Evidence that ceramide mediates the ability of tumor necrosis factor to modulate primitive human hematopoietic cell fates

Normal adult human bone marrow contains a small population of transplantable totipotent hematopoietic stem cells, each with the capacity to generate a large and diverse population of mature progeny over prolonged periods of time in vivo1-3. These stem cells are closely related to cells that can sustain the production of hematopoietic cells in cultures provided with certain types of fibroblasts for at least 5 weeks2,4,5—hence the designation of the original input cells as long-term culture–initiating cells (LTC-ICs).6 In mice and humans, LTC-ICs and transplantable stem cells can be physically separated from most colony-forming cells (CFCs) that proliferate and differentiate in semisolid media in response to appropriate growth factor stimulation.7-9 Interestingly, the distinction between repopulating stem cells, LTC-ICs, and CFCs is not fully explained by differences in the particular growth factors to which these different cell types can respond because a number of these, either alone or in combination, can stimulate all 3 types of cells.2,10-14One key difference lies in the inability of primitive hematopoietic cells to proliferate in semisolid as opposed to liquid media.11 15-17 

Tumor necrosis factor (TNF) has been shown to elicit a variety of responses, inhibitory and stimulatory, on primitive human and murine hematopoietic progenitor cells, depending on the cell population exposed and the other growth factors present.18-22 Some of these effects may be indirectly mediated owing to the ability of TNF to induce the production and release of other cytokines.20,22-24 However, direct inhibitory and stimulatory effects on primitive hematopoietic cells have also been demonstrated.21,25,26 Production of intracellular ceramide is a well-described consequence of TNF stimulation,27,28and the apoptosis-inducing effect of TNF on many cells, including hematopoietic cells, is mimicked by exposure to C2-ceramide, a diffusible analog of ceramide that enters cells passively.29 30 

In previous studies, we showed that TNF in the presence of Flt3 ligand (FL), Steel factor (SF), interleukin-3 (IL-3), IL-6, and granulocyte colony-stimulating factor (G-CSF) can, at low concentrations, markedly decrease the output of cells with LTC-IC activity. This was seen in short-term cultures of normal human CD34⁺CD38⁻marrow cells under conditions that did not appear to affect the production of CFCs.12 These observations raised the possibility that TNF, like suboptimal concentrations of FL, SF, IL-3, IL-6, and G-CSF or excessive concentrations of IL-3,31 may activate intracellular events that strongly promote the differentiation of primitive human CD34⁺CD38⁻ hematopoietic cells without influencing their viability or proliferative activity. The studies described here were designed to investigate this hypothesis and to determine whether increased levels of intracellular ceramide might be involved in mediating the effects observed.

Heparinized aspirate cells were obtained with informed consent from healthy donors of marrow for allogeneic bone marrow transplantation at our center or were from cadaveric sources (Northwest Tissue Center, Seattle, WA). Low-density (less than 1.077 g/cm³) cells were stored overnight at 4°C in Iscove medium containing 50% fetal calf serum (FCS; StemCell Technologies, Vancouver, BC, Canada) for fractionation by fluorescence-activated cell sorting (FACS, see below) or were frozen in 90% FCS and 10% dimethyl sulfoxide and stored at −135°C before further use. Highly purified (more than 99% pure) CD34⁺CD38⁻ populations were isolated from fresh or thawed cell samples by FACS and collected in bulk in tubes or as single cells in the individual wells of a 96-well microtiter plate using the single-cell deposition unit of the FACS, as previously described.15 31 Briefly, cells to be sorted were washed once in phosphate-buffered saline (StemCell) and once in Hanks HEPES-buffered salt solution. In some cases, they were then resuspended in Hanks HEPES-buffered salt solution containing 0.1 μg/mL fluorescein isothiocyanate (FITC)–conjugated Annexin V (Bio-Whittaker, La Jolla, CA) in addition to 2 μg/mL propidium iodide (PI; Sigma Chemical Co, St Louis, MO) to improve the yield of viable cells. Annexin V⁻ PI⁻ cells were isolated using a FACStar Plus (Becton Dickinson, San Jose, CA) equipped with a 5-W argon laser and a helium-neon laser before they were stained with antibodies specific for CD34 (8G12-Cy5) and CD38 (Leu17-phycoerythrin; Becton Dickinson). After one wash in Hanks HEPES-buffered salt solution containing 2% FCS (HF) and another in HF containing 2 μg/mL PI (Sigma), the CD34⁺CD38⁻ subset of Annexin V⁻ PI⁻ cells was isolated. In other cases, the first staining with Annexin V was omitted, and PI⁻CD34⁺CD38⁻ cells were isolated in a single step. Alternatively, the cells were first incubated with a 5–growth factor cocktail (see below) for 2 days, then restained with Annexin V–FITC to allow newly appearing Annexin V⁺ (apoptotic) cells to be removed. In all cases, cells were kept at 4°C throughout the labeling procedures, and positive cells were selected as those exhibiting a level of fluorescence that excluded 99% of cells incubated with control antibodies of the same isotype and labeled with the same fluorochrome.

Sorted cells were cultured in suspension in 96-well plates at 1 or 100 to 200 cells/well in 100-μL volumes (or in 24-well plates at 500 cells/well in 500 μL volumes) of serum-free Iscove medium supplemented with 10⁻⁴ mol/L 2-mercaptoethanol, 10 mg/mL bovine serum albumin plus 10 μg/mL human insulin and 200 μg/mL human transferrin (BIT; StemCell), 40 μg/mL low-density lipoproteins (Sigma), and, as indicated, 100 ng/mL FL (Immunex, Seattle, WA), 100 ng/mL SF (Terry Fox Laboratory), 20 ng/mL IL-3 (Novartis, Basel, Switzerland), 20 ng/mL G-CSF (StemCell), 20 ng/mL IL-6 (Cangene, Mississauga, ON, Canada), and TNF (R&D System, Minneapolis, MN) with or without D-erythro-MAPP (Kamiya Biomedical Co, Seattle, WA), C2-ceramide (Sigma), C6-ceramide (Calbiochem, La Jolla, CA), or dihydro-C2-ceramide (Calbiochem). Cells were incubated undisturbed at 37°C in a humidified atmosphere of 5% CO₂ in air before they were harvested and assayed at the times indicated in each experiment.

CFC numbers were determined by plating 1.1-mL aliquots of washed cells suspended in Iscove methylcellulose medium (MethoCult H4330; StemCell) supplemented with 3 U/mL human erythropoietin (StemCell), 50 ng/mL SF, and 20 ng/mL each of IL-6, GM-CSF (Novartis), G-CSF, and IL-3 in 35-mm Petri dishes. Erythroid, granulocyte-macrophage, and mixed-lineage colonies were enumerated in situ after 16 days of incubation at 37°C using standard scoring criteria to identify pure and mixed erythroid and granulocyte-macrophage colonies (from BFU-E, CFU-GM, and CFU-GEMM). Total CFC counts refer to the sum of all of these.

Cells to be tested were co-cultivated, as described previously,9 with irradiated (8000 cGy) mouse fibroblast feeders (3 × 10⁵ cells/35-mm dish) engineered to produce 10 ng/mL human IL-3, 130 ng/mL human G-CSF, and 10 ng/mL human SF, with subsequent weekly replacement of half the medium (human long-term culture medium, HCC-5100, StemCell; supplemented just before use with 10⁻⁶ mol/L freshly dissolved hydrocortisone sodium hemisuccinate, Sigma). Cultures were then maintained at 37°C, usually for 6 weeks, at the end of which both the nonadherent and the adherent cells were harvested, combined, washed, and assayed for their CFC content. These 6-week CFC output values provided a relative measure of the number of LTC-ICs in the original input suspension and could thus be used directly to compare the effects of different treatments on LTC-IC activity. Although limiting-dilution LTC-IC assays were not performed to allow formal distinction between the effects on LTC-IC frequency and the effects on their CFC-producing activity, previous experiments have shown that the latter parameter is not affected when CD34⁺CD38⁻ cells are cultured under the serum-free conditions described here.9 31 In one set of experiments, some LTCs were harvested at the end of 3 weeks of incubation, and their CFC contents were similarly determined.

Differences between progenitor numbers in different treatment groups were assessed using the Student t test.

In a first series of experiments, 1000 to 2000 PI⁻/CD34⁺CD38⁻ cells isolated from normal adult human marrow were placed in culture at 500 to 1000 cells/mL in serum-free medium supplemented with FL (100 ng/mL), SF (100 ng/mL), IL-3 (20 ng/mL), IL-6 (20 ng/mL), and G-CSF (20 ng/mL) and varying concentrations of TNF. After 3, 10, and in some cases 20 days, the LTC-IC, CFC, and total cell numbers were evaluated. Under these conditions, TNF inhibited the total cell expansion (data not shown) and the expansion of CFCs and LTC-ICs in a time- and dose-dependent fashion (Figure 1). However, the loss of LTC-IC activity was the most TNF-sensitive response. Thus, exposure of CD34⁺CD38⁻ cells for 10 days to a concentration of TNF as low as 0.1 ng/mL caused a rapid and marked reduction in LTC-IC activity (more than 10-fold below input values;P < .05) in the absence of any effect on CFC production. The addition of 0.1 ng/mL TNF also had little effect on the total number of cells generated in these cultures (less than 2-fold reduction; P > .05; data not shown). Moreover, the selective effect of TNF on LTC-IC activity was not abrogated if the addition of the TNF was delayed for 2 days, as indicated by the similar results obtained in 2 such experiments, one of which one is shown in Table 1. On the other hand, a similar but an even more rapid effect (within 24 hours) could be demonstrated when the input cells were exposed to higher concentrations of TNF (20 to 100 ng/mL, data not shown, but see also Figure 3, discussed below).

To determine whether the effects of TNF pretreatment on LTC-IC activity might influence CFC production within the first 3 weeks in culture, additional experiments were performed in which the LTCs were harvested at this earlier time point. In addition, in these experiments, some cells were incubated with TNF for only part of the pretreatment phase in suspension culture with FL, SF, IL-3, IL-6, and G-CSF and were then cultured for another 2 or 4 days in the same medium but without TNF (removed by washing). The purpose of this protocol was to determine whether the effect of exposing LTC-ICs to TNF for a given period could be reversed by further incubation in the TNF-free medium with high concentrations of stimulatory cytokines before assessing residual LTC-IC activity. The results of these experiments (performed in triplicate, ie, with CD34⁺CD38⁻ cells isolated from 3 different normal bone marrows) are shown in Table2. It can be seen that, in spite of the expected minimal effects of TNF (0.1 ng/mL) pretreatment on directly assayed CFCs (groups 1 and 4 vs groups 3 and 6, respectively, in Table2; P > .05), the CFC content of 3-week-old LTCs initiated with cells cultured in the presence of TNF was significantly lower than the CFC content of 3-week-old LTCs initiated with cells cultured in the absence of TNF (groups 1 and 4 vs groups 3 and 6, respectively, in Table 2; P < .05). Moreover, there was no evidence that the loss of CFC-generating potential in LTCs could be reversed by further culture in TNF-free but growth factor–supplemented medium before plating the cells in LTC (group 3 vs group 5 in Table 2). Indeed, the data indicate that the TNF effect is fixed fairly rapidly—exposure for 3 versus 5 days (group 2 vs group 3 in Table 2) or for 5 versus 9 days (group 5 vs group 6 in Table 2) gave similar decreases in 3-week LTC-IC activity, with little effect on directly assayed CFCs in the same experiments.

Another series of experiments was undertaken to examine whether the differential TNF-induced loss of LTC-IC activity (but not CFC activity) might reflect modulation of the functional repertoire of the responding CD34⁺CD38⁻ cells rather than selective killing of those that initially possessed LTC-IC activity. In these experiments, CD34⁺CD38⁻ cells were cultured in serum-free medium plus FL, SF, IL-3, IL-6, and G-CSF plus 0.1 ng/mL TNF as before but as isolated single cells as well as in bulk cultures initiated with 500 cells each. Direct visualization of the single-cell cultures indicated no effect of this TNF dose on the viability of the input CD34⁺CD38⁻ cells (proportion of wells with 1 or more refractile cell/well) or the ability of the input cells to divide (proportion of wells with 2 or more refractile cells/well), when the cultures were assessed after 3 or 10 days of incubation (Table 3).

A comparison of the number of LTC-ICs, CFCs, and total cells measured in the single-cell and parallel-bulk cultures at the end of the 10-day period of incubation is presented in Figure2. As can be seen, the presence of TNF had little effect (P > .05) on either CFC or total cell expansion, regardless of whether 1 or 500 cells were used to initiate the cultures. However, the negative effect of TNF on LTC-IC yields (P < .05) was more pronounced in the single-cell cultures. These results establish the ability of TNF to act directly on LTC-ICs to eliminate their function without concomitantly affecting their ability to survive and divide.

To investigate whether TNF might affect the cellular attributes required for LTC-IC function as a result of activation of the ceramide pathway, human marrow CD34⁺CD38⁻ cells were cultured for 24 hours in serum-free media supplemented with the same 5–growth factor cocktail as before, in the presence (or not) of 20 ng/mL TNF with or without D-erythro-MAPP (an inhibitor of the alkaline ceramidase that breaks down ceramide to sphingosine), 30 μmol/L C2-ceramide, C6-ceramide (another active analog of ceramide whose structure is more closely related to the natural ceramide than C2-ceramide29), or dihydro-C2-ceramide, an inactive ceramide analog.32 After 24 hours, the cultures were assayed for CFCs and LTC-ICs. As shown in Figure 3, there was a similarly marked and selective decrease in LTC-IC activity in cultures to which TNF, C2-ceramide, or C6-ceramide had been added (P < .05), and the effect of TNF was slightly enhanced when D-erythro-MAPP was present. In contrast, exposure of the cells to dihydro-C2-ceramide had no significant (P > .05) effect on LTC-IC (or CFC) numbers by comparison with control cultures containing the standard 5–growth factor cocktail only.

Of interest, it should be noted that in all the overnight cultures, including the control cultures, the number of CFCs detectable was twice the input value, even though CD34⁺CD38⁻ cells from adult marrow do not begin to divide until the third day of incubation under these conditions (Petzer et al15 and Table2). This shows that the 5–growth factor cocktail can “prime” CD34⁺CD38⁻ cells in liquid culture to overcome their inability to proliferate in semisolid medium, even in the absence of cell division. A similar effect of mitogenic growth factors on primitive Lin⁻ CD34⁻CD38⁻ human marrow cells has also been demonstrated.17 Although the mechanism responsible has not yet been investigated, we speculate that an effect on the organization of the cellular cytoskeleton may be involved.

In this study, we have demonstrated an ability of nontoxic concentrations of TNF to promote the loss of certain functional properties of primitive (CD34⁺CD38⁻) human hematopoietic cells without altering their viability or capacity to become detectable as CFCs either before or after they proliferate in response to FL, SF, IL-3, IL-6, and G-CSF. This 5–growth factor cocktail was chosen because it has been shown to optimize the expansion of both LTC-IC and CFC populations from the input population studied here.12,31 Experiments with single cells showed that the response to TNF is directly mediated and can occur in the absence of effects on cell survival or proliferation. The increased LTC-IC activity obtained when CD34⁺CD38⁻ cells are stimulated by FL in combination with SF, IL-3, IL-6, and G-CSF is due to the response of a substantial proportion (more than 30%) of the initial viable population.15 Therefore, any substantial loss of these cells, such as by apoptosis, should have been readily detectable by direct visual examination of single-cell cultures. The lack of such an observation argues strongly in favor of an alternative explanation—that, under the conditions of TNF exposure used here, CD34⁺CD38⁻ cells are induced to alter certain biologic functions associated with their original undifferentiated state, including those that allow their detection as LTC-ICs. Convincing examples of a cytokine-determined alteration in primitive hematopoietic cell differentiation status, independent of effects on survival or mitogenesis, are rare and are largely restricted to studies with murine cells.33-35 However, we have recently reported similar findings in cultures of human marrow CD34⁺CD38⁻ cells exposed either to suboptimal concentrations of FL, SF, IL-3, IL-6, and G-CSF or to excess concentrations of IL-3 (in the presence of low levels of FL and SF).31 Evidence of extrinsic determination of hematopoietic cell differentiation decisions has also been obtained in various transformed or immortalized hematopoietic cell lines of mouse and human origin.36-39 

The current observations are of particular interest because they suggest the involvement of an intracellular signaling pathway that is activated by TNF and appears able to affect hematopoietic stem cell fate decisions. This is further supported by the observation that a closely related subpopulation of cells, the Thy-1⁺ subset of CD34⁺ cells in human cord blood, contains transcripts for both TNF receptor species (p55 and p75) and can be stimulated directly by TNF to enhance their proliferative response in combination with FL, SF, G-CSF, and GM-CSF.21 However, in that study, effects on LTC-IC activity were not analyzed.

Previous investigations have identified a number of downstream targets of ceramide, including a proline-directed serine–threonine protein kinase (ceramide-activated protein kinase),40 the proto-oncogene Vav,41 a cytosolic ceramide-activated protein phosphatase,42 and protein kinase C-ζ.43 Some of these are also known to be regulated by interactions with 1,2-diacylglycerol (DG). Interestingly, the outcome of raising the intracellular level of ceramide in T cells appears to be subject to modulation according to whether the cells are costimulated by receptors that activate intracellular DG.44 More recently, evidence of ceramide-induced phosphorylation of the cyclic adenosine monophosphate response element binding protein (and a related transcription factor, ATF-1) by p38 mitogen-activated protein kinase has been reported.45 Moreover, the inhibition of this pathway had no effect on the induction of apoptosis by ceramide or TNF. This is consistent with the hypothesis that these agents may elicit different biologic responses by the activation of different intracellular intermediates. Accordingly, the cellular context in which TNF receptor activation occurs is likely to be an important determinant of the final biologic outcome it elicits in a given target cell. The ability to control over time the in vitro cytokine environment of highly purified populations of primitive hematopoietic cells should facilitate the future identification of events downstream of sphingomyelin degradation that may precipitate the accelerated differentiation of these cells and allow the potential relevance of the current findings to leukemic stem cell responses to be investigated.

We thank the members of the Clinical Stem Cell Assay Service of the British Columbia Cancer Agency for assistance in accessing and initial processing of the human marrow samples, the staff of the Flow Cytometry Laboratory for assistance in cell sorting, and T. Palmater for manuscript preparation. We also thank Dr P. Lansdorp (Terry Fox Laboratory) and Amgen, Cangene, Immunex, Novartis, and StemCell for providing reagents.