Ligand Passing: The 75-kDa Tumor Necrosis Factor (TNF) Receptor Recruits TNF for Signaling by the 55-kDa TNF’ Receptor*

TO understand the role of the 75-kDa tumor necrosis factor (TNF) receptor in non-lymphoid cells, the cyto- toxic signaling and ligand binding activities of the 55-kDa (TNF-R1) and 75-kDa (TNF-R2) TNF receptors were investigated using agonist and antagonist antibodies specific for the two receptor types. This study indicates that although TNF-R2 can significantly reduce the TNF concentration required for cell killing, the mechanism by which this is accomplished is not through the generation of an intracellular signal by TNF-R2. Instead, TNF-R2 regulates the rate of TNF association with TNF-Rl, possibly by increasing the local concentration of TNF at the cell surface through rapid ligand association and dissociation. We propose that other cell-surface re- ceptors, such as the low affinity p75 nerve growth factor receptor, may utilize an analogous “ligand passing“ mechanism.

Rl, possibly by increasing the local concentration of TNF at the cell surface through rapid ligand association and dissociation. W e propose that other cell-surface receptors, such as the low affinity p75 nerve growth factor receptor, may utilize an analogous "ligand passing" mechanism.
Tumor necrosis factor (TNF),l a potent cytokine produced primarily by activated macrophages, elicits a large number of biological effects including hemorrhagic necrosis of transplanted tumors, cytotoxicity, and inflammatory, immunoregulatory, proliferative, and antiviral responses (Beutler and Cerami, 1988;Fiers, 1991;Goeddel et al., 1986;Old, 1988). The first step in the induction of the various cellular responses mediated by TNF is its binding to specific cell-surface receptors. Two distinct TNF receptors of -55 (TNF-R1) and 75 (TNF-R2) kDa have now been identified (Brockhaus et al., Hohmann et al., 1989), and human and mouse cDNAs corresponding to both receptor types have been isolated and characterized (Goodwin et al., 1991;Lewis et al., 1991;Schall et al., 1990;Smith et al., 1990).
A number of studies investigating the individual roles of the two TNF receptors have been reported. Both polyclonal and monoclonal antibodies (mAbs) directed against human TNF-R1 have been shown to behave as receptor agonists and to elicit several TNF activities such as cytotoxicity, fibroblast proliferation, resistance to chlamydiae, and synthesis of prostaglandin Ez (Engelmann et al., 1990;Espevik et al., 1990;Shalaby et al., 1990). Anti-TNF-R1 mAbs have also been described that effectively antagonize the TNF induction of many of these responses ; Thoma et al., 1990). In addition, polyclonal antibodies to both murine TNF-R1 and TNF-R2 have * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
f To whom correspondence should be addressed Dept. of Molecular cisco, CA 94080. Tel.: 415-225-1080;Fax: 415-225-6127. Biology, Genentech, Inc., 460 Point San Bruno Blvd., South San Fran-1 The abbreviations used are: TNF, tumor necrosis factor; TNF-R1, 55-kDa TNF receptor; TNF-R2, 75-kDa TNF receptor; mAbs, monoclonal antibodies; mTNF, murine TNF; hTNF, human TNF; CHX, CYcloheximide; IL-2, interleukin-2; NGF, nerve growth factor. been developed, and each has been shown to behave as a receptor-specific agonist and to induce a subset of mTNF activities (Tartaglia et al., 1991). Studies with the murine receptor agonist antibodies have demonstrated that the two receptors signal distinct TNF activities. TNF-R1 is responsible for signaling cytotoxicity and the induction of several genes, whereas TNF-R2 is capable of signaling proliferation of primary thymocytes and a cytotoxic T cell line (Tartaglia et al., 1991).
Paradoxically, several reports have described mAbs directed against human TNF-RZ that can partially antagonize the same TNF responses (e.g. cytotoxicity and NF-KB activation) that can be signaled through TNF-R1 (Hohmann et al., 1990;Naume et al., 1991;Shalaby et al., 1990). Furthermore, mTNF is much more effective than hTNF at killing some murine cell lines (Heller et al., 1992). This is also suggestive of TNF-R2 function since hTNF can bind murine TNF-RI, but not murine TNF-RZ (Lewis et al., 1991). The authors of several of these studies therefore concluded that there is redundancy in the function of the two TNF receptors. This model in which TNF-R2 signals many of the same activities as TNF-R1 is in marked contrast to the nonredundant signaling model used to explain data with receptor-specific agonists. To help reconcile these seemingly conflicting data, we have analyzed in detail the role TNF-R2 plays in influencing TNFs cytotoxic activity. Our findings indicate that TNF-R2 does not generate a signal that potentiates the cytotoxicity of TNF-R1, but instead passes TNF to TNF-R1, where signaling occurs.
EXPERIMENTAL PROCEDURES Reagerzts-Recombinant hTNF and recombinant mTNF (specific activity of >lo7 units/mg) were provided by the Genentech Manufacturing Group. The rabbit anti-murine TNF-R2 antibodies have been described previously (Tartaglia et al., 1991). Hamster mAbs against murine TNF-R1 (mAbs 170 and 176) and TNF-R2 (mAbs 32.4 and 54.7) were generated against the corresponding soluble receptor extracellular do-mains2 and supplied by K. Sheehan and R. Schreiber. Anti-TNF receptor mAbs 984 and 1040 inhibit the binding of TNF to human TNF-R1 and TNF-RZ, respectively, and have been described previously (Pennica et al., 1992a, 199213;Tartaglia and Goeddel, 1992). lZ6I-TNF was purchased from Amersham Corp.
Cytotoxicity Assays-Murine L929 cells (2 x lo4 celldwell) were seeded into 96-well microtiter plates in 100 pl of medium (low glucose Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 1% L-glutamine, 100 unitdm1 penicillin, and 100 pg/ml streptomycin (Life Technologies, Inc.)) and incubated for 24 h at 37 "C in a 5% C 0 2 atmosphere. The medium was then brought to 10 pg/ml cycloheximide (CHX), and the anti-TNF receptor antibodies or TNF was added to the wells and serially diluted. The plates were incubated for an additional 16 h (or for the indicated time period), and the viable cells were stained with 20% methanol containing 0.5% crystal violet. The dye was eluted with 0.1 M sodium citrate, 0.1 M citric acid in 50% ethanol, and absorbance was measured at 540 nm. HeLa.R2-1 cells (2 x lo4 cells/well) were seeded into 96-well microtiter plates in 100 pl of medium (5050 low glucose Dulbecco's modified Eagle's mediudam's F-12 medium supplemented with 10% fetal calf serum, 1% L-glutamine, R. F. Weber and D. V. Goeddel, unpublished data. 100 unitdml penicillin, 100 pg/ml streptomycin, and 400 pg/ml G418) and incubated for 24 h at 37 "C in a 5% CO, atmosphere. The medium was then brought to 10 pg/ml CHX, and the anti-TNF receptor antibodies or TNF was added to the wells and serially diluted. The plates were incubated for an additional 24 h, and the viable cells were assayed as described above. Reported values are the means of triplicate determinations.
TNF Association and Dissociation Experiment-For association kinetics, lo6 U937 cells in phosphate-buffered saline, 0.1% bovine serum albumin, 0.02% sodium azide (PBSA) were incubated with 80 PM lZ6I-TNF in a final volume of 200 pl on ice. At predetermined times, samples were centrifuged and then washed once with PBSA and recentrifuged. Nonspecific binding was determined in the presence of 0.5 p~ unlabeled TNF. All values are the means of duplicate determinations. Additional details of the binding assays are as described elsewhere (Pennica et al., 199213).
For dissociation kinetics, 2.5 x lo7 U937 cells in 5 ml of PBSA were incubated with 80 PM 1261-TNF on ice. After 3 h, 100 m unlabeled TNF was added, and 2 0 0 4 aliquota were removed at various times, centrifuged, washed, and recentrifuged. The time at which unlabeled TNF was added was taken as time 0. Nonspecific binding was determined through an identical assay in which 100 n~ unlabeled TNF was added at the beginning of the 3-h incubation. All values are the means of duplicate determinations. Additional details of the dissociation assay are as described elsewhere (Pennica et aL, 1992b).

Inhibition of TNF-mediated Cytotoxicity by TNF Receptor-
specific Antibodies-% help clarify the hnctional roles of the two TNF receptors, we examined the ability of anti-TNF receptor-specific mAbs to inhibit the killing of murine L929 cells by mTNF. Antibodies specific for murine TNF-R1 were found to completely inhibit cytotoxicity at all TNF concentrations tested. This result was seen both in a 9-h assay in the presence of CHX (Fig. lA) and in a 24-h assay in the absence of CHX mAb 54.7 (m), followed by treatment with the indicated concentrations of murine TNF for 24 h. Cell viability was determined as described previously (Tartaglia and Goeddel, 1992).
signaling cytotoxicity. Interestingly, saturating concentrations of anti-murine TNF-R2 antibodies partially inhibited cytotoxicity ( Fig. 1, A and B ) . Significant inhibition by the anti-TNF-R2 antibodies was not observed at high TNF concentrations, but nevertheless, these antibodies caused a significant shift in the TNF dose response. This ability of TNF-R2 to reduce the TNF concentrations required for killing can also be seen in a comparison of the cytotoxicity dose-response curves of murine and human TNFs (Fig. 2). Human TNF binds only murine TNF-R1 and does not recognize murine TNF-R2 (Lewis et al., 1991). Therefore, the hTNF killing dose-response curve is shifted rightward relative to the mTNF killing curve. Similar effects of the receptor-specific antibodies and ligand species specificity were also observed in the NIH 3T3 cell line (data not shown).
The partial inhibition of cytotoxicity by anti-TNF-R2 antibodies and the complete inhibition by anti-TNF-R1 antibodies are consistent with either of two possible models. In the first model, both receptors are capable of activating signal-transducing pathways that ultimately lead to cell death. At high receptor occupancy, signals from TNF-R1 alone can result in cell killing, whereas signals from TNF-R2 alone are insufficient.
However, at lower receptor occupancy, signals from TNF-R2 could synergize with signals from TNF-R1 and result in increased TNF sensitivity. In the second model, only signals generated by TNF-R1 result in cell killing, and signals from TNF-R2 play no role. However, in this model, TNF-R2 facilitates the binding of TNF to TNF-R1, resulting in the triggering of TNF-R1 at lower TNF concentrations.

Agonist Antibodies to TNF-R2 Behave as Antagonists in Cy-
totoxicity Assay-% begin to distinguish between a signaling or a binding accessory role of TNF-R2, we first tested whether antibodies capable of specifically activating TNF-R2 would act as TNF antagonists or agonists in the L929 cytotoxicity assay (Fig. 3). We have previously shown that these antibodies are capable of potently activating TNF-R2 and inducing the proliferation of murine thymocytes and a T cell line (Tartaglia et al., 1991). However, in a cytotoxicity assay, these TNF-R2-activating antibodies do not act as agonists and do not shift the doseresponse curve to the left. Instead, they act as antagonists and push the dose-response curve to the right. The effect of the anti-TNF-R2 polyclonal antibodies in this assay again points to a role of TNF-R2 in affecting cytotoxicity. However, the fact that antibodies capable of activating TNF-R2 are acting as antagonists argues against a signaling role.

Occupancy of TNF-R2 by TNF Does Not Enhance the Killing Signal Generated by a Weak TNF-R1 Agonist-% test whether occupancy of TNF-R2 by TNF can generate a signal that syn-
ergizes with a weak TNF-R1 signal, we examined the effect of increasing concentrations of murine TNF on L929 cells in which access to TNF-R1 was blocked by a saturating concentration of a weak TNF-R1 agonist antibody (Fig. 4). If TNF-R2 can generate a cytotoxicity signal, then mTNF would be ex- tions of murine or human TNF and 10 pg/ml CHX for 10 h. Cell viability was determined as described previously (Tartaglia and Goeddel, 1992). of L929 cells were treated with the indicated concentrations of hTNF for 1 h at 4 "C. The cells were then further treated for 1 h at 4 "C with an excess of an antagonist antibody to TNF-R1 (mAb 170, 1:lO dilution of hybridoma supernatant). 1 ng/ml mTNF (a dose sufficient to achieve 100% killing in the absence of blocking antibody) was then added to one plate (O), whereas the other was not further treated (0). 10 pdml CHX was added to all cells, and the temperature of the experiment was shiRed to 37 "C for 10 h. Cell viability was determined as described previously (Tartaglia and Goeddel, 1992). were performed in the presence of CHX and that were >12 h in duration). Cell viability was determined as described previously (Tartaglia and Goeddel, 1992). pected to enhance the incomplete killing induced by the weak TNF-R1 agonist. However, if the role of TNF-R2 is to facilitate binding to TNF-R1, no effect of TNF-R2 occupancy should be seen since access to TNF-R1 is blocked. Under the conditions of this assay, the weak TNF-R1 agonist alone was found to kill -50% of the cell population (Fig. 4). However, further addition of mTNF (which has access to TNF-R2) does not enhance this partial killing, arguing against the existence of a TNF-R2 signal that synergizes with a signal from TNF-R1 in the induction of cell killing.
TNF"R2 Does Not Enhance Cytotoxicity if TNF Is Forced to Occupy the Tiuo TNF Receptors Independently-If TNF-R2 contributes to cytotoxicity by facilitating the binding of TNF to TNF-R1, then it should not enhance cytotoxicity when TNF is forced to bind the two TNF receptors independently (only one receptor type is accessible to a given TNF molecule). However, if TNF-R2 mediates cytotoxicity through a signaling mechanism, independent occupation of the two receptors by TNF should not eliminate TNF-R2's enhancing function. To achieve this independent occupation of the two TNF receptors, L929 cells were treated with increasing concentrations of hTNF for 1 h at 4 "C. As hTNF cannot bind mouse TNF-R2 (Lewis et al., 1991), this treatment results in the occupation of only TNF-R1. Further treatment of the cells with an excess of an antagonist antibody to TNF-R1 blocks further access to TNF-R1. Murine TNF (which now has access to only TNF-R2) was added, and the temperature was shifted to 37 "C. The addition of mTNF under these conditions, in which transfer from TNF-R2 to TNF-R1 cannot take place, did not potentiate the killing that is obtained by occupancy of TNF-R1 alone (Fig. 5). This result indicates that the binding of TNF to TNF-R2 does not enhance cytotoxicity unless the same TNF molecules that bind to TNF-R2 also have access to TNF-R1 and thus argues against a direct signaling role of TNF-R2. TNF"R2 Increases Association Rate of TNF Binding to TNF-R1-The biological effects described above are consistent with TNF-R2 facilitating the binding of TNF to TNF-R1. In an attempt to provide direct biochemical data for this phenomenon, the association rates of TNF binding to the two TNF receptors were examined.
The association of 12sI-TNF with TNF-R1 alone, with TNF-R2 alone, and with the two receptors simultaneously was determined using U937 cells treated with blocking antibodies to TNF-R2, with blocking antibodies to TNF-R1, and in the absence of blocking antibodies, respectively (Fig. 6A). The experimentally observed rate constants (koba) were derived from the slope of semilogarithmic transformed plots of the initial association data and were calculated to be 0.002 and 0.037 min" for TNF-R1 and TNF-R2, respectively (Fig. 6B). Thus, when the two TNF receptors are isolated, TNF associates nearly 20 times as rapidly with TNF-R2 as compared to TNF-R1. The association rate of TNF binding to TNF-R1 in the presence of TNF-R2 was derived from the binding that is inhibited by the anti-TNF-R1 antibody (Fig. 6 A ) (differential between TNF binding to both receptors simultaneously uersus TNF-R2 only) and is also shown in Fig. 6B. A comparison of the slope of this plot (0.023 min") with the slope of the association plot for TNF-R1 alone (0.002 min-l) shows that accessible TNF-R2 increases the rate of association of TNF with TNF-R1 -10-fold. This result is consistent with a previously unexplained observation by Hohmann et al. (1990), who noted that the Kd for all TNF-binding sites on HL-60 cells was -7-fold lower than that measured for isolated TNF-R1 sites. It is likely that this ability of TNF-R2 to enhance the binding of TNF to TNF-R1 explains the enhanced cytotoxicity observed at low TNF concentrations since TNF-R2 is not capable of generating a synergistic signal.
Possible Mechanism of TNF-R2 Accessory Role-The mechanism by which TNF-R2 can affect the association rate of TNF binding to TNF-R1 is also of interest. To address this, we have attempted to distinguish between two different models. In the first model, TNF can form heterocomplexes between the two different TNF receptors. This model is attractive because it is easy to conceptualize, and also there is a well characterized precedent for it in the case of interleukin-2 (IL-2) binding to its receptors (Lowenthal and Greene, 1987;Wang and Smith, 1987). IL-2 binding to the p55 IL-2 receptor is rapid, whereas IL-2 binding to the isolated p75 IL-2 signaling receptor is slow. When both receptors are present on the same cell, IL-2 binds rapidly to the p55 chain, thereby restricting 1L-2 to a twodimensional search for the p75 chain that ends in rapid IL-2.p55ep75 heterocomplex formation. In the alternative model, TNF cannot form heterocomplexes between the two different receptors. In this case, TNF-R2 would function to increase the local concentration of TNF at the cell surface by rapid ligand association and dissociation. We and others have demonstrated previously that TNF can induce the formation of homocomplexes of both TNF-R1 and TNF-R2 (either two or three receptordI'NF trimer) (Loetscher et al., 1991;Pennica et al., 1992aPennica et al., , 1992b). However, cross-linking and immunoprecipitation experiments have so far been unable to demonstrate the formation of TNF receptor heterocomplexes on the cell surface even though homocomplex formation is readily detectable under these same condition^.^.^ These negative data have forced us to consider the model of TNF-R2 accessory function in the absence of heterocomplex formation. Although this model is difficult to test directly, it does have certain predictions. One prediction is that the dissociation of TNF from TNF-R2 is rapid enough to account for the observed facilitation of binding to TNF-R1. An example of the rapidity of this binding facilitation i s shown in Fig. 6A. Very little TNF associated with TNF-R1 in the absence of TNF-R2 at the 30min time point, yet when TNF-R2 was accessible, binding to TNF-R1 was almost saturated at the 30-min time point (as judged by the binding inhibited by a TNF-R1 antagonist antibody). Another prediction of this model is that the dissociation rate of TNF from TNF-R2 is independent of the presence of TNF-R1. This would not be a prediction of the heterocomplex formation model, in which heterocomplex formation is likely to D. Pennica  alter the dissociation rate of TNF from TNF-R2, as has been observed for the dissociation of IL-2 from the p55 chain of the IL-2 receptor system (Lowenthal and Greene, 1987;Wang and Smith, 1987).
To test these predictions, we examined the dissociation of TNF from the TNF receptors on HeLa, 293/R2, and U937 cells. HeLa cells have been shown to express exclusively or predominantly TNF-R1 , and we have therefore used TNF binding to HeLa cells as a model of TNF-R1 binding. 293/R2 cells have been engineered to express TNF-R2 to 100fold higher levels than endogenous receptors (Pennica et al., 1992b) and were used as a model for TNF binding to TNF-R2. Dissociation of TNF from 293/R2 cells (TNF-R2) was found to be surprisingly rapid ( t y 2 s 10 min) for a high affinity receptor (Fig. 7A). Therefore, the rapid rates of both association with and dissociation from TNF-R2 could account for its observed TNF-R1 binding facilitation. In contrast, dissociation of TNF from HeLa cells (TNF-R1) was found to be quite slow, with a half-life of >3 h (Fig. 7A).
The dissociation of TNF from U937 cells for the two TNF receptors simultaneously and for both receptors individually was also examined (Fig. 7B). Again, dissociation from TNF-R2 was found to be rapid, whereas dissociation from TNF-R1 was slow. It is also important to note the shape of the dissociation curves of TNF bound to only TNF-R2 versus TNF bound to both receptors. The parallel nature of these dissociation curves (whose absolute values are shifted as a result of TNF-R1 binding) indicates that the dissociation of TNF from TNF-R2 is unaffected by TNF-R1 and therefore argues against heterocomplex formation. (0) for 1 h at 0 "C. 0.08 m '=I-TNF was then added at 0 "C for 3 h. A 1000-fold excess of unlabeled TNF was then added, and aliquots were removed at the indicated time points. Aliquots were washed as described under "Experimental Procedures," and radioactivity was determined.

Effect of TNF-R2 Expression Level on Binding Facilitation
"Most cell lines in which we have detected enhanced triggering of TNF-Rl via TNF-R2 express TNF-R2 on the order of a few thousand receptordcell. To determine whether high level expression of TNF-R2 can further accentuate its accessory effects, we examined a HeLa cell line (HeLa.R2-1) (Tartaglia et al., 1993) that expresses high levels of cell-surface TNF-R2 (52,000 TNF-R2 moleculedcell) from a transfected cDNA. HeLa cells were chosen for this transfection analysis because parental HeLa cells do not express significant levels of TNF-R2 , and their sensitivity to TNF is not affected by TNF-R2 antagonist antibodies (data not shown). Surprisingly, high level expression of TNF-R2 actually reduced sensitivity at low TNF concentrations, as determined by a comparison of TNF-induced cytotoxicity in HeLa.R2-1 cells in the presence and absence of a TNF-R2 antagonist antibody (Fig.   a). The inhibitory effects of TNF-R2 at low (and not high) TNF concentrations would be consistent with TNF-R2 titrating the TNF from the assay medium, thus leaving less available to reach TNF-R1. We therefore examined how narrow the window of TNF-R2 expression levels must be to effectively concentrate TNF at the cell surface and yet not reduce the available TNF concentration by titration. HeLa.R2-1 cells were treated with a low TNF concentration (0.01 ng/ml), and the effective concentration of TNF-R2 was incrementally increased by addition of decreasing concentrations of a TNF-R2 antagonist antibody (Fig. 8B) were then further treated for 24 h with 10 pg/ml CHX and the indicated viously (Tartaglia and Goeddel, 1992). Data shown are the means of concentrations of TNF. Cell viability was determined as described prewith the indicated concentrations of TNF-R2 antagonist mAb 1040. triplicate determinations. B, HeLa.R2-1 cells were pretreated for 1 h Cells were then further treated with a combination of 10 pg/ml CHX and 0.01 ndml TNF. Cell viability was determined as described previously (Tartaglia and Goeddel, 1992). Data shown are the means of triplicate determinations.
trations. Together, these results indicate that low level expression of TNF-R2 can facilitate triggering of TNF-R1, but high level expression of TNF-R2 is detrimental to TNF-R1 triggering at low TNF concentrations. Thus, cells must carefully regulate the level of TNF-R2 expression to optimize enhanced triggering of TNF-R1. DISCUSSION Although not yet systematically investigated, the majority of cell types and tissues appear to express both TNF receptors (Lewis et al., 1991;Schall et al., 1990;Smith et al., 1990). Therefore, it is important to understand the individual roles of these two receptors in cell signaling, both to assess whether greater clinical specificity of TNF actions can be realized at the level of receptor activation and to better understand the biology of this important cytokine. To help address this question, several groups have generated antibodies that are specific for either TNF-R1 or TNF-R2. However, despite the availability of these powerful immunological tools, the individual signaling roles of the two TNF receptors are still under considerable debate.
A significant breakthrough in our understanding of TNF receptor function came from the observation that antibodies against TNF-R1 can act as specific agonists for this receptor and can signal several diverse TNF activities such as cytotoxicity, fibroblast proliferation, resistance to chlamydiae, and synthesis of prostaglandin Ez (Engelmann et al., 1990;Espevik et al., 1990;Shalaby et al., 1990). These studies demonstrated that activation of TNF-R1 alone is sufficient to mimic many TNF activities in diverse cell types. Agonist antibodies specific for TNF-R2 can also mimic TNF and induce a small subset of TNF activities including the enhancement of T cell proliferation (Tartaglia et al., 1991). However, TNF-R2 agonists have so far been demonstrated to directly signal only in some lymphoid cell types and do not induce the same set of activities induced by TNF-R1 agonists even in cells expressing similar levels of both receptors (Tartaglia et al., 1991;Wong et al., 1992aWong et al., , 1992b. The role of the relatively high levels of TNF-R2 in non-lymphoid cells is therefore still an important and unresolved issue. Several reports have described mAbs directed against TNF-R2 that can partially antagonize the same TNF responses that are induced by TNF-R1 agonists (Hohmam et al., 1990;.-These reports suggested that there is redundancy in the function of the two receptors. However, a redundant signaling model is clearly in conflict with the demonstration, USing receptor-specific agonist antibodies, that the two receptors signal distinct and largely nonoverlapping sets of activities. Central to this controversy is the role of the two TNF receptors in mediating cytotoxicity. In cell lines examined so far, only anti-TNF-R1 (and not anti-TNF-R2) antibodies are able to mimic TNF"s cytotoxic activity, even in cell lines expressing both receptors (Tartaglia et al., 1991(Tartaglia et al., , 1993Wong et al., 1992a). Nevertheless, there are clear data showing that antibodies that block binding of TNF to TNF-R2 partially inhibit TNF killing .6 In addition, several groups have identified murine cell lines that are killed much more readily by murine TNF than by human TNF (Fransen et al., 1986;Heller et al., 1992;Rice et al., 1990), which might also be indicative of some role of TNF-R2 in cell killing.
To help resolve these conflicting viewpoints, we have analyzed in detail the role of TNF-R2 in influencing cytotoxicity. We observed that mAbs to TNF-R1 could completely inhibit cytotoxicity at all tested TNF concentrations, indicative of an absolute requirement of the binding of TNF to TNF-R1. HOW-R. Schreiber, personal communication. ever, in agreement with previous reports using the human U937 cell line (Shalaby et al., 19901, we observed that monoclonal antibodies that block binding of TNF to TNF-R2 partially inhibit TNF toxicity at low TNF concentrations. Interestingly, we found that anti-TNF-R2 polyclonal antibodies capable of activating TNF-R2 also exhibited antagonistic activity with respect to cytotoxicity. This result suggests that while the binding of TNF to TNF-R2 may be important for facilitating cytotoxicity, the activation of TNF-R2 and the generation of a TNF-R2 signal are unlikely be involved. It is therefore possible that the binding of TNF to TNF-R2 serves only to facilitate the binding of TNF to TNF-R1. To rule out the possibility that TNF initiates a signal through TNF-R2 that agonist antibodies cannot mimic, we tested whether the binding of TNF itself to TNF-R2 could potentiate the partial killing mediated by a weak TNF-R1 agonist antibody or a TNF-R1-specific ligand. Under these conditions, in which the TNF molecules binding TNF-R2 do not have further access to TNF-R1, no effects of TNF-R2 occupancy were observed. Taken together, these results argue against a Signaling role of TNF-R2 in cytotoxicity and suggest that the role of TNF-R2 is to channel TNF to TNF-R1, which is then responsible for all signal generation. To examine the biochemical mechanism by which TNF-R2 might facilitate the activation of TNF-Rl by TNF, we examined the binding of 1261-TNF to its two receptors. This study showed that TNF associates much more rapidly with TNF-R2 than with TNF-R1. However, when TNF has access to both receptors on the same cell, the presence of TNF-R2 greatly enhances the rate of association of TNF with TNF-R1. This enhancement is not likely to be a result of an intracellular signal by TNF-R2 since it occurs at 0 "C and is not mimicked by TNF-R2 agonist antibodies. The ability of TNF-R2 to enhance association of TNF with TNF-R1 is likely to account for the many inhibitory effects observed with anti-TNF-R2 antibodies in non-lymphoid cell types. It has also been proposed, based on ligand species specificity arguments, that TNF-R2 contributes to the in vivo toxicity of TNF (van Ostade et al., 1993). The ability of TNF-R2 to facilitate TNF-R1 triggering may be the primary mechanism for this enhanced toxicity.
An important feature of this accessory role of TNF-R2 is that it does not require intracellulai signaling by TNF-R2, despite the observations that antagonists to either TNF receptor can inhibit the same cellular response. This model is therefore also consistent with data on the nonoverlapping activities of TNF-R1 and TNF-RP agonists (Tartaglia et al., 19911, the completely neutralizing activities of TNF-R1 antagonists observed with cell lines expressing both receptors (Tartaglia et al., 1993;Thoma et al., 1990;this study), and the lack of homology between the two TNF receptor intracellular domains.
The mechanism by which TNF-R2 facilitates association of TNF with TNF-R1 may also be novel as it does not appear to involve the stable association of the two receptor types with one another to form heterocomplexes. Chemical cross-linking and immunoprecipitation experiments, which can readily demonstrate TNF-induced aggregation of multiple TNFs-R1 or TNFs-R2 in homocomplexes, have not detected heterocomplexes on the cell surface under these same In addition, the kinetics of dissociation of TNF from a cell expressing both TNF receptor types is a simple sum of the dissociation from the two receptors individually. This is in marked contrast to receptor chains that form heterocomplexes such as the IL-2 receptor system (Lowenthal and Greene, 1987;Wang and Smith, 1987). Why TNF receptors seem only to form homocomplexes and not heterocomplexes is also of interest. One possibility is the 57-amino acid spacer sequence between the transmembrane region and the presumed cysteine-rich ligandbinding domain of TNF-R2 (Smith et al., 1990). This spacer region is heavily 0-glycosylated and forms an extended structure (Pennica et al., 1993), which may also be quite rigid (Jentoft, 1990). The ligand-binding domain of TNF-R2 may therefore be further from the membrane surface than the ligand-binding domain of TNF-R1, which contains no such spacer region. This would make it unlikely that the compact TNF trimer could simultaneously interact with both receptors. It is possible that preventing heterocomplex formation is critical to achieve proper TNF responses since a complex between TNF-R2 and TNF-R1 would likely result in a dominant negative effect on TNF-R1 signaling, as has been observed with complexes between truncated TNF-R1 and full-length TNF-R1 (Tartaglia and Goeddel, 1992).
The kinetic properties of TNF binding to TNF-R2 suggests a mechanism by which TNF-R2 might increase the apparent association rate of TNF binding to TNF-R1. In this model, the rapid association of TNF with and the dissociation of TNF from TNF-R2 would serve to increase the local concentration of TNF at the cell surface, resulting in more rapid association with TNF-R1. This "passing" model does not require heterocomplex formation. It is therefore consistent with data in several studies (Engelmann et al., 1990;Loetscher et al., 1991;Tartaglia and Goeddel, 1992) that suggest that TNF, which exists as a trimer, must cross-link three molecules of TNF-R1 to deliver an effective signal. However, the possibility that TNF-R2 facilitates association of TNF with TNF-R1 by formation of a transient heteroreceptor complex cannot yet be ruled out.
The physiological importance of the ability of TNF-R2 to regulate TNF-R1 triggering is also of interest. In vivo, TNF is often present at rather low concentrations and is also rapidly cleared. It is therefore possible that many cells are exposed to TNF under conditions in which the ligand concentrating effects of TNF-R2 are essential for proper TNF-R1 responses. Therefore, the expression of a receptor capable of rapid ligand association and dissociation provides a novel mechanism for increasing a cell's sensitivity to a transient flux of a regulatory molecule. Rapid shedding of the TNF-R2 extracellular domain (Pennica et al., 1992b) might then be a mechanism for a cell to reduce sensitivity. In some cases, dramatically reduced TNF sensitivity might be achieved by expression of TNF-RP at higher than optimal levels (as suggested by studies on high level expression of TNF-R2 in HeLa cells (Fig. 8)). Such mechanisms may, in fact, be superior to regulating the level of TNF-R1, where very high expression would be required to obtain cellular responses at TNF levels >lOO-fold below its Kd for TNF-R1, as is the case with many cellular responses to TNF.
Whether similar mechanisms of regulating ligand binding are used by other receptor systems remains to be established. However, it seems unlikely that the TNF system alone would utilize this mechanism to control signaling sensitivity. Several known membrane receptors, whose exact functions remain obscure, could potentially fulfill a ligand passing or concentrating role analogous to that of TNF-R2. The function of the low affinity p75 NGF receptor is currently unresolved since most of the direct signaling of the neurotrophic factors appears to be mediated by the trk family of receptors (Jing et al., 1992). Nevertheless, the p75 NGF receptor still plays an important role in neural development (as demonstrated with targeted mutations in mice) (Lee et al., 1992) and, under at least some conditions, can increase the binding of NGF to trk (Hempstead et al., 1991). Interestingly, the binding kinetics of NGF to the p75 NGF receptor is very fast relative to trk binding (Hartman et al., 1992). In addition, there is no evidence to indicate that the p75 NGF receptor forms a heterocomplex with the trk receptors (Jing et al., 1992). It is therefore possible that rapid dissociation from the p75 NGF receptor concentrates neurotrophic factors near the cell surface. It has also been suggested that the transforming growth factor+ type I11 receptor serves a ligand presentation rather than a signaling role (Lopez-Casillas et al., 1991;Wang et al., 1991). Although information on binding kinetics and complex formation is not yet available for this receptor, ligand passing is a potential mechanism of action. In the case of the IL-1 receptor system, only the type I receptor has been shown to be active in signal transduction (Stylianou et al., 1992). The IL-1 type I1 receptor, which dissociates from IL-1 more rapidly ( H o d and McCubrey, 1989), may therefore serve to regulate the binding of IL-1 to the type I receptor. Further studies on the properties of these and other receptor systems are needed to establish the generality of regulating ligand binding in this manner.