These results reveal that angiogenic potency is not simply dependent proportionally on EPC numbers, as proposed in many prevailing interpretations, but is complex and may involve multiple cell types with different roles, as has been reviewed [11, 12]. In our model the angiogenic potency appears to comprise two distinct activities mediated by different cells, 1) true endothelial progenitor activity with cell incorporation into vessel walls provided by EOC, and 2) proangiogenic amplification of vascularization provided by CD34+ HPC from mobilized peripheral blood and cord blood, but not from bone marrow. Different sources of HPC that may have essentially similar clinical hematopoietic potency can have quite different angiogenic properties, and HPC content does not predict angiogenic potential. Other cells, such as monocytes or mesenchymal stromal cells, seem irrelevant in this model.
This distinction was not apparent when we proposed this study in 2006 in anticipation of a need to supply cells for therapeutic vascular repair in myocardial infarction or severe peripheral ischemia, which was and still is focussed primarily on the use of autologous HPC from bone marrow or mobilized blood, with or without enrichment of HPC by selection of CD34+ or CD133+ cells. During our studies a number of reports have appeared, which show 1) that HPC have a proangiogenic effect, particularly CD133+ HPC ; 2) that EOC probably represent true endothelial progenitors that incorporate into neo-vessels and are derived from CD34+CD133- cells and not CD34+CD133+ cells [13, 15]; 3) that early CFU-EPC are derived from monocytes and while these may express some endothelial markers, they are unrelated to EPC, do not proliferate, and do not incorporate into vessels [26–28]; and 4) that mesenchymal stromal cells do not contribute directly to angiogenesis but may support it . It is possible to review these individual reports and come to essentially the same conclusion that we do from our study, so our conclusions are not original. However, papers contradicting these findings continue to be published, so it is important to provide support for these conclusions. Also, instead of focussing on a particular cell, our work systematically compares under the same conditions, a number of different proposed sources and subpopulations of cells for therapeutic vascular repair and shows that they are not equivalent. We provide important support for the various individual studies above, for which there is not yet a consensus to judge from the range of interpretation and opinion in many current publications, and may help provide support for informed redesign of cellular therapy for vascular repair.
The sponge model, in which the implanted sponge is spontaneously encapsulated and vascularized by the host, is often called an inflammatory angiogenesis model . It is an established, widely used model which allows investigation of angiogenesis in vivo and which was well suited for the comparison of the effects of different implanted cells on new vessel formation [23, 34]. It is unlikely that the model replicates exactly the process of reparative angiogenesis in ischemic tissue , and these results may not necessarily be extrapolated to pathophysiological angiogenesis (for example, in response to myocardial infarction or in limb ischemia). In our hands the sponge model adapted well for the comparison of the effects of different implanted cells on the angiogenesis that occurs within these sponges, where it discriminates between the direct incorporation of implanted cells into neo-vessels and the proangiogenic stimulation of vessel number by different implanted cells. However, it may be that more subtle effects could be masked by the inflammatory response to the sponges .
In our experience, EOC derive only from a subset of CD34+/CD133- MNC, confirming previous reports [13, 15]. This is contrary to a number of publications stating that CD133 is found on EPC. This inconsistency in definition confounds an objective assessment of many published reports of relationships between EPC and angiogenic potential that employ different phenotype bases to quantify or enrich EOC, as reviewed recently . The subset of CD34+/CD133- cells recently described by Estes et al.  as defining the true EPC phenotype are extremely rare in circulation and also rare in accessible sources of HPC such as umbilical cord blood or mobilized peripheral blood. We find a similar very rare (fewer than one in a million) MNC subpopulation giving rise to EPC that are CD34+/CD133-/CD45-/CD146+. If CD146 is an alternative to CD31 for discriminating EPC then these cells may be the same CD34+/CD133-/CD45-/CD31+ subpopulation described by Estes et al.  as EPC with in vivo vessel-forming capacity. These CD34+/CD133-/CD45-/CD146+ cells can be detected in normal peripheral blood and cord blood MNC, both of which give rise to EOC, but are below our limit of detection in bone marrow or mobilized blood MNC, neither of which give rise to EOC (Tura et al., manuscript submitted). Thus, detection of this cell subpopulation appears to coincide with our ability to grow EOC from these different sources. We find from outgrowth studies that the frequency of EOC colonies (endothelial colony-forming cells, ECFC) in cord blood MNC is approximately one in five million, and is approximately one in thirty million in normal peripheral blood MNC. ECFC frequency remains low even in CD34-enriched MNC, indicating that EPC are a very small component of total CD34+ cells, but no ECFC are found in CD34-depleted MNC . Therefore, apart from homogeneous EOC from culture, the cell populations implanted in this study will contain very few or no true EPC as indicated by the very low ECFC frequency, even when CD34+-enriched cord blood cells are used . It is not surprising, therefore, that no human cell incorporation into sponge vessels was found in these cases.
It may be that adult EPC have been ascribed a close relationship to HPC by their coincidental expression of CD34, which is a marker of both hematopoietic progenitors and mature endothelial cells. However, HPC and EPC might not be as closely related in adults as inferred, and circulating EPC might not derive from bone marrow. The endothelial stem cell niche could be principally located elsewhere, such as in the vasculature itself. Ingram et al.  have described a hierarchy of endothelial lineage cells, including progenitors that can be recovered from vessel walls. This may be why we consistently fail to recover EOC from G-CSF mobilized blood MNC, where HPC leave the bone marrow and appear in the circulation but are not apparently accompanied by EPC, which give rise to EOC . Furthermore, existing circulating EPC could be diluted to below detection by the leukocytosis which accompanies G-CSF mobilization. This could also explain why we and others  were unable to grow EOC from bone marrow MNC and instead obtain mesenchymal stromal cell outgrowth under conditions that generate EOC when other cell sources are used. These bone marrow-derived MSC had no effect in our sponge vascularization model, showing neither incorporation into vessels, nor promotion of vascularization. Similarly, purchased MSC derived from human adipose tissue had no effect on sponge vascularization (not shown).
Melero-Martin et al.  have shown that implanted MSC can contribute to vascularization in implanted Matrigel plugs by synergizing with implanted human EOC in co-transplants. Here it was proposed that the MSC may contribute by secretion of vascular endothelial growth factor (VEGF) to induce implanted human EOC to form vessels, and the MSC may also stabilize vessels by forming perivascular structure around the endothelial-derived lumen. However, they found that EOC implanted alone did not result in vascularization of the Matrigel plug and neither is it spontaneously vascularized by the host. In contrast in our sponge implant model sponges are spontaneously vascularized by the host, and MSC implant did not apparently affect this spontaneous vascularization whereas mobilized or cord blood CD34+ cells significantly promoted it. From this we can infer that these CD34+ cells are more potent than MSC in promoting angiogenesis. We did not study co-implantation of different cell types, but human EOC implanted alone without accompanying MSC incorporated into growing vessels while MSC alone did not, and neither EOC nor MSC promoted increased vessel density. It seems that the sponge model and the Matrigel plug model differ. The sponge model has been used extensively as a model of angiogenesis and to measure effects of exogenous factors on this angiogenesis [37, 23, 34, 25]. It may be that since no spontaneous host vascularization takes place in the Matrigel plug, that model may reflect de novo vasculogenesis. It may also be that vasculogenesis and angiogenesis depend differently on EPC and accessory cells, accounting for the differences between our findings and those reported by Melero-Martin et al. . Since the sponge model indicates that available EOC are incorporated when new vessels form, it may be that the Matrigel plug model says more about the ability of the cells that are co-transplanted with EOC to promote vessel formation than about the inherent capacity of EOC to form vessels.
Melero-Martin et al.  also used anti-α-SMA antibody to discriminate MSC and perivascular cells from luminal endothelial cells. In our studies we employed cross-reactive anti-α-SMA which did not discriminate mouse (host) from human (implant) cells. This antibody stained brightly and in most cases it appeared to stain the endothelial lumen of vessels (Figure 2 and Additional file 1, Figures 2a and 2b) as has been reported for some human and mouse vessels . It also stained human EOC and HUVEC grown in vitro (not shown), which have also been reported to express α-SMA, so anti-α-SMA is not specific for perivascular cells and is expressed on luminal endothelial cells, at least in some cases. In some areas of most sponge sections acellular (non-nucleated) deposits of α-SMA were identified by this antibody (Additional file 1, Figure 2a, row B and Figure 2b, row C). This anti-α-SMA also stained perivascular cells in some vessels (see Additional file 1, Figure 2a, rows A and C) in that such cells appeared in an adjacent layer peripheral to the lumen, and often with their long axis vertical to the lumen. Similarly, apparent perivascular layers peripheral to the lumen in mouse-only vessels were also identified with anti-CD31 (Additional file 1, Figure 3a, row A), anti-endoglin (CD105) (Additional file 1, Figure 4a row C) and anti-CD146 (Additional file 1, Figure 5a, row B). Thus, all the cross-reactive antibodies, including anti-α-SMA, did not discriminate between endothelial or perivascular cells. This also differs from the report by Melero-Martin et al. , where α-SMA is claimed to be associated only with MSC and not endothelial cells. The occurrence of vessels with perivascular layers was confined to a minority of mouse-only vessels, was never found where human cells incorporated into vessels, and had no discernible association with any implanted cell type. We draw no conclusions from this but note their presence and their prominent vivid staining, especially by anti-α-SMA, but also by other antibodies.
While HUVEC, expanded routinely in vitro through two to three passages, showed normal growth and tubule formation in Matrigel, they did not incorporate into vessels when implanted in sponges. In contrast, freshly isolated HUVEC, not yet passaged, incorporated into host vessels (Table 3). According to Ingram et al., freshly isolated vessel endothelial cells should contain EPC , but it appears that they quickly lose the ability to incorporate into vessels on expansion in vitro. EOC also eventually lost the ability to incorporate after extensive passage (Table 3). We have not characterized the kinetics of this loss of incorporation ability either in EOC or HUVEC, but it could reflect a change from progenitor to mature endothelial cell. We have not identified anything within growth characteristics, expression of a wide range of surface markers, or patterns of expression of various genes, that discriminates EOC from mature HUVEC other than this ability to incorporate into vessels in the sponge model of angiogenesis (Tura et al., manuscript submitted). This should be examined further. We also note that while EOC were cultured on type-1 collagen, HUVEC were cultured on uncoated plastic, and the possible role of extracellular matrix in maintaining immature status or promoting maturation should be investigated. Cells for therapeutic use should be capable of incorporation into vessels, so it is important to retain this property during in vitro expansion if EOC are to be used for vascular regenerative therapy. The sponge model can examine this property.
Neither adherent monocytes, nor MNC cultured on fibronectin, influenced sponge vascularization by incorporation or proangiogenic activity. Indeed, of all the cells studied, adherent monocytes slightly, but not significantly, inhibited vascularization (Table 2). However, isolated human cells expressing endothelial markers were found in the sponges in which monocyte-enriched cells were implanted. The derivation of early CFU-EPC from monocytes on fibronectin, and numerous other reports of expression of endothelial markers by monocytes when cultured on fibronectin [26, 27, 29], suggest that the cells observed in the sponges are monocytes, or were derived from them. The culture of MNC on either fibronectin or collagen to produce putative EPC has caused some confusion between true EOC and differentiated monocytes. Our experience (unpublished) suggests that culture on fibronectin produces monocyte-derived cells that can mimic endothelial cells by phenotype, whereas, consistent with other reports , true EPC are produced only on collagen type-1. However, others have described production of functional EOC on fibronectin  or even on uncoated plastic under certain circumstances, so it may be that the collagen effect is quantitative rather than mandatory. However, since monocytes appear to require fibronectin for endothelial-like differentiation and formation of early CFU-EPC, the exclusive use of type-1 collagen for endothelial outgrowth from MNC might avoid any confusion between true EOC and differentiated monocytes. Since monocytes did not apparently influence vessel growth in this study, they may not be an important component of cellular therapy for vascular regeneration, and may be irrelevant when considering acquisition of cells for that purpose. However, since there is a non-specific host inflammatory response to the sponge and migration of host inflammatory cells, including monocytes, into the sponge, it may be that this masks any stimulatory effect provided by the addition of human monocytes. If monocytes or their derivatives have a masked proangiogenic effect, it must be different from the proangiogenic effect of HPC, which is evident. It has also been suggested that monocytes may have an ability to infiltrate ischemic tissue and provide pathways for vascularization by endothelial cells [42, 43], which may not have been tested in this model.
While EOC have been proposed as an option for therapeutic vascular regeneration [44, 41], to date all clinical trials have employed HPC from patients' autologous bone marrow or G-CSF-mobilized peripheral blood. Some recent studies show that bone marrow-derived HPC-like cells do not contribute to angiogenesis by incorporation into regenerating vascular endothelium [45–47]. Our study shows that the prominent, and possibly the only, angiogenic potential of HPC-like cells is proangiogenic activity, not as sources of EPC. This proangiogenic activity is not seen at low HPC frequency in unfractionated MNC and is only found when HPC are enriched through CD34+ cell selection, so is evidently quantitatively related to CD34+ cell numbers. Neither EOC nor HUVEC showed a proangiogenic effect despite EOC and early HUVEC showing incorporation into vessels, so their role seems passive, dependent only on availability. It is notable that the strongest, most significant proangiogenic effect came from G-CSF-mobilized CD34+ cells, followed by cord blood CD34+ cells, and that surprisingly bone marrow-derived CD34+ cells were not proangiogenic in this model. While HPC may be characterized by expression of CD34 or CD133, these markers are not equivalent or homogeneously expressed by HPC, and different subpopulations exist which are CD34+CD133-, CD34+CD133+ and CD34-CD133+ . The observed proangiogenic capacity follows the proportions that we have found in different HPC sources of CD34+ cells that co-express CD133, which are greatest in mobilized blood HPC (around 80%), intermediate in cord blood HPC (around 53%) and smallest in bone marrow HPC (around 13%). It has been reported that CD133+ HPC are proangiogenic, without incorporating into neovessels [19, 48]. Recently it has been shown that in subpopulations of CD34+ HPC that differ in complex phenotype only by CD133 expression, the CD133+ cells show proangiogenic activity, while the CD133- cells do not . Thus, it may be that the proangiogenic effect is mediated only by (some or all) CD133+ cells, but not by CD133- cells. This might explain the differences we found between the various sources of HPC, reflecting the proportions of CD34+ cells that co-express CD133+. If so, this could indicate that expression of CD133 may be better than expression of CD34 for selection for proangiogenic HPC. Further studies are needed to determine and characterize subpopulations of HPC-expressing proangiogenic activity, and whether they can be expanded or enhanced in vitro for therapeutic use.
This study suggests that any observed clinical benefit reported in trials of HPC therapy for vascular regeneration has predominantly been mediated by indirect proangiogenic effects rather than through direct incorporation of any EPC contained within these HPC sources. This may especially be the case in trials that have employed CD133+ cells enriched from HPC sources, since these would be depleted of CD34+CD133- cells recognised as EPC. Since the main therapeutic targets remain myocardial and severe peripheral ischemia, and the therapeutic options are confined to use of autologous cells because of requirements for histocompatibility, the current clinically available HPC sources are either bone marrow or mobilized peripheral blood HPC. Neither of these contain EPC detectable through endothelial outgrowth, and their CD34+ cells appear at opposite ends of the spectrum in their proangiogenic activity. Since neither source seems to provide EPC, it seems from this study that since mobilized blood HPC are the more potent proangiogenic cells, they could be preferred over bone marrow for autologous therapeutic vascular regeneration. With regard to provision of EPC for therapeutic use, the best available source for autologous EOC would seem to be normal peripheral blood, not bone marrow or mobilized peripheral blood HPC. It should be possible to produce autologous EOC for therapeutic use [41, 50], and clinical trials are needed to evaluate the effect of EPC distinct from or in synergy with the proangiogenic effects of HPC therapies. Since the endothelial progenitor cell has been the primary focus of most studies, new studies are now required to clarify the identity of the proangiogenic effector cell and the mechanisms by which it delivers its effect, which may have been the only clinical effect observed until now. It may be that since the proangiogenic cell does not integrate into tissue, it may have less stringent histocompatibility requirements for clinical use, or the proangiogenic effect could be delivered by cell-free systems. The precursor and proangiogenic activities appear to be the main components required to improve the design of cellular therapy for vascular regeneration. The role of other cells such as MSC or monocytes may be relatively marginal, but might provide some added benefit once the main progenitor and proangiogenic components are optimized.