CD34 Hematopoietic Stem Cells: Current Concepts and Controversies



Recent data have suggested that human CD34 hematopoietic stem cells (HSCs) exist, challenging the concept that HSCs necessarily and exclusively express the CD34 antigen. In mice, quiescent HSCs have been shown to be mostly CD34, but as a consequence of 5-fluorouracil treatment or cytokine stimulation, differentiate into CD34+ cells. Of particular interest is a novel, specific marker to identify HSCs, namely the Hoechst dye efflux property, with which a distinct side population (SP) is identified. These SP cells are mostly CD34, highly enriched for long-term repopulating cells, and durably engraft in sublethally irradiated non-obese diabetic/severe combined immunodeficient mice. Using a semiquantitative reverse transcription-polymerase chain reaction, one of the ATP-binding cassette (ABC) transporters, the breast cancer resistance protein (Bcrp) or ABC transporter G2 (ABCG2), was found to be highly expressed in SP cells as well as other primitive HSCs and to sharply drop with hematopoietic differentiation. Enforced expression of the ABCG2 cDNA resulted in a robust SP phenotype and a reduction in hematopoietic maturation. These data suggest that the Bcrp/ABCG2 gene contributes importantly to the generation of the SP phenotype, which allows for the selection of immature, pluripotent HSCs. The isolation of Bcrp/ABCG2+ cells appears to be an attractive tool to analyze and characterize HSCs, and may eventually allow for the purification of these cells for clinical purposes. In this review, current concepts on murine and human CD34 HSCs and their relationship with CD34+ HSCs are discussed.


Hematopoietic stem cells (HSCs) have the potential for self-renewal and differentiation into all lineages of blood cells. Transplantation studies with injection of HSCs into myeloablated recipients allow for the analysis of the presence of donor cells and their capacity to proliferate and repopulate, thereby indicating HSC activity [1]. In these animal models, murine HSCs have been demonstrated to be negative for lineage markers (Lin), including myeloid and B and T cell lineages, and positive for c-kit, Sca-1, and CD34 [15].

Human and Murine HSCs Express CD34

The cell-surface marker, CD34, was first identified as a hematopoietic cell-surface antigen using the early human myeloblastic cell line KG-1a [6, 7], which highly expresses CD34 and displays a strong potential for myeloid colony-forming cells [8]. Since autologous bone marrow (BM) CD34+ cells have been proven to engraft in baboons [9], human CD34+ cells have been used both for autologous and allogeneic transplantations, resulting in a rapid reconstitution of all blood lineages and a lower incidence of graft-versus-host disease compared with nonselected BM or peripheral blood (PB) cells [1012]. Moreover, human CD34+ and LinCD34+ cells have been found to engraft in fetal sheep and non-obese diabetic/severe combined immunodeficient mice [1315], which led to the concept that human HSCs are positive for the CD34 antigen.

Similarly, murine CD34+ cells have been shown to contain both functional progenitors and HSCs, indicating that CD34 is also an HSC marker in mice [16, 17]. This has also been demonstrated in embryoid bodies derived from CD34-null embryonic stem cells, where erythroid and myeloid differentiation are significantly delayed and colony-forming progenitors in BM and spleen are notably fewer [18]. Nevertheless, these CD34-null mice developed normally and showed a regular hematopoietic profile. These data suggest that, although CD34 is not essential for murine survival, it plays an important role in the differentiation of HSCs during both embryonic and adult hematopoiesis [18].

Identification of Murine CD34HSCs

In 1996, Osawa et al. first reported on CD34 HSCs. In that study, they isolated pure CD34low/-c-kit+Sca-1+Lin cells from murine BM. Transplantation of single CD34 cells resulted in multilineage repopulation, contributing to 85% of PB cells in host mice, whereas CD34+c-kit+Sca-1+Lin cells revealed early, but unsustained, multilineage hematopoietic reconstitution. The purity of this CD34 population was also examined via reverse transcription-polymerase chain reaction analysis, with lack of CD34 mRNA expression, indicating that the repopulating activity of CD34 cells was unlikely to be due to the contamination of CD34+ HSCs [19]. That report provided the first experimental evidence of the existence of CD34 HSCs.

A novel procedure to isolate HSC recently has been described by Goodell et al. [20]. Murine BM cells were stained with a fluorescent DNA-binding dye, Hoechst 33342, and analyzed with two emission wavelengths. A distinct side population (SP) of cells was identified, which had the capacity to rapidly expel the dye and, thus, were Hoechstlow. Interestingly, murine SP cells can repopulate sublethally irradiated recipients 1,000-fold better than unfractionated cells, leading to myeloid and lymphoid reconstitution in the PB of recipients [20]. Most importantly, those authors observed that murine SP cells were CD34low or CD34 [21], further supporting the concept that CD34 HSCs may indeed exist.

Morel et al. recently described a small Thy-1lowLinSca-1+ (TLS) CD34 population from murine BM that possessed a high long-term repopulating efficiency when compared with CD34+ TLS cells [22]. Donnelly et al. reported that murine BM LinCD34 and LinCD34+ cells could generate long-term engraftment, but that CD34+ cells were 100-fold more abundant than CD34 HSCs [23]. The different results of these two groups seem related to the distinct cell populations (TLS versus LinCD34) and different transplantation procedures used. Nevertheless, these observations challenge the former concept that HSCs necessarily express the CD34 antigen.

Relationship Between CD34+and CD34HSCs in mice

Sato et al. observed that the majority of long-term repopulating BM cells in adult mice were CD34 [24]. After in vivo 5-fluorouracil (5-FU) treatment, high-level engraftment was detected from both CD34+ and CD34 HSCs 5 months posttransplantation, suggesting a phenotypic change due to an altered activation state post-5-FU exposure. To further test this hypothesis, CD34 HSCs were cultured in the presence of early-acting cytokines, such as interleukin-11 (IL-11) and stem cell factor (SCF), resulting in a bulk cell population of CD34+ cells. Injection into lethally irradiated mice generated long-term and multilineage engraftment. This phenotypic change may indicate that CD34 HSCs can develop into CD34+ cells upon culture and still preserve their HSC capacity [24]. Similarly, they showed that HSCs in G-CSF-mobilized adult mice mostly expressed CD34 [25], suggesting that the CD34 expression on murine HSCs reflects their activation state caused by 5-FU or cytokine treatment. Therefore, it is reasonable to assume that the phenotype of HSCs is restored to CD34 when cell activation is terminated, and murine hematopoiesis returns to steady-state conditions. Indeed, CD34+ HSCs from G-CSF-mobilized or 5-FU-treated mice reverted into CD34 cells when restored to steady state [24, 25]. Moreover, the CD34 analysis on murine HSCs at different developmental stages has demonstrated that the antigen expression on murine HSCs declines with age: long-term repopulating HSCs from BM, liver, and spleen in fetal and newborn mice are CD34+, which is also true for HSCs in young mice, whereas CD34 HSCs emerge in 7-week-old mice and increase thereafter [26, 27].

Characterization of Human CD34HSCs

In contrast with the rapid progress in the characterization of murine CD34 HSCs, the analysis of human HSCs has proceeded more slowly. In 1997, Goodell et al. reported on human and monkey SP cells that showed a rapid Hoechst dye efflux activity. These SP cells were lineage-marker negative, revealed a highly enriched long-term culture-initiating cell (LTC-IC) capacity, and developed into CD34+ cells after stroma-supported cell culture [21], thereby suggesting that these HSCs were CD34.

In a NOD/SCID mouse model, Bhatia et al. first described a novel HSC population from human cord blood (CB) samples with a LinCD34CD38 phenotype [28]. These cells had low clonogenic activity in vitro, but remarkably, regenerated multilineage hematopoiesis in NOD/SCID mice. Of interest was that these CD34 HSCs could differentiate into CD34+ cells, resulting in a greater repopulating activity in cytokine-supported short-term cultures (2-4 days), whereas CD34+ HSCs lost their stem cell potential in culture. These data show an increase of CD34 HSCs in short-term suspension culture in conditions that do not maintain HSCs from CD34+ populations, providing in vivo evidence that human CD34 HSCs are biologically distinct from CD34+ HSCs. Using the in utero human/sheep competitive engraftment model, Zanjani et al. observed engraftment of human hematopoietic cells with LinCD34 BM cells [29]. Human CD34+ cells have also been detected in animals transplanted with human CD34 cells, again suggesting that CD34 HSCs can become activated and/or differentiate to CD34+ cells in vivo [29]. In the same animal model, serially repopulating human CD34 HSCs were identified from other specimens, such as PB and BM in normal G-CSF-mobilized donors [30].

Since ex vivo culture plays a key role in the characterization of human HSCs [3137], highly purified human LinCD34CD38 cells from normal BM and G-CSF-mobilized PB have been studied. These do not primarily grow in methylcellulose, but rapidly proliferate and differentiate into erythrocytes, granulocytes, and megakaryocytes in serum-free culture, turn into CD34+ cells after 10 days of culture, and significantly increase their colony-forming potential [38]. These results imply that primitive human CD34 HSCs can be induced to rapidly proliferate and differentiate upon cytokine stimulation in suspension culture. The observation that human LinCD34 cells contain low colony-forming activity before culture has recently been supported by Nakamura et al. [39]. Human LinCD34 CB cells were cultured with the murine BM stroma cell line HESS-5 and thrombopoetin, Flt3-ligand (Flt3), SCF, G-CSF, IL-3, and IL-6. Similar to the results from Eaves's group [38], LinCD34 CB cells contained low colony-forming potential before culture, but gained the ability to form colonies during 7 days of culture, coinciding with the formation of CD34+ cells. Interestingly, these cells were able to engraft NOD/SCID mice and to generate donor-derived CD34+ cells, whereas low repopulating activity was observed with noncultured CD34 cells. These data suggest that murine BM stroma supports and preserves human CD34 HSCs, that colony-forming and NOD/SCID-repopulating properties can be increased upon ex vivo culture, and that CD34 HSCs may be more primitive than CD34+ HSCs [39].

Using the human/sheep xenograft model, Zanjani et al. recently reported that, with both human CD34+ and CD34 cells isolated from T-cell-depleted BM cells by fluorescence-activated cell sorting, long-term engraftment was obtained [29]. However, CD34 cells showed significantly higher levels of engraftment than CD34+ cells 15 months posttransplantation [40]. Of note was that human CD34 HSCs have been detected in sheep transplanted with CD34+ cells and vice versa [40]. These data indicate that human HSCs display reversible expression of the CD34 antigen in vivo, although the mechanisms by which CD34 expression is regulated need to be further explored. Also, although the in vivo repopulation potential of CD34 cells seems unlikely due to contamination with CD34+ cells [28, 29], the purity of CD34 cells needs to be further addressed. Most ideally, a specific marker for the selection of CD34 HSCs needs to be identified before definite conclusions on human CD34 HSCs are made.

Novel HSC Markers

From currently available data, both in mice and humans, the CD34 antigen does not seem to be expressed on all HSCs. Therefore, alternative HSC markers are currently being evaluated, such as the cell surface marker, AC133. Gallacher et al. found that both human CD34+ and CD34 HSCs expressed AC133, and that AC133+ cells, both before and after culture, engrafted NOD/SCID mice, indicating that AC133 may also be an HSC marker [41]. In contrast, the cytokine receptor, Flt3, recently has been identified as a marker for murine stem cell differentiation, where low numbers of LinFlt3Sca-1+c-kit+ cells in murine BM supported long-term multilineage reconstitution, whereas transplantation of Flt3+ cells resulted in only short-term engraftment, suggesting that upregulation of Flt3 is associated with a loss of self-renewal capacity [42, 43].

In addition, and beyond these cell surface markers, the Hoechst dye efflux property appears to be a promising marker to isolate HSCs from various sources. Murine SP cells from skeletal muscle can reconstitute the BM of irradiated mice. Conversely, murine BM SP cells can repair muscle cells in mdx mice, an animal model of Duchenne's muscular dystrophy [44, 45]. These data illustrate the potential stem cell plasticity of SP cells, both for muscle and blood cell reconstitution. Others, as well as ourselves, have also detected SP cells from human BM, CB, and apheresis products and have described that these are mainly CD34 and significantly enriched for LTC-ICs [21, 4649]. Furthermore, SP cell numbers significantly increase after short-term suspension culture when replated into methylcellulose [46, 47], supporting the previous observation that suspension culture leads to the maturation of less primitive progenitors and, thereby, increases the colony-forming unit potential [38, 39]. Hoechst efflux studies with CB cells also have revealed a distinct LinCD34CD7+ population of cells, which differentiate into natural killer cells, again providing evidence for CD34 progenitors [48]. Recently, Uchida et al. presented data on human fetal liver SP cells, where, despite being mostly CD34, the HSC activity was confined to LinCD34+CD38 SP cells. In this fraction, a 10-fold enriched HSC activity was observed compared with LinCD34+CD38 cells, suggesting, at least in that particular cell source, that selection for SP cells provides a more powerful tool for identification of HSC [49].

The SP Phenotype Serves as a Potential HSC Marker

Others, as well as ourselves, have shown that SP cell numbers decrease significantly when costained with verapamil [20, 21, 4648]. This may be due to the multidrug-resistance-1 (MDR1) gene product, P-glycoprotein (P-gp), as the molecular basis of the Hoechst efflux phenomenon. Transduction of murine BM cells with the MDR1 gene results in a substantial increase in SP cell numbers, both in vitro and in vivo [50]. Nevertheless, SP cells are found in P-gp-knockout mice [51], so that other ATP-binding cassette (ABC) transporter members are likely to be responsible for the energy-dependent transport of a wide spectrum of substrates across membranes and contribute to the Hoechst efflux. High levels of ABC transporter breast cancer protein (Bcrp) mRNA expression are detected in HSCs from murine BM, which decrease with cell differentiation. Enforced expression of Bcrp on normal BM cells results in expansion of transplantable SP cells and reduction in lineage differentiation [51]. These data suggest that the ABC transporter, Bcrp, is a dominant Hoechst pump and may serve as a potential marker for HSCs. The human Bcrp gene (also termed ABCG2) has similar properties, as reported by Scharenberg et al. [52]. Since Hoechst 33342 displays cell toxicity [49], isolation of primitive HSCs via nontoxic monoclonal antibodies against Bcrp may be valuable and is currently being pursued.

In conclusion, the notion that HSCs are necessarily CD34+ may have to be revised: from our current knowledge, quiescent murine HSCs appear to be mostly CD34, but acquire the CD34+ phenotype upon 5-FU treatment or cytokine stimulation. The fact that a definite number of human HSCs are CD34 should be taken into consideration when CD34 selections are performed for clinical uses, including transplantation, purging, and gene therapy procedures. Hoechst sorting is currently used to enrich for HSCs. Further isolation and characterization of human CD34 HSCs, including use of a nontoxic monoclonal antibody against Bcrp, are currently awaited and will give us further insight into the biology of HSC development.


We appreciate Dr. Roland Mertelsmann's continuous support and thank Dr. Jan Burger for his valuable comments and critical reading of the manuscript.

This work was supported by grant #C6 from the Center for Clinical Research (ZKF), University of Freiburg, Germany.