New Directions in Therapeutic Angiogenesis and Arteriogenesis in Peripheral Arterial Disease

Abstract

The prevalence of peripheral arterial disease (PAD) in the United States exceeds 10 million people, and PAD is a significant cause of morbidity and mortality across the globe. PAD is typically caused by atherosclerotic obstructions in the large arteries to the leg(s). The most common clinical consequences of PAD include pain on walking (claudication), impaired functional capacity, pain at rest, and loss of tissue integrity in the distal limbs that may lead to lower extremity amputation. Patients with PAD also have higher than expected rates of myocardial infarction, stroke, and cardiovascular death. Despite advances in surgical and endovascular procedures, revascularization procedures may be suboptimal in relieving symptoms, and some patients with PAD cannot be treated because of comorbid conditions. In some cases, relieving obstructive disease in the large conduit arteries does not assure complete limb salvage because of severe microvascular disease. Despite several decades of investigational efforts, medical therapies to improve perfusion to the distal limb are of limited benefit. Whereas recent studies of anticoagulant (eg, rivaroxaban) and intensive lipid lowering (such as PCSK9 [proprotein convertase subtilisin/kexin type 9] inhibitors) have reduced major cardiovascular and limb events in PAD populations, chronic ischemia of the limb remains largely resistant to medical therapy. Experimental approaches to improve limb outcomes have included the administration of angiogenic cytokines (either as recombinant protein or as gene therapy) as well as cell therapy. Although early angiogenesis and cell therapy studies were promising, these studies lacked sufficient control groups and larger randomized clinical trials have yet to achieve significant benefit. This review will focus on what has been learned to advance medical revascularization for PAD and how that information might lead to novel approaches for therapeutic angiogenesis and arteriogenesis for PAD.

THE PERVASIVE AND PERSISTENT PROBLEMS OF PERIPHERAL ARTERIAL DISEASE

Systemic atherosclerosis remains the number one cause of morbidity and mortality in the western world and is the major cause of peripheral arterial disease (PAD), causing obstructions in blood flow in one or more of the major leg arteries.^1–4 Estimates place the prevalence of PAD at over 14 million in the United States and >200 million people worldwide.^1–3,5 In patients, PAD is still defined by the finding of an ankle-brachial blood pressure index of 0.90 or less.^5–7 Although the risk factors for the development of PAD overlap those of coronary artery and cerebrovascular disease, diabetes and smoking have been and remain the 2 strongest age-adjusted risk factors for the development of PAD.^1,5,7 Patients with PAD continue to suffer excessively high risks of heart attack and stroke when compared with patients without PAD.^8–10

The 2 major symptomatic manifestations of PAD are intermittent claudication (IC) and critical limb ischemia (CLI).^1,8,11 IC is the most prevalent clinical manifestation in patients with PAD and is defined by the presence of calf or buttock muscle discomfort produced by exercise that is relieved by rest.¹² CLI is the most severe manifestation of PAD and is characterized by one or more of the following: leg pain occurring at rest, ulceration, and gangrene.^11,13 Of critical importance, classic claudication symptoms actually occur in only a minority of patients with PAD, and this has been shown in those with a confirmed low ankle-brachial blood pressure index.^6,7,10 Despite its high disease prevalence, the low ankle-brachial blood pressure index is not sufficiently used, and because symptoms may be atypical or absent in many patients, PAD often goes unrecognized and undertreated.^2–4 Of critical importance is that all patients with PAD, even those who lack symptoms, have significant disability with limitations in walking capacity, daily functional activities, and exercise performance.^14,15 The clinical course of patients with PAD is strikingly different based upon whether the patients present with IC or CLI. Patients with CLI suffer rates of death and major amputation of 30% or more by 1-year, whereas over a 5-year period, only 1.0% to 3.3% of IC patients will require amputation.^1,8–10 Even in the absence of CLI, the functional limitations seen in patients with PAD are similar to the impairment seen in patients with advanced congestive heart failure.^14,15

Currently, medical therapies used to treat patients with PAD include antiplatelet and antithrombotic agents, statins to lower cholesterol, antihypertensive therapy with ACE (angiotensin-converting enzyme) inhibitors or ARB (angiotensin receptor blocker) or β-blockers, blood glucose control, and smoking cessation. These therapies were largely established from treatments used for patients with coronary artery disease. In patients with PAD, these agents may prevent heart attack and stroke, but clinical cardiovascular event rates in patients with PAD remain high, and even new anticoagulant therapies and intensive lipid-lowering therapies have not consistently altered the progression of disabling claudication or limb loss.^1,9,10 These medical therapies do not improve lower limb perfusion in patients with PAD.¹⁶ Surgical or endovascular procedures may be suboptimal in relieving symptoms and may not be indicated because of the extent of disease or because of comorbid conditions. Finally, even after successful large vessel revascularization, residual microvascular disease may well limit the effectiveness of therapies in patients with PAD.^11,13,16 Still today PAD effects both legs and life. Few areas in cardiovascular medicine are more in need of new therapeutic modalities than PAD.

ANGIOGENESIS, ARTERIOGENESIS, VASCULOGENESIS, AND PAD

Within the adult, the overriding purpose of the vasculature is to regulate blood flow (delivery of oxygen and nutrients along with the removal of toxins) to organs to meet the metabolic demands of active cells. The central problem in PAD is the limitation of perfusion to the muscle and soft tissue of the leg(s). It would, therefore, seem rational to attempt to improve perfusion in PAD by enhancing vascular regeneration and function. Angiogenesis is defined as the growth and proliferation of blood vessels from existing vascular structures.¹⁷ Angiogenesis is, of course, critical in embryonic development. In adults, angiogenesis is required for tissue repair or remodeling, as well as restoration of tissue perfusion.¹⁸ Angiogenesis can be physiologically regulated such as in the response to exercise training. A major process in angiogenesis is the sprouting of endothelial cells (ECs) from preexisting capillaries under the regulation of angiogenic factors such as VEGF (vascular endothelial growth factor) generated by ischemic tissue (Figure 1). The sprouting ECs migrate, proliferate, and form lumens while other processes that may help to generate a functional microvasculature include intussusception of existing capillaries,^17,18 the incorporation of circulating endothelial progenitors,^19,20 and the generation of cytokines by circulating angiogenic cells.²¹

Figure 1. Processes to restore perfusion in peripheral arterial disease (PAD).

In angiogenesis, new microvasculature is derived from the sprouting of preexisting capillaries under the influence of VEGF (vascular endothelial growth factor), FGF (fibroblast growth factor), and other angiogenic cytokines. Arteriogenesis is the positive remodeling of preexisting collaterals when blood flow is redirected through these channels. The positive remodeling of these channels is due, in part, to the mobilization of monocytes by G-CSF (granulocyte-colony stimulating factor) and vascular expression of monocyte chemotactic protein, the ligand for the monocyte CCR2 (C-C chemokine receptor type 2). Circulating angiogenic cells (CACs) are also mobilized by G-CSF to participate in adult vasculogenesis. These cells express CXCR-4 (C-X-C chemokine receptor type 4), the receptor for SDF-1 (stromal-derived factor 1) that is expressed by the ischemic tissue. Most CACs are of hematopoietic origin and contribute to restoration of perfusion by secreting angiogenic cytokines. Rare CACs are of endothelial lineage and can incorporate into the newly forming vessels. Transdifferentiation of fibroblasts to endothelial cells may also contribute to angiogenesis. In the setting of inflammatory signaling, increased DNA accessibility permits fluidity of cell fate, and environmental influences such as hypoxia activate transcription factors determining endothelial lineage.

Pathological angiogenesis may support another pathological process and examples would include tumor angiogenesis, atherosclerotic plaque neovascularization, or age-related macular degeneration. In each of these tobacco-related conditions, nicotinic acetylcholine receptors on the endothelium are stimulated by exogenous nicotine to proliferate, migrate, and form a microvasculature that supports the underlying pathological process.^22,23 Another form of pathological angiogenesis is when the angiogenic response to ischemia is inadequate to meet tissue demands, and this most commonly occurs in the settings of coronary artery disease and PAD. Here, hypoxia in heart and leg muscles due to poor tissue perfusion activates the transcription factor HIF-1 (hypoxia-inducible factor 1) as its subunit HIF-1α has a degradation domain that is dependent upon oxygen tension.²⁴ With low oxygen levels, HIF-1α is stabilized, translocates to the nucleus, and activates target genes including VEGF, FGF (fibroblast growth factor), HGF (hepatocyte growth factor), PDGF (platelet-derived growth factor), Del-1 (developmental endothelial locus 1), angiopoietins, and matrix metalloproteinases to facilitate angiogenesis.²⁵ Notably, exogenous administration of each of these genes or the proteins that they encode, have been shown to improve limb blood flow in animal models of PAD,^26–31 but as discussed below, have failed in randomized clinical trials to significantly ameliorate PAD.

The remodeling of preexisting collateral channels is termed arteriogenesis. In their normal state, these collateral channels are narrow, high resistance vessels, and they provide little blood flow to their distal tissue bed. However, when a major conduit becomes obstructed, blood flow is redirected through the collateral channels (Figure 1) which causes alterations in vascular wall shear stress. This hemodynamic stimulus provokes an increase in the diameter and wall thickness of the collateral channels, with proliferation of vascular cells and turnover of the vascular matrix.^32,33 Other factors involved in arteriogenesis include chemokines and adhesion molecules, such as ICAM (intercellular adhesion molecule), CCR2 (C-C chemokine receptor 2), Delta-like 1, SDF-1 (stromal-derived factor 1), which recruit monocytes that then facilitate the positive remodeling by elaborating metalloproteinases and cytokines..^34–37 Whereas macrophages are the cell most linked to arteriogenesis, recent data have also linked inflammatory (referred to as M1-like) macrophages in processes that negatively impact angiogenesis and which may be a target for therapeutics.^38–40 Inflammatory macrophages in ischemic tissue may be a source for circulating cytokines (ie, TNF [tumor necrosis factor]-α and IL [interleukin]-6) that can trigger acute cardiovascular events.^39,41

Adult vasculogenesis refers to the mobilization of circulating cells with angiogenic potential from the bone marrow into the circulation by factors, such as G-CSF (granulocyte-colony stimulating factor) and GM-CSF (granulocyte macrophage-colony stimulating factor). These CD34⁺ circulating angiogenic cells may also express CD133, VEGFR2, and CXCR-4 (C-X-C chemokine receptor type 4; Figure 1).^42–45 Under the influence of SDF-1 (which binds to the chemokine receptor CXCR-4), angiogenic cells are recruited to the ischemic tissue.^46,47 Bone marrow-derived angiogenic cells are largely of hematopoietic lineage and largely improve perfusion by secreting cytokines and metalloproteinases, although a rare fraction may differentiate into mature endothelium in vivo.^19,48,49 It is also possible that angiogenic cytokines cause mature ECs in other nonischemic organs to be mobilized into the circulation, home to ischemic tissue, and contribute to angiogenesis.^19,50

In patients with PAD, there is some degree of endogenous arteriogenesis (and angiogenesis), but the capacity of the processes even when active are limited in their ability to restore limb perfusion to normal levels.⁵¹ In patients with claudication, capillary density is often reduced and is related to a patients’ functional capacity.⁵² Notably, an increase in capillary density is observed in the limb skeletal muscle of PAD patients with exercise training, which is still today the cornerstone of therapy in patients with PAD.⁵³ Moving from concept to the clinic, we hope to harness blood vessel growth to improve perfusion to ischemic tissue, an approach which has been termed therapeutic angiogenesis or medical revascularization. Perhaps, a better term is vascular regeneration, which encompasses angiogenesis, arteriogenesis, and vasculogenesis, the combination of which processes may be necessary for substantial benefit. Investigation into therapeutic angiogenesis dates back to the work of the late Isner et al,⁵⁴ who administered by intraarterial infusion a plasmid vector encoding VEGF to treat a 70-year-old female with midfoot 3-vessel runoff occlusion and digit gangrene. Their work lacked a sufficient control group to draw conclusions but stimulated a great number of clinical trials of therapeutic angiogenesis involving thousands of patients, with varying experimental protocols, agents, and modes of delivery. Regrettably, these studies have failed to deliver significantly positive results, as previously reviewed.^16,55–57 Thus, new approaches are needed for vascular regeneration in PAD.

NEW DIRECTIONS FOR POTENTIAL THERAPEUTIC SOLUTIONS IN PAD

Angiogenesis Independent of VEGF

To date a substantial majority of the completed human therapeutic angiogenesis trials, in whole or in part, sought to activate the well described, canonical, VEGF-VEGFR2/Akt-1/eNOS (endothelial nitric oxide synthase) pathway; where increases in bioavailable nitric oxide is a critical mediator.^58–60 Promoting angiogenesis independent of this pathway may be important for humans with PAD as these patients frequently have comorbid conditions (even diabetes which is a frequently causes and modifies PAD) which can limit the generation, or bioavailability, of NO.¹⁶ Preclinical models of PAD remain an invaluable tool for testing the potential efficacy and bioactivity of putative therapeutic agents as well as to identify novel therapeutics. Using established differences in the extent of perfusion recovery and necrosis that occurs across informative inbred mouse strains and congenic lines, Dokun et al⁶¹ were the first to identify a quantitative trait locus on the short arm of mouse chromosome 7 (termed LSq-1) that contained potential gene(s) that were sufficient to determine the extent of perfusion recovery and tissue loss after hindlimb ischemia (HLI). This was an unbiased approach and had the advantage of allowing us to look for genes that modified the course of PAD in the setting of comparable occlusions.^16,61 More importantly, this strategy had the advantage of not requiring that we preselect for known angiogenesic or arteriogenic genes or pathways. One of the first candidates to emerge from this approach was the IL-21R (IL-21 receptor) which, resides at/very near the peak of genetic linkage to variations in perfusion recovery and necrosis in the mouse model of PAD and the potential role of this receptor in modulating angiogenesis in the setting of PAD has supporting human data.^62,63

The IL-21R was identified about 2 decades ago in the laboratory of Dr Warren J. Leonard at the National Institutes of Health/National Heart, Lung, and Blood Institute and has emerged as a key molecule in innate and adaptive immunity, with IL-21 serving as its sole ligand.^64,65 Early work with IL-21R indicated that it was angiostatic, but these observations were obtained using mouse and human ECs cultured in serum supplemented with bFGF (basic FGF, a potent angiogenic agent) under normoxic conditions.⁶⁶ By contrast, skeletal muscle tissue has low levels of VEGF, bFGF, and other cytokine growth factors.^67,68 Accordingly, we altered the in vitro conditions to be more similar to those seen in ischemic skeletal muscle (with ECs subjected to hypoxia and serum starvation, ie, HSS). Under HSS conditions, human ECs upregulate IL-21R and exogenous ligand-induced angiogenesis.⁶⁹ The differences between the angiogenic outcomes from the in vitro experiments conducted under HSS versus the angiostatic outcomes seen with ECs in growth factor rich, normoxic conditions were associated with differences in downstream signaling. Specifically, under HSS, ligand-mediated IL-21R activation activated STAT3, not STAT1 nor Akt.^62,69

In wild-type C57BL/6 mice, we found that the IL-21R was significantly upregulated within ischemic muscle following HLI, and the increases were largely within the EC population. By contrast, the IL-21R showed no increase in expression in Balb/C mice, a mouse strain that develops a sustained perfusion deficit and greater tissue loss following induction of HLI.⁶² These results suggested that IL-21R on ECs might play an important role in the response to ischemia. The IL-21R seemed to have little role in non-ECs in muscle, as it was not detectable or present at very low levels in vascular smooth muscle and skeletal muscle cells in both normoxic and HSS conditions, and IL-21 administration had no effect on vascular smooth muscle nor skeletal muscle cells survival under HSS. Furthermore, IL-21 (ligand) levels do not change following HLI in C57Bl/6 nor Balb/C mice; in human umbilical vein ECs IL-21 protein was not detectable in cells before or after HSS.⁶² Silencing the IL-21R receptor blocked the angiogenic effect of exogenous ligand, in vitro and in vivo, and when angiogenesis occurred there was no change in the Akt/eNOS pathway.⁶² Consistent with its modulation of perfusion changes following vascular injury, Lee et al⁷⁰ later demonstrated that IL21R^−/− mice had greater cerebral infarct volumes when compared with wild-type C57/BL6 mice in a stroke model. Furthermore, within that study, a coding variant between C57BL/6 and Balb/c mice at amino acid 200, in the Balb/c allele of IL-21R was linked to greater neuronal injury and poorer stroke outcomes. At 200 AA, humans express the C57BL/6 sequence, with only very rare alternative coding variants. Ongoing work is focused on the IL-21R mediated, nitric oxide-independent, angiogenesis in PAD.

Refining Angiogenesis and Modulating Antiangiogenic Factors

Within the well-studied VEGF receptor-ligand family, VEGFR2 is widely regarded as the primary receptor whose activation drives hypoxia-dependent and post-natal angiogenesis.⁷¹ VEGF-A is one of several structurally related ligands for the VEGFR tyrosine kinases: VEGFR1, R2, and R3.^71–73 Within the VEGF-A gene, there are major alternative splicing sites resulting in proteins of varying lengths of 121, 165, and 189 amino acids. There is a 206 splice variant whose expression is limited to development. The longer transcripts have a stronger capacity to bind to the extracellular matrix, which alters their bioavailability. Strategies are available to activate multiple isoforms, and these include activating transcription factors and the use of single plasmid encoding multiple transcripts. Rebar et al⁷⁴ reported that the administration of a single isoform of VEGF-A led to the formation of leaky dysfunctional vessels, whereas the coordinated expression of all of the splice variants led to the formation of more functional blood vessels in a mouse ear angiogenesis model. Dai et al⁷⁵ demonstrated that a VEGF-A activating transcription factor could increase expression of multiple VEGF isoforms in ischemic muscle in vivo and promote angiogenesis following HLI. An additional and less appreciated splice variant occurs with alternate splicing of the eighth exon of VEGF-A results in a 6 amino acid switch that changes the pro-angiogenic VEGF_xxxa (xxx =the number of amino acids) to the anti-angiogenic VEGF_xxxb (VEGF₁₆₅b for 165 amino acids) isoform.^40,76–78 This splicing can occur for any of the amino acid length transcripts but has been best studied for the 165 amino acid isoform.⁷⁸

Kikuchi et al⁷⁷ showed that peripheral blood monocytes (isolated by gradient centrifugation) from PAD versus control patients expressed significantly higher VEGF₁₆₅b by quantitative polymerase chain reaction and then in the mouse HLI model showed that intramuscular delivery of a VEGF₁₆₅b specific antibody improved perfusion recovery in experimental PAD.⁷⁷ Their conclusion and indeed the conclusion in similar studies of the VEGF system focused on the activation the VEGFR2 pathway as the mechanism of action.^77,79,80 Indeed, even studies performed with VEGFR1 selective ligands in ischemic heart or PAD models concluded that the mechanism for the angiogenic effects of these factors was through occupying VEGFR1, thus allowing VEGF₁₆₅a and other ligands to preferentially activate VEGFR2.^81–86 This is often referred to as a decoy effect.⁸⁷

Using a similar approach as Kikuchi et al,⁷⁷ in Balb/C mice, we found therapeutic benefit following HLI with delivery of the VEGF₁₆₅b antibody.⁴⁰ In these in vitro and in vivo studies, we consistently found that VEGF₁₆₅b impaired endothelial VEGFR1 (not VEGFR2) activation.⁴⁰ Administration of the VEGF₁₆₅b antibody reduced binding of VEGF₁₆₅b to VEGFR1 and increased the activation of VEGFR1, as indicated by its level of phosphorylation at Y1333, and this occurred with coimmunoprecipitation with activated STAT3. In this report, we concluded that when VEGF₁₆₅b reduces the activation of VEGFR1, the antibody against VEGF₁₆₅b augments VEGFR1-mediated angiogenesis. This is not meant to imply that VEGF₁₆₅b completely silences VEGFR1. Notably, in wild-type Balb/C and C57/Bl6 mice after HLI, both VEGFR1 and VEGF₁₆₅b are increased, and therefore, the level ofVEGFR1 activation is dependent upon the levels of VEGF₁₆₅b and other VEGFR1 ligands that are competing for binding.^88,89 This is consistent with published data that kinase-dead knockout mice and VEGFR1^+/− knockout mice have impaired angiogenesis and perfusion recovery after HLI when compared with wild type.⁹⁰ Within macrophages, VEGFR1 activation promotes an M2-like angiogenic state in macrophages which is enhanced by inhibition of VEGF₁₆₅b.⁴⁰ Thus, in the setting of PAD, the VEGFR1 pathway is regulated differently than VEGFR2 and may activate angiogenic pathways by themselves or ones that can have therapeutic synergy with the VEGFR2/eNOS pathway.

Role of Noncoding RNAs (miR and Long Noncoding RNA)

Micro-RNAs are 15 to 23 nucleotide (noncoding) RNAs that have emerged as key regulators of the response to hemodynamic forces, vascular injury, and hypoxia.^91,92 A miR is thought to primarily exert its effects by incorporating into the RNA-induced silencing complex which then binds to a target mRNA, usually in the 3′ UTR of its target mRNA.^93,94 The binding of the miR to the mRNA can suppress a single gene target or the miR can have the capacity to regulate multiple genes within a common pathway simultaneously by promoting degradation of the mRNA or by inhibiting mRNA translation to protein.^94,95

Although a full review of noncoding RNA in angiogenesis and arteriogenesis is beyond the scope of this review, herein, we will provide examples of noncoding RNAs and their potential application for future vascular therapeutics in PAD. The first reports of miRs that were identified as candidates for modulation of neovascularization in PAD were of miR-92a⁹⁶ and miR-100.⁹⁷ Both of these miRs were identified based upon the response of EC to injury. Subsequently, the role of each in PAD-related angiogenesis was tested in preclinical models where the inhibition of the miR, by its antagomirs, enhanced perfusion recovery.^96,97 Antagomirs (antisense miRs) have been shown to have a high degree of specificity and beyond having little affect in a cell or tissue where the target miR is not expressed; antagomers even in target tissue can expression levels of closely related miR’s unchanged.⁹⁸ Studies that focused on the ability of miRs to modulate angiogenesis in PAD conditions appeared.^98–103 The response, that is, the gene or gene pathway, modulated by a miR can be expected to be different between cells, or target tissues, based upon the presence of an injury.¹⁰³ Moving from angiogenesis to arteriogenesis, Landskroner-Eiger et al¹⁰⁴ focused on miR(s) that regulated arteriogenesis. In vitro studies under shear stress, identified candidate miRs that could regulate arteriogenesis, and in murine PAD models, approaches that targeted miR-199a and miR-146a demonstrated improved perfusion recovery following femoral artery ligation.^105,106 Using the same approach described in the identification of the IL-21R, we used unbiased genetic, computational, and experimental approaches and in vivo HLI models where we examined the target tissue in ischemic skeletal muscle and identified miR-93 as a potential therapeutic target to modulate PAD.⁹⁸ For miR-93, although overexpression has beneficial effects on ECs and macrophages in PAD relevant conditions, the target gene(s) for ECs was shown to be p21, p53, and E2F1, while in macrophages, it was immune response factor-9.^39,98

Predicting the net physiological effect of miR modulation in PAD conditions is not always straightforward. Although miR-155 has antiangiogenic effects in targeting eNOS, and VEGF levels are inversely correlated with miR-155 expression, the inhibition of miR-155 attenuates blood flow recovery and leukocyte recruitment in vivo following HLI.^107–109 Indeed, the totality of data suggests a negative role for miR-155 in the response to HLI. Thus, in considering miRs for therapeutic modulation, relevant disease models can be informative. In this regard, however, not all miRs, or the gene targets for a given miR, are conserved through evolution and miRs that have retained homology and function from rodents to humans may be preferential for clinical development.

Another class of noncoding RNAs is the long noncoding RNAs (lncRNAs) which are typically defined as RNAs of >200 nucleotides.¹¹⁰ Reports are beginning to emerge on the potential role of lncRNAs in PAD.¹¹¹ When compared with miRs, lncRNAs are far more complex. In general, lncRNAs are known to be far more species-specific and they contain numerous potential splice variants. A lncRNA can serve numerous potential regulatory functions from acting as epigenetic modifiers, to regulating gene expression, to modifying protein translation, and acting as protein scaffolds.^112,113 For example, the lncRNA LEENE facilitates the recruitment of RNA Polymerase II to the eNOS promoter to enhance eNOS RNA transcription.¹¹⁴

SENCR is a lncRNA that is highly expressed in EC and is necessary for EC proliferation, migration, and tube formation in vitro.^115,116 Intriguingly, SENCR is expressed at lower levels in the vascular tissue of patients with CLI.¹¹⁶ For reasons that include the potential to examine rodent models, lncRNAs that are conserved from mouse to human are far easier to study than are those that are not conserved. Human-specific lncRNA cannot readily be used in preclinical models of PAD to examine effects on angiogenesis, arteriogenesis, perfusion recovery, and potentially tissue loss. Studies have sought to identify candidate lncRNAs based on their association with angiogenesis and have, for example, characterized the response of ECs to hypoxia in vitro where lncRNAs, including LNC00323, MIR503-HG, GATA-AS, have all been shown to be differentially regulated in human EC with after hypoxic injury.^117–120 Further analysis showed that a targeted knockdown of LINC00323 impaired EC tube formation while MIR503-HG knockdown reduced EC migration, although both were necessary for proliferation.¹¹⁷ Voellenkle et al¹²⁰ found H19, MIR210HG, MEG9, MALAT1, and MIR22HG to be elevated by hypoxia in human umbilical vein ECs in vitro and because these were evolutionarily conserved, they were able to be examined in mouse gastrocnemius muscle following HLI. MALAT1 inhibition impaired angiogenesis in vitro and blood flow recovery and reduced capillary density following HLI in vivo.¹¹⁸ In aged mice, MEG3 inhibition was shown to improve blood flow recovery following HLI.¹²¹ The systematic examination of human tissue will likely offer the best way to identify lncRNAs that may play a role in, or serve as therapeutic targets for, PAD.¹²²

Endothelial Metabolism in Angiogenesis.

Research in therapeutic angiogenesis, certainly from the cardiovascular perspective, has largely focused on the use of cytokine growth factors to bind to and activate receptors on the EC surface to drive receptor signaling within in vitro systems to move EC into a proliferative state. Even when in a quiescent state, energy from ATP is required for ECs to maintain physiological function and homeostasis. In ECs, the majority of ATP is produced via glycolytic metabolism, and this phenomenon occurs despite the location of ECs being adjacent to the bloodstream and in contact with the sufficient levels of oxygen.¹²³ In ECs, the 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase enzymes PFKFBs product is an activator of PFK1 (phosphofructokinase 1), a rate-limiting enzyme of glycolysis.¹²⁴ Although there are 4 isoforms of PFKFB, PFKFB3 has a kinase/phosphatase activity >700, the kinase to phosphatase activity in PFKFB2 and PFKFB4 are 1.8 and 0.9, respectively, and the expression of PFKFB1 in ECs is low.^124,125 Therefore, PFKFB3 is critical for converting F6P to F2,6BP, with F2,6BP being the main allosteric activator of PFK1; the most important regulatory enzymes of glycolysis. Taken together, PFKFB3 is critical for the regulation of endothelial glycolysis.^123–126

Beyond homeostasis, ECs need ATP for angiogenesis. Tumor angiogenesis is perhaps the most extensively studied angiogenesis systems and one in which the growth/proliferation of ECs requires an increase in glycolytic metabolism.¹²⁷ Exposing ECs to hypoxia alone has been used to model conditions such as tumor angiogenesis and retinopathy.¹²⁸ The changes in EC metabolism observed under hypoxia are similar to what is seen with exposure of ECs to VEGF₁₆₅a, where EC growth and proliferation is accompanied with an increase in glycolytic metabolism that occurs with, and is dependent on, an upregulation of PFKFB3 to increase the glycolytic flux.^129–131 In human umbilical vein ECs challenged with hypoxia only, PFKFB3 knockdown using siRNA or pharmacological inhibition has resulted in decreased cell proliferation.¹²⁴ Tumors which are known to use enhanced glycolysis for angiogenesis when embedded in EC-specific PFKFB3 knockout mice show slower growth rates than tumors transplanted in control mice.¹²⁵ Attempts to promote blood vessel growth in the ischemic muscle bed to improve leg blood flow with agents such as VEGF₁₆₅a has been met with limited therapeutic success in humans.^56,132 The angiogenesis that occurs in ischemic leg muscle, when compared with angiogenesis in a tumor, is occurring in a tissue that is relatively low in both cytokine growth factors and, of course, oxygen tension.^68,98 In the state of hypoxia with serum starvation, the addition of VEGF₁₆₅a to ECs induces angiogenesis by reducing cell death and increasing EC proliferation in association with an increase in glycolysis.^133,134 However, in VEGF-mediated tumor angiogenesis, there is vascular permeability and leaky blood vessels.^133,134 Whether all forms of angiogenesis use the same EC metabolism pathways as tumors is not completely understood nor are the potential consequences of using metabolic pathways other than glycolysis.^133–135

Generating Better Autologous Cell Therapies

As described briefly above, large randomized, well-conducted clinical trials of cell therapies for PAD, such as the National Heart, Lung, and Blood Institute-funded PACE study, have not shown benefit of adult stem cell therapy. It is possible that the difference between the preclinical models (where many cell therapies have efficacy), and the clinical trials, is related to the number and the quality of the autologous stem cells derived from individuals with cardiovascular disease.¹³⁶ In this regard, the ability to provide sufficient numbers of high-quality therapeutic cells is likely to be critical in obtaining a reproducible and significant benefit. In any event, different approaches are needed. One novel approach to cell therapy for PAD is to generate therapeutic cells using autologous skin fibroblasts.¹³⁷ In this case, a field of fibroblasts are exposed to retroviral vectors encoding the master regulators of pluripotency (eg, Oct 4 [octamer-binding transcription factor 4], Sox2 [SRY (sex-determining region Y)-box 2], KLF4 [Kruppel-like factor 4], and cMyc [cellular myelocytomatosis]). These transcriptional factors direct the activation of a cascade of factors required for pluripotency, and over a period of several weeks, many of the fibroblasts will become induced pluripotent stem cells (iPSCs). Subsequently, the iPSCs can be differentiated into any of the 3 germ layers, and further differentiated into the lineage of choice, using growth factors and small molecules that facilitate the differentiation into a particular lineage, such as ECs. Rufaihah et al¹³⁸ have previously shown that iPSC-derived ECs manifest all of the expected characteristics and functions, expressing endothelial surface markers (eg, VE-cadherin), generating nitric oxide and angiogenic cytokines, and forming vascular networks in Matrigel. Furthermore, when injected into the underperfused limb of the murine HLI model, perfusion was increased. Interestingly, when delivered systemically the ECs derived from pluripotent stem cells can migrate to the ischemic hindlimb and increase perfusion as assessed by bioluminescence and laser doppler spectroscopy.¹³⁹ These findings suggest that autologous iPSC-derived ECs might be useful as a cell therapy in patients with PAD. However, iPSC derivatives will need to be rigorously tested for the fidelity of reprogramming and the exclusion of pluripotent cells from the therapeutic product.

Another approach to generating autologous ECs is to reprogram fibroblasts directly into ECs.^140–142 This form of nuclear reprogramming from one somatic cell lineage to another is also known as direct reprogramming or transdifferentiation so as to distinguish it from the process where the generation of an iPSC is an intermediate step. Transdifferentiation is a change from one somatic cell type to another. In addition, the term transdifferentiation may also apply to radical changes in cell form or function within a cell lineage, such as when a senescent EC is transformed into a juvenile phenotype, as with the forced expression of telomerase.¹⁴³ Whether transdifferentiation is across somatic lineages, or across time as in the latter example, the cell must overcome epigenetic barriers that maintain cell phenotype, and must rearrange chromatin configuration so as to support the global change in transcriptional profile that underlies a new cell identity.^144,145

In the case of direct reprogramming, retroviral vectors or mRNA encoding the master regulators of endothelial identity (eg, ETV2, FLI1 [Friend leukemia integration 1 transcription factor], GATA2 [GATA-binding factor 2], and KLF4) can induce fibroblasts to become ECs. These induced ECs are similar to iPSC-ECs in endothelial functions, as well as the ability to restore perfusion when injected into the ischemic limb of the mouse.¹⁴⁶ In either case, the generation of the induced pluripotent stem cell or the direct reprogramming of a fibroblast to an EC requires the activation of cell-autonomous innate immune signaling.¹⁴⁷ The activation of innate immune signaling increases DNA accessibility to facilitate the action of lineage determination factors to change the cell phenotype.^147–149 The process by which inflammatory signaling increases DNA accessibility to facilitate changes in cell fate is termed transflammation (Figure 2).

Figure 2. Damage-associated molecular patterns (DAMPs) generated by hypoxic injury stimulate PRRs (pattern recognition receptors) which activate signaling pathways that alter transcription or posttranslational modifications of epigenetic enzymes to promote DNA accessibility.

In addition, a glycolytic shift supplies metabolites to the nucleus for histone modifications that further facilitate an open chromatin configuration. Together with environmental influences that favor an endothelial phenotype (eg, VEGF [vascular endothelial growth factor]), fibroblasts undergo a transdifferentiation to angiogenic cells that generate vascular growth factors and incorporate into the microvasculature.

Specific caveats regarding the use of reprogrammed cells for therapy relates to concerns regarding residual pluripotent stem cells that could generate abnormal tissues (in the case of iPSC-ECs); and the fidelity of the reprogramming to functional ECs.¹⁵⁰ Other general concerns of cell therapy remain, for example, dosing, duration of treatment, delivery methods, and cellular persistence in a hostile ischemic environment and to what extent the environment can be modulated before therapy.

Generating Autologous Therapeutic Cells In Situ

If autologous therapeutic cells could be generated in situ, it might solve the issues regarding delivery of the cells and could potentially reverse processes contributing to disease, such as endothelial-to-mesenchyme transition, which may contribute to scarring in the setting of ischemic injury. Indeed, there is evidence that fibroblast-to-endothelial transdifferentiation (MEndoT) contributes to the recovery from ischemia in vivo, although this concept remains controversial.^{137,141,142,147–149} Ubil et al¹⁵¹ have presented evidence that resident fibroblasts may transdifferentiate into ECs, to increase capillary density and to enhance perfusion in the murine myocardial infarction model which is mediated partially through p53. However, a subsequent paper supported by extensive lineage tracing found no evidence of MEndoT in the murine myocardial infarction model.¹⁵²

The contribution and extent of potential modulation of MEndoT in the response to limb ischemia is still in the process of being understood. Because there are significant differences in peripheral versus myocardial tissues, in their vascularity and their response to disease, lineage tracing and single-cell RNA sequencing were used in an HLI model, to detect evidence of MEndoT. Surprisingly, 8 distinctly different subpopulations of fibroblasts in the mouse hindlimb, confirmed by their high expression of genes known to be preferentially expressed in fibroblasts (eg, Tcf4/Tcf7l2, fibronectin, and collagen IV) yet distinct by their global transcriptional profile. Two subpopulations expressed some EC genes at baseline and expanded dramatically during ischemia. Genetic or pharmacological suppression of inflammatory signaling abrogated the expansion of this population and impaired vascularity, recovery of perfusion, and preservation of ischemic tissue. Notably, surface markers for these 2 populations permitted their isolation from the ischemic hindlimb by fluorescence-activated cell sorting. Whereas most fibroblasts will form clusters of cells in Matrigel, one of the subpopulations reproducibly formed networks and displayed surface markers characteristic of ECs when placed into Matrigel.¹⁵³

The expansion or transdifferentiation of resident fibroblasts to ECs depends upon inflammatory signaling, an observation consistent with our prior work showing that inflammatory signaling promotes DNA accessibility and phenotypic plasticity.^148,153 Hypoxic injury (generating damage-associated molecular patterns or DAMPs) and inflammatory cytokines released in the setting of ischemia activate inflammatory signaling (Figure 2). This triggers global changes in epigenetic modifiers (such as upregulation of histone acetylases and downregulation of histone deacetylases) which favor an open chromatin configuration.¹⁵⁴ In addition, activation of iNOS (inducible nitric oxide synthase) facilitates S-nitrosylation of the polycomb repressive complex 1 and NURD1 (nucleosome remodelling and deacetylase 1), causing the dissociation of these repressive complexes from the chromatin.^148,149 Finally, inflammatory signaling induces a glycolytic shift, which supplies the nucleus with metabolites that participate in epigenetic modification. For example, the glycolytic shift is associated with citrate export from the mitochondria to the nucleus, where it is converted to acetylcoA for histone acetylation.¹⁵⁵

To summarize, single-cell profiling and lineage tracing has revealed that there may be some contribution of specific populations of tissue fibroblasts to angiogenesis in limb ischemia. It may be possible to enhance cellular plasticity to enhance therapeutic transdifferentiation of fibroblasts to angiogenic cells. Drugs which favorably affect the activity of epigenetic modifiers or the glycolytic state are worthy of further study for PAD.

EPILOGUE

The past 25 years of research into angiogenic and cell therapies for PAD have not yet translated into significant changes in PAD management. The discordance between what we have learned and what we have accomplished therapeutically raises important questions. Are the animal models that we are using misdirecting our therapeutic trials? The classic hindlimb ischemia model in mice is typically an acute femoral artery ligation, in the absence of cardiovascular risk factors such as hypercholesterolemia, hypertension, tobacco exposure, diabetes, or senescence. It is critical to consider the effects of risk factors (in particular diabetes and tobacco exposure) in preclinical models. Better animal models are available (eg, hypercholesterolemic cynomolgus monkeys) but are prohibitively expensive and low throughput. A more feasible approach may be to preferentially use genetic mouse models that combine cardiovascular risk factors and use modifications to the classic ligation/excision to more closely mimic human conditions. An animal model that develops peripheral arterial lesions would be ideal but those will need to reach a state of hemodynamically significant occlusion. Conversely, multiorgan chip models that combine microfluidics and physiological hemodynamic forces with human cells could be useful. Finally, the application of contemporary molecular technologies to human samples may be most useful in generating novel therapeutic avenues. For example, genome-wide association studies have provided support for antithrombotic treatment of PAD,¹⁵⁶ a finding that is consistent with the improved limb outcomes in patients with PAD with low dose rivaroxaban in the Compass study. Ongoing studies using single-cell multiomics (eg, single-cell RNA sequencing, genome sequencing, and ATAC sequencing) applied to human tissues will also provide novel insights into disease mechanisms. Such studies may provide useful information such as target receptor expression in subgroups of patients or better validate molecular targets identified in murine models.

It is critical to remember that in patients with PAD, there is substantial patient heterogeneity in the response to similar degrees of atherosclerotic disease. Patients with the same hemodynamic compromise may have different presentations. Why one patient has IC, whereas another has CLI is not entirely explained by the large vessel disease, suggesting that there are microvascular or end-organ differences. In which case, the pathophysiology, and the therapeutic reversal of disease, may be different in these 2 sets of patients.

Is the failure of angiogenic and cell therapies related to our tendency to do single-agent studies? Would a combinatorial approach be more successful? Rather than just facilitating angiogenesis, should we also be considering other processes such as endothelial permeability and pericyte association, which may translate into less permeable, more functional, and more stable vessels? What about other processes that may contribute to angiogenesis, such as clonal EC expansion?¹⁵⁷ What is the role of the skeletal muscle itself in vasculogenesis? Certainly, metabolic products during exercise affect vasomotion. More studies of how such metabolic products effect maintenance or repair of the microvasculature are needed.

Are iPSC-derived cells a better form of cell therapy? Can we ensure the fidelity of reprogramming and the stability of cell fate? Will such cells be more resistant to the hostile environment into which they are placed? And how will we deliver such cells, and in what dose and duration? Is it possibly better to transdifferentiate resident mesodermal cells into ECs that can expand the microvasculature? Can we restore tissue plasticity and regeneration by epigenetic-directed therapies? Or is any of this going to be useful without first restoring flow through the large conduit vessels?

We may also need to rethink how we test novel PAD therapies in clinical trials. For example, in trials of angiogenic agents, hemodynamic, and anatomic assessments, such as low ankle-brachial blood pressure index and angiography, are poor surrogate measurements for perfusion. Treadmill studies to assess function may be complicated by many factors including the fibro-fatty replacement of skeletal muscle that attends chronic limb ischemia. Thus, improved measures to assess improvements in muscle perfusion and function will be advantageous. For cell therapy trials, do we have the proper tools, for example, imaging, cell tracking, etc for conducting more informative early phase human studies? Inflammation plays a major role in coronary artery disease, and likely plays an important role in PAD. However, emerging evidence indicates that inflammatory signaling is important for vascular regeneration.¹⁵⁸ Methods are needed to assess inflammatory signaling in patient tissue, so as to optimally modulate this process to enhance vascular anatomy, perfusion, and skeletal muscle function. For patients with CLI, the major criteria for Food and Drug Administration approval are amputation-free survival. However, for these patients, should not pain relief and reduction in narcotic use be approvable end points?

CONCLUSIONS

These are just a few of the difficult questions that face investigators and regulators in the field of vascular regeneration. However, it is our feeling that the answers to these questions are within reach. In recent years, there have been significant advances in our understanding of the genetic, epigenetic, proteomic, and cellular regulation of angiogenesis, arteriogenesis, adult vasculogenesis, and cell fate decisions regulating vascular recovery. The answers lie in the refinement of our models, in the application of new molecular tools to patient tissues, and in the persistence of our efforts to benefit our patients.

Nonstandard Abbreviations and Acronyms

ACE

angiotensin-converting enzyme

ARB

angiotensin receptor blocker

CCR2

C-C chemokine receptor 2

CLI

critical limb ischemia

CXCR-4

C-X-C chemokine receptor type 4

Del-1

developmental endothelial locus 1

EC

endothelial cell

eNOS

endothelial nitric oxide synthase

FGF

fibroblast growth factor

G-CSF

granulocyte-colony stimulating factor

GM-CSF

granulocyte macrophage-colony stimulating factor

HGF

hepatocyte growth factor

HIF-1

hypoxia-inducible factor 1

HLI

hindlimb ischemia

HSS

hypoxia and serum starvation

IC

intermittent claudication

ICAM

intercellular adhesion molecule

IL

interleukin

iNOS

inducible nitric oxide synthase

iPSCs

induced pluripotent stem cells

lncRNA

long noncoding RNA

PAD

peripheral arterial disease

PCSK-9

proprotein convertase subtilisin/kexin type 9

PDGF

platelet-derived growth factor

PFK-1

phosphofructokinase 1

SDF-1

stromal-derived factor 1

TNF

tumor necrosis factor

VEGF

vascular endothelial growth factor