Amplified expression of tumor necrosis factor receptor in cells transfected with Epstein-Barr virus shuttle vector cDNA libraries.

As an approach to isolate the cell-surface receptor for tumor necrosis factor (TNF), we have developed transfectants of human B-lymphoblastoid cells (UC cells) that overexpress the TNF receptor. These transfectants were isolated from UC cells transfected with cDNA libraries of HeLa or NG108 cells constructed in the mammalian expression vector EBO-pcD. This vector contains the Epstein-Barr virus origin of replication (ori-P) plus the EBNA-1 gene conferring replication function to ori-P and, therefore, the ability to replicate autonomously within the transfected cell (Margolskee, R.F., Kavathas, P., and Berg, P. (1988) Mol. Cell. Biol. 8, 2837-2947). Cells overexpressing the TNF receptor were identified and separated by the binding of fluoresceinated TNF and flow cytometric selection. Scatchard analysis of 125I-TNF binding data revealed a single class of high affinity receptors with a dissociation constant (Kd) of 0.2 to 2 nM and a receptor density of about 150,000 per cell, an increase of approximately 150-fold over UC cells. Cross-linking of receptor-ligand with bis-sulfosuccinimidyl suberate followed by polyacrylamide gel electrophoresis gave estimates of 87 and 104 kDa for the size of the complex. Based on its ability to bind TNF, a 68-kDa receptor protein was identified in cell extracts enriched for the receptor by using immobilized wheat germ agglutinin and TNF affinity chromatography. The difference in the estimated size of the receptor and the receptor-ligand complexes demonstrates that TNF binds to the receptor as a monomer or a dimer. Analysis of cDNA sequences conferring receptor amplification in transfectants revealed that plasmid DNA was present at 30 or more copies per cell, most likely integrated into the genomic DNA or organized into high molecular weight catenanes, and autonomously replicating units could not be recovered. Therefore, while this vector was useful in generating stable receptor-amplified cells, it was not maintained as a recoverable episome.

secreted by macrophages and a member of the immunoregulatory class of molecules termed cytokines that includes factors such as interleukins and interferons. Like other cytokines, TNF exerts a diversity of biological responses, including inflammation (l-3), endotoxin shock (4), cytotoxicity to certain tumors and tumor cell lines (5, 6), and the catabolic state termed cachexia associated with chronic disease (7,8).
TNF interacts with a wide variety of cell types by binding to high affinity receptors present on the cell surface. The availability of radioiodinated derivatives of TNF has made it possible to measure the number of surface receptors on many different cell types. Scatchard analyses of the ligand-receptor binding data indicate that the number of receptors is small, usually in the range of 500 to 10,000 per cell (7,(9)(10)(11)(12). Higher numbers (around 50,000 per cell) have been reported for HeLa-Ss (11) and a subcloned human histiocytic cell line U937 (13), although these have not been confirmed by others.' There appears to be a single class of specific high affinity receptors with affinity constants for binding in the 0.1 to 10 nM range (11,12,14,15). Ligand-receptor complexes are thought to be internalized, processed, and degraded in the lysosomes (11,(16)(17)(18).  and tumor promoters such as phorbol esters (19)(20)(21)(22) down-regulate receptor numbers while interferon-y enhances receptor expression (15,17,23). Although little is known about post-receptor events leading to the production of intracellular messengers, Gproteins have been implicated in the signal transduction pathway by the demonstration that pertussis toxin inhibits TNF-induced cytotoxicity (24). TNF also stimulates arachidonic acid release and subsequent production of prostaglandins (25).
One difficulty in the isolation of the receptor protein has been the low abundance of TNF receptors on cells. The protein has been partially purified from human histiocytic cells (U937) and estimated to have a size of 65 + 32 kDa (13). However, extremely low recovery plus the presence of other proteins make this size estimate tentative.
Covalent crosslinking of radioiodinated TNF to receptor with bis-sulfosuccinimidyl suberate generates products ranging from 74 to 105 kDa (11,13,16), but with 1,5-difluoro-2,4-dinitrobenzene (a different cross-linking reagent) two additional complexes of 54 and 138 kDa were reported (26). The size of the receptor itself from this ligand-receptor complex is difficult to derive because of the uncertainty of the form of TNF associated with the receptor. TNF self-associates predominantly as a trimer (27, 28), but its interaction with the receptor as a monomer, ' HeLa-S3 cells were obtained from the American Type Culture Collection, Rockville, MD, and when used to measure TNF receptor number had approximately 3000 receptors per cell but no more than the HeLa cells used in this investigation. dimer, or trimer has not been established. A full understanding of TNF-receptor interaction requires the isolation and identification of the receptor, characterization of its ligand binding properties, and elucidation of the intracellular mechanisms by which it asserts its cytotoxic action and other biological responses.
As an approach to the isolation of the receptor, we report the development and use of human B-cell lines that overexpress TNF receptor (the receptor is defined here as a cellsurface protein that binds TNF with high affinity). These cells were isolated from a pool of B-cells transfected with cDNA libraries, derived from HeLa or NG108 cells, constructed in the vector EBO-pcD. This vector permits expression of insert cDNA in mammalian cells, transforms human cells with high efficiency, and has the potential to exist as an autonomous plasmid at Z-10 copies within the transfected cell. Recovery of introduced cDNA clones from mammalian cells should therefore be facilitated (29). However, in this case the EBO-pcD vector appears either to integrate into cellular genomic DNA or to exist as a high molecular weight concatamer not recoverable as plasmids. We have nevertheless utilized these cell lines to identify the TNF receptor in detergent extracts of cells after enrichment for the receptor using affinity chromatography procedures. Based on the size of the receptor and the receptor-ligand complexes we conclude that TNF associates with the receptor primarily as a monomer and to a lesser extent as a dimer.
-" Cell Culture-NG108 cells (NG108-15 rat neuroblastoma x mouse glioma hybrid cells) were obtained from Dr. James Eberwine, Department of Psychiatry, Stanford University; 2938 human embryonic transformed kidney cells were from Dr. Michele Calos, Department of Genetics, Stanford University: EBV-transformed human lvmphoblastoid UC 729-6 cells (UC cells), HeLa cells (human epithelioid cervical carcinoma cells), and COS-7 (SV40-transformed CVl African green monkey kidney cells) were obtained from Dr. Paul Berg's laboratory, Stanford University. HeLa cells, NG108 cells, COS-7 cells, and 293s cells were maintained in Dulbecco's modified Eagle's medium while UC cells were cultured in RPM1 1640 medium. Media were supplemented with penicillin and streptomycin plus 10% fetal calf serum. Hygromycin was added to 200 pg/ml concentration. Cells were maintained in 5% CO,, 95% air in 37 "C. EBO-pCD HeLa and NG108 Expression Libraries-Expression libraries and their size-selected fractions were prepared in the pcD expression plasmid developed by  with modifications for insertion of the Epstein-Barr virus sequences as described by Margolskee et al. (29). Radioiodination of TNF-Purified recombinant TNF was nrovided by Cetus Corp., Emeryville, CA. A 4-pg aliquot in 100 ~1 of 0.1 M NaH2P04 buffer, PH 7.4, was mixed with 1 mCi of carrier-free Na"? (Du Pont-New England Nuclear), in 3 pg of IODO-GEN-coated glass tube for 5 min at room temperature. The reaction was brought up to 1 ml with the buffer containing 0.1% gelatin (PBS buffer), and the unreacted iodine was removed by gel filtration on a Sephadex G-25 PD-10 (Pharmacia LKB Biotechnology Inc.) column equilibrated with PBS buffer. The column was washed with 1.9 ml of buffer and the flow-through discarded. Radioiodinated TNF was eluted with the next 1.5 ml of buffer. The specific activity of the product was 20 @.Zi/ Fg TNF. More than 95% of the iPhI was incorporated into TNF as determined by trichloroacetic acid precipitation of total radioactivity and by SDS-PAGE in which a single band of TNF at 17 kDa was detected as radiolabeled.

125Z-TNF-Receptor Binding
Assay-Binding assays on intact cells in suspension were performed in 200 ~1 of RPM1 1640 medium supplemented with 10% fetal calf serum. One X lo6 cells were incubated on ice for 2 h with various concentrations of ""I-TNF with or without loo-fold excess of cold TNF. Cells were washed three times with ice-cold culture medium, and radioactivity in the cell pellet was determined in a gamma counter. Specific binding is defined as the amount of ""I-TNF that can be displaced by loo-fold excess cold TNF. Soluble receptor was identified by a filter binding assay to immobilize proteins on a nitrocellulose membrane held in a Minifold II apparatus (Schleicher & Schuell). The membrane was air-dried after sample application and blocked in a 5% non-fat dry milk solution made in phosphate-buffered saline (PBS) for 2 h at room temperature. Immobilized receptor was incubated with ""I-TNF at a 10 rig/ml concentration in 10% fetal calf serum in PBS for 1 h. The filter was washed three times for 15 min each at room temperature with PBS followed by autoradiography. This assay is described here as a method of tracing receptor activity. Details of it as a quantitative receptor assay will be published elsewhere.

Fhorescein or Biotin
Conjugation of TNF-TNF (10 mg/ml) dissolved in sodium bicarbonate buffer, pH 9.5, was conjugated to fluorescein isothiocyanate (FITC) dissolved in dimethyl sulfoxide. A 290.fig aliquot of TNF was added to 66 pg of FITC in a total volume of 35 ~1, and the vial containing the reagents was wrapped in aluminum foil and rocked gently at room temperature for 2-3 h. A PD.10 column was equilibrated with PBS buffer and the FITC-TNF solution brought to 106 ~1 with PBS buffer was loaded onto the column. The column was washed with 1.9 ml of buffer and the flow-through discarded. FITC-TNF was collected with the next 1.5 ml of buffer eluate. NHS-LC-biotin (10 mg/ml) in dimethyl sulfoxide was reacted with TNF (1 mg/ml) in 0.1 M bicarbonate buffer, pH 8.5, at a 3O:l molar ratio at room temperature for 2 h. Dimethyl sulfoxide in the reaction mixture was less than 10% of the total reaction volume. Biotinylated TNF was recovered by gel filtration as described above for FITC-TNF.

Transfection of Human Lymphoblastoid
Cells-UC cells (1 X 10R) were electroporated separately with 50 pg of DNA from each library and 20 pg of EBO-pcD Leu 2. Conditions for cell preparation, electroporation, and selection of hygromycin-resistant populations were as described previously (29).
Flow Cytometry and Sorting-Transfected UC cells (2 x 106) were prepared for flow cytometry by staining with FITC-TNF at a concentration of 850 rig/ml in 800 ~1 of RPMI-deficient medium at 4 "C for 2 h. Propidium iodide (Calbiochem) was added at a final concentration of 5 @g/ml prior to analysis to label dead cells. Cells were analyzed and sorted on a FACStar Plus sorter (FACS II, Becton Dickinson, San Jose, CA) by illuminating them with 500 milliwatts of 488 nm laser light and measuring forward and wide angle light scatter, 530 nm FITC fluorescence, and propidium iodide fluorescence greater than 630 nm. Sort windows were generated to eliminate debris and to sort the brightest 1% of FITC fluorescence distribution which was propidium iodide-negative. Early selection was accomplished using Becton-Dickinson's "enrich" mode, which accepts coincident particles in positive sort envelopes. Approximately 50,000 cells were grown in culture to l-2 x lo6 cells in the presence of 200 rg/ml hygromycin, and this cycle of selection and expansion was repeated as described.
Cross-linking of lz5Z-TNF to Receptors-Cells (l-2 x 106) were incubated for 2 h at 4 "C in RPM1 1640 plus 10% fetal calf serum (culture medium) containing 300 rig/ml of '*"I-TNF with or without loo-fold excess of unlabeled TNF. After incubation, the cells were washed three times by centrifugation through ice-cold culture medium to remove unbound TNF and resuspended in 1 ml of PBS. To this suspension 20 ~1 of 50 mM bis-sulfosuccinimidyl suberate in PBS was added, and the cells were left at room temperature for 30 min. The cells were washed three times with PBS and lysed in 0.5% Nonidet P-40. Undissolved debris was removed by centrifugation at 10,000 x g for 5 min, and the protein in the supernatant was precipitated with 10% trichloroacetic acid. The pellet was resuspended in 50 ~1 of Laemmli sample buffer and the proteins were analyzed on 10% polyacrylamide gels according to the method of Laemmli (31). After electrophoresis, the gels were dried and subjected to autoradiography.

RESULTS
By using lZ51-TNF we found that a variety of cell lines (e.g. HeLa and NG108) contain approximately 3000 TNF receptors per cell whereas the UC, B-lymphoblastoid line, contains about 1000 receptors per cell (Fig. 1). In each case a single class of high affinity receptors was identified with a Kd for TNF between 0.1 and 1.0 nM. Attempts to isolate a functional cDNA encoding the TNF receptor were initiated based on the notion that UC cells expressing additional receptor from an exogenously introduced cDNA could be identified by one of several techniques. Consequently, we prepared cDNA expression libraries in the Epstein-Barr virus shuttle vector, EBO-pCD (Fig. 2), using RNA from HeLa and NG108 cells (29). The HeLa libraries were made from five different sizes of cDNA in which the cDNA inserts were size 1, <0.4 kb; size 2, 0.4-0.8 kb; size 3, 0.8-1.6 kb; size 4, 1.6-2.9 kb; and size 5 with greater than 2.9-kb cDNA inserts. Fractions 2 through 5 alone or a pool of these four fractions referred to as HeLa 2-5 or the NG108 total cDNA library were introduced separately into UC cells by electroporation of 1 X 10' cells with 50 pg of library DNA. EBO-pcD Leu 2, harboring a cDNA encoding a human T-cell surface marker, was transfected separately into a population of UC cells to monitor the efficiency of electroporation, hygromycin selection, and flow cytometric analysis (29) Selection of UC Transfectants Expressing Elevated Levels of TNF Receptor-In order to detect the rare cells expressing higher receptor number, two derivatives of TNF were prepared for fluorescent staining of cells for FACS analysis and tested for their specificity of staining. These derivatives were biotinylated TNF for use with a second step reagent such as FITC-avidin or phycoerythrin-streptavidin and fluoresceinconjugated TNF alone. Class A reagents yielded nonspecific staining, which was 5-fold higher than unstained UC cells and therefore could not be used to measure specific binding of TNF (data not shown). However, FITC-TNF showed little or undetectable levels of nonspecific staining and was useful in detecting receptors on HeLa cells (Fig. 3) were stained with FITC-TNF and analyzed flow cytometritally. Initially, the spectrum of fluorescence intensity in a population of FITC-TNF-stained UC transfectants compared with nontransfected cells was indistinguishable. Nevertheless, about 0.05% of the most intensely staining UC transfectants (about 50,000 cells) were collected asceptically and cultured to grow to a population of l-2 X lo6 cells. These cells were restained with FITC-TNF and resorted. By the end of the fifth such selection, a population of cells with distinctly higher FITC-TNF binding was seen in transfectants from the entire HeLa library cDNA, and in HeLa size cut 4 cDNA (representing the 1.6-to 2.9-kb size), and in transfectants from the NG108 cDNA library. HeLa cDNA in the size range of 0.4-0.8, 0.8-1.6, or larger than 2.9 kb showed little enhancement of TNF receptor expression as did the EBO-pcD-Leu 2 plasmid DNA after equivalent repeated rounds of cell sorting (Fig.   4).

The transfectants
were characterized further by several independent methods to verify the increase in receptor expression. Measurements made by '*"I-TNF binding confirmed flow cytometric observations (Fig. 5). Scatchard plots of the steady-state binding of '251-TNF to the three positive cell lines revealed the presence of approximately 150,000 TNF receptors with & in the 0.2 nM (Fig. 6)  Preparations of receptor amplified cells showed the complex as two bands on autoradiographs of the gel: a dense band at 87 kDa and a fainter band at 104 kDa when cells were incubated with lZ51-TNF alone, but not when excess unlabeled TNF was included during the binding reaction (Fig. 7). The difference in size between the two bands is approximately 17 kDa and is pos-   sibly a reflection of the binding of an additional molecule of TNF to the receptor (see "Discussion").
Two faint bands of approximately the same size were also observed from extracts of UC cells but only after 3 weeks of autoradiography (data not shown).
Identification and Size Determination of TNF Receptor by Ligand-blot Analysis-Detergent extracts of UC cells and UC/ HeLa 2-5 receptor amplified cells were prepared and enriched for the receptor by sequential binding to lectin and TNF affinity columns. Fractions were tested for the presence of the soluble receptor using a filter binding assay (described under "Experimental Procedures"), and aliquots were simultaneously analyzed by nonreducing SDS-PAGE and ligand blotting to determine the size of the protein associated with receptor activity. An approximately 6%kDa band with a high level of ""I-TNF binding activity was recognized readily in pH 2.5 eluates from the TNF affinity matrix when using UC/ HeLa 2-5 cell extracts (Fig. 8A). Parallel samples obtained from unamplified UC cells processed through the lectin and TNF affinity steps did not bind an observable amount of Y-TNF. However, a d-day exposure of a ligand blot to x-ray film did produce a faint band with TNF-binding activity corresponding in position and size to the 6%kDa protein seen in amplified cells (Fig. 8B). This binding of ""I-TNF was specific since it was eliminated in the presence of excess unlabeled TNF.
Identification of a 68-kDa TNF Receptor Protein Band-Parallel preparations from UC cells and UC/HeLa 2-5 cells were processed through the same affinity chromatography procedures, and the TNF affinity eluates were subjected to reducing and nonreducing SDS-PAGE; proteins were detected by silver staining (Fig. 9, A and B) and also analyzed by ligand blotting (Fig. 9C). A 6%kDa protein band was recognized only in receptor amplified cells and was not detectable under these conditions in UC cells. Furthermore, this band corresponded in size and mobility with the 68-kDa protein exhibiting high TNF-binding activity as assessed by ligand blotting. Based on this evidence we infer that the 68-kDa protein is the TNF receptor.
EBO-pcD Plasmids in Receptor Amplified Cells and Their Rescue-Low molecular weight DNA was isolated by the method of Hirt (33) or its modification described previously (29) from cells enriched in TNF-receptor after four, five, or six serial flow selection cycles. No plasmids were recovered in E. coli from any of these cell lines after many repeated attempts. A conclusion to be drawn from these experiments was that the introduced plasmids no longer existed as recoverable episomes. The plasmids may have integrated or rearranged such that they could not be propagated in E. coli. Analyses of total genomic DNA from positive and negative cell lines were made after digestion with EcoRI (two sites within the vector) or BamHI (to release insert cDNA). Southern blots analyzed with radiolabeled probes containing EBO-pcD sequences of the EBO-pcD Leu 2 plasmid (Fig. 10)  Detergent extracts of'UC/HeI,a 2-5 cells were subjected to SDS-PAGE following chromatography on WGA-Sepharose (lanes I), WGA-Sepharose followed by TNF-Sepharose (Innes .Z), and WGA-Sepharose followed by two cycles of' TNF-Sepharose (lanes 3). Prior to electrophoresis the molecular weight standards and the receptor samples in A were heated for 5 min at 90 "C in the presence of 35~ (v/v) 2-mercaptoethanol and in H and (' applied without heat or 2-mercaptoethanol treatment. Alter electrophoresis acrylamide gels A and R were stained with silver nitrate and C was ligand blotted with "'I-TNF as described under "Experimental Procedures." reductase cDNA inserted into the cloning site of the pcD vector (30)), or only SV40 promoter sequences (data not presented) identified plasmid sequences. Different restriction patterns were obtained from the various transfectants indicating that the integration had occurred at different sites in the genome or that rearrangements and concatameric forms were generated. The number of copies of integrated vector sequences in UC/HeLa 2-5 was estimated (by comparison of autoradiographic band intensity with known amounts of DNA) to be approximately 30 copies per cell. One consideration for the absence of recoverable episomal DNA in transfected cells may be the duration of time between transfection and the selection of receptor enriched cells, normally a period of 6-8 weeks, during which time the plasmid could rearrange or integrate into genomic DNA. One way to reduce the probability of integration or recombination would be to decrease the time taken for selection. One such experiment is shown in Fig. 11. Here, 100 pg of EBO-pcD NG108 DNA were transfected into 1 x 10' UC cells; the cells were immediately transferred to hygromycin-containing medium and analyzed in the FACS 4 days later to measure expression from the introduced plasmids. This 4-day period was selected arbitrarily to allow cells to recover from the shock of electroporation and is comparable with a transient expression assay.
Contrary to our experience with stably transformed pools of UC cells, a distinct population of cells with high TNF receptors was observed (Fig. llB, arrow). These cells were recovered and are referred to as sort 1B (SlB). They were transferred to growth medium and expanded into a larger population.
From these, 2 x 10' cells were removed for the preparation of episomal DNA while 1 X 10" cells were cycled through repeated flow selection of the brightest l-2% of cells. These cells remained amplified through subsequent sorts (Fig. 11,S3B and S5B). SlB cells increased in receptors 30-fold over UC/uns cells and S3B and S5B another 3-fold amounting to a total go-fold amplification.
From 2 x 10' cells of SlB low molecular weight DNA was recovered and used for bacterial transformation to yield about 1 x lo" bacterial colonies. These plasmids were expected to represent a population enriched in genes conferring TNF receptor abundance. Plasmid DNA was prepared from a pool of these transformants, and 100 Fg of DNA was transfected into a fresh population of 1 X 10' UC cells which were flow cytometrically analyzed 4 days later. No increase in TNF receptor was seen in these cells after 4 days of recovery or in cells selected for high FITC-TNF binding even after seven cycles of sorting and expansion. Low molecular weight DNA was prepared from the selected cell populations and, except for SlC (see Fig. 9E) which yielded 20 transformants, no plasmids were recovered from DNA of any other sort upon repeated attempts. Restriction endonuclease digests of plasmid DNA from SlB and SlC with BarnHI of XhoI to release insert cDNA showed that the insert size was either very small (less than 500 base pairs) or absent. Transient expression assays have also been performed with other cells such as COS-7 and 2938 as recipients for HeLa and NG108 cDNA libraries.
In these experiments no receptor enriched cells were obtained upon initial transfection nor repeated flow selection of the brightest 2% of cells. Plasmid recovery was low and repeated transfections with rescued plasmid DNA failed to generate positive cells. Plasmids can be recovered from recipient cells but their abundance declines rapidly within 2-3 weeks after transfection, and, when plasmids are recovered, they have either very small cDNA inserts or have no detectable inserts (data not shown).

DISCUSSION
The diversity of biological effects induced by TNF is initiated by its interaction with cell surface receptors. The nature of the signals transduced by this interaction is not yet known, but, as a step toward understanding the signal transduction pathways, a knowledge of the receptor is required. Receptors for TNF are found on most cell types but their number is usually in the range of a few thousand per cell (7,(9)(10)(11)(12). Because of its low abundance, purification of the receptor is a difficult task. To overcome this limitation we have combined the expression of transfected cDNAs with a powerful selection using the FACS to identify rare cells expressing high levels of receptor. Using this approach we have succeeded in isolating UC cell lines from both HeLa cDNA and NG108 cDNA transfected cells that express TNF receptor at levels 150-fold higher than untransfected UC cells. The results of the HeLa cDNA transfections were particularly interesting since only one size fraction of the cDNA library, that with 1.6-2.9-kb inserts, conferred increased receptor density in two separate sets of independent transfections. These observations suggest the presence of a specific cDNA sequence in the HeLa cell library responsible for the increase in receptors. However, the mechanism by which receptor numbers are increased is not clear. The selection pressures that were applied to select and maintain stable transformants were maintenance on 200 pg/ ml hygromycin and flow cytometric sorting for high TNF receptor levels. The cDNA responsible for conferring this phenotype could possibly be a full length sequence encoding the receptor, integration and/or amplification of which causes receptor numbers to increase. It could also be due to a partial cDNA sequence whose insertion into the resident TNF receptor gene followed by amplification could lead to higher levels of expression. Alternatively, we may not be selecting a receptor encoding gene but one whose protein product could upregulate receptor numbers. Such a cellular product, e.g. interferon-y, has been reported, although the magnitude of increase by it is a modest 2-3-fold (15,17,23).
Scatchard analysis of steady-state receptor binding with ""I-TNF shows that indeed the dissociation constant of the over-expressed receptor is characteristic of high affinity receptors and similar to that of HeLa and UC cells with a Kd in the range of 0.9-2 x lo-" M. Cross-linking studies were performed with the receptor-enriched transfectants using "'I-TNF and the cross-linking reagent bis-sulfosuccinimidyl suberate. Cross-linked complexes were observed when "'I-TNF was incubated with the cells in the absence of cold TNF and followed by treatment with the cross-linking reagent. Two protein bands were observed in gels at positions corresponding in size to 87 and 104 kDa consistent with previously observed bands (11,13,16,23). Neither band was seen if cross-linking was performed in the presence of cold TNF. It appears that the molecular mass of the receptor observed in our studies is approximately 68-70 kDa if these complexes represent the binding of TNF monomer and dimer, respectively. However, if the complexes represent binding of a dimer and timer of TNF (27,28) then the receptor itself may be 50-55 kDa.
Given the high concentration of the receptor in amplified cells we have undertaken to characterize its size and ligand binding properties in order to eventually purify the protein to homogeneity.
An assay developed to measure ""I-TNF binding to soluble receptor was useful in pursuing the enrichment of cell extracts for receptor protein. The applications of immunoaffmity chromatography including WGA-Sepharose and TNF-Sepharose (details to be published elsewhere) provided a rapid and convenient method of increasing receptor content that has allowed us to identify by ligand blotting techniques a band of approximately 68 kDa with high level ""I-TNF binding activity. Nonreducing and reducing SDS-polyacrylamide gels (stained for protein and silver) showed a 68-kDa band in enriched cell extracts of amplified cells. This band was difficult to find in comparable preparations from UC cells. Therefore, our identification of the 68-kDa protein as the receptor is based on the use of the ligand binding functional assay and the presence of a protein band coincident with the activity. The usefulness of the immunoaffinity procedures employed here brings into question an earlier published report (13) describing the failure of TNF receptor to bind to TNF immobilized to agarose and its poor recovery and lack of enrichment from lectin-affinity resins. Possibly, the rich source of receptor in the amplified cells plus the development of a solubilized receptor assay to trace the receptor has helped us to optimize the use of these affinity chromatography procedures.
The sizes of the TNF-receptor complexes are approximately 87 and 105 kDa as determined by cross-linking '""I-TNF to the receptor on intact cells. The estimated size of the TNF receptor deduced from the ligand blot is 68 kDa. The subtracted difference approximates the sizes of the TNF monomer (17 kDa) and dimer (34 kDa). Although TNF reportedly self-associates into a trimer (27, 28), there has been considerable discussion as to the preferred size of the TNF molecule that associates with the receptor. Since the intensity of the 87-kDa receptor-ligand complex is greater than the 105-kDa band (Fig. 7), our data reveal for the first time that TNF binds to its receptor predominantly as a monomer and to a lesser extent as a dimer.
An important aspect of the approach we employed was the use of the EBO-pcD vector for the construction of the libraries. These libraries were initially prepared in the pcD vector containing the SV40 early region promoter and the late splic- ing function (30). The novelty of the EBO-pcD plasmid lies in a sequence of DNA encoding the Epstein-Barr virus origin of replication plus the EBNA-1 gene conferring replication function to ori-P and a selectable marker conferring resistance to the drug hygromycin B. These genes were introduced into pcD-HeLa and pC-NG108 libraries as a cassette to transform a pcD-cDNA plasmid into an EBO-pcD plasmid, thereby conferring replication and maintenance functions on plasmids introduced into the EBV-transformed UC-lymphoblastoid cells (38). Libraries made in this vector yielded hygromycinresistant transfectants within 2 weeks upon introduction into UC cells. When these transfectants were used to isolate autonomously replicating plasmids from cells maintained on drug selection the recovery of plasmids dropped dramatically over time. Initially, up to 2 weeks after transfection around 1000 plasmids could be recovered from 2 X lo7 cells. By 3 weeks post-transfection this recovery had dropped to less than 10 plasmids and at 4 to 6 weeks post-transfection plasmids were recovered rarely. These results suggest that the plasmids were not being maintained autonomously.
Restriction enzyme analyses with BumHI or X/z01 of recovered plasmid DNA showed that the cDNA insert was either missing or present as a fragment smaller than 500 base pairs (data not shown). Since the starting libraries had a wide range of insert cDNA, particularly the size selected HeLa library, this loss of cDNA indicates an inability of the plasmid to maintain structural stability or a preferential recovery of clones with short inserts. Analysis of high molecular weight DNA for vector sequences identified what appear to be integrated molecules in all transfectants analyzed. Therefore, while the EBO-pcD plasmids can be maintained as autonomously replicating units in cells for varying periods of time (29), in our experience using routine procedures they are not easily recoverable as plasmids.
Instead these DNAs appear to be preferentially maintained as integrated sequences or perhaps very high molecular weight catenates that behave like chromosomal DNA. The expression in COS-7 cells of cDNA libraries constructed with the CDM8 vector (39) and affinity selection techniques using either antibody or ligand binding have worked successfully for the cloning of some cell surface and growth factor receptors (39-42). This vector or the recently described SRa vector (43,44) offers alternative systems for cloning cDNA sequences expressing cell-surface receptors.