Purification and characterization of human adipose-resident microvascular endothelial progenitor cells

Human adipose tissue is a rich source of adipose-derived stem cells (ASCs) and vascular endothelial progenitor cells (EPCs). However, no standardized method has been established for the isolation and purification of adipose-derived EPCs (AEPCs). The aim of this study was to establish a method for the isolation and purification of AEPCs. The stromal vascular fraction (SVF) was extracted from human lipoaspirates, and the CD45−CD31+ fraction of the SVF was collected by magnetic-activated cell sorting (MACS). The CD45−CD31+ fraction was cultured for 4.5 days, followed by a second MACS separation to collect the CD31+ fraction. Purified AEPCs were expanded without being overwhelmed by proliferating ASCs, indicating that a high level (> 95%) of AEPC purification is a key factor for their successful isolation and expansion. AEPCs exhibited typical endothelial markers, including CD31, von Willebrand factor, and the isolectin-B4 binding capacity. AEPCs formed colonies, comparable to cultured human umbilical vein endothelial cells (HUVECs). Both AEPCs and HUVECs formed capillary-like networks in the tube formation assay, with no significant difference in network lengths. We are the first to establish a purification and expansion method to isolate these cells. Because adipose tissue is a clinically accessible and abundant tissue, AEPCs may have potential advantages as a therapeutic tool for regenerative medicine.

Extraction and cultivation of the SVF. After obtaining informed consent, we used the IRB-approved protocol to collect human lipoaspirates from 17 healthy donors aged 28-66 years who had undergone liposuction of the abdomen or thighs.
After natural gravity sedimentation, lipoaspirates develop three layers: oil, adipose tissue, and tumescent liquid. The floating adipose tissue layer was extracted, measured, and digested with an equivalent volume of collagenase-based enzyme solution (details below) at 37 °C for 30 min at 120 rpm of reciprocating motion (Yamato Scientific), followed by centrifugation at 800×g for 10 min. The resulting cell pellet was designated the SVF (Fig. 1A). After washing with Hanks' Balanced Salt Solution (HBSS; Thermo Fisher Scientific; #14175-103), the SVF was sequentially passed through 100-µm and 40-µm sieve cell strainers (Corning). The strained cell suspension was centrifuged again at 800×g for 5 min at 4 °C, and the obtained SVF pellet was washed with HBSS. The nucleated cell number, viability, and total cell-sized particle number were measured using a fluorescent cell counter (LUNA-STEM; Logos Biosystems) after double staining with acridine orange and propidium iodide (Logos Biosystems).
Cultured cells were dissociated with TrypLE Express Enzyme (Thermo Fisher Scientific) for 5 min at 37 °C in a CO 2 incubator. The enzymatic reaction was stopped by adding cold medium containing fetal bovine serum (FBS), and the cells were resuspended in FACS buffer.
Cell samples were analyzed with a flow cytometer (Miltenyi Biotec MACSQuant Analyzer, and MACS-Quantify software ver. 2.5) for the cell size, granularity, and expression of CD45, CD31, CD34, CD105, CD146, CD157, and CD200. Viable cells were selected using a live/dead cell staining dye (Fixable Viability Dye  Endothelial cell colony-forming unit assay. The EC colony-forming unit (CFU-EC) assay was modified from the fibroblastoid colony-forming unit assay 22 . Cultured AEPCs (P = 5) and HUVECs (P = 5) were dissociated using TrypLE Express Enzyme. The cells were seeded on cell culture plates (Corning; #353502) at 125 cells per 35-mm-diameter well (12.6 cells/cm 2 ) and cultured for 7 days. To visualize colonies, the cells were fixed with 4% PFA in PBS for 15 min at RT, stained with 0.05% (w/v) crystal violet (FUJIFILM Wako) for 30 min at RT, and washed with distilled water. Serial photographs were obtained using a light field microscope (Keyence; #BZ-X710), and the digitized images were combined to generate a single large image using analyzer software (Keyence).
Endothelial cell network formation assay. Expanded AEPCs and HUVECs (50 µL 2 × 10 5 cells/mL of each complete medium) were plated on a 96-well plate (PerkinElmer) precoated with 50 µL sol-gel transited Matrigel (Corning; #356237) and incubated for 6 h at 37 °C in a CO 2 incubator. Photographs were obtained using an inverted phase-contrast microscope with a camera (Leica Microsystems; DM IL LED and MC170 HD). Images were analyzed using the Angiogenesis Analyzer plug-in for ImageJ software 25 according to the standard protocol found on the developer's website. Although this software can analyze various parameters in constitutive elements of tubular networks, we measured the total segment length, total branch length, and total length as appropriate evaluation standards. Because AEPCs showed stronger cell-cell interactions than HUVECs during the cell dissociation step, the AEPC suspension consisted of both single cells and cell clusters containing 2-4 cells. www.nature.com/scientificreports/  www.nature.com/scientificreports/ Statistical analysis. Quantitative data were presented as means ± standard deviation (SD). We used twotailed unpaired t-test for the following analyses: immunocytochemistry, CFU-EC assay, and EC network formation assay. All statistical analyses were performed using Prism 6 ver. 6.0d (GraphPad Software).
Purification and expansion of AEPCs. As a preliminary experiment, we attempted to extract the SVF using three enzyme solutions: #1: collagenase and CaCl 2 ; #2: collagenase, CaCl 2 , and DNase1; #3: collagenase, CaCl 2 , DNase1, and Pol188. The percentage of AEPCs in the SVF tended to be higher when extracted using enzyme solution #2 (Supplemental Fig. S1A). We also examined the effects of various collagenase concentrations (0.0-2.0%) in enzyme solution #2. The numbers of total nucleated cells and ASCs in the SVF increased in a dose-dependent manner with increasing collagenase concentrations. The numbers of AEPCs in the SVF peaked at collagenase concentrations of 0.1-0.4% (Supplemental Fig. S1B). Based on these results, the same volume of the #2 enzyme formulation containing 0.2% (w/v) collagenase, 3 mM CaCl 2 , 1000 U/mL DNase1, and HBSS was added to adipose tissue in subsequent experiments. We also examined the use of several good manufacturing practice (GMP)-grade enzymes and combinations to prepare purified AEPCs for clinical use (Supplemental Fig. S1C).
Therefore, we performed a second MACS enrichment using CD31 microbeads to further purify the AEPC population. A schematic view of the procedure of AEPC purification is shown in Fig. 3A. After performing purification steps of fresh SVF (Fig. 3Ba) and the first MACS CD45 − CD31 + (AEPC-rich) fraction (Fig. 3Bb), the second MACS separation step of cultured AEPC-rich populations was performed after either 4.5 days (before the timing of ASC proportion increase) or 7 days (after the timing of ASC proportion increase) of culture (Fig. 3Bc,d,g,h). When separated on day 4.5 (steps in Fig. 3Bc,d), the respective percentages of CD45 − CD31 + cells before and after the second MACS separation were 91.6% and 97.1%, respectively (n = 2; Fig. 3C). When separated on day 7 (steps in Fig. 3Bg,h), the respective percentages of CD45 − CD31 + cells before and after separation were 70.4% and 94.7%, respectively (n = 2; Fig. 3C). AEPCs separated on day 4.5 were cultured for 6 days (Fig. 3Be,f) and Figure 3. Establishment of the AEPC purification. (A) Schematic view of the AEPC purification method from human lipoaspirates. AEPCs were purified from the SVF using two MACS sorting steps (CD45 − CD31 + sorting and only CD31 + sorting) and adherent cultures. (B,C) Optimization of the timing for the second MACS sorting step for AEPC purification. Cells were dissociated from the cell culture surface with TrypLE Express Enzyme after either 4.5 or 7 days of culture and were subjected to MACS separation using CD31 microbeads. Cell surface markers (CD45, CD31, and CD34) were examined by flow cytometry. The percentage of CD45 − CD31 + (AEPC) cells in CD45 − CD31 − (ASCs, isotype + cells, other cells, and debris) cells is shown in magenta. ASCs drastically decreased CD34 expression in vitro, which could be observed in the CD45 − CD31 − population (arrowheads). Scale bars in microscopic photographs represent 100 µm. This experiment was performed twice, independently from 2 donors. (D) Morphologies of purified AEPCs over time. The photographs were taken using a phasecontrast microscope (Leica DM IL LED with a camera MC170HD; 100× magnification). P1D6 indicates the culture period at passage 1, after 6 days in culture. The experiment was performed twice, independently from 2 donors. Bars represent 100 µm.  www.nature.com/scientificreports/ expanded with satisfactory purity through at least 5 passages (Fig. 3D). By contrast, AEPCs separated on day 7 were overwhelmed by proliferating ASCs after 6 days of culture (Fig. 3Bi,j).

Time-dependent changes in CD marker expression profiles AEPC and ASC cultures. AEPCs
(obtained as the CD45 − CD31 + fraction after two MACS separation steps) and ASCs (obtained as the CD45 − CD31 − fraction after the first MACS separation) were cultured, and time-dependent changes in CD marker expression profiles (CD45, CD31, CD34, CD146, and CD105) were examined. Cultured AEPCs maintained CD31, CD146, and CD105 expression through at least 5 passages, whereas CD34 expression decreased over time in culture (Figs. 1B and 4). ASCs began to express CD105 after seeding, despite being CD105 − before seeding, and maintained expression through at least 5 passages, although CD34 expression decreased over time in culture. CD146 expression was not observed in ASCs.
Characterization of expanded AEPCs. We characterized and performed functional analyses of expanded AEPCs (passage 5), compared against expanded HUVECs (passage 5), using immunocytochemistry, CFU-EC, EC network formation, and flow cytometry analyses. Immunocytochemistry showed similar CD31 and vWF expression profiles and isolectin-B4 binding capacity between AEPCs and HUVECs (Fig. 5A) In the CFU-EC assay, the colony forming cells in AEPCs and HUVECs were 59.7 ± 8.4% and 47.3 ± 7.9%, respectively (which did not differ significantly) on day 7. AEPCs proliferated more slowly in colonies, and the average size of AEPC colonies was smaller than that of HUVEC colonies (Fig. 5B). Even if using the same medium, EGM-2MV, AEPCs formed colonies smaller in size than HUVECs, and reached a similar size on day 12 to that of HUVECs on day 7 (Supplemental Figure S3).

Discussion
Although human microvascular AEPCs are an important cell population, two primary challenges have prevented the establishment of purification and expansion methods. First, AEPCs are easily overwhelmed by contaminating ASCs during the adherent culture process. Second, AEPCs appear to be more vulnerable to tissue enzymatic dissociation processes than ASCs. Both the total nucleated SVF yield and ASC yield increased with increasing collagenase concentrations between 0.02 and 2% (w/v), whereas the optimal AEPC yield was observed at collagenase concentrations between 0.1 and 0.4%. In addition, the SVF obtained from adipose tissue showed differential population profiles depending on the supplemental reagents used during extraction. The addition of a biocompatible amphipathic detergent, Pol188, increased the extraction efficiency of total SVF and ASCs from adipose tissue 24 ; however, the extraction efficiency of AEPCs decreased following the addition of Pol188. AEPCs and ASCs both localize in proximity to the microvasculature of adipose tissue 26,27 and are exposed to the enzymatic dissociation process in a similar manner, suggesting that AEPCs are more sensitive to artificial stimuli.
Instead of FACS purification, MACS separation was used in this study because the ultimate goal is the clinical application of AEPCs that requires a sufficiently large number of viable cells. MACS separation can process a much larger number of cells within a shorter time without damaging their viability although its purification efficiency is inferior to that of FACS. Because of the same reason, this roughly separated population was assessed instead of AEPCs collected from colony formation.
During the flow cytometric analysis, the populations of freshly isolated AEPCs and AS C s we re char a c te r i z e d as C D 4 5 − C D 3 1 + C D 3 4 + C D 1 0 5 + C D 1 4 6 + C D 1 5 7 ± C D 2 0 0 ± a n d CD45 − CD31 − CD34 + CD105 − CD146 − CD157 ± CD200 − (Fig. 1B), respectively, a finding that is partly supported by that in a previous study 28 . CD34 expression gradually decreased in AEPCs over the culture period, whereas CD34 expression was drastically reduced in ASCs as early as passage 2 (Fig. 4). The purification of AEPCs was completed when the CD31-positive population was at 97%, but not at 94% (Fig. 3C). This difference was small but changed the final result of purification. AEPCs were successfully purified using MACS separation and could be successfully expanded through passage 5, although they continued to be contaminated with a small number of ASCs (approximately 3%), suggesting that the purification rate of AEPCs is a critical factor to avoid ASC expansion. AEPCs at passage 5 showed reduced CD31 expression (< 80%) by flow cytometry, and other morphological cell types could not be observed under the microscope, as confirmed by the immunocytochemical observation of CD31 staining (Figs. 4A and 5A). Some studies have reported that ECs show decreased CD31/ PECAM-1 expression in a time-dependent manner in vitro 29,30 , and AEPCs may also decrease CD31 protein expression following extended expansion. In the CFU-EC assay, AEPC colonies proliferated more slowly than HUVECs (Figs. 5B and S3), exhibiting smaller colonies in size than HUVECs. www.nature.com/scientificreports/ EPCs are difficult to obtain from circulating peripheral blood (CD34 + CD133 + VEGFR-2 + expressed cells), with one report identifying only 1074 EPCs in 20 mL of peripheral blood 31 . In addition, bone marrow-derived EPCs have a controversial issue. Although some papers have reported that bone marrow-derived EPCs are mobilized to undergo temporarily angiogenesis in the inflammatory phase and granulation tissue construction phase of the wound healing process, others have indicated that not circulating EPCs but tissue resident EPCs, such as AEPCs, predominantly contributed to neovascularization [32][33][34] . Considering clinical applications, it is more reasonable to purify and expand AEPCs from the microvasculature of human adipose tissue 23,35 . One research group successfully purified AEPCs from large-volume human adipose tissue using Dynabeads magnetic sorting combining the CD44 − and CD90 − fractions. However, they reported that "Adipose tissue ECs may be a more practical alternative for obtaining large quantities of autologous ECs" 36 . However, our method has resolved the issue. CD157, a novel tissue-resident EPC marker, has previously been detected in the portal vein and adipose tissue blood vessels of mice 9,11 . Three populations of VE-cadherin + CD31 + CD45 − EPCs were reported: CD157 + CD200 + , CD157 − CD200 + , and CD157 − CD200 − . In addition, CD157 + cells were recognized to generate functional blood vessels in a mouse liver injury model 11 . In our study, expanded AEPCs contained CD157 + cells through at least passage number 5, although the percentage was small. In addition, expanded AEPCs exhibited clonal expansion capacity, self-renewal capacity, EC-specific characteristics, such as vWF expression, lectin binding, and network formation capacity, indicated that AEPCs feature progenitor characteristics.

Conclusions
Our study is the first to characterize human AEPCs and establish a purification and a large-scale (clinical scale) expansion method to isolate these cells. Because adipose tissue is a clinically accessible and abundant tissue, AEPCs have potential advantages as a therapeutic tool for regenerative medicine and surgical applications compared with EPCs derived from other sources, such as bone marrow and peripheral blood. The preclinical and clinical value of AEPCs to treat potential target diseases, such as ischemic pathologies, remains to be established in future studies. Characterization and functional analysis of expanded AEPCs compared with HUVECs. All the experiments were performed using AEPCs and HUVECs at passage number 5. (A) Fluorescent immunocytochemistry of the expression profiles of endothelial cell-specific markers, CD31, vWF, and the binding capacity of isolectin-B4, in AEPCs and HUVECs. Nuclear staining was assessed by 4′,6-diamidino-2-phenylindole (DAPI). Photographs were taken using a confocal microscope (OLYMPUS, FV1000 with a CCD camera DP71 and an objective lens UPlanFL N 40×/1.30 oil; 400× magnification) as a plane image. Bars represent 100 µm. The graphs show the percentage of endothelial marker-positive cells among DAPI-positive cells. The data represent two independent experiments (2 donors), and percentages were determined from cultivations performed in triplicate. The bars represent means ± SD (n = 6, **P < 0.01, two-tailed unpaired t-test). (B) CFU-EC analysis of AEPCs and HUVECs. Cells were seeded at 125 cells per 35-mm-diameter well (12.6 cells/cm 2 ) and cultured for 7 days. Each experiment was performed twice independently (two donors) in technical triplicates. The bars represent means ± SD (n = 3; n.s., not significant by two-tailed unpaired t-test). (C) Network formation capacity of AEPCs compared with that of HUVECs. Representative phasecontrast pictures corresponding to the right-hand side of the extracted network skeleton. The total segment length (magenta color), total branch length (yellow-green color), and total length were determined using the Angiogenesis Analyzer plugin for ImageJ software. Scale bars represent 100 µm. The experiment was performed independently three times (3 donors) in technical pentaplicates. Bars represent means ± SD (n = 5, **P < 0.01; n.s., not significant by two-tailed unpaired t-test). (D) Cultured AEPCs and HUVECs were characterized by the expression of CD45, CD31, CD34, CD157, and CD200 by flow cytometry analysis. This experiment was performed independently three times (3 donors) for AEPCs and independently two times (2 lots, 5-or 7-donor mixture) for HUVECs and is listed in Table 2.  www.nature.com/scientificreports/ Patent information. The preparation method used for adipose-derived endothelial (progenitor) cells and its use as a medical formulation are patent pending (Japanese Patent Application Disclosure, P2019-88279A). The medical-grade enzymatic formulation used to extract SVF to effectively purify both adipose-derived endothelial (progenitor) cells and adipose-derived stem cells using GMP-grade enzymes is patent pending (Patent Application Disclosure, WO2021-112168).

Data availability
The datasets are available from the corresponding author upon reasonable request. Ethical restriction exists on sharing the original study datasets. www.nature.com/scientificreports/