Placental growth factor reconstitutes hematopoiesis by recruiting VEGFR1⁺ stem cells from bone-marrow microenvironment

CD34⁺VEGFR1⁺ cells comprise of NOD/SCID-repopulating cells

Hematopoietic stem cells (HSCs) express the antigens CD34 and/or AC133. In unseparated cord blood (CB), 4% of mononuclear cells expressed both VEGFR1 and CD14, whereas a smaller cell fraction (<0.2%) stained for VEGFR1, but not for CD15 and CD14. VEGFR1 was expressed on 6.4 ± 0.5% and 4.3 ± 0.3% of human fetal liver and CB-derived CD34⁺ and AC133⁺ cells, respectively (Supplementary Fig. A online). VEGFR1⁺ cells were also detected in the CD38⁻ population, which comprises a phenotypically primitive population of HSCs.

To evaluate the pluripotency of VEGFR1⁺ cells, human CB-derived CD34⁺, CD34⁺VEGFR1⁺ and CD34⁺VEGFR1⁻ cells were transplanted into sublethally irradiated NOD/SCID mice. Transplantation of 10⁴ human-derived CD34⁺VEGFR1⁺ cells into NOD/SCID mice resulted in long-term engraftment of human/mouse chimeric BM cells 4 months after transplantation (Supplementary Fig. B online). Eight times more cells were necessary to engraft NOD/SCID mice with CD34⁺VEGFR1⁻ cells. In the BM of NOD/SCID mice transplanted with VEGFR1⁺CD34⁺ cells (and CD34⁺ cells), 70 days after transplantation, 12.6% ± 2.0 (20.7% ± 0.7) of BM cells expressed the human B-cell CD19, 0.3% ± 0.1 (0.4% ± 0.2) stained for the human T-cell CD3 and 4.9% ± 0.9 (8.0% ± 7.4) expressed the human myeloid CD33 and co-expressed human CD45, respectively. These data indicate that CD34⁺VEGFR1⁺ cells comprise cells with long-term hematopoietic repopulating capacity, capable of multi-lineage cell differentiation.

Short- and long-term BM-repopulating cells express VEGFR1

Murine lineage-negative Sca-1 positive (Lin⁻Sca-1⁺) cells and side population cells (SP cells)²⁶ are enriched for pluripotent HSCs with BM-repopulating capacity. FACS analysis revealed that 1.4% of murine unseparated BM mononuclear cells (BMMCs) stained for VEGFR1 (Fig. 1a). Following Lin⁻Sca-1⁺ isolation, 73.5 ± 3.1% (n = 6) of these cells expressed VEGFR1, 29.5% expressed c-Kit and 2.5% double-stained for VEGFR1 and VEGFR2. VEGFR1⁺ cells were also detected in the SP cells of murine BMMCs (data not shown).

In BALB/c mice, 2 days after 5FU treatment, 5.1 ± 0.3% (n = 6) of BMMCs expressed VEGFR1 as compared with 1.4% prior to treatment (Fig. 1b). Within the VEGFR1⁺ population, 31.4 ± 0.3% of the cells co-expressed Sca-1 and 35.7 ± 0.7% stained for c-Kit, both of which are stem cell-associated markers. To determine the BM-repopulating capacity of VEGFR1⁺ cells, post 5FU-purified VEGFR1⁺ BMMCs from male mice were transplanted into lethally irradiated syngeneic female mice. All mice transplanted with VEGFR1⁻ and VEGFR2⁺ cells died within 14 days, whereas 38%, 63%, and 100% of mice transplanted with 1 × 10², 1 × 10³ and 1 × 10⁵ of VEGFR1⁺ BMMCs from male donors survived beyond 150 days, respectively (Fig. 1c). At 5 months, more than 80% of BM cells were Y-chromosome positive donor cells in female recipients transplanted with VEGFR1⁺ BMMCs.

Similarly, congeneic transplantation of 1 × 10³ VEGFR1⁺Sca-1⁺ or VEGFR1⁺Sca-1⁻ BMMCs obtained from 5FU-treated C57BL/6-Ly5.2 mice rescued lethally irradiated C57BL/6-Ly5.1 mice (Fig. 1d). Five months after transplantation, 88% of myeloid (CD11b, Gr-1) and lymphoid (B220, Thy-1) cells in the circulation were of VEGFR1⁺ donor (Ly5.2) origin (Fig. 1e), indicating that murine VEGFR1⁺ BM cells have short- and long-term BM-repopulating capacity.

Activation of VEGFR1 is essential for BM reconstitution

5FU-injected mice treated with IgG (controls) or monoclonal antibody to VEGFR2 had complete hematopoietic reconstitution within three weeks (Fig. 2a). However, 67% of mice treated with neutralizing antibody to VEGFR1 succumbed within 3 weeks after 5FU-induced BM-suppression (Fig. 2b). The surviving mice had profound delay in hematopoietic recovery (Fig. 2a). Treatment with either anti-VEGFR2 or anti-VEGFR1 for 4 weeks in non-5FU treated mice did not cause hematopoietic toxicity (Fig. 2a). These data suggest that signaling through VEGFR1, but not VEGFR2, plays a role in hematopoietic reconstitution after myeloablation.

If VEGFR1 inactivation following myelosuppression delays hematopoietic reconstitution, administration of PlGF, the ligand signaling through VEGFR1 but not VEGFR2, should improve hematopoietic recovery. Mice were treated with 5FU (Fig. 2c) or a combination of carboplatin and total body irradiation (TBI) (Fig. 2d). Following myelosuppression, elevated PlGF levels delivered by adenoviral-vector expressing PlGF (Ad-PlGF), resulted in faster WBC recovery in the early (day 0–6) and later (day 7–16) phases of BM suppression. Extent and duration of WBC recovery of less than 2000/μl was shorter in the Ad-PlGF-treated group compared with the Ad-null-treated group (Fig. 2c and d). Plasma elevation of PlGF improved WBC recovery similar to that seen after G-CSF treatment (Fig. 2d), except in the early phases after myelosuppression, at time points where G-CSF is proven to be ineffective^27,28.

Inhibition of VEGFR1 impairs hematopoietic recovery

The delayed hematopoietic recovery following myelosuppression was associated with a significantly reduced number of BMMCs in mice injected with anti-VEGFR1, as compared with mice treated with antibody to VEGFR2 or IgG (Figs. 2e, f and g and 3a). The BM of 5FU-treated mice injected with anti-VEGFR1 showed a decrease in lineage-committed cells, particularly in the megakaryocytic lineage (Fig. 2h–m).

FACS analysis of BM cells confirmed the tri-lineage suppression after treatment with blocking anti-VEGFR1. The frequency of BMMCs expressing erythroid (TER119), myeloid (CD11b, Gr1) and lymphoid (B220) markers was significantly decreased on day 6 after VEGFR1 antibody treatment as compared with IgG controls (Fig. 3b). There was also a delay in the erythroid and lymphoid recovery in mice receiving anti-VEGFR2, but this did not translate into increased mortality in these mice. The decreased BM cellularity in anti-VEGFR1-treated mice was due to fewer Lin⁻Sca-1⁺ (Fig. 3c) and Sca-1⁺ cells (Fig. 3d and e) in S and G₂/M phase of the cell cycle as compared with anti-VEGFR2-and IgG-treated mice. Because 70% of the mice treated with anti-VEGFR1 succumbed to fatal complications afforded by BM failure, timely activation of the PlGF/VEGFR1 pathway is important for HSC and progenitor recruitment, cell differentiation and restoration of multi-lineage hematopoiesis to avoid life-threatening complications.

PlGF enhances recruitment and migration of VEGFR1⁺ HSCs

How does PlGF enhance early phases of hematopoietic recovery? PlGF increased mature WBC counts 2-fold above baseline, with a significant increase in circulating monocytes (Fig. 4a). Mobilization of hematopoietic cells into circulation followed the kinetics of PlGF plasma elevation (Fig. 4a, insert). Increased numbers of hematopoietic progenitors capable of forming CFU colonies (CFU-Cs) (Fig. 4b) were found within the PlGF-mobilized cells. PlGF augmented the number of circulating pluripotent hematopoietic cells (CFU-Ss) by 20-fold, as compared with controls three days after adenoviral injection (Fig. 4c). PlGF also mobilized VEGFR1⁺ BM-repopulating stem cells that engrafted and rescued lethally irradiated recipients (Fig. 4d). More than 80% of BM-donor cells were detected in mice transplanted with PlGF-mobilized cells. We found that PlGF and VEGF-A stimulate the release of pro-MMP-9 (Fig. 4e, insert) and induce migration of human CD34⁺ cells and cells with stem-cell potential (cobblestone-forming cells, CAFC). This migration was blocked by anti-VEGFR1 (Fig. 4e). These data suggest that PlGF is a chemoattractant that mediates its effect through VEGFR1, because blockade of VEGFR1 completely abolished HSC and progenitor mobilization in vitro and in vivo.

PlGF restores late phases of hematopoietic recovery

In the absence of exogenous growth factor(s), BM reconstitution is complete within 20–30 days following myelosuppression. Immediately post-BM suppression, quiescent Lin⁻Sca-1⁺ HSCs undergo rapid cell cycling to replenish the progenitor-cell pool. Blocking VEGFR1 signaling following myeloablation resulted in a decrease in immature Sca-1⁺ and Lin⁻Sca-1⁺ BM cells in S phase of the cell cycle (Fig. 3c, d and e). In vitro exposure of CD34⁺ cells to PlGF in the presence of sKitL, thrombopoietin and Flt3/Flk2 ligand had no effect on cell proliferation as determined by cell number, CFU-C and long-term culture-initiating cell assay (data not shown).

Therefore, the mechanism for the differential behavior of HSC in vitro and in vivo could be indirectly mediated through the release of a stem cell–active cytokine and/or through the activation of a protease regulating the bioavailability of cytokines. We demonstrate that PlGF induced matrix metalloproteinase-9 (MMP-9) expression in BM cells (Fig. 5a). If MMP-9 upregulation is the critical step during PlGF-mediated HSC mobilization, PlGF should not mobilize HSCs in MMP-9^−/− mice. Indeed, plasma elevation of PlGF in wild-type, but not in MMP-9^−/− mice, mobilized mature WBCs (Fig. 5b), CFU-Cs (Fig. 5c) and CFU-Ss (Fig. 5d), especially in later phases of PlGF-induced mobilization.

To confirm the role of PlGF-induced MMP-9 activation in BM recovery, we injected Ad-PlGF in 5FU-treated MMP-9^−/− and MMP-9^+/+ mice. In the early phase of hematopoietic recovery, higher WBC counts were found in Ad-PlGF-treated mice, an MMP-9 independent process (Supplementary Fig. C online). In the late phase of hematopoietic recovery, Ad-PlGF improved WBC recovery in MMP-9^+/+, but not in MMP-9^−/− mice. The improved WBC recovery in Ad-PlGF-treated MMP-9^−/− mice translated into a slight reduction in the mortality following myelosuppression (Supplementary Fig. C online).

We recently showed that MMP-9-mediated release of soluble Kit-ligand (sKitL) is essential for hematopoietic reconstitution⁴. Notably, we detected a 10-fold increase in sKitL plasma levels in Ad-PlGF-treated mice as compared with controls (data not shown). PlGF-mediated MMP-9 activation is not critical during the early, but is necessary during the later phases of hematopoietic recovery following myelosuppression. As PlGF was not able to rescue 5FU-treated MMP-9^−/− mice, we hypothesized that the late phases of BM recovery are indirectly regulated by PlGF-mediated release of sKitL. Therefore, if the release of sKitL after PlGF administration is mediated through MMP-9, significantly lower sKitL levels should be found in PlGF-treated MMP-9^−/− as compared with MMP-9^+/+ mice, which was indeed the case (Fig. 5e). Because PlGF-mediated upregulation of MMP-9 augmented sKitL plasma levels, we predicted that administration of sKitL would rescue 5FU-myeloablated mice treated with anti-VEGFR1. Plasma elevation of sKitL restored survival (Fig. 6a), and hematopoietic recovery (Fig. 6b–d) of myeloablated mice treated with anti-VEGFR1. Remarkably, 5FU-treated mice injected with neutralizing anti-VEGFR1 showed significantly decreased sKitL plasma levels as compared with VEGFR2-treated or control mice (Fig. 6e). These data suggest that during late phases of BM recovery PlGF-mediated upregulation of MMP-9 facilitates the release of sKitL, thereby promoting cell differentiation and accelerates hematopoietic reconstitution (Fig. 6f).

Discussion

We demonstrate here that human and murine HSCs with long-term BM-repopulating capacity express VEGFR1, conveying signals regulating cell cycling and motility. Inhibition of VEGFR1, but not VEGFR2, blocked recruitment of HSCs from their resting microenvironment, thereby retarding hematopoietic reconstitution after myelosuppression. PlGF augmented early phases of hematopoietic recovery by promoting recruitment, chemotaxis and mobilization of VEGFR1⁺ HSCs and progenitors, while during late phases of BM recovery PlGF-mediated upregulation of MMP-9 facilitated the release of sKitL thereby accelerating hematopoietic reconstitution.

Our data suggest that PlGF activates non-angiogenic pathways and thus serves functions other than regulating angiogenesis. Based on our data, the accumulation of angioblasts within the yolk sac in VEGFR1^−/− mice may not only be due to enhanced commitment to angioblasts¹⁷, but may be due to impaired cell migration. This appears to be the case in Drosophila, where VEGF-A activation is directly linked to hemocyte motility²³.

How does VEGFR1 activation increase HSC motogenic potential? Chemotaxis of VEGFR1⁺ HSCs and progenitors accounts for the WBC increase in the early phase of PlGF mobilization. PlGF mobilized CFU-Cs and CFU-Ss in MMP-9^+/+, but to a lesser extent in MMP-9^−/− mice. In later phases, PlGF-induced cell mobilization was impaired in MMP-9^−/− mice, suggesting that PlGF recruits HSCs and progenitors by a different mechanism. PlGF induces the release of sKitL through MMP-9 activation. sKitL increases motility by promoting cell cycle transition through interaction with its receptor c-Kit^29,30. Activation of this pathway accounts for the hematopoietic reconstitution during later phases of BM recovery. Indeed, neutralizing anti-VEGFR1 profoundly diminished plasma elevation of sKitL in myelosuppressed mice during BM recovery, while introduction of sKitL into myelosuppressed mice treated with anti-VEGFR1 completely restored hematopoiesis.

The increase in cell cycling and proliferation of VEGFR1⁺Sca-1⁺ cells after BM suppression is most likely not a direct effect of PlGF, because activation of VEGFR1 did not change HSC survival or colony formation in vitro. These data support other reports in which VEGF-A or PlGF had no major effect on the proliferation or survival of HSCs or progenitors^31–33. However, these studies do not rule out the possibility that VEGF/VEGFR internal autocrine signaling pathways may synergize with other cytokines to convey survival signals for repopulating cells.

Although VEGFR2 is expressed on human NOD/SCID-repopulating cells¹⁴, the functional role of VEGFR2 in the regulation of post-natal hematopoiesis remains unclear. Transplantation of murine VEGFR2⁺ BM cells failed to reconstitute hematopoiesis in lethally irradiated recipients¹⁵. These data suggest that murine BM-derived VEGFR2⁺ cells may mark endothelial progenitors^1,34 rather than repopulating HSCs. Supporting these observations, we showed only a transient delay in the recovery of lymphoid and erythroid precursors in mice treated with neutralizing anti-VEGFR2. This delay in hematopoietic recovery was not sufficient to induce life-threatening complications, since all mice treated with 5FU and anti-VEGFR2 survived. Anti-VEGFR2 may delay the recovery of certain lineages through interference with the production of lineage-specific cytokines, such as GM-CSF and IL-6 (ref. ³⁵).

In contrast, mice treated with 5FU and neutralizing anti-VEGFR1 showed a striking impaired tri-lineage cell recovery with prolonged pancytopenia, which resulted in the demise of 70% of treated mice kept under germ-free conditions. The lack of a more profound defect in myelosuppressed mice treated with anti-VEGFR2 seems to be in disagreement with the dramatic defect in blood-island formation observed in VEGFR2-null mice. It is difficult to compare the role of VEGFR2 signaling in the regulation of hematopoiesis in the embryo and adult mice, as there are fundamental differences between the established post-natal BM microenvironment and fragile embryonic blood islands. Adult BM stromal cells provide a fortified cellular scaffold that may be resistant to effects of anti-VEGFR2. Since stromal cells express VEGFR1³⁶, it is conceivable that anti-VEGFR1 block MMP-9 expression leading to failure of the recruitment of HSCs. This may explain why anti-VEGFR1 compromises hematopoietic recovery, while inhibition of VEGFR2 has only a marginal effect on stress hematopoiesis.

We have shown that plasma elevation of VEGF₁₆₅ with or without angiopoietin-1 promoted mobilization of HSCs and CFU-Cs (ref. ²²). Here, we show that, under steady-state conditions, chronic inhibition of VEGFR1 or VEGFR2 had no major effect on hematopoiesis. This is in sharp contrast to the failure of hematopoietic reconstitution and demise of the treated mice by inhibiting VEGFR1 during BM suppression. These data suggest that during steady-state hematopoiesis, VEGFR1 inhibition does not result in a profound impairment in hematopoiesis, because the demand for HSC recruitment is low. In contrast, during stress hematopoiesis activation of VEGFR1 is necessary for hematopoietic reconstitution. This phenomenon is similar to other physiological processes requiring rapid tissue vascularization, where stress angiogenesis driven by the mobilization of VEGFR2⁺ BM-derived cells is necessary to accelerate neo-angiogenesis¹. These data suggest that in stress hematopoiesis the molecular switch requires activation of VEGFR1, whereas during stress angiogenesis the molecular switch is driven by VEGFR2 activation.

Based on our data, the clinical use of VEGFR1 inhibitors, but to a lesser degree VEGFR2, delivered in combination with myelosuppressive agents, may result in a life-threatening prolongation of BM suppression. In this respect, suppression of hematopoiesis through VEGFR1 inhibition has two ramifications. On one hand, blocking VEGFR1 may inhibit tumor angiogenesis and growth, but on the other hand, it may introduce unwanted BM toxicity. Nonetheless, since VEGFR1 toxicity is dose-dependent, VEGFR1 inhibitors can be delivered with a manageable toxicity profile. Simultaneous administration of sKitL may reduce BM toxicity caused by blocking VEGFR1 and therefore rescue patients from infection or hemorrhage.

Stress promotes mobilization of BM-derived stem cells that ultimately incorporate, although at very low levels, into specific organs. This low efficiency of incorporation may be reflected in the paucity of readily motile stem cells capable of being recruited from the BM to circulation. PlGF, endowed with a low toxicity profile, allows for mobilization of stem cells to the circulation, facilitating recovery of many stem cells that may ultimately be used for organ restoration.

Our findings introduce the novel paradigm that determination of VEGFR1 and MMP-9 expression should be considered as surrogate markers for evaluating the efficacy and success of stem cell mobilization and engraftment in transplantation settings. It is conceivable that dysregulated MMP-9, VEGFR1 or PlGF expression may be responsible for defects in the stem-cell microenvironment leading to BM failure. In this regard, therapeutic strategies to upregulate VEGFR1, PlGF and MMP-9 may provide an effective means to restore motogenic potential of BM-repopulating cells and help to replenish the stem cell pool after BM transplantation.

Methods

Supplementary Material