Distinct hematopoietic progenitor compartments are delineated by the expression of aldehyde dehydrogenase and CD34

Hematopoietic stem cells (HSCs) and progenitors maintain and replenish the blood in steady state and after blood or marrow transplantation.^1,2  The HSCs themselves provide long-term hematopoietic repopulation of all of the blood cell lineages. Progenitor cells are more restricted in their ability to proliferate and in their capacity to generate multiple cell lineages.^3,4  The CD34 cell surface antigen is expressed by human HSCs and progenitors and has been widely used to characterize these cells.^5-13  However, CD34⁺ cells are heterogeneous,^10-13  and recent evidence suggests that not all HSCs or progenitors express CD34 at all times.^14-18  One promising complementary strategy for identifying and studying hematopoietic stem cells and progenitors is to rely on the expression of intracellular enzymes that may be important during development. One such enzyme is aldehyde dehydrogenase (ALDH). ALDH may play an important role in retinoid metabolism, and it appears to be relatively highly expressed in primitive hematopoietic cells in a number of species.¹⁹  Viable cells that express ALDH can be isolated by flow cytometry using fluorescent aldehyde substrates that freely diffuse across cell membranes.^20,21  Polar fluorescent products accumulate when these dyes are oxidized in cells that express ALDH. Consequently, cells with high levels of ALDH activity stain more brightly than cells with lower ALDH expression and can be easily distinguished.

Proof of principle for this approach was first demonstrated in mice using a fluorescent ALDH substrate termed dansyl aminoacetaldehyde (DAAA).²²  In those studies, the cells were first enriched from total bone marrow by counter-current elutriation (Fr25) and were then stained with DAAA to detect cells expressing high levels of ALDH (ALDH^br cells). Phenotypically, the murine Fr25 ALDH^br cells were CD34^neg Sca-1^neg cKit^neg and were depleted of lineage-committed cells (lin^neg).²²  These cells did not form colonies in vitro and did not establish splenic foci in short-term transplantation assays (splenic colony-forming units; CFUs-S). However, in long-term bone marrow transplantation assays, as few as 10 cells were sufficient to establish complete, multilineage hematopoietic repopulation. Thus, the Fr25 ALDH^br cells appeared to be a highly purified population of hematopoietic stem cells.

More recently, we developed an alternative fluorescent substrate for ALDH, termed BODIPY aminoacetaldehyde (BAAA) (Aldefluor; StemCo BioMedical, Durham, NC), to select human hematopoietic progenitors that express ALDH.^21,23  In those studies, progenitors from human umbilical cord blood (UCB) and from human mobilized peripheral blood (mPB) were readily identified based on low orthogonal light scatter and high ALDH expression (SSC^lo ALDH^br cells).^21,23  Hematopoietic progenitors within this cell fraction included clonogenic progenitors (CFUs) and cells with the capacity to initiate long-term culture (LTC). Subsequent studies indicated that SSC^lo ALDH^br UCB cells were also highly enriched for cells that engraft nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice (SCID-repopulating cells [SRCs]).^24-26  Thus, in contrast to the murine Fr25 ALDH^br HSCs, human cells that expressed ALDH exhibited a broad range of developmental potentials.

In all the previous studies on ALDH^br cells, it was noted that the SSC^lo ALDH^br UCB and mPB were enriched for CD34⁺ cells and for CD34⁺ CD38^neg cells.^21,23,25,26  However, though there was considerable overlap between cells that expressed CD34 and those that expressed ALDH, there were also cells that expressed one or the other determinant. In the current study, we sought to determine whether the differential expression of ALDH and CD34 could be used to define and segregate developmentally distinct progenitor populations. To pursue this, ALDH^br CD34⁺, ALDH^neg CD34⁺, and ALDH^br CD34^neg hematopoietic cells were purified from SSC^lo UCB, and their developmental potential was tested in a variety of in vitro and in vivo assays. The most primitive hematopoietic progenitors were highly enriched within the ALDH^br CD34⁺ cell fraction. In addition, short-term myeloid progenitors were most abundant within the ALDH^br CD34⁺ cell subset. By comparison, the ALDH^neg CD34⁺ cells contained few primitive progenitors and were reduced in short-term myeloid progenitors. However, when measured in short-term culture assays, the ALDH^neg CD34⁺ cells contained higher frequencies of committed NK progenitors than did the ALDH^br CD34⁺ cells. The ALDH^br CD34^neg cells were not consistently supported by any of the assays used, and their function remains undefined. Together these observations demonstrated that functionally distinct human hematopoietic progenitors could be resolved based on their relative expression of ALDH and CD34. In particular, these studies suggested that, within the CD34⁺ progenitor compartment, ALDH expression was significant for primitive multilineage progenitors and for the early myeloid compartment. In contrast, decreased ALDH expression was associated with enhanced short-term NK development by CD34⁺ progenitors.

BAAA (Aldefluor) was provided by StemCo Biomedical, Inc. Hydrocortisone hemisuccinate, verapamil, dimethylaminobenzaldehyde, Zn₂SO₄, Cu₂SO₄, Se₂O₃, and ascorbic acid were purchased from Sigma-Aldrich (St Louis, MO). 7-Amino actinomycin D (7AAD) was purchased from Molecular Probes (Eugene, OR). All cytokines were purchased from R&D Systems (Minneapolis, MN). All cell culture media and media supplements (2-mercaptoethanol [2-ME], pyruvate, glucose, Na₂CO₃, and HEPES [N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid]) were purchased from Invitrogen Life Technologies (Carlsbad, CA). Fetal calf serum (FCS) and equine serum were purchased from HyClone (Logan, UT).

UCBs were obtained from the Carolina Cord Blood Bank under protocols approved by the Institutional Review Board at Duke University Medical Center. UCB cells were stained with BAAA, as previously described.²¹  Briefly, cells were resuspended at 10⁶ cells/mL in Iscove modified Dulbecco medium (IMDM) supplemented with 2% FCS (IMDM+2%) and equilibrated to 37°C. Cells were stained with 1 μM BAAA for 30 minutes in the presence of 50 μM verapamil (Sigma-Aldrich). Baseline fluorescence was established based on control cells stained with BAAA in the presence of 10 μM dimethylaminobenzaldehyde (DEAB), a potent inhibitor of ALDH.²⁷  After 30 minutes, the cells were centrifuged, placed on ice, and stained with fluorescence-conjugated antibodies. These included monoclonal antibodies directed against CD3, CD19, CD56, CD235a (all phycoerythrin [PE] conjugated) and CD34 (allophycocyanin [APC] conjugated) (all from BD Pharmingen, San Jose, CA). Dead and dying cells were excluded by staining with 7AAD (1 μg/10⁶ cells; Molecular Probes). The cells were washed and resuspended in chilled IMDM+2% containing 50 μM verapamil and were kept on ice before fluorescence-activated cell sorting.

Cell purification was performed using a FACS Vantage Cell Sorter (BD Biosciences). Samples were chilled throughout the cell sorting. During initial preparative sorts, the cells were captured in IMDM+2% containing 50 μM verapamil. For purity, the cells were immediately resorted into complete culture media that did not contain verapamil. UCB cells were fractionated to derive SSC^lo ALDH^br CD34⁺ cells, SSC^lo ALDH^neg CD34⁺ cells, and SSC^lo ALDH^br CD34^neg cells. In most studies, these cell fractions were derived from lin^neg UCB (Figure 1F-G; n = 25). In some studies, lin^neg UCB cells were enriched before sorting by depleting the PE-labeled CD3⁺, CD19⁺, CD56⁺, and CD235a⁺ cells using immunomagnetic beads (PE-specific EasySep beads; StemCell Technologies, Vancouver, BC, Canada). Residual lin⁺ (PE⁺) cells were excluded during the cell sorting. For some NOD/SCID reconstitution studies, control CD34⁺ cells were isolated from lineage-depleted UCB prepared using high-density magnetic enrichment (StemSep enrichment, human progenitor cocktail; StemCell Technologies).

NOD/SCID (ltz-ltz) mice were purchased from Jackson Laboratories (Bar Harbor, ME). Hematopoietic transplantations were performed according to protocols approved by the Institutional Animal Use and Care Committee at Duke University (Durham, NC). The mice underwent irradiation before transplantation (325 cGy). Each mouse underwent transplantation with purified cells derived from a single UCB. Cells were prepared in IMDM supplemented with 10% FCS and were transplanted by tail-vein injection. To prevent cell loss during the injections, carrier cells were added to purified cells before the transplantations. Carrier cells were total blood cells from irradiated NOD/SCID mice (10⁶ cells per transplant; approximately 2 μL total blood).

NOD/SCID mice were killed 6 to 7 weeks after transplantation (short term) or 18 weeks or longer after transplantation (long term). Bone marrow cells were prepared in IMDM+2% and were concentrated to 10⁷ cells/mL. Before antibody staining, murine Fc receptors were blocked by 15-minute incubation with an antibody specific for murine CD16 and CD32 (BD PharMingen). Dead and dying cells were excluded from analyses using 7AAD (Molecular Probes). The presence of human cells was detected with antibodies directed against CD45 and was further confirmed with antibodies directed against human class 1 HLA antigens (BD PharMingen). In addition, murine hematopoietic cells were excluded with antibodies directed against Ly5 (murine CD45; BD PharMingen). Human hematopoietic engraftment was defined by the presence of CD45⁺ Ly5^neg cells (Figure 2A). Lineage-specific differentiation was determined using antibodies directed against human antigens (all BD PharMingen), including CD19 (B cell) and CD13 or CD15 (myeloid). Human progenitor cells were detected by their expression of CD34. In some animals, the total bone marrow cell preparations were also depleted of murine lineage-committed cells to enrich human cells (Figure 2B) (SpinSep isolations, murine progenitor enrichment cocktail; StemCell Technologies). Cells were stained with antibodies for 20 minutes at 4°C, washed with Dulbecco phosphate-buffered saline (PBS) supplemented with 2% FCS (PBS+2%), and fixed in PBS+2% containing 1% formaldehyde (methanol-free, EM grade; Polysciences, Warrington, PA).

Hematopoietic progenitor colony assays were performed in MethoCult GF⁺ H4435 (StemCell Technologies). Myeloid colonies (more than 50 cells) were scored between 14 and 18 days.

LTC assays were performed on AFT024 fetal liver stroma (kindly provided by Dr Ihor Lemischka, Princeton University, NJ).²⁸  AFT024 stromas were maintained in DMEM/F12 medium supplemented with 10% FCS. For LTCs, the stromas were established in 24-well cell-culture dishes. Stromas, maintained at 33°C for normal passage, were irradiated (20 Gy) and transferred to 37°C before the establishment of LTCs. Primary and secondary LTCs were grown in MEMα supplemented with 12% FCS, 12% equine serum, 55 μM 2-mercaptoethanol (2-ME), 100 nM hydrocortisone, recombinant human interleukin-3 (IL-3; 5 ng/mL), and recombinant human macrophage inflammatory protein alpha-1 (MIP-1α) (100 pg/mL).^29,30  Cultures were fed weekly by replacing half the medium in each culture. After 5 weeks, primary LTCs were harvested with trypsin-EDTA (ethylenediaminetetraacetic acid) to establish clonogenic assays in MethylCult GF⁺ H4435 (StemCell Technologies), to establish secondary LTCs on AFT024 stroma, and to establish secondary NK cell development assays on AFT024 stroma (see “Secondary NK assays”).

In some LTC assays, cell proliferation was monitored by fluorescence-activated cell sorter (FACS) analysis. For these analyses, a portion of each culture was concentrated to less than 100 μL, and the cells were stained with antibodies specific for human CD45, CD13, and CD34 (all BD PharMingen). Dead and dying cells were excluded with 7AAD (Molecular Probes). The cells were stained for 20 minutes at 4°C, washed in PBS+2%, and fixed in PBS+2% containing 1% formaldehyde (Polysciences). For accuracy, the CD45⁺7AAD^neg cells were enumerated relative to fixed numbers of 6-μm fluorescent reference beads that were added to the reactions (Polysciences).

After primary LTCs, the potential for NK cell development was monitored by replating a fraction of the culture on irradiated AFT024 cells (20 Gy) in a DMEM/F12 medium supplemented with 10% FCS that was also modified in a manner similar to a media formulation previously described.³¹  This medium was adjusted to 20 mM HEPES, 2.5 g/L Na₂CO₃, 4.5 g/L glucose, 2 mM sodium pyruvate, 55 μM 2-ME, 5 μMZn₂SO₄, 100 nM Cu₂SO₄, 100 nM Se₂O₃, 100 nM ascorbic acid, and 0.3 μM biotin. NK cell development was encouraged with recombinant IL-2 (5 ng/mL), IL-7, IL-15, kit ligand, and Flt3 ligand (each at 10 ng/mL). Cultures were fed twice weekly by replacing half the medium in each culture.

After 5 to 6 weeks, the cultures were monitored for NK (CD56⁺) and myeloid (CD13⁺) cells by flow cytometry. Cultures were harvested with trypsin-EDTA and were stained with fluorescence-conjugated antibodies and 7AAD. The expansion of viable hematopoietic cells (CD45⁺7AAD^neg) was calculated relative to 6-μm fluorescent reference beads (Polysciences). Anti-CD45 (fluorescein isothiocyanate [FITC], clone J33) was purchased from Coulter-Immunotech (Miami, FL). All other antibodies were purchased from BD PharMingen.

Short-term culture assays for lymphoid development were monitored on STO murine embryonic fibroblasts (CRL-1503; American Type Culture Collection, Manassas, VA). For these assays, the STO cells were established as monolayers in 24-well dishes in DMEM/F12 media supplemented with 10% FCS. At confluence, the cells were irradiated (20 Gy). Purified cells were cultured on STO feeders in DMEM/F12 medium that was supplemented as described for secondary NK cultures. In short-term assays, NK cell development was encouraged with recombinant IL-3 (10 ng/mL), IL-7 (20 ng/mL), and IL-15 (10 ng/mL). After 21 days, NK cell development was monitored by flow cytometry. The cultures were harvested with trypsin-EDTA and were stained with fluorescence-conjugated antibodies and 7AAD. The expansion of viable hematopoietic cells (CD45⁺7AAD^neg) was calculated relative to 6-μm fluorescent reference beads. NK cells were monitored as CD56⁺ CD3^neg cells. The total output of lymphoid cells was calculated as [(total cell expansion) × (percentage CD56⁺ cells)]. Myeloid cells were monitored as CD13⁺ cells. The total output of myeloid cells was calculated as [(total cell expansion) × (percentage CD13⁺ cells)]. The relative cell output was calculated as [(total output, ALDH^neg CD34⁺ cells)/(total output, ALDH^br CD34⁺ cells)] or as [(total output, ALDH^br CD34⁺ cells)/(total output, ALDH^neg CD34⁺ cells)]. Anti-CD45 (FITC, clone J33) was purchased from Coulter-Immunotech (Miami, FL). Anti-CD3 (PE) was purchased from Becton Dickinson. All other antibodies were purchased from BD PharMingen.

For most studies, the mean (± SD) and median values were presented. After development in vitro, the performances of lin^neg SSC^lo ALDH^br CD34⁺ and lin^neg SSC^lo ALDH^neg CD34⁺ cells were compared in paired 1-tailed nonparametric analyses (Wilcoxon signed rank tests). Frequencies for NOD/SCID repopulating cells, for secondary LTCs, and for lymphoid progenitors were estimated by limiting dilution analyses using L-Calc (StemCell Technologies). For NOD/SCID studies, the frequencies of SRCs were calculated based on the mice that demonstrated multilineage engraftment, in which not more than 90% of human cells were CD19⁺. LTCs for secondary NK progenitors were considered negative when fewer than 10 CD56⁺ cells were generated per cell used to initiate the primary LTCs.

The overall goal of this study was to determine whether distinct hematopoietic progenitors could be delineated within UCB based on their expression of ALDH and CD34. As had been previously noted,²¹  the SSC^lo ALDH^br UCB was highly enriched for CD34⁺ cells and contained 79.7% ± 16.8% of the SSC^lo CD34⁺ CD38^neg cells (n = 8; Figure 1C-D [compare panels]). To characterize these cells, SSC^lo ALDH^br CD34⁺, SSC^lo ALDH^neg CD34⁺, and SSC^lo ALDH^br CD34^neg UCB cells were purified for functional analyses. Our previous studies had indicated that SSC^lo ALDH^br UCB contained few cells that expressed lineage-specific antigens.²¹  However, to ensure this, lineage-committed cells were excluded to yield a lin^neg population (depicted in Figure 1F-G). Each fraction contained small cells with minimal cytoplasm that were not morphologically distinguishable (data not shown). The SSC^lo ALDH^br cells represented 0.96% ± 0.5% of the UCB, and 50.9% ± 18.3% of these cells expressed CD34. Similarly, 0.9% ± 0.6% of the total UCB was lin^neg SSC^lo CD34⁺, and 47.9% ± 14.3% of these cells expressed ALDH. Absolute percentages of the lin^neg SSC^lo ALDH^br CD34⁺, lin^neg SSC^lo ALDH^neg CD34⁺, and lin^neg SSC^lo ALDH^br CD34^neg UCB cell subsets are summarized in Table 1. In addition, it was noted that CD34 expression was significantly more intense on lin^neg SSC^lo ALDH^br CD34⁺ cells than on lin^neg SSC^lo ALDH^neg CD34⁺ cells (Table 1; Figure 1F-G). Together these observations suggested that discrete populations of cells could be defined based on their expression patterns of ALDH and CD34.

To confirm that primitive progenitors could be distinguished on the basis of ALDH and CD34 expression, ALDH^br CD34⁺, ALDH^neg CD34⁺, and ALDH^br CD34^neg UCB cells were transplanted into NOD/SCID mice. In 3 initial studies, these cell fractions were purified from SSC^lo UCB. In all subsequent studies, the cells were purified from lin^neg SSC^lo UCB, as depicted in Figure 1. Because no differences were observed between these studies, the results were pooled. Control mice underwent transplantation with CD34⁺ lin^neg cells. In all studies, human hematopoietic engraftment was monitored by the content of human CD45⁺ ly5^neg present in NOD/SCID bone marrow (Figure 2A) and was routinely confirmed by their expression of class 1 HLA antigens (data not shown). When limiting cell doses were used, some mice achieved only limited (less than 0.1%) engraftment. Therefore, in some mice, low-level human hematopoietic engraftment was also confirmed by enriching the relative proportions of human cells by depleting lineage-committed murine cells (n = 7; Figure 2B). To test for multilineage engraftment, the NOD/SCID marrow was also stained for human B cells (Figure 2D), myeloid cells (Figure 2E), and CD34⁺ progenitors (Figure 2F).

Long-term human hematopoietic engraftment was reliably detected at varying levels in NOD/SCID mice in which 1000 to 3000 ALDH^br CD34⁺ cells were transplanted (Figure 2G). Human hematopoietic engraftment in these mice was 64.5% ± 30.7% B-lymphoid and 29.7% ± 34.7% myeloid cells. In addition, 5 of 9 mice maintained discernible CD34⁺ cells (9.16% ± 9.07%). SRCs were defined as cells that repopulated multiple human hematopoietic lineages in NOD/SCID mice. Limiting dilution analyses estimated the frequency of SRCs to be 1 in 4687 ALDH^br CD34⁺ cells (95% confidence interval [CI]: 1/2180-1/10 077). In control animals, the SRC frequency was estimated to be 1 in 31 486 CD34⁺ lin^neg cells (95% CI: 1/12 497-1/79 332). Therefore, the ALDH^br CD34⁺ cell fraction was enriched for SRC approximately 7-fold when compared with the CD34⁺ cell fraction. In contrast, the ALDH^neg CD34⁺ cells exhibited very low levels of SRC activity. Only 2 of 10 mice that received 10 000 ALDH^neg CD34⁺ cells had discernible human hematopoietic engraftment. With this cell fraction, estimates for the frequency of SRCs would not be accurate. However, limiting dilution analyses performed on these data suggested that SRCs were likely to represent less than 1 in 45 000 ALDH^neg CD34⁺ cells. Higher cell transplant doses would be required to accurately resolve this issue. Finally, the ALDH^br CD34^neg cells did not engraft NOD/SCID mice at any of the cell doses used (Figure 2G). These studies suggested that long-term multilineage NOD/SCID repopulating cells were enriched in cells that expressed both ALDH and CD34.

To continue these studies, short-term transplantation assays were performed using cell doses that approached the limiting dilution for the ALDH^br CD34⁺ cells. Purified lin^neg SSC^lo ALDH^br CD34⁺, lin^neg SSC^lo ALDH^neg CD34⁺, or lin^neg SSC^lo ALDH^br CD34^neg cells were transplanted into NOD/SCID mice, and human hematopoietic engraftment was monitored at 6 to 7 weeks after transplantation (Figure 2H). Using 3000 cells per transplant, the ALDH^br CD34⁺ cells reliably engrafted NOD/SCID bone marrow, which recapitulated what had been observed in long-term hematopoietic transplantation assays. In contrast, the ALDH^neg CD34⁺ and ALDH^br CD34^neg cells did not achieve detectable human hematopoietic engraftment at these low cell doses, even in short-term assays. These data suggested that short-term NOD/SCID repopulating cells were enriched within the ALDH^br CD34⁺ cell fraction.

LTC assays were established as an alternative assay for primitive progenitors. These assays were used to test a potential for long-term multilineage growth without the more stringent requirement for cell homing that is present in NOD/SCID transplantation assays. lin^neg SSC^lo ALDH^br CD34⁺, lin^neg SSC^lo ALDH^neg CD34⁺, and lin^neg SSC^lo ALDH^br CD34^neg UCB cells were purified as in Figure 1. LTCs were initiated on AFT024 stroma using 50 to 500 cells/culture (n = 8).^29,30  After 5 weeks, the LTCs were harvested, and the collected cells were monitored for total hematopoietic cell expansion, content of myeloid CFU, content of secondary LTC, and content of secondary NK progenitors. As with the primary LTCs, secondary LTCs were maintained in a standard Dexter-type medium supplemented with human IL-3 and MIP-1α. Secondary NK cultures were established without hydrocortisone in the presence of human Kit ligand, Flt3 ligand, IL-2, IL-7, and IL-15.

The distributions of cells that initiated LTCs closely paralleled the data for SRCs. In primary LTCs, the ALDH^br CD34⁺ cells expanded 4-fold more than the ALDH^neg CD34⁺ cells and yielded approximately 60-fold more CFUs (Table 2). Similarly, the ALDH^br CD34⁺ cells yielded approximately 100-fold more CFUs after secondary LTCs when compared with ALDH^neg CD34⁺ cells (Table 2). Limiting dilution analyses estimated that 1 in 43 ALDH^br CD34⁺ cells initiated secondary LTCs (95% CI: 1/18-1/103). In contrast, the ALDH^neg CD34⁺ cells were nearly devoid of secondary LTCs, and only 3 of 8 ALDH^neg CD34⁺ cell preparations initiated these cultures. Limiting dilution analyses estimated that 1 in 1130 ALDH^neg CD34⁺ cells initiated secondary LTCs (95% CI: 1/364-1/3513). Finally, the ALDH^br CD34^neg cell fraction failed to initiate primary or secondary LTCs.

Secondary NK culture assays were established from primary LTC cultures and were used to measure long-term lymphoid development in vitro. In these assays, the lin^neg SSC^lo ALDH^br CD34⁺ cells consistently gave rise to high percentages of CD56⁺ cells in secondary cultures (Figure 3A-B). Because all these cultures contained NK cells, the frequencies of the long-term NK progenitor cells could not be estimated. However, under these same conditions, only half the primary LTCs that had been initiated with lin^neg SSC^lo ALDH^neg CD34⁺ cells yielded any NK progeny in secondary cultures (Figure 3C-D). Limiting dilution analyses estimated that 1 in 709 lin^neg SSC^lo ALDH^neg CD34⁺ cells initiated secondary NK cultures (95% CI: 1/258-1/1947). Collectively, these studies indicated that long-term multilineage growth in vitro was primarily associated with cells that expressed ALDH and CD34, similar to what had been observed in SCID repopulation studies.

The data for SRC and LTC activity suggested that primitive, multilineage progenitor cells were highly enriched within the ALDH^br CD34⁺ cell fraction. To determine whether this was also true for short-term myeloid progenitors, lin^neg SSC^lo ALDH^br CD34⁺, lin^neg SSC^lo ALDH^neg CD34⁺, and lin^neg SSC^lo ALDH^br CD34^neg UCB was monitored for clonogenic cells (CFUs) in hematopoietic progenitor colony assays (HPCAs). Both the ALDH^neg CD34⁺ and the ALDH^br CD34⁺ cell fractions contained clonogenic myeloid progenitors (Table 3). However, myeloid CFUs were more than twice as frequent within the ALDH^br CD34⁺ subset as they were among the ALDH^neg CD34⁺ cells. The ALDH^br CD34^neg cells did not contain detectable clonogenic progenitors. These studies indicated that myeloid CFU activity was enriched in cells that coexpressed CD34 and ALDH but that some CFU activity remained in CD34⁺ cells that did not express ALDH. In contrast, no clonogenic myeloid progenitors were detected in cells that did not express CD34.

To complete these studies, we next sought to determine whether short-term lymphoid progenitors could be discriminated based on the expression of ALDH and CD34. Purified lin^neg SSC^lo ALDH^br CD34⁺, lin^neg SSC^lo ALDH^neg CD34⁺, and lin^neg SSC^lo ALDH^br CD34^neg cells were cultured on STO murine embryonic fibroblasts in the presence of recombinant human IL-3, IL-7, and IL-15 to encourage lymphoid progenitors to develop into NK cells. This culture system has been previously used to support the rapid emergence of NK cells from CD7⁺ CD34^neg cells (R.W.S., unpublished observations, December 2003). After 21 days, the cultures were analyzed for their total expansion of hematopoietic cells (CD45⁺7AAD^neg) and for their relative frequencies of CD56⁺ and CD13⁺ cells.

In these short-term cultures, the ALDH^br CD34⁺ cells expanded more efficiently than did the ALDH^neg CD34⁺ cells. Both cell fractions gave rise to CD56⁺ progeny that, based on their phenotype, appeared to be NK cells. However, the ALDH^neg CD34⁺ cells consistently gave rise to higher percentages of CD56⁺ lymphoid cells than did the ALDH^br CD34⁺ cells (Table 4; Figure 4A-B [compare panels]). Furthermore, even with lower total cell expansion, the ALDH^neg CD34⁺ cells produced 3-fold more NK progeny than were observed in cultures initiated with ALDH^br CD34⁺ cells (Table 4). Indeed, in 9 of 10 cultures established, the ALDH^neg CD34⁺ cells produced more CD56⁺ progeny (Figure 4J). This suggested that the ALDH^neg CD34⁺ cell fraction was enriched for short-term NK progenitors relative to ALDH^br CD34⁺ progenitors. In contrast, the ALDH^br CD34⁺ cells predominantly gave rise to CD13⁺ myeloid cells (Table 4; Figure 4D-E [compare panels]). Taken together, the total output of myeloid progeny by ALDH^br CD34⁺ was approximately 7-fold greater than that of ALDH^neg CD34⁺ cells (Table 4). This was consistent with their clonogenic potential as monitored in myeloid CFU assays (Table 3). Furthermore, relative to ALDH^neg CD34⁺ cells, the ALDH^br CD34⁺ cell fraction always produced more CD13⁺ progeny (Figure 4K). Finally, of 9 culture assays performed, only 2 preparations of ALDH^br CD34^neg cells proliferated to any degree. Furthermore, within those 2 assays, the cells expanded less than 10-fold and gave rise to progeny that were nearly entirely CD56⁺ (Figure 4C-F). Given that these events were sporadic, no conclusions were drawn as to the developmental potential of the ALDH^br CD34^neg cells in these short-term assays.

We and others have previously demonstrated that the SSC^lo ALDH^br UCB was enriched for clonogenic progenitors, progenitors that initiate LTC and NOD/SCID repopulating cells.^21,24-26  In all these studies there was substantial, but incomplete, overlap between cells that expressed ALDH and CD34. Therefore, the major focus for the current study was to determine whether the expression of ALDH and CD34 could be used to delineate functionally distinct progenitor cell subsets. To address this question, a variety of assays was used to explore the developmental potentials of 3 distinct and nonoverlapping cell populations: ALDH^br CD34⁺, ALDH^neg CD34⁺, and ALDH^br CD34^neg cells.

In total our analyses suggested that substantial functional distinctions did exist between ALDH^br CD34⁺, ALDH^neg CD34⁺, and ALDH^br CD34^neg cells. ALDH^br CD34⁺ cells had a high frequency of short- and long-term SRCs. In addition, this cell fraction was highly enriched for cells that initiated primary and secondary LTCs. In contrast, by these same assays, the ALDH^neg CD34⁺ cell subset contained few primitive progenitors. Furthermore, the ALDH^br CD34^neg cells did not repopulate NOD/SCID mice or initiate LTC. Thus, the ALDH^br CD34⁺ cells contained multilineage lymphomyeloid progenitors with an extensive capacity for proliferation.

ALDH^br CD34⁺ and ALDH^neg CD34⁺ cells contained progenitor cells with more limited capacities that could be detected in short-term in vitro assays. ALDH^br CD34⁺ cells had more than a 2-fold higher content of clonogenic myeloid progenitors when compared with ALDH^neg CD34⁺ cells. Furthermore, ALDH^br CD34⁺ cells expanded efficiently into myeloid progeny under conditions designed to primarily support NK cell growth. Under the same in vitro culture conditions, ALDH^neg CD34⁺ cells consistently gave rise to more lymphoid progeny than ALDH^br CD34⁺ cells. Taken together, these studies suggested that, within the CD34⁺ cell compartment, ALDH is not maintained among short-lived NK progenitors. Finally, as in the studies that characterized primitive hematopoietic cells, the ALDH^br CD34^neg cell fraction contained no consistent short-term myeloid or NK progenitor activity. Thus, though it is possible that this cell fraction may contain an NK progenitor, this phenomenon was too inconsistent to draw firm conclusions regarding the developmental potential of ALDH^br CD34^neg cells.

Collectively these observations suggest that the expression of ALDH and CD34 can be used to distinguish developmental stages of human hematopoiesis. As suggested by the model depicted in Figure 5, the most primitive hematopoietic cells expressed ALDH and CD34. CD34 expression and ALDH expression are maintained through early myeloid differentiation. However, during lymphoid differentiation, ALDH expression apparently diminishes before the loss of CD34 expression.

An important caveat must be considered with this model. It is possible that the most primitive hematopoietic cells failed to grow in the assay systems that were used. In our studies, the ALDH^br CD34^neg UCB cells contained no SRC, LTC, or clonogenic myeloid progenitors and exhibited only minimal, sporadic growth in short-term NK cultures. However, murine ALDH^br CD34^neg HSCs repopulate mice in long-term hematopoietic transplants with as few as 10 cells but do not contain clonogenic myeloid progenitors or more primitive multilineage CFUs-S.²²  In addition, human CD34^neg cells do not engraft NOD/SCID mice efficiently without direct injection to the bone marrow.^14,17,32  Preliminary studies by Hess et al³³  suggest that the ALDH^br CD34^neg cells might represent a primitive progenitor. A related question is how these various progenitors might contribute to actual human engraftment and immune reconstitution after clinical transplantation. Our studies with long-term growth and NOD/SCID reconstitution suggest that the cells most likely to contribute to bone marrow reconstitution would coexpress CD34 and ALDH. The ALDH^neg CD34⁺ cells had only limited capacity for long-term NOD/SCID reconstitution, and the lymphoid potential of these cells was elucidated only in short-term in vitro cultures. However, it is possible that some ALDH^neg CD34⁺ cells may contribute to short-term immune reconstitution after transplantation. Future studies will more thoroughly compare the relative contributions of ALDH^br CD34⁺ and ALDH^neg CD34⁺ cells toward short-term reconstitution in vivo. Last, our studies to date have only been performed using conditions that favor NK cell development. Future studies will also more thoroughly explore the relative capacities of ALDH^br CD34⁺, ALDH^neg CD34⁺, and ALDH^br CD34^neg cells to differentiate toward the B- and T-cell lineages in vitro.

We thank Mrs Adelyn Luther for her continuing generosity in support of this work. We also thank Lynn Martinek and Dr Michael Cook of the Duke University Cancer Center Flow Cytometry Core for their valuable expertise in cell sorting and Drs John Chute and Andrew Balber for their critical reading of the manuscript.