CD31⁺ cells represent highly angiogenic and vasculogenic cells in bone marrow: novel role of non-endothelial CD31⁺ cells in neovascularization and their therapeutic effects on ischemic vascular disease

Abstract

Rationale

Bone marrow (BM) cells play an important role in physiologic and therapeutic neovascularization. However, it remains unclear whether any specific uncultured BM cell populations have higher angiogenic and vasculogenic activities. Moreover, there has been controversy regarding the vasculogenic ability of BM cells.

Objective

Preliminary flow cytometric analysis showed that CD31, traditionally a marker for endothelial cells, is expressed in certain non-endothelial BM mononuclear cells (MNCs) in both human and mouse. Based on the conserved CD31 expression in the axis of hematopoietic stem/progenitor cells (HSC/HPCs) to endothelial cells, we further sought to determine the comprehensive vasculogenic and angiogenic characteristics of human and mouse BM-derived CD31⁺ cells.

Methods and Results

Flow cytometric analysis demonstrated that all CD31⁺ cells derived from BM were CD45⁺ and expressed markers for both HSC/HPCs and ECs. Comprehensive gene expression analyses revealed that BM-CD31⁺ cells expressed higher levels of angiogenic genes than CD31⁻ cells. Endothelial progenitor cells, as well as HSC/HPCs, were almost exclusively confined to the CD31⁺ cell fraction, and culture of CD31⁺ cells under defined conditions gave rise to endothelial cells. Finally, injection of CD31⁺ cells into ischemic hindlimb repaired ischemia, increased expression of angiogenic and chemoattractive factors, and in part directly contributed to vasculogenesis, as demonstrated by both three-dimensional confocal microscopy and flow cytometry.

Conclusions

These data indicate that BM-CD31⁺ cells represent highly angiogenic and vasculogenic cells and can be a novel and highly promising source of cells for cell therapy to treat ischemic cardiovascular diseases.

Introduction

Formation of new blood vessels (neovascularization) consists of two processes, vasculogenesis and angiogenesis. Vasculogenesis refers to the de novo development of blood vessels from endothelial progenitor cells (EPCs) or angioblasts which differentiate into endothelial cells (ECs). In contrast, angiogenesis is the formation of new vasculature from preexisting blood vessels through proliferation, migration, and remodeling of differentiated ECs. The identification of circulating EPCs in adult vertebrates suggested a role for BM cells in postnatal vasculogenesis^1–3, and led to trials of BM cells for therapy for ischemic cardiovascular diseases. However, conflicting results from recent clinical trials^{4, 5} suggests a need for the discovery of new cell types⁶ and more thorough investigation of the therapeutic mechanisms.

Two of the most important questions in current EPC biology are whether the reported cultured EPCs or similar BM cells have true vasculogenic potential and whether a specific marker can prospectively identify true EPCs or vasculogenic cells. The endothelial differentiation or vasculogenic potential of early EPCs has been questioned^{7,3, 8–11}. Since various studies have used different cell types and animal models, it is difficult to directly compare the results. Interestingly, a majority of the papers reporting the non-differentiation of EPCs used MI models. Unique conditions of the heart, including the constant motion of the myocardium and the high oxygen consumption of cardiomyocytes make sustained survival of the transplanted cells more difficult. This could reduce the observed differentiation, possibly masking true differentiation potential. While early EPCs are isolated by relatively short-term culture of peripheral blood cells. a newer type of EPCs, known as outgrowing ECs¹², late EPCs¹³, or endothelial colony-forming cells (ECFCs)¹⁴, can be derived from relatively long-term culture (more than 10 days). Cultured early EPC and EPC-colony forming unit (CFU) are indeed angiogenic cells co-expressing myeloid and endothelial markers in contrast to late EPC” or ECFC ^{14, 15}. Moreover, to date, no markers are unanimously accepted to define uncultured EPCs in circulation or BM in mice or humans¹⁵.

It now seems that paracrine, or humoral, effects are the main mechanism responsible for the therapeutic effects of BM-derived cell therapy for ischemic cardiovascular disease^{7, 16}. Since most of the cells used for previous studies were selected based upon their stem cell-like or vasculogenic potential^{3, 17, 18}, identification of cells with higher angiogenic or paracrine activities is now necessary to improve therapeutic efficacy and to develop next generation cell therapy for ischemic cardiovascular diseases. However, little research has been conducted to identify cells having higher angiogenic effects. Based on the above concerns and newly developed mechanistic findings, we sought to identify and isolate a specific cell population which possesses higher angiogenic or paracrine activities, includes vasculogenic cells and does not need ex vivo culture for identification or therapeutic use. In pursuit of this goal, we found that CD31 appears to be a good marker for this purpose. CD31 (also known as platelet endothelial cell adhesion molecule-1) is a 130-kDa, type I transmembrane glycoprotein, and is a member of the immunoglobulin gene superfamily. CD31 is highly expressed on ECs and to different degrees on several hematopoietic cells, including monocytes, granulocytes, and platelets¹⁹. Furthermore, CD31 is expressed in lineage⁻ Sca-1⁺c-Kit⁺ (LSK) mouse HSCs²⁰, side population cells in mouse BM²¹, mouse embryonic stem cells²², and human early EPCs⁷. However, the expression of CD31 in mouse HPCs and human HSCs has yet to be clearly investigated. Interestingly, HSCs/HPCs were already known to have angiogenic and/or vasculogenic potential as well as favorable cardiovascular regenerative effects^{18, 23, 24}. This conserved expression of CD31 in the axis of HSC/HPCs, EPCs and ECs has led us to investigate whether this group includes all the necessary cells associated with neovascularization and may be beneficial for treating ischemic cardiovascular diseases.

Methods

Animals

C57BL/6J mice, GFP-transgenic mice (C57BL/6J background) and athymic nude mice (The Jackson Laboratory) were used. All protocols for animal experiments were approved by the Institutional Animal Care and Use Committees of Caritas St. Elizabeth Medical Center and Emory University.

Harvesting BM-mononuclear cells and CD31 selection

Human BM (hBM) from healthy adults was purchased from Lonza, and mouse BM (mBM) cells were harvested as previously described ¹⁶. Mouse and human BM-MNCs were isolated by density gradient centrifugation using Histopaque 1083 (Sigma) and Histopaque 1077 (Sigma), respectively. Mouse and human BM-CD31⁺ and CD31⁻ cells were isolated using magnetic columns (MACS^®, Mitenyi Biotec).

Transplantation of the cells into ischemic hindlimb

Hindlimb ischemia was generated by ligation of the femoral artery and cauterization of large branches¹⁷. One million mBM-CD31⁺ or mBM-CD31⁻ cells in 100μl PBS were intramuscularly injected into the ischemic hindlimb of C57BL/6J mice, and cells in 200 μl PBS were intravenously injected into nude mice with ischemic hindlimb^{17, 25}.

Details on the materials and methods, including the following items, can be found in the online supplemental materials: microarray and gene set enrichment analysis²⁶, colony forming unit assay, BM transplantation, EPC culture assay^{1, 17, 27}, induction of differentiation of CD31⁺ cells into ECs in vitro, blood flow measurement in ischemic hindlimbs, measurement of capillary density ¹⁶, histologic analysis, cell adhesion test, and real-time RT-PCR¹⁶.

Results

Dual hematopoietic and endothelial cell characteristics of BM-CD31⁺ cells

To characterize the phenotype of mBM-CD31⁺ cells, we first isolated mBM-MNCs from C57BL/6J mice and carried out flow cytometric analysis (Figure 1A). About 34% of mBM-MNCs were found to expr ess CD31. Next, we determined which subsets of mBM-MNCs express CD31 using antibodies specific to B- and T-lymphocytes, myelomonocytic cells, and erythroid cells. FACS analysis showed that, on average, CD31 was expressed in 73% of lineage-depleted (Lin⁻) cells, 53% of B- (B220⁺) lymphocytes, 42% of T- (CD3e⁺) lymphocytes, 12% of erythroid cells (Ter-119⁺), 17% of CD11b⁺ myelomonocytic cells and 17% of Gr-1⁺ myelomonocytic cells (Figure 1B). Among myelomonocytic cells which express CD11b or Gr-1, the rate of CD31 expression was higher in Ly-6g⁻ cells (mainly monocytes, 32%, data not shown) than in Ly-6g⁺ cells (mainly granulocytes, 4.4%, Figure 1B). Among CD31⁺ cells, 5% were Lin⁻ cells, and 63% expressed B-220, 10% CD3e⁺, 2% Ter-119⁺, 28% CD11b⁺, and 29% Gr-1⁺, respectively (Figure 1C). Virtually all (> 99%) mBM-CD31⁺ cells expressed CD45 (Figure 1C), suggesting that the CD31⁺ cells are hematopoietic cells, but not ECs, existing in BM. Among myelomonocytic CD31⁺ cells (CD11b- or Gr-1-expressing CD31⁺ cells), only 9% were Ly-6g⁺ (Figure 1D), suggesting that a majority of myelomonocytic CD31⁺ cells are monocytes.

Figure 1. Expression of CD31 in mouse BM-MNCs.

BM-MNCs were isolated from 8–12 week-old C57BL/6J mice and subjected to flow cytometric analysis. Green color represents isotype control. (A) Expression of CD31 in mouse BM-MNCs (mBM-MNCs). (B) Each bar represents the percentage of CD31⁺ cells among the specified cells. (C) Expression of lineage markers in mBM-CD31⁺ cells. (D) Expression of Ly-6g in CD31⁺ myelomonocytic cells identified by the expression of either CD11b or Gr-1. Most of the CD31⁺ myelomonocytic cells were Ly-6g⁻, suggestive of monocytes. (E) Each bar represents the percentage of CD31⁺ cells among the EC marker-expressing cells. (F) The majority of cells double positive for CD31 and EC markers were also positive for CD45, indicating hematopoietic origin. N = 3 (A – F).

Next, we explored whether mBM-MNCs express other endothelial cell markers, and if so, to determine whether these cells express higher levels of CD31, thereby representing more highly angiogenic or vasculogenic cells. FACS analysis showed that mBM-MNCs which expressed various EC markers also co-expressed CD31 at the following rates: VE-cadherin (VE-cad), 77%, Flk-1 (VEGFR-2), 63%, Tie2, 76% and von Willebrand factor (vWF), 87% (Figure 1E). Triple staining with CD45 antibody, a pan-hematopoietic cell marker, showed that the majority of cells double-positive for CD31 and any of the EC markers also expressed CD45 (Figure 1F), suggesting that these cells are hematopoietic in origin rather than circulating mature ECs. Human BM-derived CD31⁺ (hBM-CD31⁺) cells also had similar hematopoietic and endothelial characteristics. FACS analysis showed that about 35 to 50% of hBM-MNCs expressed CD31 (Online Figure IA), and that CD31 was expressed in 42% of CD19⁺ cells (B cells), 46% of CD3⁺ cells (T cells), 67% of CD14⁺ cells, and 76% of CD11b⁺ cells (Online Figure IB). Among CD31⁺ cells, 19% expressed CD19, and 39%, 33%, and 29% were CD3⁺, CD14⁺, and CD11b⁺, respectively (Online Figure IC). Similar to mBM-CD31⁺ cells, almost all hBM-CD31+ cells expressed CD45 (Online Figure IC), indicating their hematopoietic, but not EC, identity. Again, similar to the mouse data, a majority of hBM cells expressing EC markers expressed CD31: 90% in VE-cadherin⁺ BM cells, 65% in KDR⁺ cells, 55% of Tie2⁺ cells, and 96% in vWF⁺ cells (Online Figure ID). As in the mouse, most of the human BM cells double-positive for CD31 and any of the EC markers expressed CD45 (Online Figure IE), indicating hematopoietic origin.

BM-CD31⁺ cells are enriched with higher levels of angiogenic and HSC/HPC genes

To determine whether mBM-CD31⁺ cells are enriched with angiogenic and hematopoietic stem/progenitor genes, we carried out microarray analysis. We compared the gene expression among CD31⁺, CD31⁻, Lin⁻CD31⁺, and Lin⁻CD31⁻ cells using microarray and gene set enrichment analysis (GSEA)²⁶ to determine the enriched gene classes in each population (Figure 2). The GSEA showed that pro-angiogenic, but not anti-angiogenic genes were significantly upregulated in mBM-CD31⁺ cells as compared to mBM-CD31⁻ cells (Figure 2A). Furthermore, the expression levels of HSC/HPC-specific genes were also significantly higher in mBM-CD31⁺ cells than in mBM-CD31⁻ cells. Higher expression of pro-angiogenic genes and HSC/HPC genes was also observed in Lin⁻CD31⁺ cells compared to Lin⁻CD31⁻ cells. These data suggest that CD31 is a marker for angiogenic cells and HSCs/HPCs in both total BM cells and Lin⁻ BM cells. Next, we further carried out leading-edge subset analysis to identify the core pro-angiogenic genes that confer angiogenic characteristics of CD31⁺ cells, and found that high expression levels of 22 pro-angiogenic genes comprised the angiogenic phenotype of mBM-CD31⁺ cells (Figure 2B). The higher expression levels of most of these genes were also confirmed by real-time RT-PCR analysis (Figure 2B).

Figure 2. Gene expression pattern reveals the angiogenic characteristics of BM-CD31⁺ cells.

(A) Total RNA was isolated from mouse BM-CD31⁺, BM-CD31⁻, Lin⁻CD31⁺, and Lin⁻CD31⁻ cells and subjected to microarray. N=3. (B) The leading-edge subset of pro-angiogenic genes upregulated in mBM-CD31⁺ cells compared to mBM-CD31⁻ cells. (C) Total RNA was isolated from human BM-CD31⁺ and CD31⁻ cells and analyzed similarly (D) The leading-edge subset of pro-angiogenic genes upregulated in hBM-CD31⁺ cells compared to hBM-CD31⁻ cells. In B and D, red characters represent the values with statistically significant difference. N=3.

Similar experiments were performed for human BM-CD31⁺ and CD31⁻ cells. GSEA showed significantly higher expression of pro-angiogenic genes and HSC/HPC genes in hBM-CD31⁺ cells as compared to hBM-CD31⁻ cells (Figure 2C, D), indicating that angiogenic cells and HSC/HPCs are enriched in the CD31⁺ fraction. The leading-edge subset analysis identified high expression levels of 24 core pro-angiogenic genes, including VEGF-A and angiopoietin-1, in hBM-CD31⁺ cells.

HSCs and HPCs are almost exclusively restricted to the BM-CD31⁺ cell population

We explored the expression of CD31 in HSCs and HPCs. Approximately 91% of LSK cells in mouse BM, which include HSCs and multipotent progenitor cells (MPPs), express CD31 (Figure 3A)²⁰. Since it appears that the level of CD31 expression decreases as HSCs undergo differentiation into mature hematopoietic cells, we examined the expression of CD31 in cells of intermediate status, i.e. oligopotent progenitor cells such as common lymphocyte progenitors (CLPs) and common myeloid progenitors (CMPs). As in LSK cells, most CLPs and CMPs identified as Lin⁻IL-7Ra⁺c-Kit^lowSca-1^lowThy-1.1⁻²⁸ and Lin⁻IL-7Ra⁻c-Kit⁺Sca-1⁻FcγR^lowCD34⁺²⁹ cells, respectively, expressed CD31 (Figure 3A), suggesting that CD31 expression is maintained from HSCs and MPPs to oligopotent progenitor cells. To confirm that CD31⁺ cells include HSCs and HPCs, we performed an in vitro clonogenic assay and an in vivo BMT experiment. As expected, colony forming units, including BFU-E, CFU-GM, and CFU-GEMM, were almost exclusively observed among mBM-CD31⁺ cells (Figure 3B,C). Transplantation of 1 × 10⁵ mBM-CD31⁺ cells into lethally irradiated mice rescued all of the mice (10/10), whereas transplantation of mBM-CD31⁻ cells saved only 20% of the mice (2/10) (Figure 3D). In experiments with human BM-MNCs, we found that cells expressing CD34 and CD133 were enriched in the CD31⁺ fraction (Online Figure IIA). Furthermore, CD31 was expressed in 99.8% of CD34⁺CD133⁺ cells and 89% of CD34⁺CD133⁻ cells (Online Figure IIB), suggesting that HSCs and most HPCs are included in the CD31⁺ cell population. Taken together, these results indicate that HSCs and HPCs in both mice and humans exist almost exclusively in the CD31⁺ fraction.

Figure 3. Enrichment of hematopoietic stem and progenitor cells in the CD31⁺ fraction of mouse BM.

(A) FACS analysis was performed using mBM-MNCs. N = 3. (B, C) Colony forming units in mBM cells were determined by methylcellulose culture. (B) Representative photomicrographs of a colony forming unit assay. (C) The number of colony-forming units was compared between mBM-CD31⁺ cells, mBM-MNCs or mBM-CD31⁻ cells (**P < 0.01, ***P < 0.001 vs. CD31⁻ cells; ^†P < 0.05, ^††P < 0.01, ^†††P < 0.001 vs. MNCs, n = 3). (D) 1 × 10⁵ mBM-CD31⁺, CD31⁻ cells, or PBS were injected into lethally irradiated mice and the survival was recorded (n = 10). Bars: B, 100 μm.

EPCs are cultured from the CD31⁺, but not CD31⁻ cell fraction

We determined whether mBM-CD31⁺ cells show endothelial differentiation potential in vitro. First, conventional early EPC culture conditions were applied to CD31⁺ and CD31⁻ cells, and EPCs were identified by DiI-ac-LDL uptake and binding to BS-1 lectin^{1, 30}. Surprisingly, such EPCs were exclusively cultured from mBM-CD31⁺ cells, but not from mBM-CD31⁻ cells (Figure 4A, B). When the CD31⁺ cells were cultured for 8 or 9 days in EGM-2 media supplemented with additional VEGF (50 ng/ml), the adherent cells expressed vWF, Flk-1, VE-cadherin, and CD31, indicating an EC phenotype (Figure 4C). Similarly, the early EPC assay using human BM cells showed that EPCs were cultured almost exclusively from hBM-CD31⁺, but not hBM-CD31⁻ cells (Online Figure IIIA). When hBM-CD31⁺ cells were cultured in EGM-2 with 20% FBS for 6 days, the cells also underwent EC phenotypic changes such that a majority of cells expressed endothelial-specific proteins such as vWF, KDR, VE-cadherin, and CD31 (Online Figure IIIB). The CFU-Hill assay²⁷ also showed that CFU-Hill colonies were higher in hBM-CD31⁺ cells than hBM-CD31⁻ cells (Online Figure IIIC).

Figure 4. Differentiation of mBM-CD31⁺ cells into an endothelial cell phenotype in vitro.

(A,B) mBM-CD31⁺ or -CD31⁻ cells were cultured for four days in EGM-2 media, and cells were incubated with DiI-acLDL and stained with FITC-conjugated BS-1 lectin and DAPI. Cells triple-positive for DAPI, acLDL, and BS-1 lectin were counted as EPCs. (B, ***P < 0.001, n = 6 – 7). (C) mBM-CD31⁺ cells were induced to differentiate into ECs. The cultured CD31⁺ cells exhibited multiple EC markers including vWF, Flk-1, CD31, and VE-cad. Bars: A, 100 μm; C, 20 μm.

BM-CD31⁺ cells improve hindlimb ischemia and increase vascularity

To examine the therapeutic effects and vascularizing capacity of mBM-CD31⁺ cells in vivo, we intramuscularly transplanted mBM-CD31⁺ cells into mice in a hindlimb ischemia model. Nude mice were initially used, because 60 to 100% of nude mice lose all or part of their hindlimbs following surgery, providing unambiguous results on the therapeutic efficacy of transplanted cells¹⁷. In contrast, C57BL/6 mice spontaneously recover their blood flow to near normal levels within four weeks, making it difficult to evaluate the therapeutic effects of the treatment. mBM-CD31⁺ cells injected into nude mice prevented limb loss in 6 out of 7 mice (86%), whereas mBM-CD31⁻ cells and PBS treatment rescued limbs in only 1 out of 7 and 0 out of 7 mice, respectively (Figure 5A). LDPI analysis also revealed that mBM-CD31⁺ cell transplantation resulted in 1.8- and 2.4-fold enhanced blood flow compared to mBM-CD31⁻ cell- and PBS-treated controls, respectively (Figure 5C, D, mBM-CD31⁺ vs. mBM-CD31⁻ and PBS, 33 ± 2 vs. 43 ± 7 and 79 ± 5%, P < 0.01). We tested the therapeutic effects again in C57BL/6 mice by injecting cells intramuscularly to increase the engraftment. Again, mBM-CD31⁺, but not mBM-CD31⁻ cell-injected mice showed significantly improved perfusion in the hindlimb at day 14 compared to the PBS treated mice (P < 0.01) (Online Figure IVA, B). In addition, injection of human BM-CD31⁺ cells into nude mice with hindlimb ischemia showed similar therapeutic effects represented by higher limb salvage at 3 weeks (Online Figure VA, B) and 1.5- to 2.4-fold higher blood flow in the hindlimb over 1 to 3 weeks than hBM-CD31⁻ cells and PBS (Online Figure VC, D).

Figure 5. Increase in blood flow and capillary density in ischemic hindlimb treated with mBM-CD31⁺ cells.

Two weeks after intravenous injection of mBM-CD31⁺ or -CD31⁻ cells into athymic mice with ischemic hindlimbs, the limb salvage (A, B), blood perfusion (C, D), and capillary density (E, F) were investigated. (A - B) Representative photographs of the limbs (A) and limb salvage rate (B)(n = 7) (C - D) Representative laser Doppler perfusion images (C) and quantitation of blood perfusion (D, **P < 0.01, n = 3). (E, F) Capillary density counted after ILB4 staining. Representative photomicrographs showing capillaries in the different treatment groups (E). Quantitative analysis of capillary density (F, **P < 0.01, n = 5). Bars: 100 μm.

We also compared the therapeutic effects of the CD31⁺ cells with well characterized other BM cell subpopulations on the recovery of hindlimb ischemia and found that there was no difference in the capability to recover hindlimb ischemia among the CD31⁺ cells, CD34⁺ cells^{18, 24}, CD34⁻CD31⁺ cells, and cultured EPCs^{10, 17} (Online Figure VI). The therapeutic effects of CD31⁺ cells were significantly higher compared to total mononuclear cells (Online Figure VII). To determine the therapeutic effects of HSC/HPCs included in CD31⁺ cells, mouse Lin⁻CD31⁺ cells and Lin⁺CD31⁺ cells were compared. The therapeutic effects of these two subpopulations was not significantly different (Online Figure VIII), suggesting that the therapeutic effects of CD31⁺ cells are attributable to both subpopulations.

To determine whether the CD31⁺ cells are still effective in subjects having cardiovascular risk factors, we tested the therapeutic effect of the CD31⁺ cells isolated from ApoE-KO mice fed a high-fat diet, a well known atherosclerosis model. We found no difference in the therapeutic effects between the CD31⁺ cells derived from normal mice and ApoE-KO mice, whereas the effects of CD31⁻ cells derived from ApoE-KO mice were significantly lower compared to CD31⁺ cells derived from normal or ApoE-KO mice (Online Figure IX).

To determine the effects of the CD31⁺ cells on neovascularization, we counted the number of capillaries in the hindlimb muscles. mBM-CD31⁺ cell-injected nude mice showed 2.2-or 2.3-fold higher capillary density than mBM-CD31⁻- and PBS-injected mice, respectively (CD31⁺ vs. CD31, PBS; 947 ± 114 vs. 436 ± 53, 408 ± 47 capillaries/mm², P < 0.05)(Figure 5E, F). The beneficial effects of mBM-CD31⁺ cell injection on capillary density were also observed in C57BL/6 mice but with less prominence (Figure IVC, D). Similarly, injection of hBM-CD31⁺ cells into nude mice with ischemic limbs resulted in 1.5- and 2.1-fold higher capillary density compared to hBM-CD31⁻ cells and PBS, respectively (P < 0.01 vs, hBM-CD31⁻ cells and PBS)(Online Figure VE, F). The capillary density was not significantly different between the groups injected with CD34⁺ cells and CD34⁻CD31⁺ cells (Online Figure X).

BM-CD31⁺ cells have higher engraftment potential than BM-CD31⁻ cells

We examined the engraftment of the transplanted cells into ischemic muscles. For this, we isolated C0D31⁺ and CD31⁻ cells from GFP-transgenic mice and injected them into a hindlimb ischemia model. Fluorescent microscopic examination showed that two weeks after cell injection, GFP-positive cells were more abundant in the CD31⁺ group than in the CD31⁻ group (Figure 6A, Online Figure XI, Online Video 1, 2). To quantify the number of engrafted cells, in a separate series of experiments, we enzymatically digested the hindlimb muscles and determined the number of GFP-positive cells by flow cytometric analysis. This analysis showed that the number of engrafted cells was 3.4-fold higher in the CD31⁺ group than in the CD31⁻ group (P < 0.05) (Figure 6B). To further investigate the mechanism underlying this higher engraftment of mBM-CD31⁺ cells, we carried out in vitro adhesion studies, in which we investigated the adhesion capability of the cells to representative extracellular matrix (ECM) proteins of the muscle tissue. This adhesion assay revealed that mBM-CD31⁺ cells displayed 5-, 10-, and 2-fold higher adhesion to collagen, laminin, and vitronectin, respectively, compared to the mBM-CD31⁻ cells (Figure 6C, D), indicating that higher adhesion capacity may underlie the higher engraftment of mBM-CD31⁺ in vivo.

Figure 6. Higher engraftment and adhesion capacity of mBM-CD31⁺ cells.

(A, B) Microscopic and FACS analyses of the engrafted cells. The CD31⁺ and CD31⁻ cells isolated from GFP mice were intramuscularly injected into ischemic hindlimbs. Two weeks later, the engrafted cells in the hindlimbs were examined by confocal microscopic (A) and flow cytometric (B) analysis (*P < 0.05, n = 5). (C, D) The representative images and quantitative analysis of the adhesion assay. CD31⁺ and CD31⁻ cells were plated onto dishes with various ECM proteins and the number of adherent cells was determined by fluorescent microscopy (*P < 0.05, n = 3 – 5). Bars: A, 50 μm; C, 100 μm.

BM-CD31⁺ cells induce vasculogenesis in vivo

We determined the vasculogenic potential of mBM-CD31⁺ cells in vivo, i.e., whether the transplanted CD31⁺ cells can give rise to ECs. As the vasculogenesis of BM cells is disputed, we used rigorous criteria to judge the incorporation of cells into blood vessels as ECs. To track the transplanted cells, we isolated CD31⁺ and CD31⁻ cells from the BM of GFP-transgenic mice (Figure 7) and confirmed these findings again with DiI labeled cells (Online Figure XII). To identify functional ECs, we intravenously injected ILB4, which stains the luminal side of ECs in functioning blood vessels⁹. We also used high resolution confocal microscopy and analyzed the images by Z-stacked three-dimensional (3D) reconstruction and FACS analysis. The images showed that mBM-CD31⁺ cells were incorporated into functional capillaries and displayed typical EC morphology (Figure 7A, B, Online Video 3), indicating true vasculogenic potential of mBM-CD31⁺ cells. Then, we quantified the number of vasculogenic cells derived from the transplanted mBM-CD31⁺ cells using flow cytometric analysis after enzymatic digestion of the hindlimb tissues⁹. ECs were identified by in vivo perfusion of ILB4 and the transplanted cells were identified by their intrinsic GFP fluorescence. The number of injected cells exhibiting ILB4 was 3.7-fold higher in the CD31⁺ group than in the CD31⁻ group (engrafted cells/total ILB4⁺ ECs, mBM-CD31⁺ vs. mBM-CD31⁻, 3.3 ± 0.8 vs. 0.9 ± 0.3%, P < 0.05) (Figure 7C, D), suggesting that vasculogenic cells are significantly enriched in the CD31⁺ population. As CD31⁺ cells include HSCs, HPCs, and a fraction of myelomonocytic cells, all of which were reported to have vasculogenic effects, we further investigated what fractions contribute to the vasculogenic effect of the CD31⁺ cells. For this, we quantitatively evaluated the vasculogenic effect of CD34⁺ cells (HSCs and HPCs) and CD14⁺ cells (myelomonocytic cells) with CD34⁻CD31⁺ cells and CD14⁻CD31⁺ cells, respectively. The vasculogenic effects of CD34⁺ and CD34⁻CD31⁺ cells were similar (Online Figure XIIIA, B), suggesting that the CD34⁻CD31⁺ cell fraction also contains a significant number of vasculogenic cells. Intriguingly, the vasculogenic effect of CD14⁻CD31⁺ cells were significantly higher than CD14⁺ cells (Online Figure XIIIC, D), indicating that a considerable portion of vasculogenic cells are enriched in non-myelomonocytic CD31⁺ cells. Together these results suggest that CD31⁺ cells include novel vasculogenic cells that do not belong to known vasculogenic populations: hematopoietic stem and progenitor cells (CD34) and myelomonocytic cells (CD14).

Figure 7. Vasculogenesis in the ischemic hindlimb tissue.

mBM-CD31⁺ or -CD31⁻ cells were isolated from GFP mice and intramuscularily injected into ischemic hindlimbs. One (A-B) or two weeks (C,D) after the transplantation, mice were intravenously injected with Alexa 647-conjugated ILB4. (A-B) Representative confocal images. Blue, DAPI (nuclei); green, GFP (CD31⁺ cells); red, ILB4 staining. (A and B) Orthogonal (A) and 3D-projection (B) views with corresponding single color images. Transplanted CD31⁺ cells incorporated into functional blood vessels as ECs. (C) Representative FACS analysis of digested hindlimb tissue to determine ECs derived from the transplanted cells. (D) The number of ECs derived from injected cells. (*P < 0.05, n = 5). Bars: A and B, 10 μm.

Transplantation of BM-CD31⁺ cells upregulates angiogenic, anti-apoptotic, and chemoattractant factors in ischemic hindlimbs

To investigate the effects of mBM-CD31⁺ cells transplanted into ischemic hindlimb, the biological factors associated with neovascularization were measured by real-time RT-PCR on adductor muscles harvested 1 week after treatment. Various angiogenic factors, such as Fgf2, Plgf, Angpt1, Angpt2, Tgfb, and Il6, were more highly expressed in the mBM-CD31⁺ group compared to the CD31⁻ and PBS groups (Figure 8). Interestingly, Angpt2 increased by 400 fold, whereas Angpt1 increased by only 13 fold in the CD31⁺ group compared to the PBS group. The relative predominance of Angpt2 over Angpt1 has been previously reported to be an indicator of a pro-angiogenic milieu³¹. Furthermore, anti-apoptotic and angiogenic factor Igf1 was significantly upregulated in the CD31⁺ group compared to the other groups. A representative chemoattractant molecule, Cxcl12 (Sdf1a) was also significantly upregulated in the CD31⁺ group compared to the other groups. Taken together, these findings indicate that many representative biological factors associated with angiogenesis, anti-apoptosis, and BM cell chemotaxis were significantly upregulated in the ischemic muscles injected with mBM-CD31⁺ cells, supporting a paracrine role for CD31⁺ cells in the improvement of hindlimb ischemia.

Figure 8. Upregulation of various paracrine or humoral factors in ischemic hindlimbs injected with mBM-CD31⁺ cells.

The hindlimb tissues were harvested 7 days after cell transplantation and subjected to qRT-PCR. Gene levels were normalized to GAPDH. Data are presented as fold differences compared to the value of the PBS group (n = 4 per group, *P < 0.05, **P < 0.01, ***P < 0.001).

Discussion

In this study, we report that CD31 is a unique marker of a full spectrum of non-endothelial hematopoietic BM cells that are tightly associated with neovascularization, as evidenced by high angiogenic properties, high adhesion capacity, exclusive HSC/HPC activity, and vasculogenic ability. These versatile functions enable CD31⁺ cells to induce higher therapeutic neovascularization and greater therapeutic effects on the repair of limb ischemia.

CD31⁺ cells are highly pro-angiogenic cells both in mouse and human BM. To our knowledge, this is the first study to demonstrate, with the use of multiple objective analyses, that a specific, unmanipulated cell population in BM is more highly angiogenic than others. In GSEA analysis, we used a priori-defined gene sets to analyze the data since post hoc examination of top-ranking genes can be subjective and prone to bias. The histologic analysis of hindlimb tissues showed that incorporation of transplanted CD31⁺ cells into the vasculature as ECs was significant, but accounted for only 3.3% of the total functional ECs. However, CD31⁺ cell injection increased capillary density by 1.3-fold, implying that non-vasculogenic effects of the transplanted CD31⁺ cells are a more dominant mechanism for neovascularization. Compatible with these data, transplantation of CD31⁺ cells resulted in global upregulation of multiple angiogenic, anti-apoptotic, and chemotactic factors in the ischemic limb.

With regard to the vasculogenesis of BM cells, we provide unambiguous evidence for the physical contribution of injected BM cells to endothelial cell generation. Some recent studies have argued against the vasculogenic potential of externally injected BM cells, particularly in cardiovascular models¹¹. We assumed that much of this controversy originated from the technical difficulties of confirming vasculogenesis or from the difference in the experimental settings such as type of injected cells, the recipient environment (animal models), or route of cell administration. Supporting our notion is the fact that the data regarding the absence of differentiation of BM cells into endothelial cells and cardiomyocytes came from studies using specific HSC populations and specific animal models such as MI⁸, in which long-term engraftment and survival of transplanted cells is expected to be low due to constant motion of the heart and little oxygen and nutrients available for the transplanted cells. In fact, another paper refuting the transdifferentiation of similar HSCs to cardiomyocytes did not mention the potential of endothelial transdifferentiation of the same cells³². Other recent studies showing the direct contribution of BM cells to new vasculature used several tumor models and BM transplantation^{3, 9}. Based on these observations, we hypothesized that postnatal vasculogenesis can occur through BM cells, although it may not occur frequently and may not be the main cause of the therapeutic effects of BM cells on cardiovascular disease. With the following extensive and meticulous analyses, the use of both 3D reconstruction of confocal images, flow cytometric analysis of the digested tissues, use of both chemical dye and GFP-transgenic mice for cell targeting, and application of rigorous criteria, we were able to confirm that BM cells, especially CD31⁺ cells, can generate significant numbers of ECs in ischemic tissues. Thus, we were able to conclude that when an appropriate BM cell is introduced into a specific environment (here, hindlimb ischemia), genuine vasculogenesis from BM cells can occur. This idea can be further expanded to describe the differentiation into ECs as a context-dependent phenomenon rather than an “all or none” event. This study provides convincing evidence for the occurrence of vasculogenesis or endothelial differentiation by externally injected BM cells in at least an ischemic vascular disease setting.

This study further revealed another novel mechanism underlying the observed therapeutic effect, cell engraftment. The higher paracrine and vasculogenic effects of CD31⁺ cells may be, at least in part, due to higher engraftment of the transplanted CD31⁺ cells. The mechanisms underlying the higher engraftment of CD31⁺ cells may be stronger adhesion of CD31⁺ cells to the ECM, which prevents the transplanted cells from both anoikis (apoptosis due to lack of adhesion to ECMs) and cell removal by the vascular system. The importance of cell engraftment and cell survival for increasing therapeutic effects has been demonstrated by previous studies which employed overexpression of the Akt pathway in transplanted cells or application of heat shock prior to transplantation^{33, 34}. This study demonstrates, for the first time, that a certain native population of BM cells, CD31⁺ cells, possesses higher engraftment capability upon transplantation. We also do not exclude possible favorable effects of an immunomodulatory role of the CD31⁺ cells on therapeutic effects. Given that CD31⁺ cells include B and T lymphocytes and other potential immune related cells, CD31⁺ cells may exert an immunomodulatory function to reduce tissue damage and augment tissue regeneration. We also showed that mouse and human HSCs and HPCs are almost exclusively restricted to the CD31⁺ fraction. Although mouse HSCs have been reported to express CD31²⁰, it was not known whether human HSCs and mouse HPCs do so as well. Our flow cytometric analysis, in vitro clonogenic assay, and short-term BM transplantation studies demonstrated that almost all HSCs and HPCs both in mouse and human BM are CD31⁺, and we propose CD31 to be a universal marker for HSCs and HPCs. Since there have been reports showing the favorable effects of HSCs for ischemic cardiovascular diseases^{18, 23, 24, 35}, the enrichment of HSCs in the CD31⁺ fraction may provide another advantage of using CD31⁺ cells for therapeutic purposes.

Our study provides compelling evidence for the therapeutic utility of CD31⁺ cells for regenerating ischemic tissues. In earlier studies, the criteria to select a specific population for therapeutic use was based on the two premises that HSC/HPCs or EPCs would have higher differentiation potential to give rise to ECs, and that therapeutic effects are mainly attributable to the vasculogenic effect of the transplanted cells. Thus, HSC/HPC or EPC markers, such as Lin⁻c-kit^{+ 23}, CD34^{+ 18, 24}, CD133^{+ 35}, and Hoechst dye exclusion (side population)³⁶, have been used to select therapeutically effective cells. However, given the recent reports that therapeutic effects of BM cells in ischemia repair are mainly attributable to humoral or angiogenic effects ^{16, 37}, isolating more highly angiogenic and/or paracrine cells may be a better option for cell therapy. In this respect, BM-CD31⁺ cells could serve as a favored option for second-generation cell therapy, as CD31⁺ cells have several advantages over other cell types. First, transplantable CD31⁺ cells can be prepared without ex vivo culture, which is required for EPCs and mesenchymal stem cells. Second, CD31⁺ cells are relatively abundant. Unlike CD133⁺ and CD34⁺ cells, it is unnecessary to collect a large volume of BM or to mobilize BM cells. Third, CD31⁺ cells are safer; we did not observe any of the adverse effects, such as aggravation of hindlimb ischemia³⁸, calcification or tumor formation, reported in the use of BM-MNCs³⁸, whole BM cells³⁹ or mesenchymal stem cells⁴⁰, respectively. However, given the recent reports showing a contribution of CD31 to the development of atherosclerosis in abdominal aorta in an animal model⁴¹, close monitoring is required when using CD31⁺ cells in atherosclerotic disease. Another concern is the effects of the disease state on the function of CD31⁺ cells. Although CD31⁺ cells derived from ApoE KO mice are therapeutically comparable to wild-type CD31⁺ cells, we do not exclude the possibility that CD31⁺ cells derived from chronically ill patients with multiple cardiovascular risk factors may be less efficacious. Together, this study suggests that CD31⁺ cells could serve as a novel therapeutic option for treating ischemic vascular diseases.

Novelty and Significance.

What is known?

• Bone marrow (BM)-derived endothelial progenitor cells (EPCs), which are cultured cells, were reported to have blood vessel-growing capability.

• There has been controversy regarding the vasculogenic ability or direct (trans)differentiation of BM-derived cells.

• CD31 is a traditional marker for endothelial cells and is also expressed in a fraction of BM cells.

What new information does this article contribute?

• CD31 is a comprehensive marker to identify highly angiogenic and vasculogenic cells among uncultured heterogeneous BM mononuclear cells (MNCs) both in human and mouse.

• A fraction of BM-MNCs that express CD31 can contribute to blood vessel formation by differentiation into endothelial cells.

• Human and mouse BM-derived CD31⁺ cell transplantation ameliorated the effects of ischemia in a hindlimb ischemia model.

Cell therapy with BM cells has emerged as a new therapeutic option for treatment of ischemic cardiovascular diseases. However, several important questions remain regarding the therapeutic mechanism and choice of cells: whether BM cells directly contribute to the generation of endothelial cells (vasculogenesis), and whether there is a marker to identify cells possessing high blood vessel forming capabilities among uncultured heterogeneous BM cells. In this study, we show for the first time that CD31 can serve as a marker to identify highly angiogenic and vasculogenic cells among uncultured BM-MNCs. We also provide evidence for the occurrence of adult vasculogenesis from BM-CD31⁺ cells injected into ischemic tissue. We discovered that CD31⁺ cells showed higher adhesion and engraftment properties and paracrine activities than CD31⁻ cells and are effective for treating ischemic vascular disease; however, the major therapeutic mechanism(s) underlying the effects of CD31⁺ cells are independent of endothelial differentiation of the transplanted cells. This study suggests that BM-CD31⁺ cells can be a highly promising cell source for cell therapy to treat ischemic cardiovascular diseases.

Non-standard Abbreviations and Acronyms

CLPs

common lymphocyte progenitors

CMPs

common myeloid progenitors

ECM

extracellular matrix

ECs

endothelial cells

EPCs

endothelial progenitor cells

GSEA

gene set enrichment analysis

HPCs

hematopoietic progenitor cells

HSCs

hematopoietic stem cells

LSK

lineage⁻ Sca-1⁺c-Kit⁺

MI

myocardial infarction

MNCs

mononuclear cells

MPPs

multipotent progenitor cells