological analysis 6 days after implantation. As we have previously shown using capillaries constituted of bovine carotid arterial ECs,11 a capillary structure engineered using PKH26-labeled HOMECs appeared to contain blood-like cells (Figure 6A). Moreover, some of the capillaries kept an open lumen in ischemic hindlimbs after 6 days of implantation (Figure 6B and 6C). These results suggest that some of the implanted capillaries survived and remained open at least for 6 days after surgery. In our previous report,12 we detected intravital blood flow through the implanted capillaries of bovine carotid arterial ECs and into host blood vessels by in vivo imaging analysis. Therefore, we expected that the implanted capillaries could function as blood vessels.
Implantation of Engineered Capillary Structures Increases Blood Flow in Hindlimb Ischemia
To investigate whether the engineered capillary network could be used for the treatment of ischemic disease, a nude mouse model of hindlimb ischemia was used. Furthermore, for potential future clinical application we used DC-AMs as a pathogen-free and low immunogenic transplantation scaffold. The morphology of the capillaries on DC-AMs was not different from that on conventional, untreated AM (data not shown). Implantation of capillary structures on DC-AMs significantly increased blood perfusion compared with control from as early as 5 days and up until 10 days after implantation (Figure 6D and 6E).
We have previously demonstrated that capillary structures engineered ex vivo using printing technology11 could improve blood perfusion in a mouse hindlimb ischemia model.12 Implantation of capillary structures constituted of bovine carotid artery ECs on AM increased blood perfusion significantly earlier than control (AM only). In contrast, injection of the same number of single cells as used for implanted capillaries had no effect on perfusion in the ischemic limb compared with AM only.12 These results indicate the superiority of the implantation of 3-dimensional capillary structure to the injection of the single cells. Moreover, we confirmed that implanted capillaries constituted of HUVECs remained open in the subcutaneous tissue (data not shown). However, HUVECs may cause allograft rejection at the moment when the autologous cells are not applicable. With the aim of clinical application, we tested capillary structures that consisted of HOMECs as an autologous cell source. HOMECs were isolated from omental tissue obtained from patients undergoing abdominal surgery. As an alternative cell source, ECs of stromal vascular fraction from subcutaneous adipose tissue have been used for angiogenic cell therapy of the mouse ischemic model.26,27 These cells might also be useful for our technology.
Our preliminary experiments showed that DC-AM, a pathogen-free and low immunogenic membrane, is useful as a transplantable scaffold for bovine and human ECs (bovine carotid arterial ECs and HUVECs, respectively) (data not shown). However, HOMECs that were aligned on the substrate scattered when transferred not only to DC-AMs but also to untreated AMs (Figure 1A). As a result, tube formation on AM ended in failure. Previous reports indicated that S1P enhances cell-cell contacts by induction of VE-cadherin localization at cellular junctions,19 suggesting the possibility that S1P treatment might prevent cell scattering and align the HOMECs on AMs. Indeed, although there was no significant increase in VE-cadherin localization at adherence junctions of HOMECs, S1P could keep HOMEC alignment on AMs and enabled HOMECs to form a tube (Figure 2B). To clarify the mechanism of the S1P effect on HOMECs, we performed a cell migration assay. A stimulatory effect of S1P on EC migration has been observed in HUVECs,20,21 bovine aortic ECs,21 human aortic ECs,22 and fetal bovine heart ECs.23 Moreover, these cells,20–23 as well as human dermal microvascular ECs28 and human pulmonary microvascular ECs,29 express S1P1 and S1P3 but not S1P2. Furthermore, increasing evidence has demonstrated that S1P induces cell migration by activating Rho and Rac via S1P1 and inhibits migration by suppression of Rac activity via S1P224,25. Therefore, S1P is believed to induce EC migration through the endothelial specific S1PRs, S1P1 and S1P3. However, surprisingly, S1P inhibited HOMEC migration (Figure 2C). Analyses of S1PR expression, cell migration using anti-S1P-siRNA, and the effect of a ROCK inhibitor on cytoskeletal rearrangements indicated that S1P inhibits HOMEC migration through an S1P2-Rho-ROCK signaling pathway. Sanchez et al demonstrated that S1P inhibition of cell migration is mediated by S1P2-Rho-dependent activation of the PTEN (phosphatase and tensin homolog deleted on chromosome 10) phosphatase.30 Recently, reports of S1P2-expressing ECs have been published including an SV40-infected mouse vascular EC line, SVEC4-1028, ischemic mice retinal endothelium,31 senescent HUVECs, and human pulmonary artery ECs.32
The use of S1P treatment allowed for successful formation of an ex vivo capillary structure constituted of HOMECs on DC-AMs. When transplanted into ischemic hindlimbs of mice, this capillary structure significantly improved blood perfusion as early as 5 days after implantation (Figure 6D). Exercise rehabilitation consisting of treadmill walking has been shown to improve the walk performance and the claudication pain symptoms of PAD patients.33,34 Claudication pain often discourages patients from exercise rehabilitation, and an increase in blood flow of the ischemic limb is thought to reduce the pain. Therefore, implantation of this capillary structure into an ischemic region of PAD patients would be expected to facilitate patients’ exercise and to improve patient symptoms more efficiently than in the animal model.
We thank Professor Shigeki Arii (Hepato-Biliary-Pancreatic Surgery, Tokyo Medical and Dental University) and Associate Professor Tatsuyuki Kawano (Vascular and Applied Surgery, Tokyo Medical and Dental University) for providing human omental tissues.
Sources of Funding
This work was supported in part by Grants-in-Aid for Scientific Research (grants 17590235, 20592423, 20390470, and 20659306) from the Ministry of Education, Science, Sports and Culture of Japan.
Received on: October 15, 2009; final version accepted on: March 30, 2010.
Urbich C, Dimmeler S. Endothelial progenitor cells: characterization and role in vascular biology. Circ Res. 2004; 95: 343–353.
Kinnaird T, Stabile E, Burnett MS, Epstein SE. Bone-marrow-derived cells for enhancing collateral development: mechanisms, animal data, and initial clinical experiences. Circ Res. 2004; 95: 354–363.
Martin K, Weiss S, Metharom P, Schmeckpeper J, Hynes B, O'Sullivan J, Caplice N. Thrombin stimulates smooth muscle cell differentiation from peripheral blood mononuclear cells via protease-activated receptor-1, RhoA, and myocardin. Circ Res. 2009; 105: 214–218.
Tateno K, Minamino T, Toko H, Akazawa H, Shimizu N, Takeda S, Kunieda T, Miyauchi H, Oyama T, Matsuura K, Nishi J, Kobayashi Y, Nagai T, Kuwabara Y, Iwakura Y, Nomura F, Saito Y, Komuro I. Critical roles of muscle-secreted angiogenic factors in therapeutic neovascularization. Circ Res. 2006; 98: 1194–1202.
Houghton J, Stoicov C, Nomura S, Rogers AB, Carlson J, Li H, Cai X, Fox JG, Goldenring JR, Wang TC. Gastric cancer originating from bone marrow-derived cells. Science. 2004; 306: 1568–1571.
Fujisato T, Minatoya K, Yamazaki S, Meng Y, Niwaya K, Kishida A, Nakatani T, Kitamura S. Preparation and recellularization of tissue engineered bioscaffold for feat valve replacement. In: Mori H, Matsuda H, eds. Cardiovascular Regeneration Therapies Using Tissue Engineering Approaches. Tokyo: Springer; 2005; 83–94.
Miyashita H, Kanemura M, Yamazaki T, Abe M, Sato Y. Vascular endothelial zinc finger 1 is involved in the regulation of angiogenesis: possible contribution of stathmin/OP18 as a downstream target gene. Arterioscler Thromb Vasc Biol. 2004; 24: 878–884.
Lepley D, Paik JH, Hla T, Ferrer F. The G protein-coupled receptor S1P2 regulates Rho/Rho kinase pathway to inhibit tumor cell migration. Cancer Res. 2005; 65: 3788–3795.
Wang F, Van Brocklyn JR, Hobson JP, Movafagh S, Zukowska-Grojec Z, Milstien S, Spiegel S. Sphingosine 1-phosphate stimulates cell migration through a G(i) -coupled cell surface receptor. Potential involvement in angiogenesis. J Biol Chem. 1999; 274: 35343–35350.
Kimura T, Watanabe T, Sato K, Kon J, Tomura H, Tamama K, Kuwabara A, Kanda T, Kobayashi I, Ohta H, Ui M, Okajima F. Sphingosine 1-phosphate stimulates proliferation and migration of human endothelial cells possibly through the lipid receptors, Edg-1 and Edg-3. Biochem J. 2000; 348Pt 1:71–76.
Panetti TS, Nowlen J, Mosher DF. Sphingosine-1-phosphate and lysophosphatidic acid stimulate endothelial cell migration. Arterioscler Thromb Vasc Biol. 2000; 20: 1013–1019.
Miranville A, Heeschen C, Sengenès C, Curat C, Busse R, Bouloumié A. Improvement of postnatal neovascularization by human adipose tissue-derived stem cells. Circulation. 2004; 110: 349–355.
Krump-Konvalinkova V, Yasuda S, Rubic T, Makarova N, Mages J, Erl W, Vosseler C, Kirkpatrick CJ, Tigyi G, Siess W. Stable knock-down of the sphingosine 1-phosphate receptor S1P1 influences multiple functions of human endothelial cells. Arterioscler Thromb Vasc Biol. 2005; 25: 546–552.
Sanchez T, Thangada S, Wu MT, Kontos CD, Wu D, Wu H, Hla T. PTEN as an effector in the signaling of antimigratory G protein-coupled receptor. Proc Natl Acad Sci U S A. 2005; 102: 4312–4317.
Estrada R, Zeng Q, Lu H, Sarojini H, Lee JF, Mathis SP, Sanchez T, Wang E, Kontos CD, Lin CY, Hla T, Haribabu B, Lee MJ. Up-regulating sphingosine 1-phosphate receptor-2 signaling impairs chemotactic, wound-healing, and morphogenetic responses in senescent endothelial cells. J Biol Chem. 2008; 283: 30363–30375.
McDermott MM, Ades P, Guralnik JM, Dyer A, Ferrucci L, Liu K, Nelson M, Lloyd-Jones D, Van Horn L, Garside D, Kibbe M, Domanchuk K, Stein JH, Liao Y, Tao H, Green D, Pearce WH, Schneider JR, McPherson D, Laing ST, McCarthy WJ, Shroff A, Criqui MH. Treadmill exercise and resistance training in patients with peripheral arterial disease with and without intermittent claudication: a randomized controlled trial. J Am Med Assoc. 2009; 301: 165–174.