Use of a Sensitive Receptor Binding Assay to Discriminate between Full-length and Truncated Human Recombinant Tumor Necrosis Factor Proteins*

A radioreceptor assay (RRA) capable of detecting picomolar concentrations of human recombinant tumor necrosis factor (TNF) was used to compare the relative binding affinities of genetically engineered full-length and truncated TNF proteins. The specific cell-surface receptors for TNF present on the human cervical carcinoma cell line ME- 180 were characterized 8.8 having a Kd of 0.2 nM and a density of 2700 sites/cell. Conditions were then defined for an RRA that maximized the specific binding of ”‘1-TNF to this adherent cell line. Incubation of ME-180 cells with ”‘I-TNF at 37 “C in the presence of 0.02% sodium azide resulted in a 4-fold increase in assay sensitivity and a doubling of specific counts bound, as compared to binding done at 4 “C with or without sodium azide. Inhibition of recep- tor-ligand internalization under these conditions was a likely reason for the increases. This system was utilized to compare low concentrations of the full- length TNF protein and a genetically altered TNF protein (mutein) which lacks the 10 N-terminal amino acids and contains an N-terminal methionine. Previous studies showing the truncated TNF to be 2- to 3-fold lower in cytotoxic activity on a variety of tumor cell lines were corroborated by our findings that the mutein was also three and one-half times lower in relative affinity for the TNF receptor on ME-180 cells. These results suggest a possible role for these residues in receptor binding and illustrate the use of a highly sensitive RRA

* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The abbreviations used are: TNF, recombinant human tumor necrosis factor: SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis; DPBS, Dulbecco's phosphate-buffered saline; RRA, radioreceptor assay; EGF, epidermal growth factor; NaN3, sodium azide. ing normal tissues (1,2). This prompted efforts toward purification of TNF from several species (3-6), culminating in the cloning and expression in Escherichia coli of human recombinant TNF (6)(7)(8). Further studies made possible by the availability of the highly purified, cloned TNF have confirmed its i n vitro cytostatic and cytotoxic effects (6,9). In addition, effects such as stimulation of eosinophil toxicity against parasites (lo), induction of HLA expression in endothelial cells and fibroblasts (111, and suppression of lipoprotein lipase activity in 3T3-U cells (12) suggest other possible roles of TNF.
The recent availability of deletion mutants of TNF' has allowed investigation of the region(s) of the molecule that may be important for cytotoxic activity (13). We report here the development of a radioreceptor assay (RRA) for human recombinant TNF based on the competitive binding of the ligand to the human cervical carcinoma cell line ME-180. This highly sensitive RRA was used to determine the relative binding affinities of genetically engineered full-length and truncated TNF proteins and to illustrate a correlation between the cytotoxic activity and binding ability of these two proteins.

MATERIALS AND METHODS
Cell Culture-The cervical carcinoma cell line ME-180 was maintained in subconfluent cultures in 175 cm2 tissue culture flasks. The cells were grown in McCoy's 5A medium containing 10% fetal bovine serum, 2 mM glutamine, and 1% Fungi-Bact (Irvine Scientific).
TNF-Human recombinant TNF and deletion mutants were expressed in E. coli and purified to >90% homogeneity as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (6).
The sequence of the first 16 residues from the amino terminus of purified TNF was determined in a Beckman 890C Sequenator with 0.1 M Quadrol buffer. Phenylthiobydantoin-derivitized amino acids were identified by reverse-phase high pressure liquid chromatography on an Altex Ultrasphere ODS column (4.6 X 250 mm) at 45 "C. Elution was at 1.3 ml/min with the following program: 20.6 min from 100% buffer A to 40% buffer B, 3.0 min at 40% buffer B, and 8.4 min from 40 to 70% buffer B, where buffer A was 0.0115 M sodium acetate, 5% tetrahydrofuran, pH 5.11, and buffer B was 10% tetrahydrofuran in acetonitrile.
Purified TNF was radioiodinated using IODO-BEADS (Pierce Chemical Co.) (14). Briefly, 10-20 pg of carrier-free TNF was added to 2 mCi of NaIz5I (New England Nuclear) for 10 min at room temperature. The IZ5I-TNF was separated from free iodide by filtration on a Sephadex G-25 column (Pharmacia P-L Biochemicals) equilibrated with Dulbecco's phosphate-buffered saline (DPBS) containing 0.2% bovine serum albumin and 0.02% sodium azide (NaN3). Autoradiography of lz5I-TNF following SDS-PAGE showed that the iodinated protein and the unlabeled TNF migrated in the same fashion. The majority (50-100%) of the initial bioactivity was recovered, as determined by the L929 cytotoxicity assay (6).The specific activity of the labeled TNF was 0.7-7 X 10' cpm/pg (0.71-7.1 mCi/ pmol).
Binding Studies-ME-180 cells were grown to 90% confluence in T-175 flasks. The cells were then trypsinized and seeded in 24-well tissue culture plates at 2.5 X lo6 cells/well and incubated overnight at 37 "C, 5% CO,. On the day of the assay, the culture medium was aspirated off the cells and increasing concentrations of TNF were added to duplicate wells. TNF dilutions were added as a 0.5-ml volume in binding buffer (DPBS containing 0.2% bovine serum albumin). Iz5I-TNF in 0.5 ml was added to each well, and the plates were incubated an additional 2 h. The cell monolayers were washed three times with binding buffer, solubilized with DPBS containing 1 N NaOH and 1% SDS, and bound radioactivity was determined in a y counter. Duplicate determinations varied by less than 10%. The amount of bound IZ6I-TNF sensitive to acid treatment was determined by adding 1 ml/well of 0.05 M glycine, 0.15 M NaCl, adjusted to pH 3.0 with HCl (15). After 5 min at 4 "C, the solution was removed and replaced with another 1 ml of solution for 3 min. The cells were then washed, and bound radioactivity was determined.

RESULTS
Protein Sequencing-The first 16 amino acids of the -10 mutein were found to be as predicted by the DNA sequence of the full-length protein (Fig. l ) , except for the presence of an N-terminal methionine. Removal of the methionine during bacterial processing of the protein is mediated by the enzyme methionine aminopeptidase, the activity of which is dependent upon the residue following that methionine. The removal of methionine in the full-length T N F where the penultimate residue is valine and the presence of methionine in the truncated T N F where the penultimate residue is lysine are consistent with the findings of Tsunasawa et al. (16).
Equilibrium Saturation Binding-The ME-180 cervical carcinoma cell line has previously been shown to be sensitive to the cytotoxic effects of T N F (9) and to possess high affinity T N F binding sites (17). To confirm these findings and to help ascertain the potential usefulness of this cell line in a RRA, saturation binding studies and Scatchard analysis were performed (Fig. 2). lZ5I-TNF bound to cells in a specific manner, as seen by the displacement of up to 87% of the total counts by a 1000-fold excess of unlabeled ligand. Iz5I-TNF specific binding was also saturable, reaching a maximal level at 2. RRA Studies-Studies were then carried out to maximize the binding of T N F t o ME-180 cells under the nonsaturating concentrations necessary for optimal assay sensitivity. Thus, the effects of temperature and the metabolic inhibitor sodium azide were investigated. T N F dilutions were prepared in binding buffer with or without 0.02% sodium azide, and binding was carried out at 37 or 4 "C as described under "Materials and Methods." Fig. 3A illustrates the increase in assay sensitivity that resulted from the addition of sodium azide at 37 "C. There was a 4-fold decrease in the amount of unlabeled ligand required to displace 50% of the lz5I-TNF specifically bound. The increase in sensitivity has been as much 7-fold in other experiments. Kinetic equilibrium studies have shown maximum specific binding to occur in 2 to 3 h at 37 "C with sodium azide under these conditions. Sodium azide had no effect on assay sensitivity a t 4 " C 3 There was also a 90% increase in the number of specific counts bound (total minus nonspecific) a t 37 "C when sodium azide was present (4765 f 402) versus when it was absent (2509 f 65). In contrast, there was little difference a t 4 "C when sodium azide was present (2556 f 99) or absent (3122 f 166).
To investigate the possible reason(s) for the increased bind-Data not shown.
ing and assay sensitivity in the presence of sodium azide, the effects of acid treatment on specifically bound counts were studied. ME-180 cells were incubated with 0 or 1000 ng/ml T N F (approximately 400-fold excess "cold" ligand) with or without sodium azide at 37 or 4 "C for 2 h. Radiolabeled ligand was then added for an additional 2 h. Nonspecific counts were subtracted from the total to derive specific counts bound. The sensitivity of specifically bound lZ5I-TNF to removal by p H 3 treatment was then used to determine if the ligand was surface-associated (15). As seen in Fig. 3B, the percentage of acid-sensitive counts bound at 4 "C with and without sodium azide was 68 and 70%, respectively. This agrees well with the values obtained for acid-removable binding by other TNFsensitive cell lines (18). Binding in the absence of sodium azide at 37 "C resulted in a marked decrease of acid-sensitive counts to 21%, suggesting internalization of the majority of specifically bound lZ5I-TNF. However, including azide at 37 "C prevented internalization and resulted in virtually the same percentage of Y -T N F bound at the cell surface (65%) as at 4 "C.
Use of the RRA to Determine Relative Binding Affinities-This assay system was then utilized to compare the relative binding affinities of the full-length recombinant T N F used above, termed 711, and the novel mutant TNF protein shown in Fig. 1 that lacks the 10 N-terminal amino acids, termed 742. Previous comparisons of the biological activities of the two proteins had shown 742 to be up to three times less cytotoxic than 711 on ME-180 cells and eight other human cell lines (13). To determine if the binding affinity of the protein was also altered by the deletion of these 10 residues, the ability of the two proteins to compete with iodinated 711 protein was compared. In the experiment illustrated in Fig. 4, 711 displaced 50% of the specifically bound lZ5I-TNF at a concentration of 0.32 ng/ml. In contrast, 742 showed nearly a 5-fold lower ability to compete for the same binding sites on ME-180 cells, displacing 50% of the counts bound at a concentration of 1.5 ng/ml. In four experiments of this type, 50% displacement occurred at 0.33 f 0.04 ng/ml (mean ? S.D.) of full-length T N F uersus 1.16 +. 0.24 ng/ml of truncated T N F (p < 0.005), a difference of approximately 3.5-fold. In all experiments the linear range of the curve occurred over approximately a 15-to 30-fold concentration of competing TNF. Kinetic equilibrium studies showed that the apparent difference in the relative binding affinities of the TNF proteins was not due to a difference in their rate of binding. Also, in contrast to the results obtained with the -10 mutein, muteins lacking the first 4, 7, or 8 N-terminal residues were found to exhibit binding competition curves which were superimposable on that of the full-length p r~t e i n .~

DISCUSSION
The measurement of cytokines and cytotoxins such as T N F has typically begun with bioassays and then progressed to more specific radioimmunoassays and radioreceptor assays as the molecule becomes more well defined. We report here a competitive binding assay for TNF that provides greater specificity than cytotoxicity assays that in contrast cannot discriminate between TNF and other cytotoxic factors that may bind to distinct receptor sites. Since the RRA specificity is dependent upon receptor binding, we were also able to detect purified native TNF-P (lymph~toxin),~ in agreement with a recent report which indicates that these two factors share a common receptor (17). The TNF RRA using ME-180 cells does not require lengthy periods for cell growth and differentiation after plating, as with a receptor-binding assay using murine 3T3-Ll cells (22). Also, as opposed to the murine

FIG. 1. N-terminal sequence analysis of the -10 TNF mutein and comparison to the full-length TNF protein.
The first 16 amino acids of the truncated protein, determined as described under "Materials and Methods," are compared to the published sequence (6,7) of full-length human recombinant TNF. The presence of an N-terminal methionine in the truncated TNF is consistent with the activity of methionine aminopeptidase (16) during bacterial processing of the protein. 5% CO2. The media was aspirated off, and the indicated concentrations of T -T N F were added with (A) or without (0) a 1000-fold excess of unlabeled TNF to determine nonspecific and total binding, respectively. After 2 h at 4 "C, the cells were washed three times, and bound radioactivity was determined as described under "Materials and Methods." Results represent specific binding (0) (total minus nonspecific binding) of triplicate determinations. Scatchard analysis of the binding data is also shown (inset). system, the T N F RRA described here is designed to measure recombinant human T N F in a species-homologous system. The sensitivity of this RRA was equal to or greater than that possible with the L929 cytotoxicity assay, in which 1 unit/ml is the operational limitation. The initiation of human trials of T N F a s a therapeutic agent and the possible clinical significance of measuring T N F production by cancer patients (23) suggest potential applications of a sensitive RRA for TNF.
The density of T N F receptors on ME-180 cells is comparable to that of other murine and human adherent cell lines tested (18, 21); however, this is a relatively low number of sites/cell in comparison to other ligands such as colonystimulating factor-1 and epidermal growth factor (EGF), for which RRAs have been developed (24, 25). T o overcome this inherent limitation on the assay signal, specific binding was increased, along with assay sensitivity, by incubating the cells at 37 "C in the presence of sodium azide during binding. This technique has also proven useful in detecting low numbers of receptors (<500/cell) for colony-stimulating factor-2-a (26).
The increases seen here were most likely related to the inhibition of receptor-ligand internalization. This is supported by the lack of an effect by sodium azide at 4 "C, when internalization is already arrested. In addition, T -T N F bound at 37 "C was found to be acid-sensitive only when sodium azide was present. Internalization of TNF has previously been noted to occur in other cell types (18,19,27).
The TNF RRA was applied as a means of measuring the relative binding affinity of genetically altered T N F proteins. The bioactivity of a molecule may be correlated to its receptor binding ability, as seen here. The 3.5-fold difference in binding affinity between the full-length and mutant deletion proteins sodium azide at 37 "C for 2 h. lo5 cpm of Iz5I-TNF was then added for an additional 2 h under the same conditions, and counts bound were determined as described under "Materials and Methods." Solid lines represent least squares analysis of the data, and dashed lines indicate the log TNF concentration required to displace 50% of specific counts/min bound with no competing ligand. Duplicate determinations are shown. B, effects of sodium azide and temperature on acid-sensitive binding of Iz5I-TNF. ME-180 cells were incubated under the indicated conditions with 0 ng/ml TNF (total) or 1000 ng/ ml TNF (nonspecific), followed by Iz5I-TNF as described above. The percentage of specifically bound TNF sensitive to removal by glycine HCl was then determined. Values represent mean * S.D. of duplicate determinations.
is in close agreement with the 2-to %fold difference in bioactivity seen previously. Our interpretation of the data as representing a difference in relative binding affinity assumes that equivalent proportions of both muteins are capable of binding. In a similar study of insulin, deletions, substitutions, or both resulted in decreased as well as increased binding and bioactivity (28). The results seen here suggest part or all of relative binding affinities that are indistinguishable from that of the full-length protein.
Our data suggest that the residues in this N-terminal region may, for example, be directly involved in receptor binding, may be important for maintaining the tertiary structure of a separate receptor binding domain, or may be involved in stabilizing the receptor-ligand complex. Further studies are necessary to determine the full significance of this region in receptor binding and whether it may be possible to generate T N F molecules with higher relative affinities.
The availability of genetically altered proteins, as well as proteins altered by other physical or chemical means, provides unique opportunities to investigate structure-function relationships. This approach was applied in a recent report which indicated that mutant calmodulin proteins can differentially activate various target enzymes (29). Similar studies may help in developing agonists and antagonists to the various biological activities of TNF and in determining if these activities are the result of common or distinct mechanisms. A highly sensitive RRA, therefore, will be a useful method for the characterization of the properties and mechanisms of TNF.