An injection of TNF in mice induced profound thrombocytopenia, due to an increase of platelet consumption, that was evident after 1 h and lasted for 3 days. This process was evident in mice that were genetically deficient in TNFR2 (p75) but not in mice lacking TNFR1 (p55), indicating that the process is mediated by TNFR1-bearing cells. To explore the site of action of TNF, labeled platelets from TNFR1 −/− or +/+ donors were transferred to TNFR1 −/− or +/+ recipients. TNF induced the consumption of platelets from TNFR1 −/− donors when injected into +/+ recipients, while platelets from +/+ donors were not consumed when present in TNFR1 −/− recipients; this finding indicates that TNF acts on the TNFR1 of host cells but does not act on platelets. The expression of TNFRs is consistent with this interpretation, since TNFRs were not detected on platelets by flow cytometry. In megakaryocytes, the expression of TNFR1 was detected by immunohistochemistry. These results indicate that TNF induces platelet consumption by acting not on platelets directly but on the TNFR1 of other cells, presumably increasing the release of factors with agonist activity for platelets.
TNF is an important effector of inflammatory reactions that is known to exert a wide variety of responses (1). Systemic inflammatory diseases such as acute respiratory distress syndrome (2), severe malaria (3), sepsis, or the release of LPS (4) are associated with thrombocytopenia, due to an increase in platelet consumption (2) with a poorly understood pathogenesis. When injected in rodents or humans, TNF reportedly induces thrombocytopenia (5) and consequently may contribute to the thrombocytopenia associated with diseases in which this cytokine is produced in increased amounts. In various types of inflammation, the local release of TNF elicits the trapping of platelets in inflamed blood vessels by mechanisms that are not completely understood (6, 7, 8).
Platelets are produced by megakaryocytes, located mainly in the bone marrow, and released as anucleated elements which remain in the plasma for ∼2 days in mice (9). Consequently, thrombocytopenia may be due either to an increase in platelet consumption or a decrease of platelet production. An increase in platelet consumption is generally believed to be the result of the release of molecules with agonist activity for platelets; such molecules include thrombin, plasmin, serotonin, etc. (reviewed in 2 . TNF is not a strong platelet agonist when tested on isolated platelets in vitro (10, 11).
TNF exerts its effects through two glycoprotein membrane receptors that are expressed at variable degrees on the majority of cells: a 55-kDa receptor, TNFR1, and a 75-kDa receptor, TNFR2, (reviewed in 12 . These receptors are widely distributed, but their presence on cells of megakaryocyte/platelet lineage has not been described.
In this report, we explored the role of TNFR1 and TNFR2 in TNF-induced thrombocytopenia in mice as well as the mode of action of TNF, using mice that were genetically deficient in TNFR1 (13) or TNFR2 (14). TNF-induced thrombocytopenia was found to be entirely dependent upon the TNFR1. However, we did not find a detectable expression of this receptor on platelets, and platelet transfer experiments showed that TNF acts on the TNFR1 of host cells, but not on platelets directly.
Materials and Methods
CBA/J and C57BL/6 mice were purchased from Iffa Credo (L’Arbresle, France) and bred for three to four generations in our animal facility. Both mice that were genetically deficient for TNFR1 and TNFR2 and their +/+ littermates (a gift of Dr. H. Bluethman, Hoffmann-La Roche, Basel, Switzerland) were obtained in the (C57BL/6 × 129)F1 background (13, 14) and bred in our animal facility. Experiments were performed with 2- to 4-mo-old mice of either sex.
Injection of TNF
Murine rTNF-α was obtained from Dr. B. Allet at Glaxo Institute of Molecular Biology (Geneva, Switzerland). Its specific activity (1 U = 0.7 pg/ml) was determined with W13 cells; its LPS contamination, determined with the E toxate test, was estimated to be between 5 and 50 ng/mg. An injection of boiled TNF did not have a detectable effect on platelet numbers. Mice were injected i.v. in the retroorbital plexus with 10 μg of TNF diluted with NaCl or with NaCl alone as a control. Injections were administered at the same time of day (between 8 and 9 A.M.).
Blood (150 ml) was collected in 2.5 mM EDTA from the retroorbital plexus of mice that had been injected previously with 5 U of heparin (Liquemin, Hoffmann-La Roche). Platelet-rich plasma was obtained by centrifuging the blood at 480 × g for 5 min at room temperature, and platelets were washed twice in acid-citrate-dextrose (ACD)4 buffer at 1800 × g for 10 min (15). Platelets were diluted using the Unopette micro collection system for platelet determination (Becton Dickinson, Rutherford, NJ) and counted in a hemocytometer. Three independent samples of a minimum of 100 cells were counted, and the arithmetic mean of the three counts was calculated. The mean platelet volume as well as platelet number were also counted in a Sysmex cell counter (Toa Medical, Kobe, Japan).
Platelet survival and trapping evaluation
Platelets prepared in ACD buffer were 51Cr-labeled as previously described (15). A total of 40 to 50 × 106 platelets (0.5–1 × 105 cpm) were injected i.v. in the retroorbital sinus. For the trapping experiments, mice were sacrificed at 3 h postinjection, and the radioactivity in the lung, spleen, and blood (50 ml) was counted in a gamma-counter (Becton Dickinson, San Jose, CA). For survival evaluation, 0.05 ml of blood was withdrawn from the retroorbital sinus at various times after injection using a calibrated, heparinized capillary.
Rabbit anti-mouse/rat/human/CD62P (P-selectin) and a hamster anti-mouse CD61 polyclonal Ab were both obtained from PharMingen (San Diego, CA). Rabbit anti-mouse TNFR1 and TNFR2 polyclonal Abs were a gift from Dr. G. Wong, (Genentech, San Francisco, CA). Rat anti-mouse TNFR1 (HM-103, IgG2a anti-p55) and rat anti-mouse TNFR2 (HM-102, IgG2a anti-p75) mAbs were derived in our laboratory. Abs were used as purified IgG, and the appropriate nonimmune IgG fractions or monoclonal isotype were used as specificity controls. FITC-labeled goat anti-rat, goat anti-hamster (Caltag, San Francisco, CA), and goat anti-rabbit (Sigma, St. Louis, MO) IgGs were used as second-step reagents.
Fluorescence-activated cell sorter
Platelets were prepared as described above, and 0.5 to 1 × 106 cells were added to polystyrene tubes (Becton Dickinson), incubated for 20 min with normal goat serum, washed, and incubated at room temperature with the appropriate IgGs. After 30 min, the cells were washed twice in ACD containing 0.1% BSA (Sigma) and subsequently incubated with an appropriate dilution of FITC-labeled secondary Abs for 30 min. Platelets were washed twice, and the samples were analyzed using a FACScan flow cytometer (Becton Dickinson, Mountain View, CA). Side and forward angle light scatter as well as staining with anti-CD61 mAb were used to gate the platelet population. An endothelial cell line (sEnd.1 (16)) was used as a positive control for TNFR expression.
Sections from frozen spleen were prepared and stained with rabbit anti-TNFR1 and anti-TNFR2 polyclonal Abs. As a control, a nonimmune rabbit serum was also tested. These Abs were revealed with alkaline phosphatase-labeled goat anti-rabbit Ab (Dako Diagnostics, Carpinteria, CA). As an additional control of specificity, spleens from TNR1 −/− and TNFR2 −/− donors were also examined.
Statistical analyses were conducted using the Student t test for unpaired data.
TNF-induced platelet consumption
An i.v. injection of 10 μg of TNF to mice induced profound thrombocytopenia, which became evident at 1 h after injection and lasted for ∼3 days (Fig. 1). A similar decrease in platelet counts was observed in the C57BL/6, CBA/J, or (C57BL/6 × L129)F1 mouse strains (data not shown). Thrombocytopenia was mainly due to an increase of platelet consumption (see below).
TNF induces platelet consumption in TNFR2 −/−, but not in TNFR1 −/− mice
We subsequently tested the effect of TNF on the platelet counts of mice that were genetically deficient in TNFR1 and TNFR2. TNF induced thrombocytopenia of a similar magnitude in TNFR2 +/+ and −/− mice, (i.e., resulting in a decrease in platelet number of ∼30%), indicating that TNFR2 does not play an important role. In contrast, TNF did not induce significant thrombocytopenia in TNFR1 −/− mice, as seen in a representative experiment shown in Figure 2. This lack of induction was observed at 2 h (Fig. 2) or at later time points up to 24 h after injection (data not shown), indicating that TNFR1 is of major influence during this process. However, when we analyzed the values obtained in different experiments involving a total of 23 mice, the very moderate effect of TNF in TNFR1 −/− mice became statistically significant, leaving open the possibility that TNFR2 also marginally contributes to the effects of TNF on platelets.
TNF is not acting on platelets directly
TNF did markedly reduce the survival of normal (+/+), labeled platelets when these platelets were injected into a +/+ host (Fig. 3,A). Similarly, TNF also reduced the survival of platelets from TNFR1 −/− donors when the platelets were injected into +/+ hosts (Fig. 3,B), indicating that platelet TNFR1 is not required for this process. In addition, the survival of platelets from +/+ donors was markedly reduced by TNF when injected in a +/+ host, but was not markedly reduced when such platelets were injected in TNFR1 −/− hosts. This observation indicates that it is the TNFR1 of the host, but not of the platelets, which is critical in TNF-induced thrombocytopenia (Fig. 3 C).
TNF-induced platelet consumption is associated with an increase of the localization of labeled platelets in the lung: thus, the injection of TNF immediately after the injection of labeled platelets increased the localization of platelets in the lung. For platelets from either TNFR1 −/− or +/+ donors, the localization was higher in TNFR1 +/+ than in TNFR1 −/− recipients (Fig. 4), further demonstrating that it is the TNFR1 phenotype of the host, but not of the platelets, which is critical for the TNF-induced platelet adhesion.
TNFRs are detectable on megakaryocytes, but not on platelets
The presence of the TNFRs was explored on the megakaryocytes present within the red pulp of the spleen by histochemistry, using the rabbit anti-TNFR Abs. Anti-TNFR1 Ab did stain megakaryocytes, while anti-TNFR2 did not stain more than the nonimmune Ab (Fig. 5). As an additional control, anti-TNFR1 Ab was assayed on the spleen from TNFR1 −/− mice and did not stain megakaryocytes more than the nonimmune control (data not shown).
Platelets were examined by flow cytometry after staining with anti-TNFR1 or anti-TNFR2 mAbs. mAbs to mouse TNFR1 or TNFR2 did not stain platelets more than a control isotype (Fig. 6). These Abs did significantly stain positive control cells, such as an endothelial cell line (Fig. 6). Similar results were obtained with the rabbit anti-TNFR Abs (data not shown). Anti-CD62 mAb, recognizing a determinant known to be present on platelets (17), also stained platelets (Fig. 6). These experiments indicate that TNFRs are not expressed on platelets at a level detectable by immunofluorescence methods.
The presence of TNF within the plasma induces thrombocytopenia that seems to be due to irreversible platelet consumption. The half-life of platelets is ∼2 days in mice, suggesting a platelet turnover rate of ∼2% per h (9). Thus, the loss of ∼30% of circulating platelets that is seen at 1 h after a TNF injection cannot be due to an effect of TNF on platelet production. The slow recovery of the thrombocytopenia indicates that it is not due to the recirculation of temporarily sequestered platelets, but rather to a reconstitution of the platelet pool by platelet production. This interpretation is also in accord with the survival of labeled platelets, which is markedly and irreversibly reduced by an injection of TNF (Fig. 3 A).
The effects of TNF are exerted by two membrane receptors, which are expressed to variable degrees on the majority of epithelial and mesenchymal cells (18). A study of mice with deletion of these receptors indicates that both platelet consumption and trapping in the inflamed lung are primarily mediated by the TNFR1 (Fig. 2). Indeed, the TNF-induced thrombocytopenia in TNFR2 −/− mice was of an amplitude similar to that seen in +/+ mice, indicating that the TNFR1 alone that is expressed on host cells is sufficient to trigger the events leading to thrombocytopenia and platelet adhesion. These observations are related to those observations indicating that TNFR1 is the main mediator of the inflammatory effects of TNF (12), and that platelets are participating in inflammatory reactions, including those associated with fibrogenesis (7, 19). TNFRs are detected mainly on cells of lymphohemopoietic origin according to histochemical methods, while the study of cell lines suggests that low numbers of TNFRs are detectable on various cell types, including epithelial and mesenchymal cells (18). We were able to detect the TNFR1 on megakaryocytes by histochemistry, raising the possibility that TNF influences megakaryocytopoiesis. Current evidence shows that TNF inhibits megakaryocytopoiesis in vitro (20), while it increases the number of megakaryocytes in the spleen in vivo (21). In contrast, TNFRs are not detectable on their progeny, which are platelets; i.e., they are present in very low numbers or are absent altogether. In this respect, platelets resemble another anucleated cell, the RBC, on which the TNFRs are not detectable (22). This finding does not completely exclude the possibility that TNF acts on platelets, since it has been observed with other cells that the response to TNF does not seem to correlate with the number of TNFRs present (22). However, this possibility is ruled out by platelet transfer experiments, which demonstrate that platelet consumption is triggered when TNFR1 is present on host cells, while the TNFR1 phenotype of platelets is unimportant.
It is generally believed that platelet consumption is the result of platelet activation that manifests itself in vitro by an aggregation or a change in the expression of cell surface proteins such as CD62P (17). While TNF is not known to activate platelets directly in vitro (Ref. 10 and our unpublished observations), the present results suggest that TNF probably does activate platelets in vivo. Therefore, it is very likely that TNF acts on the TNFR1 of cells that are capable of releasing a mediator with agonist activity for platelets. The task of identifying the important pathways leading to platelet activation might be difficult, since TNF can act on most cell types and induce a wide variety of responses, and also because platelet agonists are numerous (2). One possible cascade is the activation of the endothelium for the production of tissue factor, which results in the generation of thrombin, a well-known platelet agonist (23). This cascade reportedly involves TNFR1 (24), but its importance is doubtful under the present conditions, since TNF-induced thrombocytopenia is not associated with fibrin consumption (25) and cannot be attenuated by heparin (our unpublished observation). Another pathway might be the production of proteases (26), notably plasminogen activators, which increase the concentration of plasmin, another documented platelet agonist (27). Finally, the release of monoamines from mast cells also appears to contribute to the effects of TNF on platelets (Ref. 28 and our unpublished observations).
We thank Dr. H. Bluethman (Hoffmann-La Roche, Basel, Switzerland) for the gift of the TNFR-deficient mice, Dr. G. Wong (Genentech, San Francisco, CA) for the gift of the rabbit anti-TNFR Abs, C. Barazzone for a critical reading of the manuscript, and G. Brightouse and Y. Donati for their help in the preparation of the manuscript.
This work was supported by Grant 32 47284.96 from the Swiss National Science Foundation.
Abbreviation used in this paper: ACD, dextrose (anhydrous)-trisodium citrate·2H2O-citric acid·H2O-H2O.