Essential Role for the P55 Tumor Necrosis Factor Receptor in Regulating Hematopoiesis at a Stem Cell Level

p55- and p75-deficient mice, originally supplied by Immunex Corp. to J.L.M. Ferrara, and C57BL/6 WT animals (The Jackson Laboratory) were bred and maintained at the Redstone Animal Facility of the Dana-Farber Cancer Institute. Congenic Ly5.1⁺ C57BL/6 mice were purchased from the National Institutes of Health. All animals were kept under microisolators and provided with food and water ad libitum.

To obtain peripheral blood (PB) samples, mice were first anesthetized by metaphane inhalation. For white blood cell (WBC) and RBC counts, the mice were terminally bled via a cardiac puncture. The counts were performed on a Coulter MaxM. A small blood sample was obtained by puncture of the venous sinus in the eye for analysis for the presence of Ly-5.2⁺ cells (see below). BM cells were obtained by flushing both tibiae and femurs of donor mice with PBS containing 2% FCS. The cell suspension was layered on a cushion of FCS and washed once in PBS, 2% FCS. Nucleated cells were counted with a hemocytometer. Assuming that the cell content of a tibia is ∼0.6 that of a femur 35, the total cell number was divided by 3.2 to calculate the cell number per femur.

To determine the presence of the different lineages in the BM, cells were incubated on ice for 35 min with the FITC-conjugated lineage (Lin) markers B220 (RA3-6B2), Mac-1 (M1/70), Gr-1 (RB6-8C5), Ly-1 (53-7.3), or PE-labeled Ter119 (PharMingen). Cells were then washed twice with PBS, 2% FBS. During the last wash, the medium contained propidium iodide (PI, 2 μg/ml; Sigma Chemical Co.) to stain the dead cells. Cells were analyzed on a single laser FACScan™ (Becton Dickinson).

For purification, cells were labeled with a cocktail of mAbs, including Sca-1, derivatized with cyanine 5 succinimidyl ester (a gift from Dr. P. Lansdorp, Terry Fox Laboratory, Vancouver, British Columbia, Canada), FITC-labeled Lin markers, and (PE)-conjugated c-Kit (2B8; PharMingen). Where indicated, CD34-FITC (RAM34; PharMingen) was used to further subdivide the Sca-1⁺Lin^−/loc-Kit⁺ subpopulation into CD34⁺ and CD34⁻ cells. A preenrichment step was included in the purification protocol when cells were sorted; total BM cells were depleted of cells that expressed mature Lin markers by magnetic separation according to the manufacturer's directions (StemCell Technologies). Multicolor analysis to identify and sort Sca-1⁺Lin^−/loc-Kit⁺ progenitor cells was performed on a Coulter Epics^® Elite ESP.

To calculate the absolute number of each cell type per femur (i.e., the different mature cell lineages and Sca-1⁺Lin^−/loc-Kit⁺ cells), the total number of cells per femur was multiplied by the fraction of each cell population in total BM cells as obtained by FACS^® analysis.

Unseparated BM cells or highly purified progenitor cells were fixed in ice-cold 80% ethanol and kept on ice overnight. The next day, cells were spun down and washed once in PBS with 0.1% Triton X-100. Cells were resuspended in 0.5 ml PBS containing 50 μg/ml PI, incubated at 37°C for 30 min, and then kept on ice until analysis, which was within 1–2 h of finishing the staining. The sample was analyzed on a Coulter Epics^® XL analyzer.

To assay for erythroid (BFU-E), granulocyte and/or macrophage (CFU-G/M), multilineage (CFU-GEMM), and pre-B clonogenic progenitors, cells were resuspended in IMDM containing 2% FCS. A portion of the cell suspension was resuspended in methylcellulose containing medium that was supplemented with the appropriate cytokines (StemCell Technologies). Pre-B cell colonies were scored in situ after 7 d of culture, the other colonies after 12–14 d of culture. The number of each clonogenic progenitor per femur was calculated by multiplying the number of nucleated cells per femur with the frequency of each progenitor, i.e., the number of respective colonies counted divided by the number of cells plated (5 × 10⁴ cells for the pre-B cell assay and 1.5 × 10⁴ cells for the erythroid/myeloid colony assay).

The BM content of CFU-S day 12 cells was determined as originally described 36. In brief, 7 × 10⁴ unseparated BM cells were injected into C57BL/6 recipients that were irradiated with 950 cGy given in two equally split doses (5 h apart) from a ¹³⁷Cs source. The spleens were harvested 12 d later, and the macroscopic surface colonies were counted. The spleens of control animals that were irradiated but not transplanted with BM cells showed none or one to two pin-point size colonies. This size of colony was not included in the colony counts of the spleens from test animals. The number of CFU-S per femur was calculated by multiplying the number of nucleated cells per femur with the frequency of CFU-S, i.e., number of colonies counted divided by 7 × 10⁴.

A slightly modified competitive repopulating units (CRU) assay was used to determine the number of HSCs in BM 37,38. In brief, varying doses of unseparated test cells (Ly5.2⁺) were injected together with 1 × 10⁵ unseparated adult BM “helper” cells (Ly5.1⁺) into lethally irradiated (as for CFU-S) recipients (Ly5.1⁺). The allelic difference at the Ly5 locus between the recipient/helper cells and the test cell population was used to follow hematopoietic reconstitution from the test cells. 4, 8, and 16 wk after the transplantation, a small blood sample was taken from each mouse and analyzed for the presence of Ly5.2⁺ cells. A mouse was considered to be reconstituted by CRU present in the test cell population (“positive”) when, 16 wk after the transplantation, >1% of the PB cells were Ly5.2⁺ and included both myeloid and lymphoid cells, as determined by their respective distinct side scatter (SSC) properties. CRU frequencies for every type of BM tested were calculated by analysis of the proportion of negative mice using Poisson statistics 39. The number of CRU per femur was calculated by multiplying the number of nucleated cells per femur with the frequency of CRU.

Some transplanted mice were used as donors for secondary BM transplantations. Marrow cells from both femurs and tibiae were isolated, and a proportion of these cells together with Ly5.1⁺ helper cells was injected into three to five irradiated Ly5.1⁺ recipients (analogous to a normal CRU assay).

The Student's t test was used for the comparison of the mean of various hematological parameters from WT animals with that from p55^−/− or p75^−/− animals, unless otherwise stated.

Several assays were performed to assess the various lineages and subpopulations of the hematopoietic system (Table). In young TNFR-deficient mice (<3 mo), the WBC and RBC counts in the PB did not differ from those of WT mice, nor did the relative contributions to the several myeloid and lymphoid lineages (data not shown). The PB indices were mirrored by similar BM cellularities in these three strains of mice. Functional analysis of marrow cell suspensions revealed a significantly higher number of myeloid clonogenic cells in p55^−/− mice compared with WT mice (P < 0.01), but no significant differences were observed in the number of erythroid and pre-B clonogenic cells, or CRU.

We hypothesized that an effect of perturbed TNFR signaling in HSCs may only be measurable with increased age. Therefore, we performed the same series of hematopoietic cell assays with BM cells from older mice (>6 mo). Neither the WT nor the p75^−/− mice showed any significant change with aging, aside from the normalization of the WBC counts in p75^−/− mice. In contrast, a profound disturbance of hematopoiesis in older p55^−/− mice could be demonstrated (Table). Nearly twice as many WBCs were found in the blood (P < 0.0005), and this was paralleled by the BM cell counts (P < 0.005). Moreover, FACS^® analysis of the various mature lineages (Fig. 1 A) showed a tendency to increased cell numbers in all lineages except the B cell lineage. The same tendency was observed for clonogenic progenitors (Fig. 1 B). When all myeloid CFCs, i.e., BFU-E, CFU-G/M, and CFU-GEMM, were combined, the total number of colonies in p55^−/− marrow was shown to be significantly higher than in WT or p75^−/− marrow (P < 0.01): 128,000 ± 38,300 vs. 85,400 ± 25,000 and 78,300 ± 28,000, respectively. To assess the status of immature progenitors, we performed CFU-S day 12 assays and immunolabeled BM cells to identify the Sca-1⁺Lin^−/lo c-Kit⁺⁺ subpopulation. Similar to the mature myeloid clonogenic progenitors, the number of CFU-S day 12 in p55^−/− BM cells was higher compared with WT (not significant) and p75^−/− (P < 0.03) BM cells: 2,700 ± 500 vs. 2,100 ± 260 and 1,700 ± 450, respectively (Fig. 1 B). Fig. 2 depicts a representative FACS^® profile of BM cells showing the parameters used to calculate the absolute number of Sca-1⁺ Lin^−/loc-Kit⁺⁺ cells per femur. There is a considerably larger, although not statistically significant subpopulation of Sca-1⁺Lin^−/loc-Kit⁺⁺ cells in p55^−/− BM compared with WT or p75^−/− BM: 11,300 ± 4,800 vs. 4,800 ± 2,200 and 5,200 ± 2,400 cells, respectively. Since this subpopulation includes long-term repopulating HSCs 40, this phenotypic finding may indicate an increase in the number of HSCs in p55^−/− BM. However, it has been clearly shown that phenotype does not always predict function 38,41. Therefore, we performed competitive repopulating experiments under limiting dilution conditions to determine the number of CRU per femur (Table). To our surprise, we found a significant decrease in the frequency of CRU among p55^−/− BM cells compared with WT (P < 0.001). When corrected for BM cellularity, this translated into a near fourfold reduction of the absolute number of CRU in p55^−/− mice (350 vs. 1,275 in WT mice). Taken together, these data show a tendency with aging in p55-deficient mice to increased sizes of most mature and immature hematopoietic (progenitor) cell populations, which could explain the overall increase in marrow cellularity found in these animals. At the same time, the number of HSCs that give rise to this variety of cell populations is reduced.

To further elucidate the extent of the stem cell defect in old p55^−/− mice, we analyzed in detail the hematopoiesis of three sets of mice transplanted with low numbers of BM cells (Fig. 3, and Table). Groups 1 and 2 represent animals transplanted with 16,000 total BM cells (WT and p55^−/−, respectively). A comparison of their donor repopulation kinetics as measured at various time points after transplant illustrates a dramatic decline in donor reconstitution over a 4-mo time period by p55^−/− BM cells, whereas the donor reconstitution by WT cells showed a slight increase over the same time period tested (Fig. 3). 16 wk after transplantation, the level of donor cells in recipients of p55^−/− BM cells (group 2) was significantly lower (P < 0.05) compared with that in recipients of WT BM cells (group 1) (1.1 ± 1.0 vs. 10.2 ± 12.2, respectively). The difference in repopulating ability between WT and p55^−/− HSCs becomes even more apparent when individual mice are analyzed (Table). Between 4 and 16 wk after the transplantation, Ly5.2⁺ donor reconstitution did not increase in any of the recipients of p55^−/− (Ly5.2⁺) BM cells. In contrast, more than half of all the animals that had received WT (Ly5.2⁺) BM cells showed an increase in Ly5.2⁺ cells, by as much as 38% in one of the recipients. Moreover, only 22% of the group 2 animals that were multilineage reconstituted at week 4 remained so at week 16, compared with nearly 80% of the animals in group 1 (Table).

Thus, when equal numbers of BM cells are transplanted, the repopulating ability of p55^−/− marrow cells is inferior to that of WT cells. However, when the CRU frequencies (Table) were used to calculate the average number of CRU received by individual transplanted animals (i.e., the total number of cells transplanted divided by the frequency of CRU), it was calculated that group 1 had received on average 1 HSC, whereas group 2 had received on average 0.3 HSC. This could potentially explain, or at least contribute to, the observed differences. Therefore, we compared group 1 WT mice with a group of mice that received on average 1 p55^−/− HSC, equivalent to 64,000 total BM cells (group 3). As expected, approximately the same number of mice remained multilineage reconstituted after 16 wk (Table; 78% in group 1 and 70% in group 3). However, although 4 wk after the transplantation the level of donor reconstitution in group 3 (p55^−/− transplant) recipients was significantly higher than in group 1 (WT transplant) recipients (P < 0.01), the percentage of donor p55^−/− cells, again, gradually decreased with each subsequent analysis (Fig. 3). 90% of all animals that received p55^−/− marrow cells showed as much as a 34% decline in donor reconstitution (Table). Thus, not only does p55^−/− marrow contain fewer HSCs, these data also indicate that the proliferative potential of p55^−/− HSCs is less than that of WT HSCs.

To further stress the proliferative response of the Ly5.2⁺ transplanted cells, BM cells of some of the primary recipients were transplanted into secondary recipients. The results of these experiments are shown in Table. Eight primary recipients of WT marrow cells and three of p75^−/− marrow cells were used in these experiments. Because we had not been able to demonstrate remarkable differences between WT and p75^−/− hematopoiesis in older mice, we grouped these mice together to compare their secondary transplantation results to the results of the serially transplanted p55^−/− marrow cells. Of 11 WT/p75^−/− primary recipients, 4 animals (36%) contained Ly5.2⁺ HSCs that generated multilineage progeny upon transplantation into secondary recipients. This proportion is in accordance with previous published data 42. Only 2 (18%) of the 11 primary recipients of p55^−/− BM gave secondary reconstitution upon serial passage. As observed in the primary recipients of p55^−/− cells, and in contrast to secondary recipients of WT donor cells, the kinetics of Ly-5.2⁺ reconstitution in the secondary recipients showed a marked decline over time (data not shown).

One possible explanation for the observed differences in HSC number and proliferative potential between WT and p55^−/− mice is a higher cycling rate of p55^−/− HSCs. Purified Sca-1⁺Lin^−/loc-Kit⁺⁺ cells still include large numbers of committed progenitor cells, as confirmed by our finding that between 20 and 30% of these cells (WT and knockout cells) were in the S/G2/M phase of the cell cycle (data not shown). Therefore, we further subdivided this population using an mAb against the CD34 antigen. The Sca-1⁺Lin^−/loc-Kit⁺⁺CD34⁻ cells, which include the long-term reconstituting HSCs 43, comprise ∼10% of all Sca-1⁺Lin^−/loc-Kit⁺⁺ cells. The sorting strategy for obtaining these cells is depicted in Fig. 4. The results of cell cycle analyses of unseparated BM cells and highly purified cells (box 3) are shown in Table. As expected, DNA staining of unseparated BM cells did not demonstrate a difference in cycling activity between the WT and either knockout mouse. Moreover, the result obtained with the highly purified CD34⁻ subpopulation of Sca-1⁺Lin^−/loc-Kit⁺⁺ cells isolated from p55^−/− marrow cells did not strongly suggest that cycling activity of p55^−/− HSCs was altered. Therefore, a difference in percentage of cycling HSCs does not seem to explain the observed qualitative and quantitative differences between HSCs from WT and p55-deficient mice.

The p55 and p75 TNFRs are ubiquitously expressed on all hematopoietic cells, with the exception of erythroid and unstimulated T cells 2,44. There is ample evidence from in vitro culture systems that, through one or both of these receptors, TNF-α can promote the expansion and differentiation of various hematopoietic subpopulations, as well as inhibit their proliferative response 10,11,12,13. However, the long-term effects of this cytokine in vivo as a possible regulator of steady state hematopoiesis, especially with regard to the stem cell compartment, are not well defined. The present studies were undertaken to address this question by analyzing the hematopoietic system of mice in which TNF-α signaling is impaired. We and others 45 have shown that most hematopoietic subpopulations tested are present in young p55^−/− and p75^−/− mice at levels comparable to those of WT mice. In older mice, no significant differences in the assays tested between p75^−/− and WT mice could be demonstrated. However, when older p55^−/− mice were compared with WT mice they showed an interesting phenotype: their HSC compartment was significantly smaller than that of their WT counterparts (approximately fourfold), whereas the Sca-1⁺Lin^−/loc-Kit⁺⁺ progenitor compartment had increased in size (approximately twofold). This was reflected in significant increases in total cellularity of the BM, total myeloid CFCs, and WBC and RBC counts in the PB. Thus, with aging of the mice, the lack of signaling through the p55 TNFR seems to affect the balance between generating new HSCs (self-renewal) and producing committed progenitors (differentiation). Interestingly, only the B cell lineage deviated from this pattern, in that the numbers of B220⁺ cells in the marrow (Fig. 1 A) and in the blood (data not shown) were similar to their WT counterparts, whereas the number of pre-B CFCs was significantly decreased in p55-deficient mice (Fig. 1 B). Previous studies have shown that the p55 TNFR is essential for the structural organization of lymphoid organs 46,47,48,49 and the number of B cells in the Peyer's patch 48. Our data also suggest that the generation of B cells in the marrow depends on p55 TNFR signaling. Whether this is mediated by TNF-α or lymphotoxin α remains to be studied.

Exhaustion of the HSC pool can be explained by loss of HSCs due to cell death or differentiation. TNF-α has been shown to both induce and suppress apoptosis in murine Sca-1⁺Lin⁻ cells. Analogous to the effects of TNF-α on the proliferation of hematopoietic cell populations, its effect on apoptosis seems to depend on which additional cytokine was present: suppression in cultures with IL-1α present, and induction when Steel factor was supplied 50. The apoptotic effects of TNF-α are well documented and mainly exerted through its p55 receptor 51,52, although an apoptotic role for the p75 receptor has been demonstrated in mature CD8⁺ cells 9. Increased apoptosis in HSCs will definitely decrease the HSC pool. However, if this were the case in p55^−/− mice, one would expect to find an accompanying decrease in the number of progenitor cells. The opposite was found in our p55-deficient mice. The mechanisms that control self-renewal versus commitment decisions in HSCs are not known. However, it is generally accepted that an increase in the number of HSC cell divisions is accompanied by an increased chance of HSCs to commit and differentiate among one of the lineages 53,54,55. TNF-α has been shown to prevent WT but not p55^−/− Sca-1⁺Lin⁻c-Kit⁺ cells from entry into S phase 31. Increased cycling of p55^−/− HSCs could explain our results. However, we could not detect a difference in cell cycle activity between Sca-1⁺Lin⁻c-Kit⁺ subpopulations (CD34⁺ or CD34⁻) sorted from WT or p55^−/− BM cells. The effects of TNF-α on the cell cycle of HSCs may be very subtle, which would explain why HSCs in young mice seem unaffected and why it is not detectable by the methods used in this study. One study has demonstrated the importance of TGF-β in maintaining HSCs in a quiescent, noncycling state in unperturbed steady state hematopoiesis 30. Since TNF-α cooperates with TGF to inhibit the proliferation of primitive progenitor cells 32, it is not inconceivable that more HSCs escape their quiescent state when one of the inhibitory signals is missing.

There are two important considerations with respect to the finding that the phenotype in p55^−/− mice manifests only with aging (>6 mo of age). First, TNF-α production and TNF-α signaling cascades seem to change with aging. Aged mice produce more TNF-α in response to LPS 56, and T cells from older humans express more p55 but less p75 TNFR subunits and are more susceptible to TNF-α–induced apoptosis 57. It is possible that TNF-α signaling in HSCs also changes with aging, and that any deficiencies in this signaling pathway have greater consequences later in life. It may be of interest to look at mice of 20 mo or older since the number of cycling HSCs increases dramatically at this age 58. Second, hematopoiesis during fetal and/or neonatal development, which is at that time expanding tremendously, not only contains HSCs with properties different from adult HSCs 42,59, but is also probably regulated differently than adult, steady state hematopoiesis 60. The regulatory mechanisms of fetal/neonatal hematopoietic cells may not depend on TNF-α signaling at all. This, in addition to the possibility that older hematopoietic cells, and HSCs in particular, may be more sensitive to TNF-α, could explain the lack of a phenotype in young p55^−/− mice.

Thus, our findings suggest that, with age, a lack of signaling through the p55 TNFR increases the number of committed hematopoietic progenitors and their subsequent progeny at a cost to the stem cell pool. The scarcity of secondary recipients with multilineage p55^−/− progeny confirms the lack of self-renewal or exhaustion of p55^−/− HSCs. However, what is surprising and contrary to our expectations is the lower level of multilineage reconstitution per HSC from p55^−/− donors compared with WT. The explanation for this reduction may be that hematopoietic regeneration in myeloablated recipients is different from both adult steady state hematopoiesis and fetal/neonatal hematopoiesis. Although little is known about the cytokines and other regulatory elements in the activation and/or recruitment of transplanted HSCs, there are some important differences compared with steady state hematopoiesis 61. One important difference is that transplanted HSCs must home to an appropriate microenvironment. It is possible that altered TNF-α signaling affects the homing properties of p55^−/− cells. In this regard, it is of interest that TNF-α can affect the expression of CXC chemokine receptor 4 62,63,64, a chemokine receptor critical for murine BM engraftment and regeneration as well as B cell development 65,66. Moreover, in our experimental design, the transplanted HSCs are influenced by stress-induced environmental changes associated with the myeloablation of the recipient. Stress can induce elevated levels of glucocorticoid hormones, which, in synergy with other cytokines, make BM cells more susceptible to IL-1 through upregulation of the IL-1 receptors, which favors myeloid differentiation 67,68. A beneficial effect of IL-1 may be the suppression of apoptosis. However, for this effect to be delivered, the combination of TNF-α may be essential 50. Therefore, it is possible that p55^−/− HSCs acquire elevated levels of IL-1 receptors after transplantation but are more likely to differentiate than their WT counterparts when they do not receive the appropriate TNF signal. Thus, the loss of HSCs and all other hematopoietic cell types after transplantation of p55^−/− marrow cells may not necessarily reflect steady state hematopoiesis, but rather the inability of these cells to adequately respond to extreme proliferative stimuli.

The need for intact TNFR signaling in HSCs during hematopoietic regeneration of a myeloablated recipient may be more crucial for its survival than during steady state hematopoiesis. The G0/G1 arrest that TNF-α has been shown to induce in Sca-1⁺Lin⁻c-Kit⁺ cells may be imperative in restraining the proliferative response of HSCs to external stimuli and thereby preventing their extinction. In this respect, it is of interest that in TNF^−/− mice it was shown that TNF is important for the generation of an inflammatory response, but its presence was far more essential in limiting the extent and duration of this response 69. This possible regulatory role in limiting the number of cell divisions in HSCs warrants further investigation, for example in retroviral gene therapy protocols with HSCs as target cells. Future studies of the role of TNF in steady state hematopoiesis should include the analysis of mice older than 20 mo and the generation of allophenic mice, which would allow the tracking of HSC progeny over time, without perturbing the hematopoietic system 70.

We thank John F. Daley for advice on flow cytometry, the Redstone Animal Care Facility for their support and excellent care of the mice, and Amy Perrault for her help in preparing this manuscript.

This work was supported by National Institutes of Health grants RO1 HL55709 (to C.A. Sieff) and PO1 CA39542 (to J.L.M. Ferrara).