KDR Receptor: A Key Marker Defining Hematopoietic Stem Cells

See allHide authors and affiliations

Science  03 Sep 1999:
Vol. 285, Issue 5433, pp. 1553-1558
DOI: 10.1126/science.285.5433.1553


Studies on pluripotent hematopoietic stem cells (HSCs) have been hindered by lack of a positive marker, comparable to the CD34 marker of hematopoietic progenitor cells (HPCs). In human postnatal hematopoietic tissues, 0.1 to 0.5% of CD34+ cells expressed vascular endothelial growth factor receptor 2 (VEGFR2, also known as KDR). Pluripotent HSCs were restricted to the CD34+KDR+ cell fraction. Conversely, lineage-committed HPCs were in the CD34+KDRsubset. On the basis of limiting dilution analysis, the HSC frequency in the CD34+KDR+ fraction was 20 percent in bone marrow (BM) by mouse xenograft assay and 25 to 42 percent in BM, peripheral blood, and cord blood by 12-week long-term culture (LTC) assay. The latter values rose to 53 to 63 percent in LTC supplemented with VEGF and to greater than 95 percent for the cell subfraction resistant to growth factor starvation. Thus, KDR is a positive functional marker defining stem cells and distinguishing them from progenitors.

The hierarchy of human hematolymphopoietic cells is defined by functional assays. HSCs with extensive self-renewal capacity are assayed in vivo for their capacity to xenograft nonobese diabetic–severe combined immunodeficiency disease (NOD-SCID) mice and sheep fetuses (1–3). These models are surrogates for a syngeneic transplantation assay. Primitive HPCs with limited self-renewal potential are identified in vitro as high–proliferative potential colony-forming cells (HPP-CFCs) (4). Lineage-committed HPCs with no self-renewal activity are also defined in vitro by clonogenic assays as colony-forming units (CFUs) or burst-forming units (BFUs) (1, 5). Dexter-type LTC, consisting of a liquid phase on irradiated BM stroma, identifies LTC-initiating cells (LTC-ICs, generating HPCs assayed in secondary culture) (6) and cobblestone area–forming cells (CAFCs, generating hematopoietic colonies recognized as “cobblestone areas” in LTC stroma) (7). Depending on the LTC duration, LTC-ICs represent primitive HPCs (5- to 8-week LTC) (8) or highly quiescent putative HSCs resistant to retroviral gene transfer (12-week LTC) (9) (see below).

Although HSC identification is still elusive, recent observations have suggested a role for VEGFR2 (Flk1 in mice) in murine embryonic hematoangiogenesis. Targeted gene disruption studies indicate that Flk1 is required for initiation of hematolymphopoiesis and vasculogenesis (10), implying that Flk1 may be required for generation of hemoangioblasts, that is, the hypothetical stem cells for both hematolymphopoietic and endothelial lineages (11). Other studies also suggested the existence of embryonic Flk1+ cells with hemoangiogenic potential but did not allow identification of prenatal repopulating HSCs (12). We identified KDR as a major functional marker for postnatal HSCs.

In postnatal life, KDR is expressed on endothelial cells (13), and the mRNA can be detected in HPCs and megakaryocytes (14, 15). We used a high-affinity monoclonal antibody (mAb clone 260.4) that binds the extracellular KDR domain to monitor KDR expression by flow cytometry on cells from BM, normal peripheral blood (PB) or mobilized peripheral blood (MPB), and cord blood (CB) (16). Only 0.1 to 0.5% of CD34+ cells ≥95% pure from these tissues were KDR+ (Fig. 1A, top), as confirmed by reverse transcriptase polymerase chain reaction (RT-PCR) analysis (Fig. 1A, bottom right). The CD34+KDR+ cells, essentially lin (<10% CD45RA+; <5% CD13+, CD33+, CD61+, and CD19+), are at least in part positive for early progenitor and stem cell markers, namely, CD38, Thy-1+, and kitlow(Fig. 1B). The published HSC-enriched fractions (for example, CD34+38, CD34+Thy-1+, and CD34+ kitlow; see below) comprise ≤1 to 2% of KDR+ cells.

Figure 1

KDR expression and distribution in CD34+ cells. (A) (top) KDR expression by flow cytometry on BM, PB, MPB, and CB CD34+ cells (16). Cells gated on physical parameters were analyzed for specific and nonspecific (isotype-matched) reactivity (>40,000 cells were acquired for analysis). Percentage of CD34+KDR+ cells is shown by numbers. The CD34+KDR+ fraction size is similar in PB (0.19 ± 0.04 of CD34+ cells, mean ± SEM from five experiments), MPB (0.25 ± 0.04), BM (0.34 ± 0.06), and CB (0.40 ± 0.04). (bottom) (left) Standard gates for sorting of KDR+, KDR+/±, and KDRCD34+ cells (16). A representative CB experiment is shown. (right) RT-PCR analysis of mRNA KDR in CD34+KDR+ versus CD34+KDR sorted BM cell populations and HEL cell line (16). KDR and β2-microglobulin (β2m) bands are shown. (B) Electronically gated CD34+ cells (top) from BM, PB, MPB, and CB cells were analyzed for expression of KDR and relevant early hematopoietic antigens (16). CD34+KDR+ fraction cells are CD38 (35.3% ± 1.6 to 45.6% ± 1.5 of cells, mean ± SEM from four experiments), Thy-1+(55.6% ± 3.7 to 83.1% ± 1.7), and kitlow (c-kit) (81.5% ± 2.0 to 91.0% ± 1.6, except for PB).

VEGF treatment of human HPC culture or normal mice inhibits or stimulates hematopoiesis (17), possibly through growth factor (GF) release by accessory cells. In our studies, VEGF addition in 90 to 95% purified HPC culture (18) mildly stimulated multipotent CFU (CFU-Mix), HPP-CFCs, and LTC-ICs (19). Thereafter, we purified CD34+cells and separated CD34+KDR+ versus CD34+KDR fractions (Fig. 1A, bottom left) (16). Both subsets were functionally analyzed. Upon addition of different GF combinations (20), we observed that PB lineage-committed HPCs (BFU-E and CFU-GM) were essentially restricted to the KDR fraction (Fig. 2A, left). Similar results were obtained for BM and CB (19). HPP-CFCs were present in both KDR+ and KDR fractions. Their frequency in secondary cultures was more elevated in the former population (Fig. 2A, right). Similar results were obtained for BM and CB (19). In Dexter-type LTC (Fig. 2B), we evaluated the HPCs generated by LTC-ICs in 5-, 8- and 12-week LTC seeded with PB (Fig. 2B, left panel), BM, MPB, and CB cells (19). In CD34+ cell LTC, HPC production declined from the fifth through twelfth weeks; in KDR LTC, HPC production similarly declined at 5 to 8 weeks, but no residual HPCs were detected at 12 weeks. Conversely, in KDR+/± LTC, the HPC number was low at 5 to 8 weeks but markedly increased at 12 weeks. An equivalent pattern was observed in BM, MPB, and CB LTC by 6-, 9- and 12-week CAFC assay (Fig. 2B, middle and right). Altogether, lineage-committed HPCs were restricted to KDR cells, whereas putative HSCs (12-week LTC-ICs and CAFCs) were restricted to KDR+ cells; the intermediate primitive HPCs (HPP-CFCs, 5- to 8-week LTC-ICs, 6- to 9-week CAFCs) were present in both fractions.

Figure 2

In vitro HPC and HSC assays of CD34+KDR+ cells. (A) HPCs in PB CD34+KDR+/± (open columns) and CD34+KDR (solid columns) cells, assayed in cultures supplemented with a restricted (left) or large (middle) spectrum of HGFs (20). Primary (open columns) and secondary (solid columns) HPP-CFC colonies in PB CD34+KDR+ and CD34+KDR cells (right). Mean ± SEM from four independent experiments. (B) (left) PB CD34+, CD34+KDR+/±, and CD34+KDR cell LTC. At 5, 8, or 12 weeks, supernatant and adherent cells were assayed for HPCs. (middle and right) BM and CB CD34+KDR+ and CD34+KDR cell LTCs analyzed for CAFC-derived colonies at 6, 9, and 12 weeks. Mean ± SEM from three experiments. (C) LDA of 12-week LTC-ICs and CAFCs. (top) (left) LTC-IC frequency in PB CD34+, CD34+KDR+/±, CD34+KDR+, and CD34+KDR cells. Mean ± SEM from five VEGF+ or three VEGF separate experiments. (right) Representative LDAs for CD34+ and CD34+KDR+ cells [LTC-ICs: 0.19% and 73%, respectively; 100 wells for the lowest cell concentration (1 KDR+ cell) and decreasing well numbers for increasing cell concentrations (50, 20, and 10 with 2, 5, and 10 KDR+ cells respectively)]. (bottom) CAFC frequency in CD34+KDR+ and CD34+KDR cells from BM (left) or CB (right). Mean ± SEM from three experiments. **, P < 0.01 when compared with VEGF group. °°, P< 0.01 when compared with other groups. (D) (Top) Starvation of PB CD34+KDR+ or CD34+KDR cells in single-cell FBS liquid phase culture supplemented or not with VEGF: percentage of cells that had survived at day 21 (mean ± SEM from three experiments). (Bottom) Minibulk (2 × 103cells/ml) PB CD34+KDR+ starvation culture supplemented with VEGF: LDA of LTC-IC frequency in the ∼25% cells surviving on days 5 and 25.

The frequency of putative HSCs was evaluated by limiting dilution analysis (LDA) (20). In VEGF-treated LTCs, the 12-week LTC-IC frequency was 0.19% in PB CD34+ cells [that is, one-sixth that of the more differentiated 5- to 8-week LTC-ICs (21)]. Twelve-week LTC-ICs were enriched by more than a factor of 300 in KDR+ fraction (63%) but absent in the KDR subset (Fig. 2C, top). Similar results were obtained for BM and CB KDR+ cells by 12-week CAFC LDA (Fig. 2C, bottom). The LTC-IC and CAFC frequency in KDR+ fraction is significantly higher in VEGF+ (53 to 63%) than in VEGF (25 to 43%) LTC (Fig. 2C), suggesting a functional role for KDR in HSCs. Specifically, VEGF protected from apoptosis 20% of PB CD34+KDR+ cells seeded in single-cell culture in liquid phase serum-free GF medium (20) (Fig. 2D, top). The LTC-IC frequency in the starvation-resistant CD34+KDR+ subset was > 95% (Fig. 2D, bottom).

To study repopulating HSCs, we transplanted irradiated NOD-SCID mice with CD34+ (50,000 to 250,000 per mouse), CD34+KDR+ (5000 to 10,000 per mouse), and CD34+KDR (10,000 to 250,000 per mouse) cells from BM, PB, MPB, or CB (22). The mice were not treated with cytokines and were killed after 3 months [the assay often involves cytokine treatment and usually lasts 1.5 to 2 months (8)]. BM, spleen, and PB were analyzed for human cells by flow cytometry with human-specific mAbs. We observed consistent engraftment with KDR+ or KDR+/± cells and essentially no engraftment with the KDR fraction.

In a representative BM experiment (Fig. 3A), 250,000 CD34+ cells showed engraftment, whereas 250,000 double-sorted KDRcells did not (Fig. 3A, top left). KDR+ cells (100 to 1600) consistently engrafted the hematopoietic lineages (CD33+CD15+ and CD14+CD45+, CD71+GPA+, and CD45+ CD41+ cells in granulomonocytic, erythroid, and megakaryocytic series, respectively; Fig. 3A, bottom) and the B and T lymphoid compartments (CD19+CD20+ and CD4+CD8+ CD3+, respectively) and natural killer (NK) cells (CD16+CD56+; Fig. 3A, bottom). A dose response was observed for all engrafted cell populations, specifically CD45+ cells (Fig. 3A, top right). Although T cell precursors require specific cognate interaction for maturation, human CD3+CD4+CD8+ and CD3+CD2+ cells have been generated in NOD-SCID mice BM after injection of CD34+CD38(2, 23) or CD34lincells (24). The in vitro BM microenvironment is permissive for T cell development and may recapitulate thymic maturation (25). Human T cell precursors may thus develop in NOD-SCID mice BM. The presence of contaminant mature T cells in the transplanted CD34+KDR+ cells can be excluded, in view of the lack of human T lymphocytes in mice receiving large numbers of CD34+KDR.

Figure 3

Engraftment of BM or CB CD34+KDR+ cells in NOD-SCID mice. (Aand B) Graded numbers of BM CD34+KDR+ cells were injected; positive and negative controls received CD34+ and CD34+KDR cells, respectively. Percentage of human cells is indicated. (A) Repopulating activity of 1600 to 100 CD34+KDR+ cells in recipient mice BM. (Top) Human CD34+CD45+ cell engraftment (left) and CD45+ cell dose response (mean ± SEM, three mice per group; r = 0.99) (Right). Dose-dependent engraftment was also observed in recipient PB and spleen (for example, 800 CD34+KDR+ cells generated 38.4 ± 5.0% and 15.0 ± 1.1% CD45+ cells, respectively). (Bottom) Expression of human hematolymphopoietic markers (see text) in a representative mouse injected with 1600 CD34+KDR+ cells. (B) HSC frequency in CD34+KDR+ cells. (Top right) Human CD45+ cells in BM of mice injected with 250, 50, 10, or 5 CD34+KDR+ cells (3, 9, 6, and 6 mice per group, respectively) (mean ± SEM, r = 0.99). (Left) LDA of HSC frequency (20%). (Bottom left) Human HPCs in the four engrafted mice injected with five cells (mean ± SEM). (Right) PCR analysis of alpha satellite DNA [867–base pair (bp) band] in all scored colonies from a representative mouse in the five-cell group (lane 13, human DNA positive control; lane 14, no template control; lane 15, DNA from BM mononuclear cells of nontransplanted mouse; M.W., molecular weight marker). (C) A representative mouse receiving 6000 CB CD34+KDR+ cells: informative human hematolymphopoietic markers (see text).

After injection of 5 to 250 BM KDR+ cells, a dose-dependent multilineage engraftment was monitored (Fig. 3B). All mice were repopulated by 250 and 50 cells, whereas five of six and four of six mice injected with ten and five cells, respectively, were engrafted, on the basis of flow cytometry analysis (Fig. 3B, top left) and HPC assay validated by PCR of human alpha satellite DNA in the scored colonies (Fig. 3B, bottom). LDA indicated 20% of repopulating HSCs in CD34+KDR+ cells (Fig. 3B, top right). This value is similar to the LTC-IC and CAFC frequency in VEGF BM LTC (see above), indicating that repopulating HSCs and 12-week LTC-ICs and CAFCs are closely related. Because CAFC and LCT-IC frequency rose from 25 to 42% in VEGF to 53 to 63% in VEGF+ LTC, the repopulating HSC frequency may similarly rise in VEGF-treated mice.

CB cells (200 to 10,000 CD34+KDR+ or 10,000 to 200,000 CD34+KDR CB cells) were xenotransplanted into NOD-SCID mice in five independent experiments. Human cells were virtually absent in mice transplanted with double-sorted KDR cells. In contrast, KDR+ cells consistently generated human CD45+ cells in BM, PB, and spleen in a dose-dependent pattern [for example, mice receiving 1000 to 10,000 cells showed 27.2 ± 7.1% (mean ± SEM) human CD45+ BM cells, whereas animals treated with 200 to 800 cells showed 3.75 ± 1.5% CD45+ BM cells]. In a representative experiment, mice transplanted with 6000 CD34+KDR+ cells (Fig. 3C) showed in BM abundant human CD34+ progenitors; precursors of the erythroid, granulomonocytic, and megakaryocytic lineages; and B and NK cells. The modest CD3 expression may reflect the low T cell generation potential of CB HSCs.

Repopulating HSCs were also assayed in fetal sheep (26–28). In a representative experiment, CD34+ cells were purified from two BM samples. The CD34+ population and the CD34+KDR+/± versus CD34+KDR fractions were then injected in the fetuses of eight pregnant sheep. Primary recipients receiving KDR+/±, KDR, or unseparated CD34+ cells (four, three, and two fetuses per group, respectively) were killed on day 60 after transplant. Human CD34+ cells from primary fetuses treated with KDR+/± cells were transplanted into secondary fetuses. Other fetuses injected with KDR+/± or KDRcells were born.

In primary recipients, transplantation of CD34+ cells (1.2 × 105 per fetus) consistently induced engraftment. BM analysis indicated the presence of differentiated (0.30% CD45+ cells, mean value) and undifferentiated (0.17% CD34+ cells) human hematopoietic precursors; in clonogenic assay, 6.8% of CFU-Mix/BFU-E and 5.2% of CFU-GM of all colonies were of human origin. A small number of KDR+/±cells (3 × 103 per fetus) consistently engrafted with impressive multilineage expression for the differentiated compartments (1.78% CD45+, 0.16% GPA+, and 0.34% CD3+ cells) and the undifferentiated one (0.32% CD34+ cells; the human HPC frequency was elevated, that is, 9.3% CFU-Mix and BFU-E and 16.2% CFU-GM). Eighty times as many KDR cells (2.4 × 105 per fetus) did not engraft, as indicated by consistent absence of CD34+ and CD3+ cells; only a few differentiated hematopoietic precursors were detected (0.7% CD45+ cells), together with a few late CFU-GM (2.4%) giving rise to small colonies. We estimated that a total of >108 CD34+ and CD3+ human cells were generated per fetus by KDR+ cells, whereas no CD34+ and CD3+ cells were generated by KDR cells (Fig. 4, top and middle).

Figure 4

Engraftment of BM CD34+KDR+ cells in fetal sheep. Total number of human CD34+, CD45+, GPA+, and CD3+ cells generated in primary fetuses injected with CD34+KDR or CD34+KDR+/± cells (top and middle, mean ± SEM). Human CD45+ cells and HPCs in BM of secondary fetuses (bottom, mean ± SEM), each injected with 4 × 105 human CD34+ cells from BM of primary fetuses.

Four secondary recipients received human BM CD34+ cells, derived from primary fetuses treated with KDR+/± cells. After 2 months the secondary recipients were killed and showed multilineage engraftment (Fig. 4, bottom).

At 3 weeks after birth, both sheep transplanted with KDR+ cells in fetal life showed persistent multilineage engraftment at BM level. One sheep featured extremely abundant human CD45+ cells (23.9%). Conversely, the sheep transplanted with KDR cells showed no engraftment.

These fetal sheep results, confirmed in other experiments, indicate that the KDR+ fraction is enriched for HSCs giving rise to multilineage engraftment in primary, secondary, and born recipients. Positive results in secondary recipients successfully compare with those observed by follow-up of primary transplanted fetuses for long periods after birth (28). Conversely, the KDR fraction does not engraft and contains only HPCs giving rise in primary recipients to differentiated precursors and a few late CFU-GM.

The in vivo assay results indicate HSC restriction to the KDR+ subfraction of CD34+ cells, thus in line with the in vitro observations. Previous studies in NOD-SCID mice and fetal sheep showed that HSCs are enriched in diverse CD34+cell subfractions [CD38 (2, 23,27), kitlow, Thy-1+, or Rhodamine (Rho)dim (28)], but engraftment was also observed at lower level for the complementing subfractions [that is, CD38+ (2, 23, 27), kit, Thy-1, or Rhobright(28)].

In conclusion, HSCs are in the CD34+ KDR+cell fraction, whereas lineage-committed HPCs are restricted to the CD34+KDR subset. The 20% repopulating HSC frequency in CD34+KDR+ BM fraction is >100 times as elevated as that in CD34+38 BM or CB cells (2), as confirmed in our laboratory (19). This difference is readily predicted by comparing HSC activity and phenotype. Because CD34+KDR cells do not comprise HSCs but represent 98 to 99% of CD34+CD38 cells (Fig. 1B), it follows that the HSC frequency in CD34+CD38 cells is far lower (that is, <2%) than that in CD34+KDR+cells. The VEGF effects on CD34+KDR+ cells suggest that KDR plays a functional role in HSCs. Preliminary observations indicate that transfer of Flk1 gene into PB CD34+KDR cells partially rescues the HSC phenotype (19).

Large numbers (≥105) of BM or CB CD34lin cells engraft fetal sheep and NOD-SCID mice with multilineage hematopoietic expression (3,24). We observed in BM and CB that ≤1% CD34lin cells are KDR+(29), the total number of CD34linKDR+ cells was ≤50% that of CD34+KDR+ cells, and a discrete number of CD34linKDR+ cells engrafted NOD-SCID mice with multilineage expression, but the relative stem cell activity of the CD34+KDR+fraction is more elevated. The stem cell activity in the CD34+KDR+ fraction was predominant, in line with studies indicating that human HSCs with long-term engraftment capacity are CD34+(30).

Murine studies suggested that embryonic Flk1+ cells have hemoangiogenic potential but did not identify prenatal repopulating HSCs (10, 12). Our observations in humans identify postnatal CD34+KDR+ and CD34KDR+ repopulating HSCs. The CD34KDR+ and CD34+KDR+ phenotypes might define the postnatal and prenatal HSC-hemoangioblast.

Purification of CD34+ HPCs has facilitated studies on early hematopoiesis (1, 5). Isolation of KDR+ HSCs offers an opportunity to elucidate the cellular and molecular phenotype and functional properties of HSCs and HSC subsets. These issues are of pivotal importance for a large array of biotechnological and clinical challenges, such as HSC transplantation, in vitro blood cell generation for transfusion, and HSC gene therapy.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: cesare.peschle{at}


View Abstract

Navigate This Article