Due to the lack of specific surface markers on EPCs, various BM cell fractions were short-term cultured in endothelial differentiation media to enrich EPCs. For example, CD133, which is displayed on immature hematopoietic stem cells (HSCs), was used for culture derivation of EPCs [69]. On the other hand, for therapeutic purposes, the total MNC population, which has been widely used to obtain EPCs, was short-term cultured for 4.7 days on vitronectin- or fibronectin-coated dishes and the attached, or adherent, cells were used as EPCs, though not all of the cells collected were considered EPCs [14, 39, 63, 70, 71]. These cells expressed endothelial-like characteristics by the uptake of acetylated low-density lipoproteins, the binding of lectins, the expression of several EC-specific proteins (VEGFR-2, Tie2, vascular endothelial [VE]-cadherin, von Willebrand factor, endothelial nitric oxide synthase [eNOS], and CD146), and a low proliferation rate. On the contrary, other studies have refuted the endothelial-like characteristics of EPCs because these cells also displayed monocyte/macrophage markers, such as CD45, CD11b, and CD11c [54, 72-74]. Recent studies have also, instead, referred these cells as angiogenic cells [74], for they contribute more to vessel formation through angiogenic effects rather than form ECs in vivo. EPCs are able to incorporate into the neovasculature and differentiate into ECs [20, 39, 63]. Also, Wang et al. [75] postulated that EPC containing cell population is inherently multipotent and may include smooth muscle progenitor cells. Additionally, other studies pointed that EPCs were able to differentiate into cardiomyocytes and smooth muscle cells in mouse [47] and rat [17, 21] models of MI; peripheral EPCs, mobilized by granulocyte-colony stimulating factor (G-CSF) administration and myocardial ischemia, have displayed cardiac-, muscle-, liver-, and neural-lineage markers [76, 77]. However, this evidence of extra-endothelial transdifferentiation is sporadic and not widely accepted after the original publications.

In one study, EPCs have been shown to transdifferentiate by fusing with cells of different lineage, such as newly formed cardiomyocytes in mice, administered with human EPCs, contained both human and mouse X-chromosomes [78]. However, only 3% of the ECs displayed any evidence of human-mouse cell fusion. In addition, most studies have shown little or no evidence of fusion between EPCs and cells of different lineages. In most recent studies, the transdifferentiation of EPCs has been shown to be possible but not prevalent.

Expression of certain factors and molecules, such as eNOS and inducible nitric oxide synthase (iNOS), which increase circulation by dilating capillaries [42], and pro-angiogenic and anti-apoptotic growth factors (i.e. hepatocyte growth factor [HGF], insulin-like growth factor 1 [IGF-1], and VEGF) [51], is one of the ways that EPCs contribute to tissue regeneration and preservation. The release of certain growth factors causes the recruitment of EPCs around the ischemic area [52] and, in turn, stimulates the cells to proliferate, migrate, and maintain the survival of EPCs and pre-existing mature endothelial cells [79].

In earlier studies, EPCs were often administered systematically; however, in recent studies, due to the low rate of retention, cells have been transplanted or injected directly onto the affected area. Transplantation of EPCs [14, 16, 20, 80, 81] facilitated neovascularization in animal models of acute myocardial ischemia, and analogous effects in humans [30, 82, 83].

Late EPCs, or endothelial colony-forming cells (ECFCs), were discovered from circulating MNCs that were collected after more than 2 weeks in conventional EPC culture. These cells may be cultured variably; however, they commonly have round morphology, fast cell proliferation rate, and express similar surface markers (EC makers) [73, 84]. ECFCs need to be studied much further and their culture systems developed more before any application to therapeutic or regenerative use, because they were able to neither convincingly show paracrine effects nor show vasculogenic effects without Matrigel™ plug assay [84, 85]. ECFCs' capability to differentiate into ECs and their regenerative and therapeutic effects on vasculature in ischemia need to be troubleshot. Furthermore, because there is no defined line between ECFCs and ECs, the possibilities that these cells are primitive circulating ECs or are ECs sloughed off from vessels still remain to be established.

CD31, also known as platelet endothelial cell adhesion molecule (PECAM)-1, is a 130-kDa transmembrane protein consisting of six extracellular immunoglobulin folds. In its cytoplasmic domain, there are two immunoreceptor tyrosine-based inhibitory motifs (ITIM) for interactions with signaling molecules. CD31 is expressed on the cell surface of ECs and hematopoietic cells, such as monocytes, platelets, neutrophils, natural killer cells, megakaryocytes, and some T cells. CD31 mediates homotypic adhesion between adjacent ECs as well as between ECs and leukocytes [114, 115]. A role of CD31 in migration through ECs has also been reported for neutrophils and monocytes [116], and subsequently for numerous other cell types, including natural killer cells [117], hematopoietic progenitor cells (HPCs) [118] and certain subsets of lymphocytes [119, 120].

Cell-cell and cell-extracellular matrix interactions are two of the essential roles during blood vessel formation and regression. CD31 has been gaining attention as it is being recognized to suppress cell death through interactions, such as homophilic interactions of CD31 that are in between ECs and monocytes [121]. The interaction of CD31 in suppressing apoptosis suggests that CD31 releases pro-survival signals, and the interaction also has been reported to induce Akt, a serine/threonine protein kinase, phosphorylation. On the contrary to apoptosis suppression characteristics, CD31 also assist ECs in apoptosis through metalloproteinase-dependent cleavage and caspase-mediated cleavage of the cytoplasmic tail [122].

Additionally, given that it is expressed abundantly in ECs, CD31 has been shown to be involved in the initial stabilization and formation of cell-cell contacts at lateral junctions of ECs, the maintenance of a vascular permeability barrier, modulation of cell migration, transendothelial migration of monocytes and neutrophils, and formation of blood vessels in angiogenesis [114, 116, 123, 124]. Moreover, earlier studies have shown that CD31 formed a functional complex with cadherin 5, β-catenin, and F-actin to control EC tube formation [125]. However, a later study showed that CD31 was involved in the adhesion/signaling events necessary for the migration of ECs and subsequent tube formation during angiogenesis, independent of VE-cadherin [126].

To advance cell therapy, it seems very crucial to target and identify cells that are specialized for humoral effects, since recent studies almost universally reported that the major therapeutic mechanism of BM-derived stem or progenitor cells is humoral or paracrine effect [52, 55-57]. Therefore, recent studies have used specific surface markers to isolate cells enriched in humoral activities from BM and/or PB [61, 62]. Using surface makers allows the cells to be isolated without cell culture, which saves costs, and safely evading any harmful effects that may come with using animal serum. In a series of studies, CD31 was used as a marker to find BM-derived cells that have multimodal effects, such as angiogenic, vasculogenic, higher adhesion, and almost exclusive hematopoietic stem/progenitor cell activities (Fig. 1).

Based on genome-wide gene expression analyses of mouse BM (mBM), human BM (hBM), and hPB, angiogenic genes were enriched in CD31⁺ cells when compared with CD31^- cells [61, 62]. Compared to their negative counterparts, hBM-CD31⁺ cells expressed higher levels of heparin-binding EGF-like growth factor (HBEGF), which is active in vascular smooth muscle cell recruitment [127] and IL-8, a macrophage-derived mediator of angiogenesis [128]; mBM-CD31⁺ cells expressed high levels of Angpt-1, an eminent angiogenic growth factor [129, 130], and GATA2, a transcription factor that proliferates angiogenesis when activated [131]. Similarly, hPB-CD31⁺ expresses extremely high levels of neuropilin-1 (NRP1), which is a necessity in vascular development and a mediator in VEGF-dependent angiogenesis [132], compared with the negative cells.

The recurring questions that need to be addressed are whether hematopoietic cells can give rise to ECs [46, 133, 134] and what is the true identity of those cells. A majority of studies, including ours, have demonstrated that BM-derived cells can generate ECs. On the other hand, several studies have refuted such notion [48, 50, 135] by using MI models to prove vasculogenic effects [48, 49]. However, those refutations are inappropriate since a MI model accepts minimal amount of engraftment from any cell, including BM-derived cells. Other studies supported our notion that the retention of ESC-derived cardiomyocytes or cardiac stem or progenitor cells was minimal in a MI model as well [136-138]. MI hinders engraftment and survival of any implanted cells, such as constant motion of the heart, inflammation in acute stages of infarction, insufficient oxygen and nutrients, and tight junctions between cardiomyocytes. One study reported that it is common for EC transdifferentiation to occur in transplanted heart samples [139]. In other independent studies, with the correct conditions, such as suitable BM-derived cells with permissive models, BM-derived cells can give rise to ECs [140, 141]. Based on the results of these studies, cell types and/or environment is very crucial to generating ECs from BM or PB cells.

In recent studies on the vasculogenic effects of CD31⁺ cells [61, 62], compared with CD31^- cells, mBM-, hBM-, and hPB-CD31⁺ cells gave rise to a large number of EPCs in culture and expressed EC-specific markers, such as von Willebrand factor, VEGFR-2, VE-cadherin, and CD31. In addition, hPB-CD31⁺ cells formed aggregates and, subsequently, underwent tubular structural changes within the round cell cluster by day 10 followed by the formation of complete linear tubular structure that mimicked in vivo vasculogenesis. This linear tubular structure had stained positive for lectin and took up acetylated human low-density lipoprotein, indicating EC characteristics. In vivo vasculogenesis studies had also been investigated. By using a mouse model of hindlimb ischemia, vasculogenic activities of CD31⁺ cells had been tested [61, 62]. However, due to the divisive notion of transdifferentiation potential of BM cells [48, 50, 54, 134, 140, 141], the confirmation methods used were rigorous and definitive: confocal microscopy with 3D reconstruction of multiple images were used to clearly demonstrate that a fraction of CD31⁺ cells were colocalized with the ECs within the vascular structure even up to 8 weeks after; flow cytometric analysis of enzymatically digested hindlimb tissues showed that up to 4% of the ECs in the ischemic tissues were derived from transplanted mBM- or hPB-CD31⁺ cells; fluorescent in situ hybridization of the digested tissues further confirmed the contribution of hPB-CD31⁺ cells into ECs [62]. This clear adaptation of definitive methods were the first to prove transdifferentiation of hematopoietic cells. Clearly, this experiment had proved that directly injected CD31⁺ cells give rise to functional ECs in ischemic tissues.

Low retention of injected cells is one of the major problems that hinder the therapeutic effects of cell therapy because a stable engraftment and a steady survival are needed for optimal cell therapy. Studies reported that less than 25% of engraftment of transplanted neonatal cardiomyocytes were present within 24 hrs of MI in animal models [142] and, similarly, only 5% of CD34⁺ injected cells remained after 1 hr in human patients with MI [143]. Additionally, a majority of the directly injected, cultured early EPCs in myocardium post-MI had disappeared within a week [52]. Cell adhesion is also another crucial factor that can increase survival. Cells that adhere to extracellular matrix have higher chance of survival since they are able to avoid anoikis, an apoptosis caused by the lack of adhesion to extracellular matrix. CD31 was first found as an adhesion molecule [115, 144] and had been shown to mediate cell-cell interaction via hemophilic interactions between CD31^-expressing cells [145]. Genome-wide gene expression data showed high levels of genes relevant to adhesion, chemokine production and reception, and extracellular matrix in CD31⁺ cells [62]. Furthermore, cell adhesion assay showed that CD31⁺ cells have the capacity to adhere to more matrix proteins, such as fibronectin, collagen, laminin, and vitronectin [61, 62], than CD31^- cells can. In vivo samples analyzed by the confocal microscopy and FACS analysis confirmed higher engraftment and survival of injected CD31⁺ cells compared to the CD31^- cells or MNCs.

Although CXCR4/SDF-1 is a recognized signaling axis to mediate cell engraftment and migration, the engraftment of CD31⁺ cells do not depend on this axis because, when the levels of CXCR4 were measured, there was no difference in the expression between CD31⁺ and CD31^- cells. On the other hand, CD31 molecules and other adhesion molecules, ICAM4 and integrin α, which were expressed more in CD31⁺ cells than CD31^- cells, could have been involved in the higher engraftment of CD31⁺ cells. The higher engraftment of CD31⁺ cells would allow mBM- and hPB-CD31⁺ cells to increase angiogenic and vasculogenic capabilities.

According to the gene expression studies, hematopoietic stem and progenitor cell gene levels were reported to be higher in mBM-CD31⁺ cells and in hBM-CD31⁺ cells than in their corresponding negative cells, suggesting that more hematopoietic stem and progenitor cells are expected to be found in CD31⁺ cell population. Concurrent with this hypothesis, according to in vitro colony-forming assay and in vivo BM cell transplantation, hematopoietic stem and progenitor cells are almost exclusively found in mBM-CD31⁺ cells; also, more than 90% of HSCs, multipotent progenitor cells, common lymphoid progenitor cells, and common myeloid progenitor cells in mBM express CD31 [61]. Similarly, in humans, CD31 was expressed in 99.8% of CD34⁺CD133⁺ and 89% of CD34⁺CD133^- hBM cells, and clonogenic HPCs are found in hPB-CD31⁺ cells [62]. These findings suggest that HSCs and HPCs are heavily enriched in the CD31⁺ cell population.

There is some possibility that HSCs or HPCs included in the CD31⁺ population could have played a role in ischemic tissue repair because CD34⁺ cells, which include most HSC and EPCs, were known to be effective for treating ischemic cardiovascular diseases in animal models and human patients. Since the therapeutic effects of hBM-CD34^-CD31⁺ cells and hBM-CD34⁺ cells were reported to be similar in improving mouse limb ischemia, it is more likely that non-HSCs could play a significant role in improving limb ischemia and/or therapeutic neovascularization.