Myeloid lineage progenitors give rise to vascular endothelium

Results

To quantify endothelial progenitor activity in distinct bone marrow populations defined by lineage and progenitor status, genetically marked cells derived from adult mice were fractionated based on cell surface marker expression and transplanted into irradiated wild-type recipients. The potential problem of variable expression of transgenic donor cell markers (19) was prospectively addressed by using two independent sources of donor bone marrow cells that were genetically marked with either β-gal (ROSA26) (20) or EGFP (21). In addition to their characteristic tissue distribution and morphology, endothelial cells were identified based on the expression of CD31 and Tie2 and the absence of hematopoietic (CD45) macrophage (F4/80) and pericyte (desmin and smooth muscle actin) marker expression. The functional status of these endothelial cells was confirmed by the production of the procoagulant von Willebrand factor (VWF). Colocalization of donor and endothelial cell markers in individual cells was evaluated by deconvolution microscopy.

To determine whether a specific hematopoietic cell surface marker was informative for identifying cells with endothelial cell potential, bone marrow cells were sorted into lineage marker-positive and lineage marker-negative subpopulations. These cells were transplanted into lethally irradiated recipients at a dose based on their frequency in 1 × 10⁶ bone marrow cells. For example, the B cell lineage marker B220 is expressed on 20% of bone marrow cells; therefore, recipients were injected with either 2 × 10⁵ B220-positive cells or 8 × 10⁵ B220-negative cells (Fig. 1a). When the incorporation of either ROSA26- or EGFP-marked donor cells into liver vascular endothelium was assayed, all endothelial cells arose from the B220-negative subset of bone marrow, indicating that cells committed to the B lymphocyte lineage (B220⁺) do not have significant endothelial cell potential (Fig. 1b). By contrast, 100% of endothelial cell activity was segregated to the c-kit-positive bone marrow subset despite a 24-fold higher input dose of c-kit-negative cells. Further evaluation revealed that both stem cell antigen-1 (Sca-1)-negative cells and cells that expressed mature hematopoietic lineage markers (Lin⁺) including Ter119 (erythrocytes) and Mac-1 and Gr-1 (myelomonocytic cells) gave rise to endothelium (Fig. 1b). Thus, bone marrow-derived endothelial cell progenitors reside outside of the stem cell population, consistent with the hypothesis that endothelial cells can arise from the differentiated progeny of HSCs. These findings, in combination with the absence of endothelial cell outcomes derived from lymphoid progenitors (data not shown), prompted us to investigate whether endothelial cell progenitors represented the normal developmental potential of myeloid lineage precursors.

Akashi et al. (22) identified CMPs and GMPs in adult bone marrow. GMPs are known to be further differentiated progeny of CMPs and have lost the capacity to produce erythrocytes and platelets. Both GMPs and CMPs have limited self-renewal capacity, and their short-lived, mature progeny typically are not detected in the peripheral blood for >3–4 weeks after transplantation (22, 23). To test the endothelial cell potential of these myeloid lineage progenitors, 5–10 × 10³ GMPs or CMPs isolated from either ROSA26- or EGFP-expressing bone marrow were transplanted into lethally irradiated recipients along with a radioprotective dose of host-type bone marrow cells (Fig. 1 c and d). Before evaluating the recipient blood vessels for donor-derived endothelium, the peripheral blood was analyzed to ensure that neither the sorted CMP or GMP populations contained any contaminating multilineage stem/progenitor cells. No donor multilineage hematopoiesis was observed in four independent experiments evaluating a total of 25 transplant recipients.

To confirm that any long-lived, donor-derived hematopoietic progeny of CMPs and GMPs in the perivascular space were distinguished from bone marrow-derived endothelial cells, a detailed phenotypic analysis was performed (Fig. 2). The endothelial cell phenotype (CD31⁺, VWF⁺, CD45⁻, F4/80⁻, and DAPI⁺) was distinct from that of platelets (CD31⁺, VWF⁺, CD45⁻, F4/80⁻, and DAPI⁻), macrophages including Kupffer cells (CD31⁺, VWF⁻, CD45⁺ or CD45⁻, and F4/80⁺), granulocytes (CD31⁺, VWF⁻, CD45⁺, and F4/80⁻), lymphocytes (CD31⁻ or CD31⁺, VWF⁻, CD45⁺, and F4/80⁻), and pericytes (CD31⁻, VWF⁻, smooth muscle actin⁺, and desmin⁺). Colocalization of these markers in individual ROSA26- or EGFP-positive donor-derived cells was confirmed by deconvolution microscopy (Fig. 2 d and e and Fig. 5, which is published as supporting information on the PNAS web site). Hematopoietic cells, tissue-resident macrophages, and pericytes were readily detected and easily distinguished from endothelium in the liver parenchyma (Fig. 2f and Fig. 6, which is published as supporting information on the PNAS web site). Liver tissue was evaluated as early as 2–3 weeks after transplantation, and the frequency of portal vein cross-sections that contained donor endothelial cells was determined. Of a total of 7,300 portal vein cross-sections from recipients transplanted with CMPs, a mean of 1.3% contained donor-derived endothelial cells. Analysis of 6,060 portal veins in the GMP transplant recipients revealed that the frequency of portal veins containing donor endothelial cells was 0.8% (Fig. 2g). GMPs are the more differentiated progeny of CMPs (24); therefore, our results indicate that endothelial cell potential persists at least until the GMP stage of myeloid differentiation. When endothelial cell outcomes were examined at 6 months after transplant, no donor-derived endothelium was detected in recipients of either ROSA26- or EGFP-marked myeloid progenitors (data not shown). This finding is consistent with the limited lifespan and the modest self-renewal potential of both the CMP and GMP populations (22). In contrast, long-lived HSCs, with a high self-renewal potential, produce a modestly increased frequency of endothelial cells over this time period (14, 17).

Although our data are consistent with the differentiation of CMPs and GMPs into endothelial cells, the results of recent studies led us to evaluate cell fusion between the transplanted myeloid cells and host endothelial cells as an alternative mechanism. We (13) and others (25) have shown that correction of the metabolic liver disease fumarylacetoacetate hydrolase (Fah) deficiency by transplantation of HSCs is the result of fusion between donor hematopoietic cells and host cells. Moreover, when the progeny of HSCs were evaluated, we found that myelomonocytic cells fused with the diseased hepatocytes (23). To exclude cell fusion as the mechanism responsible for the generation of donor-marked endothelium, we used experimental strategies that have recently been used to detect fusion-derived hepatocytes, cardiomyocytes, and Purkinje neurons (12). Briefly, HSCs isolated from Cre recombinase transgenic mice were transplanted into mice harboring a β-gal gene inactivated by a floxed stop codon (R26R) (20). Thus, only host cells that fuse with donor cells will undergo DNA recombination and express β-gal. To ensure successful recombination in endothelial cell fusion products, the Tie2 promoter was used to drive Cre expression (26). Identification of donor-derived cells that were not the product of cell fusion was facilitated by using HSCs derived from mice that ubiquitously express EGFP (19). Moreover, to simplify the detection of hepatocyte fusion events, R26R recipient mice were backcrossed to render them Fah-deficient (13).

After transplantation, both donor-derived hepatocytes and endothelial cells were observed at the expected frequencies (14, 23). None of the EGFP-positive endothelial cells expressed the β-gal reporter gene, consistent with direct endothelial cell differentiation of the transplanted cells (Fig. 3). However, EGFP-positive hepatocytes expressed β-gal, as was evident from both X-Gal reactivity and immunostaining, a finding indicative of cell fusion (Fig. 3 a–e). Because wild-type as well as Fah-deficient hepatocytes lack Tie2 expression (data not shown), this result could only be attributed to Tie2 activity in the hepatocyte’s hematopoietic fusion partner. Apart from endothelial cells, HSCs (27) and, more recently, angiogenic myeloid cells (28) have been isolated from mouse bone marrow and peripheral blood based on Tie2 expression. Fusion with these cells could result in β-gal activation in the host hepatocytes and explain this result. Thus, Tie2 expression might distinguish the hepatocyte’s hematopoietic fusion partner from other myelomonocytic cells (23).

Although Tie2 expression was readily detected in donor-derived endothelial cells (Fig. 3i), the reverse experiment was performed to rule out lack of Tie2 expression in the donor cell nucleus as the cause for failed R26R activation. HSCs expressing both EGFP and R26R were transplanted into Tie2-Cre transgenic recipients (Fig. 3 j–m). In support of our initial findings, none of the donor-derived, EGFP-positive endothelial cells expressed the β-gal reporter gene, indicating that these cells were not fusion products. Successful recombination depends not only on the activity of the promoter driving Cre, but also on accessibility of the R26R locus. Thus, recombination efficiencies might differ between hepatocytes and endothelial and hematopoietic cells. Therefore, by transplanting EGFP-expressing donor cells into ROSA26 recipients (29), a simple and independent labeling approach was additionally used to exclude cell fusion (Fig. 3 n–r). As expected, the vast majority of EGFP-positive hepatocytes also expressed β-gal, whereas none of the EGFP-positive endothelial cells were β-gal-positive. The combined results from these three experimental strategies indicate that endothelial cells derived from hematopoietic progenitors are not products of cell fusion but that these cells arise by differentiation of bone marrow progenitors.

To facilitate hematopoietic engraftment, transplantation models typically require the preconditioning of recipients with ionizing radiation. In addition to ablating stem and progenitor cells in the bone marrow, radiation exposure is toxic to vascular endothelial cells (30). This issue raises the important question of whether injury is required for the contribution of circulating progenitor cells to the endothelium. Consequently, we used parabiosis as a noninjury model to determine the contribution of circulating endothelial cell progenitors under steady-state conditions (31, 32). EGFP-expressing and wild-type mice were surgically joined for 12 weeks, separated, and analyzed 4–6 months later. Each member of the parabiotic pair showed a mean of 17% donor-derived hematopoietic cells in the peripheral blood (Fig. 4a). Circulating, donor-derived T cells, B cells, and myelomonocytic cells were found at normal frequencies (Fig. 4b). The successful transfer of self-renewing HSCs was confirmed by serial transplantation (data not shown). When the portal veins were examined in detail, the mean frequency of vessels demonstrating engraftment of EGFP⁺, CD45⁻, CD31⁺, or VWF⁺ endothelial cells was 3.9%. Donor-derived endothelial cells were also detected in other tissues including the skin and kidney (Fig. 7, which is published as supporting information on the PNAS web site). In corroboration of our findings in irradiated recipients, additional experiments with parabiotic pairs of EGFP and ROSA26 transgenic mice did not reveal any evidence for cell fusion in the vascular endothelium (data not shown). These results demonstrate that circulating endothelial cell progenitors contribute to the vascular endothelium during steady-state conditions outside of the context of radiation-induced vascular injury.

Discussion

Our results demonstrate that vascular endothelial cells can arise from a well defined population of myeloid lineage-restricted bone marrow progenitors, suggesting that endothelial cells represent a previously unrecognized differentiation potential of the myeloid lineage. This finding is consistent with results from several laboratories that have reported circulating cells with myeloid features contributing to human angiogenesis (8–11). Importantly, we have also shown that donor-derived endothelial cells differentiate from myeloid progenitors and are not products of cell fusion. Whereas transplantation of HSCs results in long-term contribution to the hematopoietic and the vascular systems, transplantation of myeloid progenitors provides only short-term engraftment of hematopoiesis and endothelial cells. This result was anticipated given the limited self-renewal capacity of these progenitor cells (22). Remarkably, the development of endothelial cells from circulating hematopoietic progenitors occurs in the absence of acute radiation injury, suggesting a role for these cells during steady-state conditions.

A notable feature of recent studies in both mice and humans is the similar frequency of long-term incorporation of bone marrow-derived cells into vascular endothelium. Studies evaluating sex-mismatched human bone marrow transplant recipients from our laboratory (6) and others (7) indicate that 2–5% of endothelial cells are donor-derived. This frequency of donor-derived endothelial cells is indistinguishable from that observed in our previously reported (14) mouse transplant model (3–4%) and is similar to recent reports by Larrivee et al. (17). Taken together, these results suggest that bone marrow-derived endothelial cells contribute to 1–5% of the endothelium in the liver, gut, and skin.

Several groups have used innovative approaches to demonstrate that growth factors and cytokines can recruit large numbers of myelomonocytic cells from the bone marrow to a variety of tissues, where they secrete factors that promote endothelial cell proliferation and differentiation (33–35). Using a tissue-specific angiogenesis model based on VEGF overexpression, Keshet and colleagues (34) showed a massive recruitment of proangiogenic myeloid cells and a relatively rare frequency of incorporation of these cells into the endothelium. These studies further extend the concept that myelomonocytic cells can serve as an important source of proangiogenic cytokines in the setting of acute ischemic injury (9).

The coexistence of myeloid progenitors capable of endothelial differentiation and proangiogenic myeloid accessory cells likely represents two complementary mechanisms by which myeloid lineage progenitors contribute to angiogenesis. The relative importance of each mechanism may depend on the type of repair or regeneration being studied. For example, our finding of endothelial cell differentiation in a parabiotic model is consistent with circulating progenitors contributing to vascular homeostasis. Our current study clearly identifies myeloid progenitors as a source of vascular endothelium. Because myelomonocytic cells are known to effectively infiltrate both normal and diseased tissues, future studies on the contribution of myeloid lineage-derived endothelial progenitors to vascular regeneration and repair are needed.

Methods

Supplementary Material