Potential of CD34 in the Regulation of Symmetrical and Asymmetrical Divisions by Hematopoietic Progenitor Cells



The control of symmetric and asymmetric division in the hematopoietic stem/progenitor cell population is critically important for the regulation of blood cell production. Asymmetric divisions depend on cell polarization, which may be conferred by location and/or interaction with neighboring cells. In this study, we sought evidence for polarization in CD34+ cells, which interact by binding to one another. In these cells, surface molecules became redistributed by mechanisms that included transport by lipid rafts, and the interacting cells were able to communicate via gap junctions. These changes were accompanied by modulation of cell cycle regulating proteins (p16Ink4a, p27kip1, cyclins D, and the retinoblastoma pathway proteins) and a reduction in progenitor cell proliferation in vitro. These results are consistent with an increase in asymmetric cell division kinetics. Accordingly, we found that interaction between CD34+ cells influenced the plane of cell division in a way that suggests unequal sharing of Notch-1 between daughter cell progeny. We conclude that interaction between CD34+ cells may coordinate cell function and participate in the control of hematopoietic stem/progenitor cell division kinetics.

Disclosure of potential conflicts of interest is found at the end of this article.


Throughout life, human hematopoietic cell production is sustained by the proliferative activity of the stem cell pool. During development, stem cells originate in the aorta-gonad-mesonephros region and then migrate and multiply in order to populate the fetal liver and, ultimately, the adult bone marrow. In the normal steady-state, stem cell numbers are maintained at a constant level but may be required to increase, for example, after treatment with myelotoxic chemotherapy. It is thought that stem cell numbers are regulated by the balance between symmetric and asymmetric cell divisions [1]. Asymmetric divisions are associated with homeostasis, whereas symmetrical divisions are required for stem cell recovery from depletion. This is because asymmetrical divisions do not allow for changes in stem cell numbers, each division producing one new stem cell and one cell destined for differentiation. In contrast, symmetrical divisions permit expansion in stem cell number by altering the proportions of stem cells that produce two new stem cells versus those that produce two differentiating cells. These considerations imply either that the probabilities of symmetric divisions producing stem cells and differentiated cells are flexible or that asymmetric and symmetric divisions coexist in variable proportions in hematopoiesis.

Asymmetric divisions are a property of polarized cells. Polarity may be conferred by location and interaction with surrounding cells, as has been demonstrated by the spatial relationship between stem cells and their niches in Drosophila and in mammalian hematopoiesis. Another mechanism that may confer asymmetric cell division is the unequal partitioning of cell fate determinants as a result of orientation of the mitotic spindle and the plane of cell division, which ensures the unequal distribution of cellular constituents between the daughter cells [2]. For example, in the Drosophila germline, it has been shown that the mitotic spindles in the germline stem cells are oriented to ensure that stem cells divide asymmetrically [3] and, in neuronal cells, horizontal and vertical divisions give rise to asymmetric and symmetric divisions, respectively [2]. Also, asymmetric neuronal cell divisions result in Notch-1, a cell fate determinant that has been associated with hematopoiesis, being inherited asymmetrically by only one of the daughter cells [2]. Notch-1 signals through the cyclin-dependent kinase inhibitor p21waf1/cip1 and the retinoblastoma protein family in human hepatocellular carcinoma cells [4]. In hematopoiesis, the cyclin-dependent kinase inhibitors p16Ink4a, p21waf1/cip1, and p27kip1 all suppress the capacity of clonogenic hematopoietic progenitor cells to proliferate [5].

Hematopoietic stem and progenitor cells are generally perceived as round, nonpolarized cells. However, CD34+ cells aggregate together when exposed to the class II anti-CD34 antibody, QBEND10, but not antibodies belonging to class III [6]. Also, CD34+ cells have been shown to bind together in vivo without antibody exposure [7]. Aggregated cells become polarized, and it has been suggested that this interaction may have a negative effect on hematopoietic cell proliferation kinetics [7]. Cell polarization probably involves the redistribution of cell surface molecules and, in this regard, “lipid rafts” have been implicated in the polarization of CD34+ cells and T cells [8, [9]–10]. Lipid rafts are distinct plasma membrane domains that are rich in cholesterol and glycosphingolipids [11]. By concentrating certain cell surface receptors and signaling molecules, they are proposed to act as platforms for cell adhesion, signal transduction, and membrane trafficking [12]. In addition, the close cell-cell interaction provides the opportunity for direct intercellular communication, which has been suggested to coordinate cellular activity [13] and plays a crucial role in regulating proliferation and differentiation processes during development [14]. Interestingly, connexin-43, a component of gap junctions that is expressed in hematopoietic cells [15], has been found in lipid rafts [16]. Gap junctions are specialized structures that form direct channels between the cytoplasms of adjacent cells. These hydrophilic pores allow ions and molecules of up to 1 kDa, including second messengers such as cyclic-AMP and Ca2+, to pass through, whereas proteins and nucleic acids cannot. By coupling the metabolic activities of adjacent cells, gap junctions facilitate coordinated responses by groups of cells to environmental stimuli.

In this study, we sought evidence for the polarization of cell division in aggregated CD34+ cells and CD34+ cells found in doublets within developing myeloid colonies in vitro, together with the effect of this interaction on hematopoietic progenitor cell kinetics. Our results suggest that aggregation between CD34+ cells may offer a potential setting for the formation of a complex multimolecular structure facilitating cell polarization and communication that may relate to the regulation of cell proliferation kinetics. We show that a multimolecular complex consisting of cell adhesion and signaling molecules forms at the contact point between aggregated CD34+ cells and that the cells are highly polarized with regard to proteins that are inside and outside the complex. Also, we found that the association between CD34+ cells resulted in orientation of the centrosomes, indicating mitotic spindle orientation, and evidence that intercellular communication occurred via gap junctions. Therefore, aggregated CD34+ cells may be considered as a functional coordinated unit that may facilitate adaptive responses to environmental changes.

Materials and Methods

Cells for Study

Peripheral blood progenitor cells were obtained by leukapheresis following mobilization with granulocyte colony-stimulating factor from donors for transplantation. The samples used were in excess of clinical requirements and were provided with informed consent and Research Ethics Committee approval.

The samples were diluted in Hanks' balanced salt solution (HBSS; Gibco, Grand Island, NY, http://www.invitrogen.com) layered over Lymphoprep (Axis-Shield, Oslo, Norway, http://www.axis-shield.com) and centrifuged for 30 minutes at 1,800 rpm. Mononuclear cells were harvested from the interface and washed twice in HBSS. CD34+ cells were isolated using MiniMACS technology according to the manufacturer's instructions (Miltenyi Biotec, Bergisch Gladbach, Germany, http://www.miltenyibiotec.com).

Cell Aggregation

CD34+ cells were suspended at a concentration of 1 × 106/ml in alpha medium (Gibco) supplemented with 15% fetal calf serum (FCS). Aliquots of 100 μl were transferred to the wells of 96-well microtitre plates and then treated with 2 μl of the anti-CD34 monoclonal antibody QBEND10 (a kind gift from Dr. L. Healy) or isotype control antibody (Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com) for 3 hours at 37°C in 5% CO2 in air. Aliquots of 30 μl were then air-dried onto Teflon-coated multiwell slides (BDH, Lutterworth, U.K.), fixed in ethanol, stained with Romanowsky dyes, and scored for aggregation by light microscopy. At least 500 cells per well were scored to give the percentage of cells in aggregates versus the percentage of single cells. The score for cells exposed to the isotype control antibody was subtracted from this value to give the specific aggregation index. Myeloid colonies were grown in standard culture conditions, harvested, and examined for the presence of CD34+ cell doublets [7].


After aggregation, cells were gently resuspended, and aliquots of 30 μl were then air-dried onto Teflon-coated multiwell slides. The cells were fixed in 3% paraformaldehyde for 10 minutes or acetone for 1 minute, rehydrated for 10 minutes in phosphate-buffered saline (PBS), and then blocked for 1 hour with 3% bovine serum albumin (PAA Laboratories, Linz, Austria, http://www.paa.at). The cells were covered with 100 μg/ml primary antibody or isotype control for 1 hour at room temperature in the dark. After four 5-minute washes in PBS, cells were incubated with 0.5% or 1% fluorochrome-conjugated secondary antibody. For dual staining experiments, the above steps were repeated for the second antibody. All slides were mounted in VECTASHIELD containing 4,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com). Immunostained cells were examined using a Zeiss Axiovert 100 fluorescent microscope with SmartCapture imaging software or a Zeiss Meta 512 confocal microscope (Carl Zeiss, Jena, Germany, http://www.zeiss.com) using the argon laser to view fluorescein isothiocyanate (FITC).

Western Blotting

Cell lysates and Kaleidoscopic prestained markers (Bio-Rad, Hercules, CA, http://www.bio-rad.com) were electrophoresed on polyacrylamide gels at 100 V and transferred to 0.45-μm polyvinylidene difluoride membrane (Millipore, Billerica, MA, http://www.millipore.com) by semidry transfer (Bio-Rad). Membranes were blocked for 1 hour at room temperature in 5% milk (Marvel, Dublin, Ireland) made up with TBS containing 0.5% Tween (BDH). Primary antibody (Santa Cruz) was applied to the membranes at 1/500 milk/Tris-buffered saline (TBS)-Tween and left at 4°C overnight. After four 5-minute washes in TBS-Tween, the relevant secondary horseradish peroxidase conjugate was diluted 1/1,000 in milk/TBS-Tween and applied to the membrane for 1 hour at room temperature with shaking. The membranes were then washed and visualized using an ECL chemiluminescent visualization kit (Amersham Biosciences, Piscataway, NJ, http://www.amersham.com).

Lipid Rafts

Lipid rafts were isolated from aggregated and nonaggregated CD34+ cells by sucrose density gradient ultracentrifugation. Cells were washed once, resuspended in 0.4 ml of ice-cold MBS-BL+ (50 mM 2-(N-morpholino)ethanesulfonic acid, 150 mM NaCl, 1 mM phenymethyl sulfonylfluoride, 1 mM sodium vanadate, 1 μg/ml protease inhibitor cocktail, 1% Triton X-100 [all from Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com]), and kept on ice for 30 minutes. Cells were then lysed after the addition of 0.4 ml of 80% sucrose and dispensed into a 5-ml ultracentrifuge tube. The lysate was gently overlaid with 2.7 ml of 30% sucrose followed by 1.5 ml of 5% sucrose. The tubes were placed into an SW55Ti rotor and centrifuged for 24 hours at 4°C. Fractions of 0.5 ml were than removed from the top of the gradient. The pellet was resuspended in 0.5 ml of 1× sample buffer, diluted to 1× with an equal volume of PBS, and boiled for 3 minutes. The fractions were analyzed by Western blotting.

To stain lipid rafts, FITC-labeled cholera toxin-B (100 μl of 10 μg/ml; Sigma) was added to slides of aggregated and nonaggregated cells followed by incubation on ice for 30 minutes in the dark. After one wash, slides were mounted in VECTASHIELD.

To deplete cells of cholesterol, they were incubated with 10 mM methyl-β-cyclodextrin (MBCD; Sigma) in RPMI for 15 minutes and washed three times in PBS. Treated cells were also incubated with cholesterol (Sigma) for 15 minutes at 37°C to reconstitute cellular cholesterol and reverse the effects of MBCD treatment.

Gap Junctions

CD34+ cells were split into two groups. One was stained with 500 nM calcein (Sigma) for 30 minutes at 37°C and the other with 4 × 10−6 M PKH26 (Sigma). The cells were washed, mixed 50:50, and aggregated by exposure to QBEND10. Then, cells were mounted on slides in VECTASHIELD and analyzed using a fluorescent microscope to document dye transfer between cells. To inhibit gap junction communication, 0.2% carbenoxalone (Sigma) was added to the cells and incubated for 18 hours.

Small Interfering RNA-Mediated Knockdown of Lymphocyte Function-Associated Antigen-1

Cells were seeded at a concentration of 2 × 105/ml in 400 μl of X-vivo medium (BioWhittaker, Walkersville, MD, http://www.cambrex.com) in 12-well plates. For each well to be transfected, 7 μl of Opti-MEM (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) was mixed with 3 μl of Oligofectamine (Invitrogen) and left at room temperature for 10 minutes. In a separate stock tube, lymphocyte function-associated antigen-1 (LFA-1) small interfering RNA (siRNA) (Santa Cruz) was diluted by adding 3 μl of 10 μM stock solution to 87 μl of Opti-MEM for each well and mixed with the diluted Oligofectamine. After 20 minutes at room temperature, 100 μl of the mixture was added to each well and incubated overnight at 37°C in 5% CO2 in air. The next day, an additional 100 μl of siRNA/Oligofectamine was added to each well. Four hours later, 60 μl of FCS was added to each well, and the plate was incubated at 37°C in 5% CO2 in air. At the relevant time point, 30-ml aliquots of cells were removed and examined by immunofluorescence and Western blotting and tested in the aggregation assay.

Colony Replating Assay

Colony-forming units-granulocyte-macrophage were cultured in standard assays for 7 days, then 120 individual colonies per group were plated into wells in 96-well microtitre plates. After an additional 7 days, secondary colonies were scored and the proliferation index was calculated [17].

Statistical Analysis

Results were analyzed using the Mann-Whitney U test or chi-squared test for statistical significance.


Homotypic Aggregation Between CD34+ Cells Involves a Multimolecular Adhesion Complex and Localization of Notch-1

CD34 is diffusely distributed in nonaggregated cells but adopts a ring-like configuration at the intercellular binding site when the cells are induced to aggregate by exposure to QBEND10 monoclonal antibody (Fig. 1A–1C). In addition to CD34, CD44, LFA-1, and F actin, but not intercellular adhesion molecule (ICAM)-1, were found to be concentrated at the contact point between CD34+ cells induced to aggregate by treatment with QBEND10 (Fig. 1D–1K). Although CD44 and LFA-1 were concentrated with CD34, antibodies to CD44 and LFA-1 did not cause CD34+ cell aggregation (data not shown). In spite of the exclusion of ICAM-1, downregulation of LFA-1 using siRNA, which was confirmed by immunofluorescence and Western blotting (data not shown), reduced the ability of the cells to aggregate (Fig. 1L, 1M). CD44 also colocalizes with CD34 in antibody-treated cells (Fig. 2A–2C), whereas CD164 locates to the opposite pole and is excluded from the adhesion site (Fig. 2D, 2F). However, unlike CD34, CD44 and CD164 are polarized without exposure to QBEND10 (Fig. 2G, 2H). This result indicates that CD34+ cell polarization may be both constitutive and induced. Interestingly, Notch-1 exhibits a patchy distribution in untreated cells (Fig. 2I) but was concentrated in the same cell hemisphere as CD34 in 67% of aggregated CD34+ cells (Fig. 2J–2L).

Figure Figure 1..

Redistribution of surface proteins on QBEND10-treated CD34+ cells. (A): Diffuse distribution of CD34 on untreated cells. (B): Polarization of CD34 between treated cells. (C): Confocal microscopy of the ring configuration of CD34 on treated cells. (D): Polarization of CD34 on treated cells. (E): Polarization of CD44 on treated cells. (F): Merge of (D) and (E). (G): Polarization of CD34 on treated cells. (H): Polarization of actin on treated cells. (I): Merge of (G) and (H). (J): Diffuse distribution of LFA-1 on untreated cells. (K): Polarization of LFA-1 on treated cells. (L): siRNA knockdown of LFA-1 evaluated by Western blotting. (M): Influence of siRNA knockdown of LFA-1 on aggregation induced by QBEND10. siRNA = LFA-1-specific siRNA; MM = mismatch siRNA control. ∗ p < .05, n = 5. Abbreviations: LFA, lymphocyte function-associated antigen; siRNA, small interfering RNA.

Figure Figure 2..

Polarization and relative localizations of CD34, CD44, CD164, and Notch. (A): Polarization of CD34 on treated cells. (B): Polarization of CD44 on treated cells. (C): Merge of (A) and (B). (D): Polarization of CD34 on treated cells. (E): Polarization of CD164 on treated cells. (F): Merge of (D) and (E). (G): Polarization of CD44 on untreated cells. (H): Polarization of CD164 on untreated cells. (I): Diffuse distribution of Notch on untreated cells. (J): Polarization of CD34 on treated cells. (K): Polarization of Notch on treated cells. Notch-1 polarization was scored on treated cells; at least 100 cells were scored per experiment, n = 3. (L): Merge of (J) and (K).

Lipid Rafts Participate in Aggregation and Complex Formation Between CD34+ Cells

LFA-1 and CD44 are lipid-raft-associated proteins [18, 19], and lipid rafts have been implicated in the polarization of CD34+ cells [8]. Staining with cholera toxin revealed some localization of lipid rafts at the contact point between aggregated CD34+ cells (Fig. 3A, 3B), and treating cells with methyl-β-cyclodextrin, which extracts cellular cholesterol and disrupts lipid rafts [20, 21], reduced aggregation of CD34+ cells incubated with QBEND10 (Fig. 3C). Incubation of MBCD-treated cells with cholesterol restored aggregation to control levels (p = NS). Sucrose density gradient ultracentrifugation was used to confirm the presence of lipid rafts in aggregated and nonaggregated CD34+ cells. Following lysis with Brij 35, lyn (a lipid raft marker), CD34, LFA-1, CD44, and CD164 were found within the lipid raft fraction (Fig. 3D). Thus, lipid-raft-mediated transfer may be involved in the relocalization of molecules that are excluded from, as well as those that are included in, the intercellular complex.

Figure Figure 3..

Role of lipid rafts in the polarization of QBEND10-treated CD34+ cells. Distribution of lipid rafts in (A) untreated cells and (B) treated cells. (C): Influence of methyl-β-cyclodextrin (MBCD) on QBEND10-induced aggregation. Black columns represent cells pretreated with 10 mM MBCD; white columns are untreated cells. The time following addition of QBEND10 is on the x-axis, ∗ p < .05, n = 5. (D): Analysis of lipid raft components in aggregated (+) and nonaggregated (−) cells by Western blotting of sucrose density gradient fractions; numbers represent gradient fraction. Lipid rafts float to the interface between the 5% and 30% sucrose layers corresponding to fractions 3 and 4. Abbreviations: K, K562 control cell lysate; LFA, lymphocyte function-associated antigen; P, probability.

Evidence for Communication via Gap Junctions Between Aggregated CD34+ Cells

Since connexin-43 has been found in lipid rafts [16], we next sought evidence for gap junction-mediated communication between aggregated CD34+ cells. Figure 4A shows that connexin-43 is detectable in CD34+ cells by Western blotting. CD34+ cells were stained with the membrane dye PKH26 (red fluorescence) or calcein, a green fluorescent dye that is cleaved to a membrane-impermeable form by cellular esterases and can only pass between cells through gap junctions (Fig. 4B, 4C). PKH26- and calcein-stained cells were mixed 50:50 and treated with QBEND10 to induce aggregation. Gap junction-mediated dye transfer, detected by the appearance of orange/yellow cells in the mixture (Fig. 4D), was expressed as the percentage of the number of potential calcein recipients (i.e., the number of PKH26-stained cells that were present in the aggregates) that turned orange/yellow. Figure 4E shows that 40% of the possible recipients received calcein via gap junctions. Figure 4E also shows that carbenoxalone, an inhibitor of gap junction transport, reduced the transfer of calcein to 10% of the control levels. Carbenoxolone was not toxic to the cells, as assessed by trypan blue dye exclusion, and did not influence the ability of the cells to aggregate when they were treated with QBEND10 (data not shown).

Figure Figure 4..

Evidence for gap junction-mediated communication between aggregated CD34+ cells. (A): Detection of connexin-43 in CD34+ cells by Western blotting. (B): Calcein-stained cell. (C): PKH26-stained cell. (D): PKH26-stained recipient of calcein via gap junction-mediated transfer. (E): Effect of carbenoxalone on gap junction-mediated transfer; white column = untreated control, black column = carbenoxalone-treated cells. At least 500 cells were scored per experiment, n = 5, ∗ p < .05.

QBEND10-Induced CD34+ Cell Aggregation Modulates the Cell Cycle Machinery

The cell cycle machinery is central to the regulation of cell proliferation kinetics [22]. We examined the expression of the cyclin-dependent kinase inhibitors p16Ink4a and p27kip1 in CD34+ cells that had been exposed to QBEND10 before lysis for Western blotting and found that they were upregulated compared to untreated cells (Fig. 5A). Consistent with this result, cyclins D2 and D3 were downregulated, and the expression of phosphorylated retinoblastoma protein (pRb) and members of the pRb pocket protein family (p107 and p130) was reduced. Hypophosphorylated proteins of the pRb family sequester transcription factors, particularly members of the E2F family, so that reduced amounts of E2F1 were detected by Western blotting. We have shown that increased CD34+ cell proliferation measured by a colony replating assay [17] is associated with reduced cyclin-dependent kinase inhibitors, increased cyclin D, and activation of the retinoblastoma pathway [5, 20]. Therefore, we tested the effect of QBEND10 pretreatment on the ability of clonogenic CD34+ cells to proliferate in culture. As may have been anticipated, the treatment reduced proliferation significantly (Fig. 5B). Antibodies to class I and class III CD34 epitopes had no effect on proliferation (data not shown), indicating that the reduction was due to aggregation itself rather than engagement of CD34 alone.

Figure Figure 5..

Influence of QBEND10-mediated polarization and aggregation on cell cycle proteins and progenitor proliferation. (A): Western blotting analysis. The x-axis shows the number of hours treated with QBEND10 prior to cell lysis. (B): Progenitor proliferation index. The black bar represents untreated cells and the white bar represents cells exposed to QBEND10 for 3 hours before culture, n = 5. Abbreviation: pRb, retinoblastoma protein.

Orientation of Cell Division

A reduction in proliferation (i.e., output of secondary colony-forming cells) could be due to increased asymmetric cell divisions [1]. Accordingly, we investigated whether aggregation between CD34+ cells might influence the plane of cell division. We used pericentrin to stain the centrosomes in interphase cells and scored their positions relative to the binding site between aggregated CD34+ cells. The centrosomes were nonrandomly distributed, since in 89.6% (199/122) of the cells they were in the cell hemisphere opposite the contact point. The distributions of the centrosomes and CD164 were similar, and comparing their localizations revealed close spatial association between them in 83.1% of the cells (Fig. 6A–6C). Similarly, when we examined cell doublets harvested from 6-day-old cultured myeloid colonies, 89.2% (173/194) of single centrosomes were located in the hemisphere opposite the contact point. Moreover, a close association between CD164 and the centrosome was seen in 95.9% of the 98 CD34+ cells harvested from myeloid colonies and scored.

Figure Figure 6..

Proposed influence of polarization on the symmetry/asymmetry of cell division by CD34+ cells. (A): Polarization of CD164 in treated cells. (B): Localization of the centrosome in treated cells. (C): Merge of (A) and (B).(D): Model relating distribution of CD34 and Notch to the plane of cell division in CD34+ cells. (E): Scheme used for scoring potential planes of division, defined by the positions of CD34 and the centrosome. We examined 149 doublets in total, 16 of which contained one cell with two centrosomes, n = 3. The chi-squared test confirmed a significant difference in the frequency of each plane of division cell from the expected values, p < .01.

The position of the centrosome in nondividing cells can indicate incipient mitotic spindle orientation, since the interphase centrosome divides, and the resultant two centrosomes may occupy opposite poles of the cell [23]. We propose the model shown in Figure 6D, in which the positions of the centrosomes may determine whether cell division is symmetrical or asymmetrical. An asymmetric division will result if the interphase centrosome lies between Figure 6Da and 6Db in the “southern” hemisphere, provided the duplicated centrosome does not migrate between 6Dc and 6Dd in the “northern” hemisphere at mitosis. Accordingly, Notch-1 and proteins at the CD34-CD34 cell-binding site may be differentially inherited at mitosis. In order to test the model, we searched for cell doublets with two centrosomes in one of the partner cells. A cell with two centrosomes was found in 16 (10.7% of the 149 doublets scored). According to the scheme in Figure 6E, the centrosomes in 87.5% (14/16) were oriented along the lines indicated by 2–3 and 5–6, which are mirror images of one another. Only one vertical (6.3%, line one) and one horizontal (6.3%, line four) arrangement was observed. These values are significantly different from those expected. Similarly, in developing myeloid colonies in vitro, no strictly vertical or horizontal arrangements were observed, although the number of cell doublets containing a cell with two centrosomes was small (3.7%). These results suggest that interaction between CD34+ cells may predispose them to asymmetric cell division, which is consistent with the reduced proliferation shown in Figure 5B, since asymmetric kinetics are consistent with a reduced growth rate [1]


Asymmetric and symmetric divisions are required for the steady-state maintenance and expansion of hematopoiesis, respectively, and the balance between the two must be tightly controlled. Although the mechanisms of this control are not well understood, certain common themes are emerging from studies in Drosophila and other systems. These themes include roles for cell interaction, communication, and polarization [24, [25]–26]. Since we had found that CD34-mediated aggregation between hematopoietic stem/progenitor cells reduced progenitor cell proliferation in vitro [7], we hypothesized that this interaction might be involved in controlling the symmetry of divisions. We attributed a role for CD34 in mediating aggregation as CD34-CD34 interactions were found within freshly isolated hematons and developing myeloid colonies in vitro [7]. The alternative theory that anti-CD34 antibodies block an antiadhesive effect of CD34, permitting engagement of other cell adhesion molecules and their ligands, therefore seems less likely [27].

In aggregated CD34+ cells, the CD34 adopted a ring-like configuration in a manner that is reminiscent of the supramolecular activation cluster (SMAC) seen in the formation of the “immunological synapse” between T cells and antigen-presenting cells [28, 29]. Another similarity between the CD34 contact zone and the immunological synapse is the presence of LFA-1, which participates in cell binding, since aggregation was reduced when LFA-1 was knocked down by siRNA. Madjic et al. [6] also reported that aggregation by CD34+ KG1a cells could be inhibited with blocking antibodies to LFA-1. CD44 has been reported to be polarized in CD34+ cells [8], and we have shown that it colocalizes with CD34 when the cells are induced to aggregate. Like Giebel et al. [8], we found CD44 polarization in nonaggregated cells. Moreover, CD164 was also polarized in nonaggregated cells but was excluded from the contact zone in aggregated cells. This is in contrast to the findings of McGuckin et al. [30], who demonstrated colocalization of CD34 and CD164 on cord blood and normal bone marrow cells. Their cells were, however, adhered to gold-positive slides, which may have different effects on the distribution of these cell surface markers. These observations suggest that CD34+ cell polarization may be both constitutive and induced, as has been noted in neuroepithelial cells [31]. Exclusion, as well as inclusion, of molecules in the SMAC has also been described in the immunological synapse, where it is related to molecular size and the efficiency of adhesion and T-cell-receptor engagement [32]

Notch is a key regulator of cell fate, and Notch-1 is expressed by hematopoietic stem cells [33, 34]. Moreover, Notch-1 is polarized in mitotic neuroblasts and activated T cells [2, 35]. Although the role of Notch in hematopoiesis is controversial [36, [37]–38], the observation that it is relocated when CD34+ cells are induced to aggregate is persuasive evidence that it may be involved in CD34+ cell regulation.

The mechanism underlying the dynamic reorganization of molecules in the cell membrane to result in the configuration seen in aggregated cells is likely to be part of the overall regulatory system. Lipid raft-mediated transport is one candidate for this, since it has been shown to participate in the relocalization of molecules in polarized T cells [9, 19, 39] and in hematopoietic stem and progenitor cells [8]. Accordingly, an accumulation of lipid rafts stained with cholera toxin was seen at the contact point between aggregated CD34+ cells. Unexpectedly, CD164, which is excluded from the CD34 contact zone, as well as CD34, LFA-1, and CD44 were found in the lipid raft fraction, demonstrating that rafts are heterogeneous and may be responsible for transporting particular molecules to specific locations on the cell surface. The importance of lipid rafts for aggregation of CD34+ cells was demonstrated by treating the cells with methyl-β-cyclodextrin to extract cholesterol and disrupt lipid raft integrity.

The close association between aggregated CD34+ cells provides suitable conditions for intercellular communication, which is known to be important for coordinating cellular responses in many contexts. In particular, cell-cell contacts may determine the symmetry/asymmetry of cell division [40]. Gap junctions permit the passage of ions and small molecules between cells and have been implicated in the synchronisation of polarization in epithelial cells [13] as well as in the interaction between hematopoietic stem cells and stromal cells. It is relevant to the present study that the gap junction components belonging to the connexin family have been found in lipid rafts [16]. Interestingly, in Drosophila, stem cell daughters destined to differentiate require gap junction-mediated cell-cell interactions in order to survive out of the stem cell niche [41]. The dye transfer studies shown in Figure 4 demonstrate clearly that gap junction communication occurs between aggregated CD34+ cells and the experiment using carbenoxalone shows that it can be regulated.

Activated Notch modulates stem and progenitor cell renewal, which is associated with upregulation of the cyclin-dependent kinase inhibitors p27kip1 and p21cip1 [4]. Moreover, expression of connexins increased p27kip1 and reduced levels of cyclins D1 and D3 [42, 43]. This evidence that two components of the interaction between CD34+ cells, namely, Notch and gap junctions, influence cell cycle-related proteins supports the view that CD34-mediated aggregation regulates progenitor cell kinetics, since the cyclin-dependent kinase inhibitors have powerful effects on renewal and differentiation [5]. Accordingly, we investigated the effects of CD34+ cell aggregation on the levels of cell-cycle-regulating proteins and found that p16Ink4a and p27kip1 were upregulated, whereas cyclin D3 was downregulated, which is consistent with the reported effects of Notch and connexins [4, 42, 43]. Also, as predicted from the results of Lewis et al. [5], these aggregation-mediated effects on the cell cycle machinery were associated with a reduction in the proliferative index of progenitor cells after aggregation.

The results so far do not provide any information about symmetry versus asymmetry of cell division. This is because symmetry/asymmetry depends on the plane of cell division, which is determined by the orientation of the mitotic spindle. Centrosomes are mitotic spindle organizers, and their position can be used to predict mitotic spindle orientation [44, [45]–46]. Interphase cells contain one centrosome, which was located opposite the contact point between aggregated CD34+ cells and in CD34+ doublets harvested from in vitro colonies in a high proportion of cases (90%). Similarly, an association between centrosomes and intercellular (adherens) junctions was observed by Chenn et al. [31]. Moreover, in CD34+ cells containing two centrosomes, the orientation was not random, but orientations consistent with strictly vertical or horizontal divisions of the type described in neuroblasts [24] were infrequent (12.6%). Rather, the observed orientations suggest that the positions of the centrosomes and mitotic spindles may oscillate within limits, as proposed for telencephalic stem cells [47] and in Caenorhabditis elegans [48], thereby allowing a range of symmetrical and asymmetrical outcomes of CD34+ cell division. Importantly, the same arrangement was observed in the aggregated CD34+ cell model and the in vitro colony model, thereby showing that the results are not an artifact resulting from exposure to QBEND10.

We acknowledge that the results reported here are related to hematopoietic progenitor cells and may not apply to the hematopoietic stem cell population. Stem cells, including hematopoietic stem cells, are frequently considered in the context of the stem cell niche. In this situation they may, like Drosophila germline stem cells, be restricted to asymmetric cell division to preserve the stem cell pool [23]. This mode of cell division is, however, incompatible with stem cell expansion [1], which can occur only in the presence of symmetrical cell division. Thus, once hematopoietic stem/progenitor cells have left the niche, the interaction between CD34+ cells may allow more flexibility so that the proportions of symmetric and asymmetric divisions can vary according to hematological demand. It is relevant that niche-independent stem cell renewal has been reported for neuronal and hematopoietic stem/progenitor cells [49, 50]. An intriguing possibility is that our observations may relate to mechanisms underlying stochastic theories of hematopoietic stem cell kinetics [51].

Disclosure of Potential Conflicts of Interest

The authors indicate no potential conflicts of interest.