What Went Wrong with VEGF-A in Peripheral Arterial Disease? A Systematic Review and Biological Insights on Future Therapeutics

Peripheral artery disease (PAD) is a progressive condition driven by gradual stenosis and ultimately occlusion of large and medium-sized arteries, other than coronary and cerebral [1-3]. It is associated with a wide range of devastating effects which include asymptomatic lumen narrowing (20–50%), classic intermittent claudication complaints (10–35%), and critical limb ischemia features, the latter comprising rest pain and tissue loss (5–15%) [4, 5]. Nonoperative therapeutic options for PAD have classically aimed to control modifiable risk factors such as nonatherogenic diet and daily activity promotion, medical management of hypertension, dyslipidemia, diabetes as well as smoking cessation. In addition, antiplatelet therapy has been adopted as secondary prevention strategy and is now recommended in patients with diabetes and a history of atherosclerotic cardiovascular disease [6, 7]. However, it seems inevitable that 16% of these patients who suffer from intermittent claudication will deteriorate so that, eventually, 7% will require a lower extremity revascularization procedure and 3–4% a major limb amputation [1, 3, 8, 9]. Literature reported rates between revascularization attempt and the requirement for a major limb amputation within the year vary between 6.3% and 16.2% [10, 11].

Despite significant advances in the endovascular and operative revascularization field, with rapidly evolving new technologies striving for reduction of immediate complications and making surgical management of the frail patient possible, feasibility and efficacy limiting caveats, such as calcification, microvascular disease, and restenosis, are yet to be overcome [9]. Modulation of intrinsic angiogenesis mechanisms has offered an attractive therapeutic target not only in patients with limited revascularization options, but also in those who have undergone endovascular or open PAD surgical management in view of stent or graft restenosis prevention. Angiogenesis and arteriogenesis are umbrella terms for highly complex physiological process which is under continuous positive and negative feedback control mechanisms [12, 13]. Although several related proteins have been identified as drivers of angiogenesis, including vascular endothelial growth factor-B (VEGF-B), VEGF-C, and placental growth factor [14], VEGF-A due to its key role in angiogenesis regulation has been the most extensively studied. Hand in hand with the VEGF family members, the VEGF receptors play a pivotal role in the modulation of the downstream intracellular signals. Dimeric VEGF binding to the VEGFR extracellular domains promotes receptor dimerization and mutual receptor subunit tyrosine transphosphorylation of the intracellular domains facilitating downstream pathway activation. VEGF ligands act through specific binding to three different cell membrane-bound receptors: VEGFR-1 (Flt-1), VEGFR-2 (Flk/KDR), and VEGFR-3 (Flt-4) [15]. Targeting, and more specifically inhibiting the action of VEGFRs, has been an attractive target especially in cancer pharmacology with vatalanib, sunitinib, and sorafenib being developed as universal inhibitors of all VEGF receptor tyrosine kinases while other monoclonal antibodies such as semaxanib and ZD6474 displaying selective receptor inhibition [15].

Pathophysiological processes that merge to create the picture of symptomatic PAD split under three main categories: vessel collateralization, plaque formation and inflammatory remodeling and subsequent lumen stenosis, and finally lumen neointima formation following surgical treatment. To successfully provide distal limb flow, without any surgical intervention, two options remain accessible, either increased collateralization and physiological bypassing of the diseased vessel section or reduction of the vessel lumen stenosis to a degree to improve flow. Despite its well-known proangiogenic function, gene therapy based on VEGF-A has had limited success in clinical trials for PAD [16, 17]. Intriguingly, patients with PAD have been shown to have impaired revascularization despite increased endogenous levels of VEGF-A isoforms, namely, VEGF-A¹⁶⁵. Recently, it has been suggested that this may be due to further alternative splicing of VEGF-A¹⁶⁵ RNA giving rise to two further isoforms in humans, VEGF-A^165a and VEGF-A^165b, and while VEGF-A^165a is a potent stimulator of vascular growth, VEGF-A^165b displays strong antiangiogenic effects by competitively binding to the VEGF-A receptor and inhibiting angiogenesis [18]. In terms of lumen stenosis, it has been stipulated that VEGF-A vessel treatment may promote vasodilation in a nitric oxide-dependent manner in peripheral artery models as well as balancing vessel permeability and endothelial cell growth promoting controlled vessel repair process in coronary artery models [19, 20].

Despite the promising evidence, VEGF-A increase has also been implicated with promotion rather than reduction of lumen stenosis in peripheral, coronary artery as well as arteriovenous fistula models [14, 21, 22]. Whether this is a result of experimental variability, inability to control endogenous VEGF levels, locoregional differences between PAD models (carotid vs. femoral vs. iliac arteries), or a fine balance between pro- and anti-angiogenic alternative splice variants as previously discussed remains to be delineated. A limited number of previous reviews have attempted to delineate the relationship between exogenous VEGF upon neointimal hyperplasia and subsequent arterial lumen stenosis, augmenting the ongoing controversy of findings in this field [23, 24]. Herein, we attempt to comparatively analyze available data regarding modulation, either increase or decrease, of VEGF family members and its effects upon neointimal hyperplasia and lumen stenosis in order to highlight the reasons behind these controversial findings with focus on peripheral arterial disease. We further aim to provide updates on our current understanding of the VEGF pathway in the context of PAD therapeutics. Thus, the present review aims to propose future research directions to clarify and enable safe VEGF applications in humans.

A systematic literature review was conducted according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA; Population Intervention Comparison Outcome (PICO) chart] (Fig. 1; Table 1). Independent literature search for relevant studies was performed up to September 20, 2021, and updated on the April 1, 2022, on six databases: Embase, Medline, CAB Abstracts, PsychINFO, Web of Science, and Google Scholar. Additional records were identified through other sources, including SSRNs eLibrary, Research Square, and medRxiv. The medRxiv search was simplified according to database search functionality. The references of the included studies were scrutinized for additional relevant studies. Search limitations included articles in English language. The following search strategy was employed: (VEGF or Vascular endothelial factor) and (*stenosis OR neointimal*).ti,ab,mh., limit to English language and adapted as per individual database requirements. The present study was registered under the international database of prospectively registered systematic reviews in health and social care (PROSPERO) with accession number CRD42021285988.

Full-text inclusion criteria were defined as below.

1.In vivo studies.

2.Assessing VEGF as a result of a molecular or chemical intervention.

3.Assessing either neointimal area proliferation (mm² or intimal: media depth ratio) or lumen stenosis (either as percentage change from control or mm²) as primary outcomes. Assessing either lymphangiogenesis or neovascularization (assessed quantitatively through imaging evidence or ankle/brachial pressure index change) as secondary outcomes with crude measurement numbers provided.

Full-text exclusion criteria were defined as below.

1.Nonperipheral vascular ischemia model (e.g., coronary or cerebral artery).

2.Not assessing any of the primary or secondary outcomes of the present study.

3.VEGF-specific protein (A, B, C, D, E) or isoform not stated.

4.Studies conducted in in vitro models.

Studies were included in the final analysis if at least one of the primary outcomes was addressed as per inclusion criteria statement. Excluded studies, with justifications, are presented in online supplementary Table S1 (for all online suppl. material, see www.karger.com/doi/10.1159/000527079).

The primary outcome of the present study was to assess the effect of VEGF sub-type modulation either increase/promotion or decrease/inhibition in protein levels upon vascular lumen stenosis and neointimal hyperplasia promotion. Secondary outcomes included the assessment of VEGF family sub-types with respect to lymphangiogenesis and neovascularization.

After removing duplicates, citations were screened by title and abstract, and then full texts were appraised to determine their eligibility by three authors (S.L.K., J.E., and M.G.) (Fig. 1). Three authors (S.L.K., J.E., and M.G.) independently conducted the abstract and full-text screening. Disagreements were resolved by the fourth author (R.M.). Study design, model organism, delivery construct, VEGF isoform sub-types assessed, effects on neointimal area proliferation, lumen stenosis, neovascularization and lymphangiogenesis, and duration of study were extracted per manuscript. Risk of bias was assessed using the Systematic Review Centre for Laboratory Animal Experimentation Risk of Bias tool (SYRCLE RoB) for animal studies and the Newcastle Ottawa scale for human studies (online suppl. Table S2, S3) [25, 26].

Statistical analysis and histogram plots were performed using GraphPad Prism (v. 9). Linear regression (best fit line and 95% confidence intervals [95% CIs]) was calculated from original data. Change in restenosis and neointimal proliferation area (mm²) was depicted as percentages where percentage change was defined as (control reported value-experimental reported value/control reported value *100) (Fig. 2). Percentage change was calculated to normalize data in comparison with control group values as time point of data collection as stated in each individual study. Best fit lines of available data were calculated with GraphPad Prism V. 9. VEGF pathway illustration (Fig. 3) was created with the BioRender.com., scientific image illustration online platform.

Following the PRISMA guidelines on systematic review search, we identified 36 eligible studies for data extraction. Full-text screening excluded 13 studies (online suppl. Table S1). A total of 22 studies remained all of which were included in this systematic review (Fig. 1; Table 1) [27-48].

Thirteen studies were conducted in rabbits [27, 30-34, 37, 39, 40, 42-45], four studies in mice [28, 29, 38, 41], one in canine models [35], and a further one in rats [36]. Finally, three studies included human subjects (Table 1) [46-48].

Of the included animal studies, nine studies used viral constructs to achieve VEGF gene transfer to native tissue [28, 30-34, 37, 41, 42]. Eight employed plasmid constructs for VEGF delivery, of which five in animals [12, 39, 40, 44, 45] and three in human participants [4, 46-48]. Two studies implemented nanoparticle delivery of VEGF [35, 43]. One further study used chemical blockade of VEGF with thalidomide [36]. One study used local application of VEGF solution [38], and finally, the last study included double gene knock-out models of the Fms-related receptor tyrosine kinase 1 gene (Flt1^−/−) (Table 1) [28].

Vast variation was also observed in the VEGF-A isoforms used as well as the individual VEGF sub-types which were examined across studies. Twenty studies effectively analyzed decrease or increase of VEGF-A [36, 38] or its isoforms (VEGF^{121, 164, 165}) [27, 28, 30, 31, 33, 35, 37, 39-48]. One of them employed membrane-bound, VEGF receptor 2 expression plasmid (binding site of VEGF-A), effectively increasing the action of VEGF-A [40]. Regarding VEGF-B, one study attempted the modulation of VEGF-B by employing a double Flt1 gene knock-out (Flt1^−/−) murine model [29], while another employed silence RNA against Flt1 (VEGF receptor-1), thus inhibiting VEGF-B activity [34]. Lastly, one study focused on VEGF-C modulation [32].

Study end time was also a significant variable across reported results. Median time across included animal studies was 28 days [min: 14; max: 84] [27-45]. Duration of human participant studies was extended to 217 days [range: 56–392 days] [46-48].

Sixteen animal model studies [27-32, 34-42, 45] reported change in arterial lumen post-VEGF modulation. Fourteen studies reported change in neointimal area proliferation (mm²) [27-29, 32-37, 39, 42-45].

Three studies were conducted on human participants and concentrated on VEGF effects upon peripheral arterial vessels [46-48]. Two studies suggested increase of neovascularization, as observed through ABPI and angiographic imaging, following intramuscular injection of VEGF¹⁶⁵ in calf/thigh muscles of ischemic limbs [46, 47]. The latter did not report significant differences among control and experimental groups, in terms of neovascularization [48], but identified significant increase (61%) of limb edema following intramuscular injection of VEGF¹²¹.

Accepting the recognized inter-species variability [49], we attempted to clarify whether previously identified contradictions among VEGF animal model derived data from overall study duration. Crude numbers (mm²) for neointimal hyperplasia and percentage of change in lumen stenosis were translated in percentage change in comparison with each individual study control group at matched time points to enable comparability across studies (Table 1; Fig. 2). Twelve studies assessed the VEGF-A effects upon lumen stenosis while eleven on neointimal area proliferation. Given the lack of adequate number of studies [3 studies] to assess the effects of other VEGF sub-types upon the defined outcomes, analysis was not feasible for VEGF-B or VEGF-C (Fig. 2).

Regarding VEGF-A treatment effects upon vessel lumen stenosis (Fig. 2a), the twelve studies that attempted to increase VEGF-A levels congruently suggested a decrease on vessel lumen stenosis even at 84 days. Equally, two studies which attempted to decrease VEGF-A levels concluded an exactly opposing result, i.e., an increase on vessel lumen stenosis. The inhibitory effect of VEGF-A increase upon lumen stenosis did not appear to reduce even in lengthier study durations (Fig. 2a). An examination of the interpolation curve indicated that the maximal effects of VEGF-A upregulation upon lumen stenosis prevention are exerted at day 51 across studies, resulting in 68.04% reduction of lumen stenosis in comparison with control (Fig. 2a).

Regarding neointimal area proliferation percentage change, nine studies employed VEGF-A upregulation to assess this outcome, one conducted both up- and downregulation of VEGF-A to augment results [42], while one focused on its negative regulation solely (Fig. 2b) [36]. All studies indicated that positive regulation of VEGF-A, regardless of upregulation mechanism, resulted in decreased neointimal area proliferation and suppressed hyperplasia. Data interpolation suggested that optimal results upon neointimal hyperplasia suppression were noted on day 21 with 71.23% prevention in a sustainable fashion in terms of duration. On the other hand, VEGF-A inhibition resulted in promotion of neointimal proliferation, albeit conclusions could not be safely drawn due to the limited number of studies and data variability (Fig. 2b) [36, 42].

Regarding the remaining VEGF sub-types, two studies assessed the effects of VEGF-B inhibition upon lumen stenosis (increase by 39.2% [29] and decrease by 41.25% [34]) and neointimal hyperplasia (8.3% increase and 47% decrease) by day 28 with contradictory results (Table 1). Of note, two different animal model organisms were employed to draw associations, respectively, mouse and rabbit, that may suggest inter-species variation of the VEGF-B downstream pathway [28, 34]. Lastly, only one study investigated the effects of VEGF-C modulation, noting a 42.85% decrease of lumen stenosis progression at day 42 [32]. Taken collectively, these data support that VEGF-A-positive modulation (increase) inhibits arterial lumen stenosis and neointimal hyperplasia, while data remain inconclusive regarding the other sub-types of the VEGF family.

In the present work, we have collectively assessed manuscripts addressing VEGF modulation in the context of peripheral arterial disease treatment to clarify the ongoing controversy concerning VEGF use in therapeutic angiogenesis and post-operative vessel restenosis [50]. To the best of our knowledge, this is the first systematic review to address this specific topic by highlighting the variability of experimental design published in the literature. Factors introducing variability across studies have been investigated, and speculations as to the sources of outcome heterogeneity have been raised [24, 31, 50]. A total of 22 studies were included in the final analysis, overall supporting the hypothesis that VEGF-A-positive modulation inhibits both arterial lumen stenosis and neointimal hyperplasia.

VEGF-A remains the most studied and pharmacologically targeted member of the VEGF sub-types for therapeutic angiogenesis and graft patency studies. Among its direct cellular actions lie endothelial cell mitogenesis and proliferation, survival promotion, control of vascular permeability, and angiogenesis regulation [14, 51]. Despite the importance of VEGF-A in cellular integrity, survival, and angiogenesis promotion, there remains at a complex interplay with the other members of the protein family, namely, VEGF-B, VEGF-C, VEGF-D, placental growth factor [52], and cross-binding cellular receptors such as VEGFR-1; VEGFR-2; VEGFR-3 (Fig. 3). Furthermore, human VEGF-A is known to undergo alternative exon splicing leading to a plethora of isoforms, namely, VEGF¹²¹, VEGF¹⁶⁵ (VEGF¹⁶⁴ in murine models), VEGF¹⁸⁹, and VEGF²⁰⁶, which nonetheless do not have the same physiological efficacy in driving VEGF-A cellular targets and biological functions [53]. Additionally, soluble forms of VEGF receptors may exert strong inhibitory effects upon VEGF-A, in contrast to their membrane-bound counterparts [54, 55]. This evidence strongly highlights the complexity associated with VEGF signaling, its essential role to vascular homeostasis, and the different affinities of its key receptors. On a clinical level, tissue ischemia due to critical vessel stenosis and subsequent hypoxia drives symptomatic PAD. These factors stimulate endogenous VEGF production which in turn drives angiogenesis. A vital step in this process is the sprouting of endothelial cells from pre-existing capillary beds and their subsequent migration, controlled proliferation, which leads to the formation of new collateral lumens and intussusception of existing capillaries [56, 57]. This process enables the incorporation of circulating endothelial progenitors to the pre-existing, diseased vessels promoting healing [58-60]. Overall, a fine balance of the proangiogenic VEGF-A splice variants (VEGFA^165a) and downregulation of antiangiogenic ones (VEGFA^165b), which commonly are found elevated in PAD patients, drives controlled lumen restoration and successful collateralization instead of deregulated growth and neointimal hyperplasia [61, 62].

Previous attempts to clarify preclinical data and inform the applicability of VEGF-A as a treatment option not only for patients noneligible for revascularization options, but also for those that have undergone endovascular or open procedures, accentuated a significant disparity between reported outcomes. Past systematic literature reviews of preclinical PAD models have produced contradictory summative outcomes regarding VEGF-A exogenous increase and its subsequent benefits upon lumen stenosis prevention and angio- and arteriogenesis. Ganta et al. [61], in their systematic review, conclude that inhibiting a VEGF-A isoform in ischemic muscle promotes perfusion recovery in preclinical PAD models whereas Gabhann et al. [63] suggested that activation of VEGF-A promotes a similar perfusion recovery. Differences between across studies may stem from a spectrum of experimental variables. The instability of preceding transfection construct expression, employed to modulate VEGF-A levels, necessitates the use of contemporary, durable expression vectors in future experiments [64]. Furthermore, soluble decoy receptors may have unpredictable effects, given the complexity of the VEGF pathway interplay and thus resulting in contradictory evidence (Fig. 3) [65]. Another important variability factor among studies may also stem from the induction of alternative VEGF-A splice variants, which as recently shown may promote or in fact inhibit angiogenesis [18].

Other sources of variability may lie within the multiple angiogenic stimuli which have been identified as positive regulators of VEGF-A in response to hypoxia, tissue trauma, and inflammation as well as tissue ischemia (Fig. 3) [66-68]. While VEGF-A mainly exerts its effects upon binding to the VEGFR-2, encoded by the Flk2 gene, other receptors accommodate VEGF-A binding with differential effects. For example, VEGFR-2 also acts as a binding site for VEGF-C, -D, and -E with variable downstream target activation (Fig. 3) [14]. For simplicity, canonical activation of the VEGFR-2 by VEGF-A results in angiogenesis, activation of VEGFR-1 by VEGF-B promotes cell survival through the PI3K/AKT cascade (Fig. 3) [69-71], while VEGFR-3 activation by VEGF-C and -D leads to lymphangiogenesis (Fig. 3) [72-74]. Of note, the complex interaction of the VEGF protein network may result in synergistic or antagonistic effects via either direct binding or initiation of feedback loops, in response to variable exogenous stimuli, thus enabling fine-tuning of cellular responses (Fig. 3) [14, 67, 68].

Moving from the molecular to an organism level, inter-species variation and overall experimental design discrepancies were notable across studies [27-45]. While differential experimental designs may be overall beneficial in assessing holistically the VEGF pathway interaction, they may cloud outcome clarity, creating controversy in the clinical applicability of VEGF-A in PAD management. The VEGF pathway differences across animal models and humans are not unexpected given the widely reported variability of other major morphogenesis and proliferation pathways such as the mammalian target of rapamycin, PI3K/Akt, and the rat sarcoma [75-77]. VEGF species variability in both structure and function emphasizes the need for identification of the optimal animal model to enable comparable VEGF modulation outcomes with human counterparts. Given the pathway transferability between swine (Sus scrofa) models and humans in muscle integrity modulation pathways, swine models prove superior for the study of VEGF effects on PAD, in comparison with murine or rabbit ones [76]. Intriguingly, among model organisms VEGF-A proteins compared against the Homo sapiens reference sequence, Sus scrofa VEGF-A isoform demonstrated the highest sequence cover (100%) with an 86.84% sequence identity in Basic Local Alignment Search Tool (BLAST) analysis [78]. Taking into account the experimental cost-effectiveness as well as all ethical and ecological considerations associated with the use of swine models [62, 79], human xenografts may also be considered as a viable, more clinically relevant alternative [80, 81].

Regarding selection of VEGF isoforms, VEGF¹⁶⁵ is the most frequently expressed isoform in tissues as well as the most physiologically active [14]. In view of the observed isoform selection variability across studies, we suggest measurement of VEGF¹⁶⁵, rather than other VEGF isoforms, in PAD-oriented experimental designs. More importantly, a human study employing VEGF¹⁶⁵ plasmid vectors as tissue delivery system reported VEGF expression in local but also distant areas at 10 weeks post-injection. Consequently, the local but also the systemic effects of VEGF¹⁶⁵ conveying plasmid vector administration should be thoroughly investigated [46]. Lastly, whether the activation of VEGF-A alone or congruent activation of downstream or parallel pathways is responsible for the beneficial effects observed upon neointimal hyperplasia diminution and consequent decrease in lumen stenosis remains to be delineated. Overall, variability of outcomes originates from the variability of experimental designs. Therefore, the applicability of VEGF-A in PAD management clinical practice remains to be clarified by a uniform body of further animal model experiments.

The present work reflects the first systematic review conducted according to the PRISMA guideline criteria, to collectively assess manuscripts addressing VEGF modulation in the context of PAD, while highlighting experimental design and model organism variability and effects upon reproducibility in human participants. Here, we also provide guidance as to selection of model organism, specific VEGF isoform assessment, and delivery system a to generate clinically relevant data.

The variability of animal model employed, overall experimental design, VEGF sub-types, and VEGF-A isoform measured as well as timing of outcome measurement was highlighted throughout this study, but inevitably may have introduced significant data bias. Only three studies were undertaken in human participants with noncomparable outcomes. Of note, among human studies, only one was randomized [48].

Herein, we demonstrated that VEGF-A-positive modulation inhibits arterial lumen stenosis and neointimal hyperplasia. We also suggested that modulation of the VEGF₁₆₅ isoform in swine models, with an extended experimental period, e.g., 58 days for neointimal proliferation or 98 days for lumen stenosis as measurement outcomes, may be the most physiologically relevant experimental option in the context of delineating the PAD and VEGF association. Further fundamental and clinical research is required to enable applicability of VEGF-A therapy in the clinical management of PAD.

The first author would like to thank Mr. Michael Sharp for his ongoing guidance and support.

The authors have no ethical conflicts to disclose. The present work reflects a systematic review of published data.

The authors have no conflict of interest to declare.

The authors declare that no funds, grants, or other support were received during the preparation of the manuscript.

Stavroula L. Kastora: conceptualization, data collection and analysis, critical appraisal, manuscript drafting and editing, final corrections, submission, and final approval. Jonathan Eley: data collection, critical appraisal, manuscript editing, and final approval. Martin Gannon and Ross Melvin: critical appraisal, manuscript editing, and final approval. Euan Munro: expert opinion, manuscript editing, and final approval. Sotirios Makris: supervision, manuscript editing, and final approval.

All supporting and associated data are available as online supplementary material.