TNF-α/Fas-RIP-1–induced cell death signaling separates murine hematopoietic stem cells/progenitors into 2 distinct populations

In hematopoietic stem and progenitor cells (HSPCs), the survival and death machinery is tightly controlled by a complex interplay between intrinsic signals and stimuli from the surrounding bone marrow (BM) microenvironment, inducing a dynamically balanced network of pro-survival and anti–survival influences. Disruption of this balance can result in hematopoietic disorders, such as myeloproliferative disorders and bone marrow failure (BMF).^1-5 

Prosurvival signaling in HSPCs has been very well studied. Genetic studies clearly demonstrate the role of the survival machinery in regulating numbers and functions of HSPCs.^1-4  Transgenic overexpression of Bcl-2, an important pro-survival gene, increases HSPC numbers and improves hematopoietic reconstitutive capacity in mice. Mice with deletions of pro-survival genes, such as Bcl-xl, Mcl-1, or survivin, develop severe BMF because of the apoptotic loss of hematopoietic cells, including HSPCs.^1-4  Importantly, the expression of these pro-survival genes is stimulated by hematopoietic cytokines in HSPCs. Hematopoietic cytokines, such as SCF, IL-3, thrombopoietin, IL-6, and IL-11 stimulate the survival of HSPCs as shown by both in vitro culture and genetic studies. Via Jak/Stat, Raf/Mapk, and PI3K/Akt, multiple signaling pathways can be activated by hematopoietic cytokines in HSPCs, resulting in self-renewal, proliferation, differentiation, and survival. Among these, the PI3K/Akt pathway has been proposed to be the mediator of the major survival signal.⁶  On activation, PI3K inactivates proapoptotic signals, such as Bad and FoxO3 via a serine/threonine phosphorylation cascade,^7,8  or stimulates the activation of pro-survival signals, such as Nf-κB,⁹  which in turn up-regulate the expression of pro-survival genes.

Anti–survival signaling in HSPCs is still not well understood. Previous studies suggested that proinflammatory cytokines, including TNF-α,¹⁰  IL-1β,¹¹  IFN-α/γ, as well as FasL (Fas ligand), exercise negative regulation over hematopoiesis and thus were implicated in a number of BMF syndromes.^12-15  However, recent studies suggested that these proinflammatory cytokines stimulate dual signals for HSPC functions.^16-19  A negative signal provokes programmed cell death and/or differentiation by inducing the production of reactive oxygen species (ROS) and by activating cell-extrinsic apoptotic signaling, mediated by caspase-8.^20,21  A positive signal, however, promotes proliferation and survival by inducing the activation of TGF-β–activated kinase-1(Tak1)–mediated signaling.^22-25  Therefore, under normal homeostatic conditions, because of high levels of Tak1 activity, HSPCs are relatively resistant to proinflammatory cytokine-induced cell death.^16,26  The role of apoptotic cell death induced by these factors in BMF syndromes remains somewhat ambiguous. In addition, recent studies have demonstrated that some of these proinflammatory factors also induce a receptor-interacting protein-1 (RIP-1)–dependent programmed cell death called necroptosis in other cell types. It was suggested that RIP-1 and caspase-8 exist within the same protein complex, called a ripoptosome, induced by inflammatory factors that functionally antagonize each other. Caspase-8 represses necroptosis by degrading RIP-1 and stimulates type I or type II apoptosis by directly activating caspase-3 or by promoting mitochondrial-dependent caspase-9 activation. Caspase-8–induced apoptosis can be converted into RIP-1-dependent necroptosis when caspase-8 activity is repressed.^27-32  The role of RIP-1–dependent necroptosis in hematopoietic cells has not been studied.

Tak1 is a member of the Map3k family of enzymes and is a key mediator of inflammatory cytokine-induced survival signals.³³  On activation, Tak1 phosphorylates downstream signaling molecules, including Ikkβ, Mkk3/6, and Mkk4/7; these in turn activate Nf-κB, p38, and c-Jun N-terminal kinase (Jnk) signaling cascades, respectively.³³  As a consequence, Tak1 signaling is involved in regulating cell proliferation and survival by inducing the production of several cytokines and the expression of many survival genes (including Bcl-xl, Mcl-1, and cIAP). Gene knockout studies done in animals suggested that Tak1 mediates a major survival signal in almost all types of cells studied, including skin keratinocytes,³⁴  lymphocytes,³⁵  intestinal enterocytes,³⁶  and mouse embryonic fibroblasts.³⁷  This Tak1-mediated survival signaling plays an essential role in protecting these cells from apoptosis induced by inflammatory cytokines and other unfavorable insults. Tak1 deletion leads to excessive activation of these cells in response to inflammatory cytokine stimulation, resulting in deregulation of the inflammatory reaction and apoptotic injury in corresponding tissues. We have found that Tak1 is also essential for the survival of hepatocytes and hematopoietic cells, including HSPCs.³⁸  Interferon-inducible Tak1-knockout mice (MxCre⁺Tak1^fx/fx, Tak1^−/− hereafter) developed severe BMF and liver failure because of the massive death of their mutant HSPCs and hepatocytes after induction of Tak1 deletion.³⁸ 

To study the molecular mechanisms underlying BMF syndromes in our inducible Tak1^−/− mice and also to investigate the factors responsible for inducing the death of Tak1^−/− HSPCs, we generated Tak1/Tnfr1, Tak1/Tnfr2, Tak1/Tnfr1/r2, Tak1/Fas, and Tak1/Tnfr1/r2/Fas compound-mutant mice to examine whether inactivation of TNF-α signaling, Fas signaling, or the simultaneous inactivation of both of these was able to prevent the BM damage seen in Tak1^−/− mice, and if so, to what degree. In addition, using in vitro assays, we also examined other potential causes of cell death in Tak1^−/− HSPCs. We found that death signaling induced by TNF-α and Fas, via activation of RIP-1, accounted for only 30% to 40% of HSPC loss in Tak1^−/− mice (based on CFU data). Other unknown factor(s), acting through the caspase-8–dependent extrinsic apoptotic pathway, induce the death of the remaining 60% to 70% of Tak1^−/− HSPCs.

All experiments using mice were performed according to the guidelines of Loyola University Medical Center and were approved by the Loyola University Institutional Animal Care and Use Committee. Mx1cre⁺Tak1^fx/fx mice^37,38  were crossed with Fas^−/− mice (B6.MRL-Fas^lpr/J from The Jackson Laboratory) and Tnfr1^−/−r2^−/− mice (B6.129S-Tnfrsf1a^tm1Imx Tnfrsf1b^tm1Imx/J from The Jackson Laboratory), respectively, to generate heterozygous compound-mutant mice. These heterozygous compound-mutant mice were further crossed to produce Mx1cre⁺Tak1^fx/fxFas^−/−, Mx1cre⁺Tak1^fx/fxTnfr1^−/−, Mx1cre⁺Tak1^fx/fxTnfr2^−/−, Mx1cre⁺Tak1^fx/fxTnfr1^−/−r2^−/− mice. Mx1cre⁺Tak1^fx/fxFas^−/− and Tak1^fx/fxTnfr1^−/−r2^−/− mice were further crossed to generate Mx1cre⁺Tak1^fx/fxTnfr1^−/−r2^−/−Fas^−/− mice. Genotypes of all mice were determined by PCR assay using primers listed in supplemental Figure 1 (available on the Blood Web site; see the Supplemental Materials link at the top of the online article). At 6 to 8 postnatal weeks, all mice, including WT controls (Cre⁻) and heterozygous compound-mutant mice, were injected with 5 μg/g weight polyriboinosinic acid/polyribocytidylic acid (poly I:C; GE Healthcare Life Sciences) every other day for a total of 3 injections to induce deletion of the Tak1 gene.

Mice were killed 7 to 8 days after the first poly I:C injection. Peripheral blood (PB), spleens, and BM were collected. PB were analyzed for WBC, platelet, and RBC counts by Hemavet 950FS (Drew Scientific). After RBC lysis, mononucleated cells (MNCs) in PB, spleens, and BM samples were counted and stained with surface markers for flow cytometric analysis as described previously.³⁸ 

Femurs from different genotypes of mice were fixed in zinc formaldehyde at room temperature for 3 days. Femurs were then decalcified by 10% EDTA buffer for 2 weeks; tissues were then transferred into 70% ethanol until embedding. Embedding and cutting of sections were performed according to standard protocols from the Pathology Department, Loyola University Medical Center. Slides were then stained with H&E.

Lipopolysaccharide (LPS), as well as ROS scavengers N-acetyl-L-cysteine and butylated hydroxyanisole, were purchased from Sigma-Aldrich. RIP-1 inhibitor necrostatin-1 (Nec-1) was purchased from Santa Cruz Biotechnology. IL-3, IL-6, and SCF were purchased from eBioscience. TNF-α, IL-1β, IFN-α, FasL, TRAIL, and IFN-γ were purchased from BD Biosciences.

High-titer retrovirus was produced by cotransfecting Phoenix cells with a retroviral vector containing the indicated genes together with packaging vectors using Calphos Mammalian Transfection Kit (Clontech). Retroviral supernatants were harvested 24 and 48 hours after transfection. Retroviral plasmids used in this study include: MSCV-Cre-GFP and MSCV-Cre-YFP vectors; MSCV-Bcl-xl-GFP, MSCV-cIAP1-GFP, and MSCV-SOD-GFP vectors, which were generated by subcloning Bcl-xl, cIAP, and SOD genes from pMIG-Bcl-xl, pcdna3.1-hciap1, and pF151-pcDNA3.1(+)SOD1WT plasmids (Addgene) into the MSCV-GFP retroviral vector; MSCV-Tak1-GFP vector, which was generated by subcloning the Tak1 gene from pCMV-HA-Tak1 plasmid³⁹  into the MSCV-GFP vector; and MSCV-Δ-caspase-8-GFP, MSCV-Δ-caspase-9-GFP, and MSCV-Δ-FADD-GFP vectors (kindly provided by Dr Strasser of the Walter and Eliza Hall Institute of Medical Research).⁴⁰ 

The c-kit⁺ HSPCs from indicated genotypes of mice were enriched by EasySep Mouse CD117 Positive Selection Kit (StemCell Technologies). HSPCs were cultured in 6-well plates in HSPC culture medium consisting of RPMI 1640 supplemented with 10% FBS, 100 ng/mL mSCF, 50 ng/mL mIL-6, and 10 ng/mL mIL-3 and were infected with high-titer retrovirus. Infected cells were either incubated in HSPC medium for analysis of changes in the percentage of GFP⁺ cells or were purified by FACS for CFU and apoptosis analysis as described in Figures 3, 5 and 6.

BM MNCs directly isolated from the indicated genotypes of mice or purified infected HSPCs were seeded into MethoCult GF M3434 medium (StemCell Technologies) and incubated at 37°C, 100% humidity, and 5% CO₂ for 10 days. Numbers of colonies were counted according to the manufacturer's instructions.

Infected c-kit⁺ HSPCs were purified by FACS and incubated in HSPC medium for 24 hours. Cells were then stained with allophycocyanin-conjugated annexin V followed by 7-amino-actinomycin D (7-AAD) staining in binding buffer following the manufacturer's instructions (BD Biosciences). Death of infected cells was examined for the percentages of annexin V⁺ and annexin V⁺/7AAD⁺ cells by flow cytometry.

All data were analyzed using the 2-tailed Student t test with GraphPad Prism Version 5.04 software to identify significant differences between groups. Values are expressed as means ± SEM. Differences were regarded as significant at P < .05.

TNF-α acts by binding to and activating its specific receptors. Two TNF-α–specific receptors, Tnfr1 (p55) and Tnfr2 (p75), have been identified.⁴¹  To investigate whether the BMF phenotype of Tak1^−/− mice can be prevented by TNF-α signal inactivation, we generated Mx1Cre⁺Tak1^fx/fxTnfr1^−/−, Mx1Cre⁺Tak1^fx/fxTnfr2^−/−, and Mx1Cre⁺Tak1^fx/fxTnfr1^−/−r2^−/− compound-mutant mice (Tak1^−/−Tnfr1^−/−, Tak1^−/−Tnfr2^−/−, and Tak1^−/−Tnfr1^−/−r2^−/− hereafter, respectively). Genotypes of the mice were verified by PCR assay, as described in supplemental Figure 1. We previously found that mice with heterozygous Tak1 deletion did not display phenotypic hematopoietic changes.³⁸  Therefore, we focused on Tak1 homozygous knockouts in our studies. Six to 8 weeks after birth, all mice, including Cre-negative Tak1^fx/fx wild-type controls, were injected with poly I:C to induce Tak1 gene deletion. High efficiency Tak1 gene deletion was confirmed in hematopoietic cells from Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice by PCR assay (supplemental Figure 1). We found that, as was the case with Tak1^−/− mice, Tak1^−/−Tnfr1^−/− mice, Tak1^−/−Tnfr2^−/− mice, and most Tak1^−/−Tnfr1^−/−r2^−/− mice died on days 8 to 10 after injection. Only 2 of 10 Tak1^−/−Tnfr1^−/−r2^−/− mice were able to survive up to 15 and 20 days, respectively, and these showed hunched backs and significantly reduced body weights. Substantial hematopoietic cell infiltration was observed in the livers of these 2 surviving mice when examined on day 15 and day 20, respectively, suggesting excessive inflammatory reactions. However, the death of hepatocytes was obviously reduced compared with Tak1^−/− mice (supplemental Figure 2). In addition, we found that BM from these 2 mice was hypocellular, as shown by a reduction in the number of total MNCs (1.2 × 10⁷and 0.96 × 10⁷ respectively, approximately one-third of WT mice). We speculated that the Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice might die of super-inflammatory reactions induced by overproduced inflammatory cytokines, such as IFN-γ (Figure 1G).

To compare changes in hematopoiesis, mice were killed on day 8 after the first poly I:C injection. Histologic sections showed a significant disruption in BM structure and a hypocellular phenotype for Tak1^−/−Tnfr2^−/− mice, which was comparable to that of Tak1^−/− mice. This Tak1^−/− BM defect was substantially prevented in Tak1^−/−Tnfr1^−/− mice and was improved still further in Tak1^−/−Tnfr1^−/−r2^−/− mice (Figure 1A). Consistent with this finding, we also found that the number of MNCs in BM and spleens of Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice was significantly increased compared to Tak1^−/− and Tak1^−/−Tnfr2^−/− mice (Figure 1B-C). As a consequence, WBC counts and reticulocyte percentages were also significantly increased in Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice compared with Tak1^−/− and Tak1^−/−Tnfr2^−/− mice (Figure 1D,F). Interestingly, the number of platelets in Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice was comparable to what was seen in Tak1^−/− and Tak1^−/−Tnfr2^−/− mice, suggesting that the defects in Tak1 deletion-related thrombopoiesis were not prevented by the inactivation of TNF-α signaling (Figure 1E). It needs to be noted here that, although the BM structure and cellularity of Tak1^−/−Tnfr1^−/−r2^−/− mice were much improved compared with Tak1^−/−Tnfr1^−/− mice, as shown by histologic section studies (Figure 1A), still, because of the relatively smaller body size of Tak1^−/−Tnfr1^−/−r2^−/− mice, the absolute numbers of MNCs in BM and spleens of these mice are still comparable with those in Tak1^−/−Tnfr1^−/− mice. These data suggested that TNF-α–induced signaling contributes significantly to the BMF phenotype of Tak1^−/− mice. Both Tnfr1 and Tnfr2 are involved in mediating TNF-α–induced repressive signaling in hematopoietic cells. Consistent with previous studies, we found that Tnfr1 is the major mediator of TNF-α–induced signaling in hematopoietic cells.^10,41-43 

To investigate whether the depletion of Tak1^−/− HSPCs could also be prevented by inactivating TNF-α signaling, we analyzed the percentages and absolute numbers of phenotypic HSCs (lineage⁻c-kit⁺Sca1⁺) and committed hematopoietic progenitors (CPs, lineage⁻c-kit⁺Sca1⁻) in the BM of mutant mice on day 8 after the induction of Tak1 deletion. We found that the percentages of HSCs and CPs were significantly increased in Tak1^−/−Tnfr2^−/− mice compared with Tak1^−/− mice. Such percentages were further increased in Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice, although they had still not returned to normal levels. However, no significant difference was observed between Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice. Consistently, the absolute numbers of HSCs and CPs were significantly increased in Tak1^−/−Tnfr2^−/− mice compared with those in Tak1^−/− mice. These numbers were significantly further increased in Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice; no significant difference was observed between the latter 2 types of mice. The absolute numbers of HSCs and CPs in Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice were only 25% to 35% of those observed in WT mice (Figure 2).

Out of concern that increased inflammatory cytokines, such as IFN-γ, might affect hematopoietic phenotype, to study the role of Tnfr1 and Tnfr2-mediated signaling in the survival of Tak1^−/− HSPCs, we isolated c-kit⁺ HSPCs (including HSCs and CPs) from Tak1^fx/fxTnfr1^−/−, Tak1^fx/fxTnfr2^−/−, and Tak1^fx/fxTnfr1^−/−r2^−/− mice and infected them in vitro with MSCV-Cre-GFP retrovirus to induce Tak1 deletion. c-kit⁺ HSPCs from Tak1^fx/fx mice were infected with MSCV-Cre-GFP and MSCV-GFP virus (empty vector only) separately and were studied in parallel as positive and negative controls, respectively. Infected cells were cultured in HSPC culture medium with daily medium changes (Figure 3A). The percentages of GFP⁺ cells (representing infected cells) were examined at the indicated time points. These percentages from each sample were first normalized to the percentages of GFP⁺ cells at 12 hours in the same sample to obtain a ratio of increased or decreased GFP⁺ cells in each sample at the indicated time points. These ratios from each sample were further normalized to the ratio from the MSCV-GFP–infected control group samples at the corresponding time points. This then set the ratio of GFP⁺ cells at 1 for the 12 hour time point for all experimental samples, and the ratio of GFP⁺ cells at 1 at all time points for MSCV-GFP control samples. The capacity of Tnfr1 and r2 deletions to prevent the depletion of Tak1^−/− HSPCs was evaluated by examining the dynamic changes in GFP⁺ cell ratios in infected cells (Figure 3B). We found that this dynamic analysis of infected cells ratio assay was very sensitive, reliable, and reproducible in studying gene transduction-related cell loss (such as Cre transduction-related depletion of Tak1^−/− HSPCs in this study). As expected, the ratio of GFP⁺ cells in Cre-infected Tak1^fx/fx HSPCs was reduced rapidly during culturing, whereas this reduction in GFP⁺ cell ratio was significantly retarded by deletion of Tnfr2, Tnfr1, or both of these simultaneously. The ability of Tnfr1 deletion to prevent depletion of Tak1^−/− HSPCs was significantly less than that observed for the deletion of both Tnfr1 and r2 but was significantly greater than for the deletion of Tnfr2 only. However, this ratio in the Tak1^fx/fxTnfr1^−/−r2^−/− group was still significantly reduced compared to the MSCV-GFP control group. These data were consistent with in vivo data, supporting the conclusion that inactivation of TNF-α signaling can only partially prevent the death of Tak1^−/− HSPCs. Both Tnfr1 and r2 contribute to the TNF-α–induced depletion of Tak1^−/− HSPCs (Figure 3B). In agreement with previous studies,^43,44  we found Tnfr1 to be the major mediator of TNF-α signaling in HSPCs. This conclusion was further confirmed by CFU assays (Figure 3C).

To study whether TNF-α signaling inactivation can rescue functional HSPCs from Tak1 deletion-induced depletion, we used FACS to purify the GFP⁺ cells from the MSCV-Cre-GFP–infected BM HSPCs with indicated genotypes 16 hours after infection. Purified cells were seeded into methylcellulose medium for CFU assay. Because exactly the same number of infected cells was sorted by FACS in each sample for CFU assay, the accuracy of experimental results (accuracy of cell numbers seeded) was assured. We found that deletion of Tnfr1, r2, or both, as well as Cre overexpression, did not affect the colony-forming ability of Tak1-intact HSPCs (supplemental Figure 3). However, deletion of Tnfr1 and r2 was able to rescue 20% to 30% and 5% to 10% of the colony-forming capacity of Tak1^−/− HSPCs, respectively. Such rescue was further enhanced by the codeletion of both Tnfr1 and r2 (30%-40% of MSCV-GFP-infected control HSPCs; Figure 3C). The phenomenon of prevention of cell death in Tak1^−/− HSPCs by inactivated TNF-α signaling is mainly accomplished through the repression of cell death, as shown by annexin V staining (Figure 3D).

To study whether HSPCs from Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− mice have multipotent differentiation and long-term engraft-ment capacities, BM MNCs were isolated from Tak1^−/−Tnfr2^−/−, Tak1^−/−Tnfr1^−/−, and Tak1^−/−Tnfr1^−/−r2^−/− mice (CD45.2⁺) before Tak1 deletion was induced. Cells for each group were mixed at a 10:1 ratio with support BM MNCs from WT mice (CD45.1⁺). These mixed BM cells were transplanted into lethally-irradiated recipient mice, each mouse receiving 2 × 10⁶ mutant MNCs and 2 × 10⁵ WT support MNCs. BM MNCs from WT (Cre⁻) littermates and Tak1^−/− mice were transplanted in parallel as positive and negative controls, respectively. One month after transplantation, the contribution of donor-derived cells (CD45.2⁺) in the PB of recipient mice was examined. Recipient mice were then injected with poly I:C (every other day) for a total of 3 injections to induce Tak1 gene deletion. The percentages of donor-derived cells in thePB of recipient mice were examined at the indicated time points (Figure 4A).

As was expected, the percentage of donor cells in WT control-transplanted mice remained consistent at > 90%, whereas this percentage was significantly reduced in mice receiving Tak1^−/− BM cells. We found that donor-derived cells in mice receiving Tak1^−/−Tnfr1^−/−, Tak1^−/−Tnfr2^−/−, and Tak1^−/−Tnfr1^−/−r2^−/− BM cells were also reduced significantly compared with WT control transplantation. However, this reduction in the Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− groups was significantly less than that for Tak1^−/−transplanted animals, whereas the Tak1^−/−Tnfr2^−/− group data were comparable with those for the Tak1^−/− group. The reduction in donor-derived cells observed in all experimental groups compared with WT controls before poly I:C injection was due to the spontaneous deletion of the Tak1 gene caused by Mx1Cre leakage. This result further supports the suggestion of a partial rescue of the depletion of Tak1^−/− HSPCs by inactivation of TNF-α signaling. The donor-derived cells of Tak1^−/−Tnfr1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− compound-mutant mice were maintained for more than 3 months in recipient mice, at which time the Tak1^−/− and Tak1^−/−Tnfr2^−/− transplantations were no long detectable (Figure 4B). Cell surface marker staining suggested that the compound-mutant HSCs had multipotent differentiation capacity with a minor T lymphocyte bias (Figure 4C).

To search for other factors that might potentially be involved in inducing depletion of Tak1^−/− HSPCs, we examined the expression levels of receptors specific for other inflammatory-related factors, including Tnfr2, Tlr-1(Toll-like receptor-1), Tlr-4, Tlr-5, Tlr-9, Dr-5 (TRAIL receptor) and IFN-γ–receptor, in normal and Tak1^−/−Tnfr1^−/− hematopoietic cells to define which of these are expressed in normal HSPCs and which are induced in Tak1^−/−Tnfr1^−/− HSPCs. We found that the expression levels of Tnfr2, Tlr-1, Tlr-4, Tlr-5, Tlr-9, Dr-5, and IFN-γ-r are higher in HSPCs and CPs than in lineage⁺ mature blood cells. We also found that the levels of these receptors are elevated in Tak1^−/−Tnfr1^−/− HSPCs compared to normal HSPCs. In addition, we found that, although Fas (CD95) is not expressed in normal HSPCs, it is expressed in Tak1^−/−Tnfr1^−/− HSPCs (supplemental Figure 4).

To study whether these inflammatory-related factors are also involved in inducing HSPC death and BMF in Tak1^−/− mice, c-kit⁺ HSPCs were isolated from Tak1^fx/fxTnfr1^−/−r2^−/− mice and infected with MSCV-Cre-GFP retrovirus. Infected cells were treated with these factors individually to test which of them could further enhance the elimination of Tak1^−/−Tnfr1^−/−r2^−/− HSPCs by examining the ratio of reduction of GFP⁺ cells. MSCV-GFP viral infection was studied in parallel as a control. We found that, among all the factors we tested, only FasL further enhanced the death of Tak1^−/−Tnfr1^−/−r2^−/− HSPCs (Figure 5A; supplemental Figure 5). IFN-γ, on the other hand, induced the death of both WT and Tak1^−/−Tnfr1^−/−r2^−/− HSPCs, which seems to be independent of Tak1 (Figure 5B). To confirm this finding further, c-kit⁺ HSPCs were isolated from Tak1^fx/fxFas^−/− mice and infected with MSCV-Cre-GFP or MSCV-GFP retrovirus separately. Infected cells were cultured in HSPC culture medium. The effects of Fas deletion on the survival of Tak1^−/− HSPCs were evaluated by examining the dynamic ratio of changes in GFP⁺ cells or by CFU assay. We found that Fas deletion could also partially prevent the depletion of Tak1^−/− HSPCs in vitro (Figure 5C-D). Interestingly, we found that the depletion of HSPCs and the BMF phenotype of Tak1^−/− mice could not be protected by Fas^−/− in vivo (Figure 5E-J). This difference in Fas^−/− phenotype in Tak1^−/− HSPCs in vitro and in vivo suggests that there might be an additional in vivo factor, which can induce the depletion of the same population of cells, rendering the effects of Fas signaling undetectable. This notion was confirmed by a comparative study of Tak1^−/−Tnfr1^−/−r2^−/− HSPCs with Tak1^−/−Tnfr1^−/−r2^−/−Fas^−/− HSPCs in vitro. We found that Fas^−/− did not further improve the survival of Tak1^−/−Tnfr1^−/−r2^−/− HSPCs, suggesting that Fas and TNF-α were operating on the same population of cells. In vivo, the death induced by high levels of TNF-α obscures the effects of Fas signaling (Figure 5K-L).

It was known that TNF-α and Fas-induced death signaling is mediated by FADD/caspase-8 (known as the traditional extrinsic apoptotic pathway) and RIP-1/RIP-3 (know as the necroptotic pathway), as shown by many cell studies.^27-32  We asked which of these 2 pathways transduces TNF-α and Fas-induced death signaling in Tak1^−/− HSPCs. We found that the death of Tak1^−/− HSPCs was significantly repressed by treatment with the RIP-1 specific inhibitor Nec-1. Interestingly, Nec-1 can only partially prevent the death of Tak1^−/− and Tak1^−/−Fas^−/− HSPCs; it failed to further improve the survival of Tak1^−/−Tnfr1^−/−r2^−/− HSPCs (Figure 6A-C). These results suggest that the RIP-1 pathway mediates TNF/Fas-induced death signaling in Tak1^−/− HSPCs.

To study whether the remaining Tak1^−/− HSPCs died of apoptosis, we examined whether caspases are activated in Tak1^−/− cells and whether inhibition of caspase activity can prevent the death of these cells. We found that the activities of caspase-8 and caspase-3 are increased in Tak1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− BM MNCs compared with WT counterparts, as shown by Western blotting (Figure 6D, data from Tak1^−/−Tnfr1^−/−r2^−/− BM MNCs only are shown; similar results were observed in Tak1^−/− cells). However, inhibition of FADD/caspase-8 signaling by transduced overexpression of dominant-negative (Δ)-FADD or Δ-caspase-8 did not repress but rather promoted the death of Tak1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− HSPCs (necroptosis occurs faster than apoptosis). Interestingly, inhibition of both caspase-8 and RIP-1 activities can nearly completely prevent the death of Tak1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− HSPCs (Figure 6F-G). These data suggested that the death of the remaining Tak1^−/− HSPCs could be attributed to caspase-8–dependent apoptosis because it can be converted into RIP-1–dependent necroptosis by the inhibition of caspase-8.^27-32 

It was known that active caspase-8 can promote either type I apoptosis by directly inducing the activation of executive caspase-3 or type II apoptosis by inducing mitochondrial-related caspase-9 activation.⁴⁵  The increased percentages of Tak1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− BM MNCs with reduced mitochondrial membrane potential (Δψm) and the activation of caspase-9 in the same cells suggested that at least a subset of Tak1^−/− HSPCs died of type II apoptosis (supplemental Figure 6A-B). This notion was further supported by a partial prevention of the apoptosis of Tak1^−/− and Tak1^−/−Tnfr1^−/−r2^−/− HSPCs by the overexpression of Bcl-xl or Δ-caspase-9, as shown by both annexin V staining and CFU assays (supplemental Figure 6C,E).

It has been shown that ROS is the major cause of apoptotic death in intestinal enterocytes and skin keratinocytes in Tak1^−/− mice.^34,36  To study whether ROS-induced cell death also contributes to the BM damage observed in Tak1^−/− mice, we first examined whether ROS levels are increased in Tak1^−/− HSPCs using CM-H₂DCFDA staining. We found that ROS levels in Tak1^−/− HSPCs are comparable to those in WT counterparts (supplemental Figure 7A). In addition, we found that repression of ROS by treatment of cells with ROS scavengers, such as N-acetyl-L-cysteine or butylated hydroxyanisole, or by overexpression of superoxide dismutase, did not protect the death of Tak1^−/− HSC/Ps in vitro (supplemental Figure 7B-E). Moreover, we found that treatment of in vivo with N-acetyl-L-cysteine can significantly prevent liver damage but not the BMF phenotype in Tak1^−/− mice (supplemental Figures 7F-I and 8). We concluded that ROS-induced cell damage might not be significantly involved in the Tak1^−/− deletion-induced apoptotic death of HSPCs.

Studies have suggested that proinflammatory factors are the major negative regulators of hematopoiesis in the BM microenvironment, inducing the depletion of HSPCs. However, studies also suggested that, under certain conditions, most of these inflammatory factors can enhance the growth of HSPCs.^16-19  The molecular mechanism underlying this phenomenon has not been explored.

TNF-α is a major proinflammatory cytokine produced by activated macrophages and lymphocytes. Physiologic levels of TNF-α can be detected in the BM microenvironment.⁴⁶  Classically, TNF-α has been understood to be an inhibitor of growth and an inducer of apoptosis in HSPCs.^5,10,43  However, increased evidence suggests that TNF-α can also have a positive stimulatory effect on hematopoiesis and is required for the self-renewal and BM engraftment of HSPCs. The significant reduction in both competitive reconstitutive capacity and success in secondary transplantation of Tnfr1^−/− HSPCs suggests that a certain physiologic baseline level of this inflammatory factor is crucial for the maintenance of proper HSC function.⁴⁷ 

The bidirectional effects of TNF-α on HSPCs observed in previous experiments have been speculated to be the result of the concentration used, effects mediated by different receptors, or other cytokines generating effects at subsequent stages of differentiation.^18,48  Recent evidence suggests that the hematopoietic system is actually maintained by multiple HSC subtypes, each possessing distinct functional characteristics. Based on the ratio of myeloid to lymphoid cells they produced in recipient animals, HSCs can be further distinguished into myeloid-biased or lymphoid-biased HSC categories.⁴⁹  Studies indicated that different subsets of HSCs might respond differently to stimulation by various environmental factors. For example, TGF-β stimulates the proliferation of myeloid-biased HSCs while repressing the proliferation of lymphoid-biased HSCs.⁵⁰  This might explain why certain lineages of hematopoietic cells become selectively activated and expand measurably in certain disease situations, as well as during the aging process.⁵⁰ 

Our studies suggested that HSPCs are also heterogeneous in terms of their responses to death signal stimulation. HSPCs can be clearly separated into 2 distinct populations based on their response to TNF-α/Fas-induced death signaling (Figure 7). TNF-α/Fas-induces a RIP-1–dependent necroptosis in 30% to 40% of Tak1^−/− HSPCs. Whether TNF-α–RIP-1 signaling selectively induces the death of lymphoid-biased HSCs or myeloid-biased HSCs needs further study. In addition, our studies suggest that the remaining Tak1^−/− HSPCs died of traditional extrinsic apoptotic pathway-mediated FADD/caspase-8 cascade–dependent apoptosis. The apoptosis-to-necroptosis conversion of Tak1^−/− HSPCs by inhibition of caspase-8 activity supports the mutual inhibition concept of FADD/caspase-8 and RIP-1/RIP-3 signaling.^27-32 

The factor(s) which induce(s) the apoptosis of these remaining cells need(s) further investigation. Nevertheless, our studies demonstrate that Fas-induced signaling is definitely not the cause. Our data suggest that Fas might be operative in the same population of hematopoietic cells on which TNF-α acts. Inflammatory factors, such as IL-1β, TRAIL, and LPS, also induce death receptor-related apoptosis and Tak1-mediated survival signaling, as demonstrated in studies of other tissues.^33,37  These factors have also been found to repress hematopoiesis and are implicated in the pathogenesis of some BMF syndromes.^11,19  Although we found that the addition of IL-1β, TRAIL, poly I:C, or LPS in vitro did not further enhance the death of Tak1^−/−Tnfr1^−/−r2^−/− HSPCs, there is a possibility that Tak1^−/− HSPCs might be super-sensitive to the apoptosis induced by these factors. Even a low level of these factors in the culture medium might be sufficient to induce the maximum elimination of Tak1^−/− HSPCs. Therefore, we cannot completely exclude the contribution of death signals induced by these factors to the apoptosis of Tak1^−/− HSPCs.

Inflammatory factors, such as IFN-α and IFN-γ, have been found to inhibit the growth of HSPCs and promote the proliferation of quiescent HSCs; this scenario has been implicated in many BMF syndromes.⁵  However, these factors do not induce death receptor-related signaling, nor do they stimulate Tak1 activity. We found that IFN-α has no significant effect on the apoptosis of WT or Tak1^−/− HSPCs, whereas IFN- γ was able to induce the death of both WT and Tak1^−/− HSPCs. This suggests that IFN-γ might induce HSPC death through a mechanism that is independent of Tak1 and TNF-α/Fas. The significant increase of IFN-γ concentration in Tak1^−/−Tnfr1^−/−r2^−/− mice suggests that Tak1 might repress the expression of IFN-γ. The mechanism behind this phenomenon needs further investigation.

The authors thank the staff of the Department of Comparative Medicine at Loyola University Medical Center for excellent animal care services, Ms Patricia Simms for flow cytometric sorting of HSCs and progenitors, and Drs Manuel Diaz, Nancy Zeleznik-Le, and Andrew Dingwall for ongoing professional collaboration and scientific suggestions and discussions that improved the present studies.

This work was supported by Department of Defense (grant PR080447 through Loyola University Chicago), the National Natural Science Foundation of China (project 81071774), and the Program for Professor of Special Appointment (Eastern Scholar) at Shanghai Institutions of Higher Learning through Shanghai Normal School. P.B. is supported in part by the Jimmy Burns Foundation.

Contribution: Y.X., H.L., Jun Zhang, A.V., Shubin Zhang, W.W., Shanshan Zhang, and P.B. performed the research, analyzed the data, and wrote the paper; and Jiwang Zhang designed and performed the research, analyzed the data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for H.L. is Department of Hematology/Oncology, Gansu People's Hospital, Gansu, People's Republic of China. The current affiliation for Jun Zhang is Department of Biology, College of Life and Environment Science, Shanghai Normal University, Shanghai, People's Republic of China.

Correspondence: Peter Breslin, Oncology Institute, Cardinal Bernardin Cancer Center, Loyola University Medical Center, 2160 South 1st Ave, Maywood, IL 60153; e-mail: pbresli@luc.edu; and Jiwang Zhang, Oncology Institute, Cardinal Bernardin Cancer Center, Loyola University Medical Center, 2160 South 1st Ave, Maywood, IL 60153; e-mail: jzhang@lumc.edu.