Multipotent self-renewing hematopoietic stem cells (HSCs) are responsible for reconstitution of all blood cell lineages. Whereas growth stimulatory cytokines have been demonstrated to promote HSC self-renewal, the potential role of negative regulators remains elusive. Receptors for tumor necrosis factor (TNF) and Fas ligand have been implicated as regulators of steady-state hematopoiesis, and if overexpressed mediate bone marrow failure. However, it has been proposed that hematopoietic progenitors rather than stem cells might be targeted by Fas activation. Here, murine Lin−Sca1+c-kit+ stem cells revealed little or no constitutive expression of Fas and failed to respond to an agonistic anti-Fas antibody. However, if induced to undergo self-renewal in the presence of TNF-α, the entire short and long-term repopulating HSC pool acquired Fas expression at high levels and concomitant activation of Fas suppressed in vitro growth of Lin−Sca1+c-kit+ cells cultured at the single cell level. Moreover, Lin−Sca1+c-kit+ stem cells undergoing self-renewal divisions in vitro were severely and irreversibly compromised in their short- and long-term multilineage reconstituting ability if activated by TNF-α or through Fas, providing the first evidence for negative regulators of HSC self-renewal.
Self-renewing hematopoietic stem cells (HSCs) are responsible for replenishing all blood cell lineages through life. Although representing only 1 in 10,000 bone marrow (BM) cells, sophisticated methods have been developed to isolate HSCs to high purity 1, allowing meaningful studies of candidate stem cells in vitro. However, HSCs cannot be purified to homogeneity, and must thus ultimately be identified through their unique ability to long-term multilineage reconstitute ablated recipients 1.
Hematopoiesis is regulated in part by a large number of hematopoietic growth factors supporting proliferation, differentiation, and survival of hematopoietic cells at different levels of differentiation 2,3. Most evidence support that the primary role of such cytokines in early hematopoiesis is to prevent programmed cell death, rather than to induce programs for growth and differentiation 4,5,6,7.
Much less is known about the potential role of negative regulators in hematopoiesis 8,9,10,11. Receptors for TNF-α and Fas ligand have been implicated to be involved in suppressing hematopoiesis. A growth-inhibitory effect of TNF-α in early murine hematopoiesis is supported by in vitro studies of purified primitive murine progenitors 11,12, but it has not been investigated whether this is associated with a negative or positive effect on HSC function. Studies in TNF receptor–deficient mice have suggested that the number of phenotypically defined (Lin−Sca1+c-kit+) stem cells is increased 11. However, more recent studies have shown that this increase does not necessarily correlate with enhanced in vivo HSC function 13, emphasizing that true long-term repopulating cells (LTRCs) only can be evaluated through their unique ability to long-term multilineage reconstitute conditioned recipients in vivo 1,14.
Fas has been implicated to play a role in regulation of myelopoiesis and as a potential tumor suppressor of myeloid leukemic development 10. However, the status of its expression and potential function on HSC remains elusive. Fas-mediated apoptosis of committed myeloid progenitors has been demonstrated to be involved in the myelosuppression associated with acute graft versus host disease (GVHD; reference 9). Accumulating evidence also support a role of TNF-α and Fas in a number of chronic disorders of blood cell formation associated with enhanced apoptosis 15,16,17,18,19,20,21,22. However, as candidate murine BM stem cells, unlike myeloid progenitors, have been reported to lack expression of cell surface Fas 10,23, HSCs have been suggested to represent unlikely targets for Fas-induced BM suppression.
We have recently, through HSC division tracking and serial transplantation experiments, demonstrated that murine Lin−Sca1+c-kit+ long-term reconstituting stem cells can undergo multiple self-renewing divisions in vitro 24,25. Others and we have demonstrated that this self-renewal process is positively affected by cytokines such as IL-3, IL-6, IL-11, c-kit ligand, and thrombopoietin 24,25,26,27. Thus, this allowed us to investigate how TNF-α and Fas might potentially affect HSC self-renewal.
In these studies we demonstrate that purified Lin− Sca1+c-kit+ repopulating BM stem cells 28,29,30 have little or no constitutive expression of Fas and fail to respond to an agonistic anti-Fas antibody. However, TNF-α efficiently induces Fas expression and responsiveness of cycling Lin−Sca1+c-kit+ HSCs, resulting in severely compromised ability to short- and long-term multilineage reconstitute the hematopoietic system.
Materials And Methods
Hematopoietic Growth Factors and Antibodies.
Recombinant rat (rr) KL (stem cell factor) and recombinant human (rh) G-CSF were provided by Amgen Corp. (Thousand Oaks, CA). Recombinant murine (rm) IL-3 was from PeproTech. rhIL-6 was a gift from Genetics Institute (Cambridge, MA). rmTNF-α was supplied by Genentech (San Francisco, CA) and rh flt3 ligand (FL) was a gift from Immunex (Seattle, WA). Unless otherwise indicated, all cytokines were used at the following predetermined optimal concentrations: rrKL and rhIL-6: both at 50 ng/ml; rmIL-3 and rmTNF-α: 20 ng/ml. When a cocktail of KL plus IL-3 plus IL-6 plus FL plus G-CSF was used, the concentrations of the cytokines were 25 ng/ml, except for IL-3 which was used at 10 ng/ml. Purified monoclonal hamster anti–mouse Fas antibody (Jo2) and irrelevant isotype-matched hamster IgG antibody (both from BD PharMingen) were used at 0.2 to 0.5 μg/ml, based on initial titration experiments showing similar growth-inhibitory effects of Jo2 at 0.05, 0.1, 0.2, or 0.5 μg/ml. Jo2 as well as the isotype-matched antibody contained no sodium azide and low endotoxin levels (≤0.01 ng/μg of protein).
Enrichment and Purification of Lin−Sca1+c-kit+ Murine BM Cells.
All animal procedures were performed with consent from the local ethics committee at Lund University. Lineage-depleted (Lin−) BM cells were isolated from normal 6–14-wk-old C57BL/6 mice (CD45.2), congenic B6.SJL-PtprcaPepb/BoyJ (CD45.1), or Fas-deficient (lpr) mice inbred to C57BL/6 background (B6.MRL-Fas<lpr> mice; CD45.2) according to previously described protocols 24,25,30,31. When phenotyping or sorting Lin−Sca1+ckit+CD34− HSCs, donor mice were always 12–14 wk old, as it has been shown that HSCs from younger mice are mostly CD34+ 32,33. HSC populations were isolated as described previously 24 and sorted on a FACSVantage™ (Becton Dickinson). Reanalysis of sorted cells reproducibly showed a purity of 96–99%.
Limiting Dilution Assay.
Lin−Sca1+c-kit+ cells were seeded in Terasaki plates (Nunc) at a density of 1 cell/well in 20 μl IMDM (BioWhittaker) supplemented with 20% FCS (BioWhittaker) containing 100 U/ml penicillin (BioWhittaker), 100 U/ml streptomycin (BioWhittaker), 2 mM L-glutamine (BioWhittaker), 0.1 mM 2-mercaptoethanol (Sigma-Aldrich), and various cytokines. Wells were scored for cell growth following 10–12 d of incubation at 37°C in a humidified atmosphere with 5% CO2 in air. In some experiments, cells were deposited by a single cell depositor coupled to a FACSVantage™ (Becton Dickinson), and subsequently carefully visualized by microscopy to only include wells containing a single cell. The cytokine response pattern was similar with these two methods.
Semisolid Clonogenic Assay.
Lin−Sca1+c-kit+ cells were plated in duplicates in 1 ml IMDM supplemented with 20% FCS, 1.2% (final concentrations) methylcellulose (Methocel™; Fluka Chemie), and cytokines at predetermined optimal concentrations in 35-mm Petri dishes. Cultures were incubated at 37°C in a humidified atmosphere with 5% CO2 in air for 10–12 d, at which time colonies (>50 cells) were visualized and scored with an inverted microscope.
Flow Cytometric Analysis.
Cell surface phenotypings was performed on freshly isolated Lin− BM cells or Lin−Sca1+c-kit+ cells, incubated for different periods of time with cytokines at predetermined optimal concentrations in a serum-free medium: X-Vivo 15™ (BioWhittaker) supplemented with 1% detoxified BSA (StemCell Technologies, Inc.), 100 U/ml penicillin (BioWhittaker), 100 U/ml streptomycin (BioWhittaker), 2 mM L-glutamine (BioWhittaker), and 0.1 mM 2-mercaptoethanol (Sigma-Aldridge) or in IMDM supplemented as described for the limiting dilution assay above. Cells were stained with a monoclonal hamster anti–mouse Fas (Jo2) PE-conjugated antibody (BD PharMingen), and monoclonal rat anti–mouse antibodies against CD34 (FITC), Sca1 (biotin, FITC, or PE), c-kit (allophycocyanin [APC]), TNF receptors p55 and p75 (biotin; HyCult Biotechnology), and lineage-specific antigens (FITC-conjugated or purified antibodies detected with a secondary goat anti–rat PE-Tricolor™ antibody; Caltag): B220, Gr-1, Mac-1, CD8, CD4, and Ter-119 (all BD PharMingen). Cells were incubated with antibodies for 15–20 min on ice after blocking unspecific binding with hamster IgG (Jackson ImmunoResearch Laboratories) or Fc block (purified CD16/CD32 monoclonal antibody; BD PharMingen). Cells were washed, incubated with Streptavidin-PE (BD PharMingen), -PerCP (Becton Dickinson), or -CyChrome™ (BD PharMingen) to visualize biotin-conjugated antibodies, and washed again. Finally, cells were analyzed for forward light scatter, orthogonal light scatter, and the four fluorescence signals on a FACSCalibur™ (Becton Dickinson). Flow cytometry analysis was performed using CELLQuest™ (Becton Dickinson) or FloJo™ (Treestar Inc.) software.
Detection of TNFR-p55 and TNFR-p75 Transcripts.
Lin− Sca+kit+34− cells were subjected to global mRNA amplification and subsequent Southern analysis essentially as described previously 34,35. Briefly, 25 Lin−Sca+kit+34− cells were transferred into PCR tubes using an automated micromanipulator device. Subsequently, cells were lysed in first strand buffer followed by reverse transcription (RT) of mRNA using an oligo(dT)-primer. In a second step the first strand cDNA was polyadenylated by terminal transferase to generate a 5′-oligo(dT)-transcript-poly(A)-3′ cDNA that finally was amplified by PCR using a sequence independent X-(dT)24 primer. The resulting cDNA (2 μg) was run on a 1% agarose gel, transferred to a nylon membrane by capillary blotting and fixed by UV irradiation. Probes used for Southern hybridization were generated by PCR from sequence verified EST clones (TNFR-p75 GenBank/EMBL/DDBJ accession no. BE635232, forward primer 5′-AGG CTG CCA TAC CTG CCC-3′, reverse primer 5′-GGT ATG TGA GTG TGT GTG C-3′; TNFR-p55 GenBank/EMBL/DDBJ accession no. AI267050, forward primer 5′-AGC CGC CGA GGA CTG CC-3′, reverse primer 5′-AGA GTC GAG GGC TCC AGC-3′). Probes (50 ng) were radiolabeled with 32P-dCTP (Random Primers DNA labeling system; GIBCO BRL/Life Technologies), purified (QIAaquick Nucleotide removal kit; QIAGEN), hybridized to the membranes in ULTRAhyb buffer, and washed according to the manufacturer (ULTRAhyb; Ambion). After washing, blots were analyzed by PhosphorImager quantitation (Fuji Imaging Plate, Fuji Imaging Analyzer BAS-2500, Image Reader V1.4E, and Image Gauge V.3.01 software; Fuji Photo Film Europe).
Competitive Repopulation Assay for Murine BM Stem Cells.
C57BL/6 recipient mice (CD45.2) were lethally irradiated by a single exposure to 9.5 Gy of gamma irradiation from a 137Cs source (Instrument AB Scanditronix). Irradiated recipients were transplanted intravenously (0.5 ml/mouse) by tail-vein injection with 1,000 freshly isolated Lin−Sca1+c-kit+ cells (from B6.SJL mice, CD45.1) or the expansion equivalent (EE) of 1,000 Lin−Sca1+c-kit+ cells cultured in the absence or presence of TNF-α and Jo2 (or an isotype-matched control antibody). In vitro cultures were performed in two to six parallels for each group of treatment. At the day of transplantation, parallels were individually counted, and subsequently pooled, washed, and transplanted. Donor cells were cotransplanted with 200,000 unfractionated CD45.2 (recipient type) BM cells to ensure survival, but also to promote a standard competitor source of HSCs, thereby allowing a better quantification of HSC activity 36. To address the effect of TNF-α and Jo2 on single HSC, a limiting number of HSCs were transplanted, in that only six Lin− Sca1+kit+CD34−CD45.2+ cells were directly sorted (by a single cell depositor) into 0.1 ml serum-free media supplemented with SCF, IL-3 plus IL-6 with or without TNF-α and Jo2 in a U-bottomed 96-well plate. After 9 d incubation, individual wells were harvested and washed with a total of 0.5 ml PBS plus 1% FCS, mixed with 240,000 unfractionated CD45.1+ competitor BM cells, and subsequently, 0.5 ml was injected into each irradiated CD45.1+ mouse (cell numbers injected corresponded to the expansion equivalent of five Lin−Sca1+c-kit+CD34− cells and 200,000 unfractionated BM cells as competitor). All mice were kept in individually ventilated cages throughout the experiment, and given sterile food and autoclaved acidified water. Mice were bled from the retroorbital sinus venous plexus and peripheral blood (PB) analyzed 4–6 wk and 14–18 wk after transplantation for donor reconstitution. PB samples were lysed with ammonium chloride before being stained with antibodies against CD45.1, CD45.2, and lineage-specific antigens (all from BD PharMingen) and analyzed on a FACSCalibur™ (Becton Dickinson). Serial transplantations were performed by pooling BM from primary recipients 16–18 wk after transplantation and injecting 0.5 femur equivalent into each new lethally irradiated secondary recipient. Analysis of secondary recipients was performed 3 mo after secondary transplantation.
All results were expressed as the mean (SEM) of data obtained from three or more separate experiments, and else as the mean (SD). The statistical significance of differences between group means were determined using the Student's t test when n ≥ 3.
Lin−Sca1+c-kit+ Candidate Stem Cells Have Little or No Constitutive Expression of Fas and Lack Responsiveness to an Agonistic Fas-activating Antibody.
Lin−Sca1+c-kit+ cells, although constituting only 0.05–0.1% of total BM cells, have been demonstrated to contain most if not all LTRCs and represent a virtually pure population of multipotent progenitors. In agreement with previous studies 23, Lin−Sca1+c-kit+ cells lacked detectable cell surface Fas expression (Fig. 1 A). In comparison, a small fraction of Lin−Sca1−c-kit+ progenitor cells expressed low levels of Fas, whereas a larger fraction of more mature Lin−Sca1− c-kit− cells were Fas+.
In vitro clonogenic growth of Lin−Sca1+c-kit+ cells cultured in the presence of KL plus IL-3 or a combination of multiple early-acting cytokines (KL plus IL-3 plus IL-6 plus FL plus G-CSF), was not affected by stimulation with a Fas-activating antibody (Jo2; Fig. 1 B). In contrast, murine thymocytes underwent apoptosis in response to Jo2 37. Thus, Lin−Sca1+c-kit+ candidate murine BM stem cells express little or no cell surface Fas, and remain unresponsive to Fas activation after activation with growth-promoting cytokines.
Effects of In Vitro Cycling and TNF-α on Fas Expression and Fas Responsiveness of Candidate Murine Stem Cells.
As Lin−Sca1+c-kit+ cells cultured in the presence of growth-promoting cytokines remained unresponsive to Fas activation (Fig. 1 B), we next investigated whether or not Lin−Sca1+c-kit+ cells remained Fas− after cytokine stimulation. Such cytokine stimulation is associated with proliferation as well as differentiation and as expected, Fas expression increased with differentiation as assessed by acquisition of lineage-specific antigens (Fig. 2 A). In contrast, cells maintaining a Lin− phenotype after cytokine stimulation were heterogeneous with regard to Fas expression. Thus, Fas expression was also specifically investigated on cells that maintained a Lin−Sca1+c-kit+ phenotype, as virtually all short- and long-term repopulating stem cells have been demonstrated to have this phenotype 28,30,38,39. After 5 d of culture in c-kit ligand, IL-3, and IL-6 (K36), cells had expanded 54-fold, of which 12% remained Lin−Sca1+c-kit+ (Fig. 2 A; means of three experiments). Whereas >50% of Lin−Sca1−c-kit+ progenitor cells expressed Fas at high levels, only a small fraction of Lin−Sca1+c-kit+ candidate stem cells expressed Fas, and at very low levels (Fig. 2). After 9 d of incubation, only a small fraction of cells remained Lin−Sca1+c-kit+, on which Fas expression was not further upregulated when compared with day 5 (unpublished data).
As TNF-α has been demonstrated to upregulate Fas expression and induce Fas responsiveness on more committed progenitor cells 18,37,40, we next investigated whether TNF-α could upregulate Fas expression on Lin−Sca1+ c-kit+ cells. After 5 d of incubation in K36 plus TNF-α, cell numbers were 57% lower than in the absence of TNF-α. However, a lower fraction of cells became lineage positive in the presence of TNF-α and 20% of the cells remained Lin−Sca1+c-kit+ (Fig. 2 B; means of three experiments). Noteworthy, most if not all (>90%) TNF-α exposed Lin−Sca1+c-kit+ cells expressed Fas at high levels (Fig. 2 B), whereas K36 plus TNF-α–treated Lin−Sca+kit+ cells from lpr (Fas deficient) mice, remained negative for Fas (unpublished data). Interestingly, the mean geometric intensity of Fas expression was higher on Lin−Sca1+c-kit+ than Lin−Sca1−c-kit+ cells (P < 0.05). Thus, TNF-α in combination with early-acting cytokines induces Fas expression at high levels on candidate murine stem cells.
Next, Lin−Sca1+c-kit+ cells were explored for their TNF-α and TNF-α plus Fas-responsiveness when cultured in KL plus IL-3 or a cocktail of early-acting cytokines (Fig. 3 A). In agreement with previous studies 11,12, colony formation by Lin−Sca1+c-kit+ cells in response to both cytokine combinations was inhibited by TNF-α. Furthermore, and in striking contrast to cells cultured in the absence of TNF-α (Fig. 1 B), KL plus IL-3 plus TNF-α- and cocktail TNF-α-stimulated colony formation was inhibited by Jo2 by as much as 69 and 59% (Fig. 3 A), respectively. Neither Jo2 in the absence of TNF-α, or an isotype-matched control antibody in the presence of TNF-α showed any effect on colony formation (Fig. 3 A).
As freshly isolated Lin−Sca1+c-kit+ cells displayed little or no Fas expression or responsiveness, we next performed single cell experiments to establish whether stimulation with TNF-α and Jo2 would be able to block the first cell divisions of Lin−Sca1+c-kit+ cells (Fig. 3 B). Again, Jo2 alone showed no ability to reduce the number or size of clones generated in the absence of TNF-α (unpublished data). Interestingly, TNF-α did not significantly reduce the number of proliferative clones, but rather reduced their size (P < 0.01). Likewise, in the presence of TNF-α, Jo2 reduced the size but not number of Lin−Sca1+c-kit+ clones (P < 0,005; Fig. 3 B), suggesting that TNF-α– and Fas-responsive Lin−Sca1+c-kit+ cells can undergo a limited number of cell divisions before becoming sensitive to TNF-α and Fas-induced growth inhibition. In further support of this, delaying addition of Jo2 for 20–92 h only slightly reduced Jo2-induced growth inhibition (Fig. 3 B). However, when Jo2 was added after 164 h, almost no Jo2-induced growth suppression was observed.
The delayed addition experiments suggested a lag time between TNF-α stimulation and acquisition of Fas responsiveness, which potentially could reflect a requirement for upregulation of TNF receptors in response to K36-induced cycling. In steady-state BM of adult mice, almost all LTRCs reside in the CD34− fraction, and Lin−Sca1+c-kit+CD34− cells represent a virtually pure population of HSCs 33,41. RT-PCR analysis of 25 highly purified Lin−Sca1+c-kit+CD34− cells demonstrated mRNA expression of both TNF receptors (Fig. 4 A). However, only a small fraction of freshly isolated Lin−Sca1+CD34- cells had detectable cell surface expression of TNF receptors (Fig. 4 B). In response to cytokine-induced expansion, LTRCs become CD34+ 42, and after 20 h of K36 stimulation (at which time point virtually no LTRCs have divided; unpublished data), TNF receptor expression was dramatically upregulated on Lin−Sca1+c-kit+CD34+ cells (86%; Fig. 4 B). Moreover, after 3 d of K36 stimulation, virtually all Lin−Sca1+c-kit+CD34+ cells expressed TNF receptors. Thus, TNF receptor expression is highly upregulated on candidate stem cells upon cytokine-induced cycling.
TNF-α and Fas Receptor Activation Compromise the Reconstituting Ability of Lin−Sca1+c-kit+ Stem Cells.
Although highly enriched in LTRC activity, Lin−Sca1+c-kit+ cells are heterogenous, and represent in part cells with only short-term repopulating activity 43. Thus, only in vivo reconstitution experiments could conclusively address whether Fas- and TNF-α–responsive progenitors included true LTRCs. Furthermore, as Lin−Sca1+c-kit+ cells can undergo a limited number of cell divisions in the presence of TNF-α and Fas receptor activation, it remained unclear whether this activation would positively or negatively affect HSC function.
First, to investigate whether Lin−Sca1+c-kit+Fas+ cells generated in in vitro cultures included LTRCs, we conducted in vivo reconstitution experiments in which Lin−Sca1+c-kit+ cells were cultured for 5 d with K36 with or without TNF-α. Remaining Lin− cells were subsequently sorted into Fas+ and Fas− subpopulations (Fig. 5A and Fig. B) and investigated for their ability to multilineage reconstitute lethally irradiated recipients. From cultures lacking TNF-α, all long-term reconstituting activity was derived from Fas− cells. This was in striking contrast to cells cultured in the presence of TNF-α, in which the Fas+ fraction was responsible for all LTRC activity (Table). Thus, TNF-α efficiently induces high levels of Fas expression on LTRCs.
We have recently demonstrated that Lin−Sca1+c-kit+ cells cultured under serum-free conditions in the presence of K36, undergo self-renewing cell divisions and preserve their ability to short- and long-term multilineage reconstitute lethally irradiated mice 24,25. Thus, Lin−Sca1+ c-kit+ cells were cultured under such self-renewing conditions in the absence or presence of TNF-α and/or Jo2. After 3, 5, and 9 d of incubation in K36, TNF-α reduced the cellular expansion by 14, 46, and 43%, respectively, whereas the addition of Jo2 only slightly reduced cell numbers further (33, 56, and 69% reduction after 3, 5, and 9 d respectively).
Lethally irradiated mice were transplanted with 1,000 freshly isolated Lin−Sca1+c-kit+ CD45.1+ cells or their expansion equivalents (EEs) after 3, 5, and 9 d of culture (in competition with 200,000 CD45.2+ BM cells; Fig. 6). 3–9 d exposure to TNF-α reduced the short-term (4–6 wk) reconstituting activity of cultured Lin−Sca1+c-kit+ cells by 40% (Fig. 6), whereas Lin−Sca1+c-kit+ cells incubated for 3, 5, or 9 d in the absence of TNF-α showed no drop in reconstituting activity in response to Jo2 (unpublished data). The level of short-term engraftment of cells cultured in the presence of TNF-α plus Jo2 was reduced by as much as 87–95% after culture (Fig. 6). Similarly, TNF-α reduced long-term (14–18 wk) reconstituting activity of cultured Lin−Sca1+c-kit+ cells by 63–71%, and TNF-α plus Jo2 by 90–98% (Fig. 6).
The ability to self-renew in vivo is an important characteristic of true LTRCs, best established through their ability to multilineage reconstitute lethally irradiated serially transplanted recipients 44,45. Thus, cells from primary recipients were transplanted into secondary recipients (Table). Reductions in reconstituting activity were observed for cells originally cultured in the presence of TNF-α, and exposure to TNF-α + Jo2 virtually abrogated the ability to reconstitute secondary recipients (Table). The relative contribution of TNF-α and/or Jo2-exposed Lin−Sca1+ c-kit+ cells toward reconstitution of various blood cell lineages (B, T, and myeloid) did not differ from that of uncultured or K36-cultured cells (Table). Thus, exposure of Lin−Sca1+c-kit+ stem cells to TNF-α and Fas activation severely compromise their ability to short- and long-term reconstitute the hematopoietic system.
Finally, to address the direct effect of TNF-α and Jo2 on individual HSC, limiting numbers of Lin−Sca1+c-kit+ CD34− HSC were used to study the effect of TNF-α and Jo2 in mice primarily repopulated by single LTRCs, also allowing a more accurate assessment of loss of LTRC activity in response to TNF-α and Jo2. Out of a total of 19 recipients transplanted with 5 Lin−Sca1+c-kit+CD34− cells stimulated for 9 d with K36, 12 (63%) were found to sustain long-term multilineage donor derived reconstitution (Table). In contrast, of mice transplanted with the same number of cells exposed to K36 plus TNF-α plus Jo2, only 1 out of 17 (6%) showed multilineage reconstitution. These data suggested that ∼20% of K36-cultured Lin−Sca1+ c-kit+CD34− cells sustained long-term reconstituting activity, which was reduced to 1–2% if cultured in the presence of TNF-α plus Jo2. Thus, TNF-α and Fas activation severely compromise the ability of Lin−Sca1+c-kit+CD34− HSC to short- and long-term multilineage reconstitute the hematopoietic system. In agreement with this, Lin− Sca1+c-kit+ cells cultured for 9 d in K36 and TNF-α showed a dramatic reduction in cells with a Lin−Sca1+ c-kit+ HSC phenotype, which were virtually eliminated upon activation with Jo2 (Fig. 7).
The negative effect of TNF-α on LTRC self-renewal might potentially be mediated through its ability to induce Fas expression on HSC, thus rendering HSCs responsive to Fas activation during/after transplantation. However, TNF-α–exposed Lin−Sca1+c-kit+ cells purified from lpr mice (defective in Fas receptor expression) were equally reduced in their long-term repopulating ability as wild-type cells, suggesting that TNF-α–induced reduction in LTRC activity occurs independently of the Fas pathway (Fig. 8).
Previous studies have demonstrated that cytokines such as TNF-α, macrophage inhibitory protein 1α (MIP-1α), and TGF-β can inhibit the in vitro growth of primitive hematopoietic progenitors 8,46,47. Such an effect of MIP-1α is however associated with enhanced rather than reduced maintenance of candidate HSCs in culture 48,49. TGF-β has been shown to accelerate depletion of HSCs under in vitro culture conditions which themselves result in loss of HSC activity 50,51, making it difficult to distinguish between effects on self-renewal and other mechanisms. Recent developments have now made it possible to study HSC self-renewal and its regulation in vitro 24,25,26,27. Through such studies, a number of growth stimulatory cytokines have been shown to promote HSC self-renewing cell divisions. Using the same approach, we here demonstrate that TNF-α and Fas can negatively regulate the self-renewal of highly purified HSCs.
TNF receptors and Fas which are members of the same receptor superfamily, have been demonstrated to be nonredundant regulators of the immune system 52,53,54. Ample evidence also implicate their involvement in the pathophysiology of a number of blood formation disorders associated with enhanced apoptosis, such as myelodysplastic syndromes and aplastic anemia 15,16,17,19,21,22,55 as well as in acute GVHD 9,56.
Whereas TNF-α and Fas have been convincingly demonstrated to mediate growth suppression of in vitro clonogenic progenitor cells 8,10,11,12,18,40, their receptor expression pattern and potential effects on LTRCs has been unknown. Recently, CD27, another member of the TNF family, was demonstrated to be expressed on HSCs 57, but the biological function of CD27 on HSCs remains to be investigated. The present studies unequivocally demonstrate that self-renewing murine BM HSCs can be targeted by TNF-α and Fas, and consequently become severely compromised in their ability to short- and long-term multilineage reconstitute hematopoiesis.
Whereas TNF-α has been clearly implicated to be involved in a number of BM failure syndromes, the potential role of TNF-α in regulating steady-state hematopoiesis is unclear. Although studies in p55 TNF receptor–deficient demonstrated increased numbers of Lin−Sca1+c-kit+ cells and colony forming unit spleen day 12 11,13, older (but not younger) mice revealed reduced HSC numbers and function, suggesting a complex role of TNF-α in regulating HSCs. The interpretation of this finding is further complicated by the existence of two TNF receptors, and the HSC compartment of mice deficient in both TNF receptors has yet to be investigated. In addition, the effects of TNF-α on hematopoiesis are pleiotropic and context dependent 8, and whereas low (physiological) levels of TNF-α in many systems have been demonstrated to be crucial, high levels can be detrimental.
The present findings suggests that TNF-α is likely to have little or no suppressive effect on quiescent HSC dominating during steady-state hematopoiesis, as supported by low levels of TNF receptor expression and lack of TNF-α suppression of noncycling HSCs. In contrast, under in vitro cycling conditions, TNF receptor expression was upregulated and HSCs became sensitive to TNF-α–induced suppression. In support of this, TNF-α has been shown to stimulate proliferation of mitotically quiescent hematopoietic cell lines, but suppress their growth when rapidly cycling 58. The negative effect of TNF-α on in vitro HSC self-renewal was found to be independent of the Fas pathway, as Lin−Sca1+c-kit+ cells from lpr mice, defective in Fas receptor expression, were equally sensitive to TNF-α–induced suppression of HSCs. Although the lpr mice have a somewhat leaky Fas deficiency, no Fas expression was observed on Lin−Sca+kit+ cells exposed to TNF-α.
In a previous study, Lin−Sca1+Thy1.1+ HSCs were found to lack expression of Fas 23. In agreement with this, freshly isolated Lin−Sca1+c-kit+ BM cells in the present studies also lacked detectable cell surface expression of Fas. When induced to cycle by combinations of early-acting cytokines, Fas expression was upregulated but only marginally on cells which maintained a Lin−Sca1+c-kit+ HSC phenotype, and in vitro clonogenic and in vivo short- and long-term reconstituting Lin−Sca1+c-kit+ cells remained unresponsive to activation by an agonistic anti-Fas antibody. This unresponsiveness was not altered by using an anti–hamster IgG antibody to further enhance cross-linking (unpublished data). The difference in Fas responsiveness of HSCs and myeloid progenitors 10 could in part be explained by HSCs expressing higher levels of apoptosis-inhibiting genes such as Bcl-2 59. In fact, when Bcl-2 is overexpressed, committed myeloid progenitors show enhanced Fas resistance 10. Noteworthy, Lin− Sca1+c-kit+ cells became highly Fas-responsive only when costimulated with cytokines promoting cycling and TNF-α, whereas TNF-α alone had little or no ability to induce Fas responsiveness (unpublished data). In support of a key role of TNF-α activation, HSC activity was retained entirely in the Fas negative fraction after in vitro expansion in the absence of TNF-α, whereas in the presence of TNF-α, LTRCs were found exclusively in the Fas-positive fraction. Thus, it appears that the ability of TNF-α to induce Fas responsiveness of HSCs is at least in part a consequence of upregulation of Fas expression on cycling HSCs, which constitutively express no Fas. The ability of TNF-α to upregulate Fas expression has been demonstrated in a number of other systems, and to be critically dependent on nuclear factor (NF)-κB activation 60,61,62. It is possible that TNF-α might also provide an important co activation signal for Fas activation in HSCs, as implicated for other cell types 63,64,65.
It is noteworthy that the reduction in HSC numbers (more than 10-fold) in response to TNF-α and Fas activation was much more dramatic than reductions in total cell numbers (threefold). A preferential targeting of HSCs was further supported by the preferential upregulation of Fas and TNF receptors on Lin−Sca1+c-kit+ cells, and the virtual elimination of such cells following prolonged TNF-α and Fas activation.
Whereas HSCs can expand in vivo, extensive efforts at promoting stem cell expansion in vitro have proven less successful 1,66. One reason for this could be the tendency of in vitro cultured stem cells to undergo apoptosis 1. Based on our findings it is obvious that HSCs will upregulate TNF receptor expression after in vitro stem cell expansion with potent growth-promoting cytokines. If such expanded stem cells are subsequently exposed to TNF-α and/or Fas ligand in vitro, or more likely in the BM microenvironment of ablated recipients, the consequence might be an engraftment defect, as frequently observed after transplantation of ex vivo expanded stem cells 67.
In conclusion, we have demonstrated that although HSCs constitutively express little or no Fas, enforced cycling of HSCs combined with TNF-α activation upregulate Fas expression on HSCs, making them susceptible to Fas-induced suppression. This combined with ample evidence of the involvement of excessive TNF-α and Fas activity in various BM failure syndromes and GVHD 9,56, suggest that self-renewing HSCs are likely targets for TNF-α– and Fas-induced BM suppression.
The authors gratefully acknowledge Yutaka Sasaki for valuable discussions, Per Anders Bertilsson, Sverker Segrén, and Zhi Ma for expert assistance with cell sorting, and Jonas Björk for providing advice with regard to statistical analysis. We thank Anna Fossum, Ingbritt Åstrand-Grundström, Ineke de Jong, Lilian Wittman, Eva Gynnstam, and Irene Persson for technical assistance and animal care. We thank Helga Björgvinsdôttir for critical reading of the manuscript. We are also grateful to Drs. Ian K. McNiece, Graham Molineux, and Stewart D. Lyman for generously providing cytokines.
These studies were generously supported by grants from: ALF (Government Public Health Grant); the Berta Kamprad Foundation; the Crafoord Foundation; the Georg Danielsson Foundation; the Greta and Johan Kocks Foundation; the Gunnar, Arvid and Elisabeth Nilsson Foundation; the Harald and Greta Jeansson's Foundation; the Thelma Zoega's Foundation; the John and Augusta Persson Foundation; the Medical Faculty, University of Lund; the Swedish Medical Research Council (MFR); the Svensson Siblings Foundation; the O and E and Edla Johansson Foundation; the Royal Physiographic Society in Lund; the Swedish Foundation for Strategic Research; the Swedish Cancer Society; the Swedish Society of Pediatric Cancer; and the Tobias Foundation.
D. Bryder and V. Ramsfjell contributed equally to this work.
Abbreviations used in this paper: BM, bone marrow; GVHD, graft versus host disease; HSC, hematopoietic stem cell; LTRC, long-term repopulating cell; PB, peripheral blood.