Improving the characterization of endothelial progenitor cell subsets by an optimized FACS protocol

The characterization of circulating endothelial progenitor cells (EPCs) is fundamental to any study related to angiogenesis. Unfortunately, current literature lacks consistency in the definition of EPC subsets due to variations in isolation strategies and inconsistencies in the use of lineage markers. Here we address critical points in the identification of hematopoietic progenitor cells (HPCs), circulating endothelial cells (CECs), and culture-generated outgrowth endothelial cells (OECs) from blood samples of healthy adults (AB) and umbilical cord (UCB). Peripheral blood mononuclear cells (PBMCs) were enriched using a Ficoll-based gradient followed by an optimized staining and gating strategy to enrich for the target cells. Sorted EPC populations were subjected to RT-PCR for tracing the expression of markers beyond the limits of cell surface-based immunophenotyping. Using CD34, CD133 and c-kit staining, combined with FSC and SSC, we succeeded in the accurate and reproducible identification of four HPC subgroups and found significant differences in the respective populations in AB vs. UCB. Co-expression analysis of endothelial markers on HPCs revealed a complex pattern characterized by various subpopulations. CECs were identified by using CD34, KDR, CD45, and additional endothelial markers, and were subdivided according to their apoptotic state and expression of c-kit. Comparison of UCB-CECs vs. AB-CECs revealed significant differences in CD34 and KDR levels. OECs were grown from PBMC-fractions We found that viable c-kit+ CECs are a candidate circulating precursor for CECs. RT-PCR to angiogenic factors and receptors revealed that all EPC subsets expressed angiogenesis-related molecules. Taken together, the improvements in immunophenotyping and gating strategies resulted in accurate identification and comparison of better defined cell populations in a single procedure.


Introduction
Over the last decades circulating EPCs have been extensively studied in the context of both health and disease. EPCs take part in neovascularization and their levels are used to monitor the effects of therapy [1][2][3][4]. Notably, the term EPC is not only used for cells with genuine endothelial lineages, but also for other cell types supporting neovascularization, including PLOS  hematopoietic progenitor cells (HPCs) [1,[5][6][7][8]. HPCs are bone marrow derived [9] and home to ischemic or neoplastic tissues that secrete chemo-attractants and, following differentiation, contribute to angiogenesis by secreting proangiogenic factors [10][11][12]. Another subset of circulating EPCs is capable of generating in vitro outgrowth endothelial cells (OECs). The in vivo equivalent of OECs is believed to contribute to vascular regeneration [7,[13][14][15][16][17]. While most circulating endothelial cells (CECs) are damaged or apoptotic mature endothelial cells with no progenitor potential [18][19][20][21], there may well be a small CEC fraction of viable endothelial progenitors from which OECs can be grown. However, the kinship of CECs and OECS has not been proven, mainly because authors used unsorted PBMCs or PBMCs enriched for specific markers using magnetic beads, instead of FACS sorting [1,7,20,[22][23][24][25][26]. The accurate identification of EPC subsets, and their subdivision, is challenged by the low frequencies of these cells in the bloodstream, the different ways of their isolation, and the discrepant immuno-phenotypical definitions used [1,5,8,23,24,[27][28][29][30][31]. The introduction of validated procedures of isolation and work-up would greatly improve accurate comparisons of the various populations and literature data on the EPC subsets, and shed more light on the genuine source of OECs [7,32].
Here we present a protocol for the accurate identification, characterization, and subdivision of HPCs, CECs and culture-derived OECs from peripheral blood samples of healthy adults (AB) and umbilical cord blood (UCB). The procedure includes the analysis of stem cell markers [32] and RT-PCR on sorted cells allows for the detection of markers beyond cell-surface expression. By following the procedures described we succeeded in demonstrating the similarities between OECs and CECs, suggestive of kinship between these populations. PCR analysis to the distinct EPC subsets and HUVECs for the detection of angiogenic factors and receptors revealed angiogenic capacities of all subsets.

Blood samples and preparation
Eighteen samples of adult peripheral blood (24-40 ml) and 15 samples of umbilical cord blood (12 ml) were used for this study. The samples were collected in BD vacutainer EDTA tubes and stored at room temperature in the dark for 18 hours. Blood was then diluted 1:1 with PBS-0,5 mM EGTA, and PBMCs were isolated using Ficoll Paque plus (GE Healthcare).

FACS analysis and sorting
PBMCs were incubated with 10% mouse serum to block unspecific antibody binding and stained 20' with specific antibodies (S1 Table). To get saturation, OECs/HUVECs were stained with 1 μg Ab/10 6 cells/200 μl, and PBMCs with 1.5 μg Ab/10 7 cells/200 μl. KDR staining was amplified using a 3-step protocol: 1) anti-KDR-APC; 2) anti-APC-biotin; 3) streptavidin-APC. After staining the cells were washed twice and re-suspended in PBS, 10% BSA, 0,1μg/ml Hoechst 3h3258 to mark dead cells. All steps were performed on ice. Live nuclear staining was performed with the cell permeant Hoechst33342 (Sigma-Aldrich), 10μM for 30' at RT. FACS analysis/sorting was performed with a BD FACS Aria III (BD Biosciences, New Jersey, US) using the parameters listed in (S1 Table). In the FSC/SSC plot, mononuclear cells were selected using a gate for high FSC cells excluding residual granulocytes, cellular debris and small particles (Fig 1 and S1 Fig)  To prevent exclusion of large cells, a large mononuclear-cell gate was applied. B. Ratios of enrichment of total live PBMCs, lymphocytes, monocytes, HPCs (as defined in Fig 2) and CECs (as defined in Fig 3) by Ficoll vs. RBC lysis. C-D. Examples of reduction of background (auto-fluorescence) by stringent exclusion of dead cells using Hoechst 33258 (C) and (residual) granulocytes (D). E. Basic procedure used, in addition to standard acquisition (1x10 6 total events), to enrich for target cells by discarding triple CD34/ CD133/KDR negative events. Upper left plot: selection of CD34 + and KDR + cells; lower left plot: selection of CD34 + and CD133 + cells. Right panels: result of a "Join (Or) gate" in FACS DIVA. KDR + CD34 + (upper plot) and CD133 + CD34 + (lower plot) populations are better visualized. F. Delineation of a scale of FACS-based fluorescence levels. Negative peak = unstained or isotype control; low (dim) = within 1 log from negative; high (bright) = more than 1 log from negative and ranging from moderately high (1 st to 2nd log above negative) to very high (3rd log or higher). "Medium" is sometimes used to define populations in between low and high. Annexin V staining Following immunostaining, 5μl of Annexin-V-FITC (BD) or Annexin-V-biotin (BD), were added to 500,000 PBMCs in Annexin Buffer and left 20' on ice, followed by one washing step. Incubation with Annexin V-biotin was followed by streptavidin-APC staining. The FACS analysis was carried out in Annexin V buffer, 0,1μg/ml Hoechst 33258.

Generation of outgrowth endothelial cells
PBMCs were re-suspended in endothelial cell medium (EGM-2 + BulletKit; Lonza), seeded in culture flasks (Corning, polystyrene) at 2.5x10 6 cells/cm 2 , and incubated at 37˚C, 5% CO2. Medium was changed daily. When OECs reached 80% confluence, cells were passaged using Accutase (Sigma-Aldrich). Gelatin, collagen-I, and fibronectin coating were tested and compared to non-coated plates. Because coating did not significantly affect the generation of OECs we used uncoated plates.

Statistical analysis
The non-parametric Mann-Whitney U test in SPSS (version 21.0.0.1) was used to determine differences in HPC and CEC frequencies and in gene expression.

Sample preparation and FACS analysis
The mononuclear cell fraction containing the EPC subsets was on average six times enriched by using Ficoll as compared to standard RBC lysis buffer, thereby leaving the ratios between the PBMCs and EPC subsets unaltered (Panels A and B in  Table). Bright antibodies or signal amplification were used to gain sufficient resolution for each marker. The visualization of rare cells was improved by discarding triple CD34/KDR/CD133 negative events, thereby reducing data overload and enabling the recording of larger sized samples. By setting a threshold around 2% positive events, the target cells were~50 times enriched (Panel E in Fig 1) and reliable visualization was accomplished for populations with a frequency as low as CECs in AB (~1/1 Ã 10 6 PBMCs) [21]. To enable direct visual comparison of fluorescent levels between populations and samples, a common scale of fluorescence intensity based on literature and our own data was applied in all FACS plots (Panel F in Fig 1) [1,33].

Characterization and quantification of HPCs
Following current definitions [5,17,34], HPCs were initially identified as a CD34 +/high cluster and refined by gating CD45 low . Based on the expression intensities for CD133, HPCs were sub-divided in CD133 high , CD133 low , and CD133 neg cells (Panela A and B in Fig 2) [1,5,17,34]. A bright antibody was necessary to properly visualize different CD133 levels (Panel A in Fig 2). HPCs were located in between lymphocytes and monocytes in the FSC/SSC plot (Panel C in Fig 2). Gating on FSC/SSC converged with CD45-based gating, and provided further purification of HPCs, mainly by cleaning the CD133 neg subpopulation. By using both CD45 and FSC/SSC, HPC gating was refined over previous protocols. To verify the identity of the HPCs we tested the expression of CD133 and c-kit by RT-PCR on sorted HPC-fractions. The expression of CD133 was verified by RT-PCR and was at low levels also found in the CD133 neg subpopulation (Panel E in Fig 2). mRNA expression for c-kit was found in all three subpopulations (Panel E in Fig 2) and appeared to be high and independent of CD133 levels. Overall, HPCs (calculated as ratio over CD45 + PBMCs) were 5.8 times more frequent in UCB than in AB (Panel F in Fig 2). The frequencies of the three CD133-based subpopulations differed between UCB and AB. CD133 high HPCs represented 72% of total HPCs in UCB but only 43% in AB.
Further HPC sub-classification was obtained by simultaneous staining for c-kit and CD133. Four sub-populations of HPCs were distinguished: CD133 neg c-kit high , CD133 low c-kit high , CD133 high c-kit high , and CD133 high c-kit neg/low (the last only observed in AB samples) (Panels A and B in Fig 3). In addition, the expression of CD34 and CD45 was found to be positively associated with CD133 levels but independent of c-kit levels. CD133 neg HPCs have the highest ckit levels (Panel A in Fig 3 and Panels A-C in S3 Fig). Expression of the endothelial markers KDR, CD144, and CD146 was only observed in sporadic HPCs (Panels C-E in Fig 3). CD105 was expressed at very low levels, only partially crossing the negative gate. Preliminary coexpression analysis revealed that KDR, CD146, CD144, CD105 do not converge to the identification of a single population of EPCs [8], but each marker mostly identifies small independent subpopulations, (Panels D and E in

Characterization and quantification of CECs
CECs were identified as CD34 + KDR + cells (Panel A in Fig 4) that are negative for CD45 (Panel B in Fig 4). Brightnesss of the staining was essential for proper CEC identification. Visualization of KDR required signal amplification (S1 Table) and the staining intensity of CD45 was crucial for separating CD45 neg CECs from sporadic CD45 low KDR + HPCs across the CD45 neg gate (Panel B in Fig 4). The frequency of CECs was significantly higher in UCB (14/10 6 CD45 + PBMCs) than in AB (1,9/10 6 CD45 + PBMCs) (Panels C and D in Fig 4). The CD34 and KDR levels encountered in UCB-CECs (mainly CD34 high KDR +/high ) differed from those in AB-CECs (mainly CD34 med KDR low ) (Panel C in Fig 4). The CECs expressed the endothelial markers CD146, CD144 and CD105 (Panel E in Fig 4 and Panel B in S4 Fig), which largely overlapped with KDR, confirming the endothelial identity of the selected population. Based on our data, CD146 and CD144 can be regarded as good substitutes for KDR in CEC identification, or as additional markers to refine the population. Immunopositivity for CD105 should be carefully evaluated since CD105-expressing HPCs (Panel C in Fig 3) that pass the CD45 negative gate may contaminate the CEC population. Notably, KDR yielded the cleanest background on HPCs followed by CD144, and CD146 (Panel C in Fig 3), and the best match (98% vs. only 90% with CD146 and CD144) in CEC-identification (Panel E in Fig 4 and   HPCs in UCB and AB are initially selected as CD34 +/hi cells and subdivided into CD133 hi , CD133 low and CD133 neg . Lower panel: discrimination between CD133 levels is partially missed using a less bright antibody. B. HPCs are refined by gating a single peak in the CD45 low region. Left plot: CD45 gating on raw HPCs (CD34 +/hi events). Right plot: CD45 gating on CD34 +/hi events pre-refined by FSC/SSC. CD45 levels slightly increase with CD133 expression (see also S3 Fig), still remaining within the "low" gate. The % of match of raw or pre-refined HPCs with the CD45low gate is indicated in the plots for each population. C. In a FSC/SSC plot, HPCs appear as a tight cluster in between lymphocytes and monocytes. Upper panel: FSC/SSC gating on raw HPCs (CD34 +/hi events). Lower panel: FSC/SSC gating on CD34 +/hi events pre-refined by CD45 levels. The % of match of raw or pre-refined HPCs with the FSC/SSC gate is indicated in the plots for each population. Comparison of B (left vs. right plot) and C (upper vs. lower panel) shows that CD45 and FSC/SSC gating converge to the identification of pure HPCs: partial overlapping in HPC cleaning indicates that the two approaches identify the same population, however each approach also provides some independent contribution to HPC purification. The less pure and most affected by cleaning gates appears the CD133neg fraction. D. FACS plots showing the distribution of pure HPCs (CD34 vs. CD133 plot) in UCB vs. AB. E. CD133 and c-kit RT-PCR in CD34 + CD133 + cells (n = 22); CD34 + CD133cells (n = 17) and CD34 -KDR -CD133 -PBMCs (n = 8) isolated from UCB. CD133 mRNA is detected in both CD133 + and CD133 -HPCs, although at different levels. c-kit mRNA further confirms the HPC identity of the CD34 + CD133cells. F. Left panel: median frequency of HPCs in UCB (5,6 in 1*10 3 CD45 + PBMCs) and AB (0,97 in 1*10 3 CD45 + PBMCs, significantly lower than in UCB, Z = -4,9; p = 1,00E -06 ). Middle panel: percentage ± SD of the three CD133-based HPC subpopulations in UCB vs. AB. Right panel: RT-PCR confirms higher CD133 expression by total UCB-HPCs than AB-HPCs. RT-PCR confirmed that KDR expression is CEC-specific and absent in HPCs (Panel C in S4  Fig). Because CD31 was expressed at high levels in other cell types, including HPCs, it cannot be considered as a selective marker for CECs. Since CD133 was found to be expressed by few CECs it cannot be regarded as a negative marker for these cells (Panel E in Fig 4). Further subdivision of CECs was obtained by the analysis of c-kit. Around 15% of CECs appeared to be ckit positive and c-kit expression was associated with higher CD34 and KDR levels. RT-PCR confirmed c-kit expression in CECs. Notably, c-kit mRNA levels were high and comparable to those of HPCs, despite much lower surface protein expression (Panel F in Fig 4), indicative of a discrepancy between mRNA and protein levels or surface expression of this marker.
In order to distinguish nucleated CECs from contaminating non-nucleated events, mainly aggregates of endothelial micro-particles, representative samples were stained with the nuclear dye Hoechst 33342 and all CECs appeared to be nucleated by FACS analysis (Panel A in Fig 5) and post-sorting microscopic inspection. Since CECs are reported to be apoptotic, apoptosis in these cells was assessed by Annexin-V staining. 78 (±4) % of CECs appeared to be apoptotic vs. only 8% of HPCs (Panel B in Fig 5). The pan-caspase inhibitor Z-VAD-FMK, added immediately after blood sampling, did not reduce the apoptotic fraction, which remained on average 80%, indicating that the cells were apoptotic from the time they entered the bloodstream. The expression of c-kit in CECs was associated with lower apoptosis (Panel B in Fig 5).    With the exception of a single outlier, CD133 expression was restricted to 5% of the OECs (Panel B in Fig 6), matching the CD133 expression in CECs (Panel E in Fig 4). The levels of c-kit, KDR and CD34 were heterogeneous and correlated positively with each other (Panels B and C in Fig 6 and  To investigate whether CECs are the source of OECs, we cultured sorted CECs in parallel with sorted and unsorted total PBMCs, using a larger nozzle (100μm) and lower pressure (20 psi) to reduce cell damage. The sorted cells (both CECs and total PBMCs) did not generate OECs, while unsorted PBMCs did. However, spindle-shaped early EPCs were successfully obtained from sorted CD14 + cells (Panel B in S5 Fig), or total sorted PBMCs. This confirmed the sound procedure of sample preparation and sorting, and suggested that FACS sorting is too damaging for the vulnerable OEC precursors to generate OECs in culture. No correlation between the yield of OECs and that of eEPCs was noticed in individual samples or in UCB vs.  [7,22,25,36,37].

Functional analysis
We analyzed a total of 10 angiogenic factors and receptors (apelin, PDGFβ, PDGFR, SCF, FGF, EGF, EGFR, VEGFA, Tie-1 and Tie-2) by RT-PCR to all distinct EPC subsets and HUVECs. Angiogenic factors appeared to be expressed in all EPC subsets. However, there was variation in the levels of expression of these factors between the various EPC subsets (S7 Fig). The levels of all angiogenic factors and receptors were lowest in HPCs with the exception of EGF. Notably, CECs showed the highest expression of Apelin, PDGFβ, SCF, EGFR, and high expression of Tie-2 as compared to the other cell types. The expressional pattern of OECs and HUVECs almost completely overlapped. Overall, the pattern of OECs was much closer to CECs than to HPCs.
In addition, we added immunohistochemistry for endothelial markers to OECs (vWF, CD31, CD105) (S8 Fig). OECs were positive for all these markers, indicative of their endothelial lineage.

Discussion
In this study we describe a procedure that allows for efficient identification, quantification, and sorting of HPCs and CECs from peripheral blood samples, and culture-generated OECs. Moreover, we improve the characterization of these populations resulting in, for instance, better definition of subpopulations, and identification of a candidate circulating OEC precursor. By using a Ficoll-based density gradient instead of a standard RBC lysis, significant enrichment for PBMCs was achieved with preservation of the relative frequencies of HPCs and CECs over total PBMCs [38,39]. Ficoll-enrichment, combined with stringent gating out of residual nonmononuclear cells, dead cells and debris, and accurate selection of markers, yielded clean and pure populations meeting the current standards of state-of-the art FACS procedures [40]. The procedure significantly reduced the sample size with preservation of the relative frequencies of HPCs and CECs over total PBMCs, allowing for significantly faster analysis and sorting of the target cells. The visualization and quantification of small populations was improved by selecting the events that were positive for the markers of interest to reduce data overload. This allowed recording the equivalent of~50 Ã 10 6 CD45 + PBMCs within 1 Ã 10 6 events (about 10-25 ml of whole blood). Moreover, by avoiding the use of toxic nuclear dyes the cells remain fully viable for subsequent culture experiments and RNA extraction. By using the outlined protocol we were able to identify HPCs and CECs more accurately, and isolate 200 to30.000 cells for RT-PCR analysis from rare populations with frequencies as low as 1/1 Ã 10 6 CD45 + PBMCs.

Improved characterization of HPCs
The current definition of HPCs is mainly based on a CD34 +/high CD45 low signature with CD133 expression as an additional marker [5,17,34]. In the present analysis HPCs were identified as a CD34 +/high CD45 low cluster and further cleaned by gating a specific region in FSC/ SSC (Panels B and C in Fig 2). Moreover, extending the definition of CD133 + and CD133 neg HPCs [17,34], we propose a subdivision of HPCs in CD133 high , CD133 low , and CD133 neg subpopulations. The HPC-identity of these subpopulations was confirmed by RT-PCR for CD133 and c-kit. By means of additional c-kit staining, HPCs were further subdivided by the demonstration of a c-kit neg/low subpopulation within the CD133 high HPCs that was found only in AB, not UCB. So far, c-kit neg/low HPCs have not been identified in human blood samples and were only described as a quiescent population in mice [41]. CD133 or c-kit alone are no markers for the identification of HPCs [42,43], since they range from negative to high. However, our data show that HPCs cannot be negative for CD133 and c-kit at the same time. Simultaneous staining of CD34+ cells (common denominator) for CD133 and c-kit leads to very accurate identification of HPCs as CD133 + and/or c-kit + (Panels A and B in Fig 3). Following CD34, CD133, and-c-kit-based identification, FSC/SSC and CD45 gating can be applied to further purify the HPCs, in particular to exclude residual doublets with other cell types, and other false positives, mainly contaminating CECs.
In line with previous data [1,27,44], the frequency of HPCs was significantly higher in UCB than in AB. We observed a prevalence of CD133 bright HPCs in UCB vs. AB. This finding fits with the notion that CD133 high HPCs are more primitive stem cells serving functions in fetal and post-natal development [43,45,46]. We also found positive correlation of CD45 and CD34 expression with CD133 levels, and a negative correlation between c-kit and CD133 levels (S3 Fig). Analysis of the endothelial markers KDR, CD146, CD144, CD105 and CD31 lead to an estimation of the risk of HPC contamination of the CEC population using each single marker. In addition, the endothelial markers highlighted the presence of potentially interesting and rare subpopulations of HPCs like the KDR + HPCs subpopulation that was described by Case et al. [5], and considered as a distinct EPC population by other authors [1,8,[27][28][29][30]. We show that the percentage of HPCs expressing endothelial markers (KDR, CD146, CD144, CD105) varies significantly and depends on the marker used. We also show that, with the exception of a few sporadic events, the markers are not concurrently expressed (Panels D and E in Fig 3 and Panels D-F in S3 Fig). Therefore, these HPCs expressing endothelial markers represent heterogeneous subpopulations requiring further exploration. Taken together, better identification of HPCs over the previous CD34 +/high CD45 dim based protocol was reached and better subdivision was achieved by means of a combined CD34, CD133, and c-kit staining.

Identification and characterization of CECs
In addition to the current criteria used for CEC-identification, we aimed at a more precise characterization of these cells, delineation of sub-populations, and sorting for RT-PCR and culture. Since CECs are generally identified as nucleated CD45 neg CD34 + events that additionally express an endothelial marker, i.e., CD146 or KDR [18,20,21,23], we used the CD34 +/high KDR + CD45profile as a common denominator (Panels A and B in Fig 4). KDR-based CEC identification was confirmed by CD146, CD105 and CD144 staining and by RT-PCR for these markers (Panel E in Fig 4 and Panel C in S4 Fig). Since a single coherent population of CECs was defined by using the endothelial markers, they may substitute KDR in the identification CECs. However, staining for KDR yields the lowest background on HPCs and the best crossmatching score and therefore, is regarded as a marker of first choice, followed by CD146 and CD144 (Panel E in Fig 4 and Panels A and B in S4 Fig). The quality of CD45 staining/gating was crucial for discriminating CECs from contaminating cells, mainly CD45 low HPCs sporadically expressing KDR, CD146, CD144, and potentially crossing the CD45 neg gate (Panel B in Fig 4 and Panel A in S4 Fig). Notably, all the CECs identified by the current procedure were nucleated (Panel A in Fig 5), i.e. not contaminated by (endothelial) micro-particle aggregates and therefore, the use of cell-permeant nuclear dyes like Hoechst or DRAQ5 was not necessary [20,47,48]. By avoiding the use of bright and toxic nuclear dyes a fluorescent channel is saved and spillover to other channels minimized, thereby improving the capacity of the detection channels and the sensitivity of the analysis. Moreover, the viability of the cells is well preserved for subsequent culture and RNA extraction.
Besides confirming higher frequency of CECs in UCB vs. AB [1,7,21,22,32,44], we found a major difference between CECs in AB as compared to UCB. The CECs in UCB express significantly higher levels of CD34 and KDR (Panel C in Fig 4). Moreover, by means of Annexin-V staining we identified two main subsets of CECs: a viable fraction of~20%, and an apoptotic fraction of~80%, equally present in AB and UCB (Panel B in Fig 5). In addition, a c-kit + subset, expressing higher levels of KDR and CD34, was identified and found to be mainly present in the viable fraction of CECs (Panel F in Fig 4 and

Characterization of OECs and prediction of their circulating precursor
The characterization of OECs was improved by various combinations of markers. In addition, comparing early vs. late cultures evidenced the alterations occurring with aging of the cultures and led to the definition of homogeneously high and stable markers (CD146, CD144, CD105, CD31) on the one hand, and heterogeneous and unstable ones on the other (KDR, CD34, ckit, CD133). The comparisons also enabled us to trace back the OEC precursors to c-kit+ CECs. By combining FACS populations and RT-PCR data we found overlap in the expression of KDR, CD146, CD105, CD144, and CD31 by OECs/HUVECs and CECs. Surface expression of the markers was generally lower in CECs than in OECs/HUVECs. This may be explained by the fact that OECs/HUVECs are derived from in vitro cultures, while CECs are in vivo single circulating cells, exposed to different environmental stimuli (lacking, for instance, the adhesion to other endothelial cells, and exposed to the blood microenvironment) and subjected to a potentially stressful purification procedure. Interestingly, the stem cell-marker c-kit was expressed by a significant fraction of early passage OECs and HUVECs and positively associated with KDR and CD34 levels. CD34 high KDR high c-kit high cells decreased with passaging in culture (Panels B and C in Fig 5 and Panels B-D in S6 Fig). This finding points to a triple +/high phenotype (CD34 +/high KDR +/high c-kit +/high ), which is lost with aging in culture, as the progenitor/founder of both cell types [24,49]. This view is supported by the recently described c-kit + KDR + CD105 + CD45 -'vascular endothelial stem cells' (VESCs) in the blood vessels of adult mice, which are capable of many cycles of replication and to generate functional blood vessels in vitro [50]. Notably, the CD34 +/high KDR +/high c-kit +/high phenotype that we trace back as potential OEC precursor well matches the viable fraction of CECs expressing c-kit and higher levels of KDR and CD34 (Panel F in Fig 4 and Panel B in Fig 5).
The current opinion on the origin of CECs is that these cells are mostly apoptotic, mature endothelial cells without progenitor potential, shed from damaged blood vessels [14,30,51]. The present data, however, show that about 20% of circulating CECs are viable and express ckit (Panel F in Fig 4 and Panel B in Fig 5). The high expression of angiogenic factors by CECs confirmed that these cells contain a fraction potentially contributing to neovascularization and in vitro generation of OECs. Unfortunately, we were unable to directly derive OECs from FACS-sorted CECs. These results are in line with results explicitly published by others [23] and with the (implicit) evidence that OECs are reported to be generated from unsorted or immuno-magnetic-bead-sorted PBMC fractions, not FACS-sorted ones [1,7,[22][23][24][25][26]. The present data, pointing to viable CECs expressing c-kit as OEC precursors, are in line with data obtained from cells purified by antibody-coated beads suggesting that the OEC-founder cells do not originate from bone marrow and express CD34 and CD146, but not CD45 or CD133, a pattern compatible with CECs [7,22,23]. The significant higher yield of OECs from UCB (containing more CECs than AB) than from AB further corroborates the view that OECs stem from CECs.

Conclusions
In conclusion, the present FACS-based analysis yields optimal identification, characterization and sub-division of cell populations meeting the current criteria for HPCs, CECs and OECs, and enabled us to propose that the circulating OEC precursors may consist of viable c-kit + CECs. , and ckit (C) levels (measured as median fluorescence intensity, MFI) in different HPC subpopulations from UCB and AB. CD34 levels are higher in CD133 high HPCs, independently of c-kit levels. CD45 levels are also positively correlated with CD133 expression, and independent of c-kit levels. The highest c-kit levels are in the CD133 neg HPCs. D-E. Additional details relative to KDR vs. CD146 expression in HPCs (see also Panel D in Fig 3). In (E) for each KDR/ CD146-based HPC-subpopulation, the frequency vs. total HPCs and total PBMCs (upper panels) is reported. FSC/SSC (medium panels) and CD45 (lower panels) based identification is also shown to confirm HPC identity. FSC/SSC analysis is from CD34+CD45dim selected HPCs, CD45 analysis is from CD34+ FSC/SSC selected HPCs. (F) Additional details relative to CD144 vs. CD146 expression in HPCs, (see also Panel E in Fig 3). The frequency of each subpopulation vs. HPCs and total PBMCs is indicated.