The 9E3 protein: immunolocalization in vivo and evidence for multiple forms in culture

The process by which cell growth is regulated in vivo remains an outstanding question in biology. We still do not understand how or why, during normal embryonic development and wound healing, cell growth is tightly regulated whereas in cancerous tissues growth is uncontrolled. However, over the last decade, considerable progress has been made in understanding the role played by oncogenes, growth factors and cytokines in the dynamic equilibrium between cellular function and replication. During this same period of time, a new group of genes has been discovered that exhibits some characteristics of each of the three molecular effectors. These genes, sometimes referred to as the gro family, are evolutionarily conserved: representatives of this family have been identified in humans (MGSA/human gro;Richmond et al., 1983, 1985, 1988; Richmond and Thomas, 1988; Anisowicz et al., 1988), mice (KC; Cochran et al., 1983), hamster (hamster gro;Anisowicz et al., 1987) and chickens (9E3/CEF4; Sugano et al., 1987; Bedard et al., 1987). Their evolutionary conservation, with only small differences in higher vertebrates (Oquendo et al., 1989), points to involvement in fundamental metabolic processes.

Although there is a considerable amount of information that has been reported on the regulation of the mRNAs and the structure of these genes, their functions remain largely unknown. Many of these studies show that this class of genes is expressed shortly after cultured cells are stimulated to grow (Gibson et al., 1986; Sugano et al., 1987; Anisowicz et al., 1987, 1988). We have shown that the 9E3 gene is induced during the G_o to G₁ transition or very early in G₁ and declines during S-phase (Martins-Green et al., 1991). Our studies on the expression of 9E3 mRNA in normal tissues in vivo are generally consistent with these observations in culture; in addition, we have shown that the expression of 9E3 is induced upon wounding and continues to be expressed in the granulation tissue of wounds, especially in areas of neovascularization (Martins-Green and Bissell, 1990). Taken together, these results point to an important physiological role and also suggest that the grò gene products may have more than one function in vivo.

Studies of the protein products of these genes are scarce, in part because of the lack of antibodies. Here, we show the production and affinity-purification of a polyclonal antibody that immunoprecipitates the 9E3 protein from the supernatant and the cell extract of normal and Rous sarcoma virus (RSV)-transformed chick embryo fibroblasts (CEFs), and we delineate the distribution of the protein in normal and wounded tissues and in RSV-induced tumors.

The last 28 amino acids of the C terminus of the molecule were synthesized by Milligen Biosearch (S. Rafael, CA) and purified by HPLC to a purity greater than 90%. The rationale for selecting this peptide is given in Results.

The procedures described here are modified from those presented by Ishigooka et al. (1992). We used the watersoluble coupling agent sulfo-maleimidobenzoyl-n-hydroxy-succinimide ester (S-MBS; Pierce, Rockford, IL) to couple the peptide to the carrier, keyhole limpet hemocyanin (KLH; Calbiochem, Inc., La Jolla, CA). The conjugation was done via the sulfhydryl group of the N-terminal cysteine of the peptide. A 40 mg sample of KLH was dissolved in 2.5 ml of 10 mM KPO4, pH 7.5, then 7.46 mg of S-MBS were added and stirred gently at RT for 30 min. The reaction mixture was applied to a PD-10 column (Pharmacia, Uppsala, Sweden) and eluted with 3.5 ml of 50 mM KPO4 buffer, pH 6.0; 2.18 ml of eluate were mixed with 40 mg peptide. The pH of the solution was adjusted to 7.0-7.5 with 1 M NaOH and the reaction allowed to proceed with constant stirring at RT for 2 h and then overnight at 4°C. The reaction mixture was dialysed against 150 mM NaCl (twice, 100 volumes) followed by prolonged dialysis in double-distilled water. The sample was then lyophilized.

Rabbits were immunized with 100 μg of the conjugate dissolved in 1 ml of 1 × PBS, pH 7.5, and emulsified with 1 ml of Freund’s incomplete adjuvant containing 5 mg heat-killed tubercle bacillus (Difco, Detroit, MI). Forty intradermal injections of 50 μl each, using a 21-gauge needle, were given along the back. The animals were boosted with 50 μg of conjugate in 1 ml of 1 × PBS and 1 ml of incomplete Freund’s adjuvant. One intramuscular injection of 250 μl was given in each limb. In addition, 10 intradermal injections of 100 μl each were given along the back.

Nitrocellulose membranes (Schleicher and Schuell, Inc), 0.2 pm pores size, were used. A dilution series of the peptide in TBS (described below) containing 10 μg/ml BSA was applied to the membrane, which was then removed from the apparatus and blocked with 10% nonfat dry milk in TBS containing 0.05% Tween 20 (TBS/Tween) for 1 h at RT. Following a quick rinse in TBS/Tween, the membrane was cut in strips representing the dilution columns and incubated with the individual antibody samples diluted in 3% nonfat dry milk in TBS/Tween overnight at 4°C. The strips were washed with TBS/Tween and incubated with 0.2 μg/ml goat anti-rabbit IgG conjugated to alkaline phosphatase (Kirkegaard and Perry Laboratories, Inc., Gaithersburg, MD) diluted in TBS/Tween containing 3% nonfat dry milk for 1 h at RT. The strips were washed with TBS/Tween followed by TBS and then incubated with substrate reaction mixture containing BCIP and NBT (Boerhinger Manheim).

Beginning six weeks after the first injection, the titer of the antibody was evaluated by immunoblot analysis every 2-3 weeks. It rose steadily, reached a plateau after 3 months and dropped slowly thereafter. After boosting at 5 months post primary injection, the titer rose quickly in 2 of 5 rabbits, to a detection level of 0.5 ng of peptide. The sera from these animals were used to affinity-purify the antibody as described in the Materials and methods. Immunoblot analysis of fractions collected during elution showed that most of the antibody eluted in fractions 3 and 4; as little as 0.1 ng of the peptide could be detected.

A 6 mg sample of peptide was dissolved in 1 ml of dimethylformamide (deionized in AG 501×8 mixed bed ion exchange resin). Then 1.2 ml of Affi-Gel 10 (BioRad, Inc.) suspended in propanol was washed briefly with doubledistilled water to remove the organic solvent and then with several volumes of dimethylformamide. All washes were done in a 10 ml BioRad column. The resin was removed from the column with a Pasteur pipette into a Wheaton vial and then centrifuged. The excess fluid was removed, the resin mixed with 1 ml of dissolved peptide and gently shaken overnight at RT. Then 100 μl of 1 M ethanolamine-HCl, pH 8.0, was added for 1 h to block any unreacted groups. The resin was put back into the column, washed extensively with 25 mM HEPES, pH 7.0, then with 10 volumes of 6 M guanidine-HCl + 10 mM CHAPS, pH 7.5, then with 15-20 volumes of 50 mM Tris-HCl, pH 7.5, and 10 volumes of 10 mM glycine-HCl, pH 2.5, containing 4 M urea + 1 mg/ml BSA (elution buffer). The peptide conjugated to Affi-Gel 10 was blocked with 1 ml of normal rabbit serum for 20 min and then washed with 10 volumes of 50 mM Tris-HCl, pH 7.5, + 150 mM NaCl (TBS). A 5 ml sample of crude antiserum was diluted 1:1 in PBS and allowed to equilibrate with the affinity-resin 10 overnight at 4°C with gentle agitation in a 15 ml centrifuge tube. The resin was centrifuged, the supernatant removed and the resin transferred back to the column. It was then washed with 20-30 volumes of 50 mM Tris-HCl, pH 7.5, + 500 mM NaCl + 0.01% NPA0, followed by washes with TBS + 0.01% NP-40 and finally with TBS. The peptide was eluted with elution buffer at a flow rate of 0.25 ml/min and collected in 1 ml fractions in sterile polypropylene Falcon tubes. The individual fractions were quickly loaded with a Pasteur pipette into dialysis tubing and dialyzed 4 times against 100 volumes of TBS + 0.02% (w/v) sodium azide and then tested by immunoblot analysis. The fractions with higher titer were pooled, dispensed into 100 μl aliquots and stored at —70°C.

The IgG content of the pooled fractions was determined by immunoblot analysis comparing serial dilutions of the pooled fractions with serial dilutions of purified rabbit IgG standards. Quantitation was performed using a Bio-Dot SF microfiltration apparatus (BioRad Lab., Richmond, CA) and a densitometer (Model 620, BioRad Laboratories).

Chick embryo fibroblasts (CEFs) were cultured as described previously (Bissell et al., 1977). The Prague C strain of RSV was used to obtain transformed cells. The Cos 7 cells were plated at 0.8×10⁶/100 mm Petri dish, allowed to grow for 24 h and then transfected with the plasmid containing the 9E3 cDNA.

The expression vector was derived from the plasmid pSV₂ and contained the Neo-resistant marker under the control of the simian virus 40 (SV40) early promoter for selection with G 418, and the mCMV promoter to drive the 9E3 gene. The fulllength 9E3 cDNA was cloned into this vector in both the sense and anti-sense direction. The transfection was done using the Ca₂PO₄ precipitation method as previously described (Gorman, 1985).

Cells were stimulated with serum-containing medium for 3 h and then metabolically labeled with [³⁵S]Met and [³⁵S]Cys for 3 h at 37°C, the condition medium collected and the cells extracted with l× RIPA (50 mM Tris, pH 7.4, 5 mM EDTA, 150 mM NaCl, 1% NP40, 1% sodium deoxicolate, 0.1% SDS). Equal numbers of counts were incubated with the antibody at a dilution of 1:100 in a 500 μl reaction mixture overnight at 4°C. This preparation was then incubated with 25 μl protein A-Sepharose for 1 h at 4°C, washed three times with RIPA and one time with 50 mM Tris-HCl, pH 7.5, resuspended in sample buffer and boiled for 5 min. The samples were loaded on a 12% urea-SDS-polyacrylamide gel. The gels were fixed in methanol/acetic acid (3:1, v/v), enhanced with Amplify (Amersham), dried and exposed to X-OMAT AR film (Kodak).

Cos 7 cells transfected with the sense or antisense 9E3 cDNA were prepared and processed for in situ hybridization with a ³H-labeled mRNA probe as previously described (Martins-Green et al., 1991).

Tissues were fixed in 4% paraformaldehyde, decalcified (Sieweke et al., 1989) and embedded in paraffin. Sections, 4 /an thick, were deparaffinized in xylene and hydrated in a descending series of ethanol concentrations to 1 × PBS. The sections were then incubated in 0.1 M glycine in 1 × PBS for 10 min, rinsed in PBS, blocked with 5% horse serum in 1 × TBS + 1% BSA for 30 min at RT and incubated with 0.7 μg/ml anti-9E3 in 1 × TBS + 1% BSA overnight at 4°C. The sections were washed with gentle shaking 3 times for 20 min each at RT in 1 × TBS + 0.1% BSA, incubated with biotinylated donkey anti-rabbit Fab fragments (Amersham) at 1:50 in 1 × TBS + 1% BSA + 5% chicken serum for 1 h at RT and then washed again as above. This was followed by incubation with streptavidin-alkaline phosphatase (Zymed) at 1:1000 in 1 × TBS + 0.1% BSA + 5 mM EDTA + 0.1% Triton X-100 for 1 h at RT and 3 washes, 20 min each in 10 mM Tris-HCl, pH 6.5, + 0.5 M NaCl + 1 mM EDTA + 0.1% BSA + 0.1% Triton X-100. The sections were then equilibrated in reaction buffer (100 mM Tris-HCl, pH 10, + 10 mM NaCl + 50 mM MgCl₂, final pH 9.5) for 5 min and then incubated with the substrate reaction mixture (10 ml reaction buffer + 45 ml NBT 4-35 ml BCIP + 100 μg/ml levamisol) in the dark for about 1 h. The reaction was stopped by incubating the sections in 10 mM Tris-HCl, pH 7.5, + 10 mM EDTA for 10 min. They were mounted in 100% glycerol and stored at 4°C in the dark. Two control experiments were performed: (1) the antibody was incubated with the 9E3 peptide, in a concentration 5 times more than that of the antibody, for 3 h at RT with gentle shaking and (2) the 9E3 antibody was replaced by rabbit IgG at the same concentration.

To enhance the probability of inducing an immune response in rabbits, we chose the last 28 amino acids of the C terminus of the protein because it is this region of the molecule that has the least homology with the mammalian gro genes and with IL-8, the proteins most closely related to 9E3 (Stoeckle and Barker, 1990). In general, C termini of proteins have been shown to be good immunogens. Furthermore, this peptide represents approximately 1/3 of the total molecule, thereby increasing the probability that it will fold properly and assume a native configuration and that the antibody will recognize the 9E3 protein in cells and tissues.

The specificity of this antibody to the peptide was demonstrated by immunoblot analysis. Peptide incubated with the affinity-purified antibody and with whole serum gave a positive reaction, whereas in the absence of the antibody or when the antibody was preincubated with the peptide-resin conjugate, no reaction was observed (Fig. 1). To establish the specificity of this antibody to the 9E3 protein, we cloned the 9E3 cDNA into a pSV2 vector containing the mCMV promotor to drive the transcription of the 9E3 gene and the neoresistance marker to allow selection with G418. The cDNA was cloned in both the sense and antisense directions. Each of these plasmids was transfected into Cos 7 cells and after selection the cells were labeled either with a radioactive antisense mRNA probe or with the antibody to 9E3. The cells containing the gene in the sense direction labeled for the mRNA (Fig. 2A) and stained for the protein (Fig. 2B), whereas the cells that had the gene in the antisense direction were negative for both (Fig. 2C,D). When sense-transfected cells were probed with pGeml riboprobe, no hybridization was seen (not shown, but see Martins-Green and Bissell, 1990; Martins-Green et al., 1991). Sense-transfected cells incubated with antibody that had been preincubated with excess peptide were also negative (not shown, but see controls shown here for all tissues stained).

Because 9E3 is highly homologous to IL-8, we tested the crossreactivity of the affinity-purified antibody to purified human recombinant IL-8 (kindly provided by Dr. J. Oppenheim, NCI; Appella et al., 1990; Baμgio-lini et al., 1989). A dot blot analysis showed that this antibody does not crossreact with hrIL-8 (Fig. 3).

Confluent starved CEFs express 9E3 mRNA at low levels (Martins-Green and Bissell, 1990) and staining for the protein could be seen in some cells and in the extracellular matrix (ECM) deposited by them (Fig. 4A). Transformed chick embryo fibroblasts (T-CEFs) contained high levels of message and stained abundantly for the 9E3 protein (Fig. 4B).

Immunoprecipitation of the protein from cell extracts of CEFs and T-CEFs and from the conditioned media of the two culture types is shown in Fig. 5. Both cell extracts contained a protein that is about 14 kDa. The supernatant from T-CEFs contained two species that were precipitated by this antibody, one about 9 kDa and the other about 6 kDa. The supernatant from confluent starved CEFs precipitated only a very weak band of the smaller of these two species and the larger form was not detectable. However, the antibody did precipitate a molecule about 30 kDa in size from the supernatant of the CEFs. All the remaining higher molecular mass bands represent nonspecific binding of the immunoglobulin molecules to proteins in the sample, because the same bands can be seen when the non-immune rabbit immunoglobulin was used in place of the antibody.

Wings of newly hatched chicks were prepared and immunostained as described in the Materials and methods. The antibody stained the cells of connective tissue and other tissues of mesenchymal origin and in some cases also stained the ECM (Fig. 6). Controls consisted of sections treated with primary antibody previously incubated with excess peptide for 3 h at RT and rabbit IgG at the same concentration as the primary antibody.

In the dermis of the skin, the label was found in fibroblasts, some of which labeled more intensely than others. Faint staining was also visible in the ECM surrounding the cells. All layers of the epidermis appeared to contain the 9E3 protein even though mRNA was not detected (Fig. 6A-C). Western blot analysis following the protocol of Woodcock-Mitchell et al. (1982) showed that the antibody for the 9E3 protein does not crossreact with the keratins of the basal layer and the stratum corneum of the skin (not shown). In the bone, the osteocytes of the outer layer, the osteoblasts and the periosteum stained intensely but the more differentiated osteocytes of the inner layers did not stain. This pattern closely approximates that of the distribution of mRNA (Fig. 6D-F). Fibroblasts of the connective tissue surrounding blood vessels labeled strongly for 9E3; in arteries (Fig. 6G), the smooth muscle layer showed light staining, which appeared to surrounding muscle blocks and fibers also stained for 9E3 but muscle fibers did not (Fig. 6H). In addition the protein was found in tendons, both intracellularly and in the ECM surrounding the cells. Again, some of the fibroblasts stained more intensely than others (Fig. 6I). Bone marrow (Fig. 6J) and endothelial cells (Fig. 6G), like muscle cells, did not stain for the 9E3 protein.

Wings of newly hatched chicks were wounded by insertion of a metal clip, as described previously by Dolberg et al. (1985) and prepared for immunostaining 10 days later when there was abundant granulation tissue present in the area of the wound. The controls were prepared as for the normal tissues.

Fig. 7 shows the staining for the 9E3 protein in the granulation tissue developed adjacent to the place where the clip was attached. Staining for the protein could be seen also in the matrix surrounding the cells, especially where scar tissue was present (Fig. 7F). In areas where blood vessels were abundant the staining was heavy in the cells around the capillaries but the endothelial cells did not stain (Fig. 7D,F). Some cells of the granulation tissue labeled intensely while others did not (Fig. 7 A,B,D,F). To test the possibility that some of these cells could be macrophages, which contain abundant peroxidases, we performed a peroxidase reaction with diaminobenzadine (DAB) on sections of wounded tissue; no staining was observed (not shown).

RSV-induced tumors did not stain for the protein (Fig. 8A) except in necrotic areas where a few cells stained very intensely (Fig. 8B). Unlike the granulation tissue, these areas of the tumor showed a number of cells staining for peroxidases (Fig. 8C). The control for the peroxidase reaction (Fig. 8D) was done in sections treated identically except that the substrate was not included in the reaction.

The results presented here show for the first time the immunolocalization of one of the gro genes in vivo. The antibody developed for this study detects as little as 0.1 ng of the 9E3 peptide on a dot blot analysis. We tested its specificity by transfecting the gene into Cos 7 cells. These experiments clearly show that the antibody is recognizing the 9E3 protein alone because only the cells containing the gene in the sense direction label for 9E3. In addition, the antibody does not crossreact with human recombinant IL-8, a molecule highly homologous to the 9E3 protein. Interleukins are highly conserved amongst species; hence, it is unlikely that the anti-9E3 antibody crossreacts with the chicken equivalent to IL-8.

Immunoprecipitation of the protein from the cell extracts and associated ECM of both CEF and T-CEF cultures yields a protein of about 14 kDa. Precipitation from the conditioned medium of T-CEF cultures, however, shows that after secretion from the transformed cells two smaller forms of the 9E3 protein (a 9 kDa and a 6 kDa form) are produced. The 9 kDa form was seen also by Sugano et al. (1987) in an in vitro translation system in the presence of dog pancreatic microsomal membranes; in the absence of dog pancreatic membranes, they observed an 11 kDa form, indicating that the protein undergoes processing. A 6 kDa form was immunoprecipitated from the conditioned medium of metabolically labeled CHO cells transfected with the MGSA gene (Derynck et al., 1990). Our inability to precipitate either of the smaller forms of the protein from the conditioned medium of the CEF cultures at more than barely detectable levels is probably because starved, confluent CEFs contain very little mRNA (Martins-Green and Bissell, 1990) and therefore would be secreting only very small amounts of protein at the time of collection of the supernatant. The presence of the protein on the cell surfaces and ECM of these cultures, when only a small fraction of cells have the mRNA, suggests that the protein was secreted previously when the cells were growing rapidly and expressing elevated levels of mRNA (Sugano et al., 1987; Martins-Green and Bissell, 1990). The protein is thus “stored” extracellu-larly; any smaller forms that may have been present in the supernatant have disappeared. In contrast, T-CEFs always show abundant mRNA and actively secrete the protein at all times. The 30 kDa species precipitated from the CEF supernatant (and not the T-CEF supernatant) could conceivably be a dimer of the 14 kDa form found in these cells. Although it is known that molecules highly homologous to 9E3 do form dimers and tetramers (Clore et al., 1989; Baldwin et al., 1991), in the 30 kDa case this explanation is not very likely because reducing conditions were used during electrophoresis. At the present time we cannot explain this band. Fortunately, it only appears in the supernatant and hence cannot be contributing to the antibody staining in cells and tissues.

Our antibody was made against the C terminus of the protein, hence both the 6 and 9 kDa secreted forms must contain the C terminus. Therefore, these two forms cannot represent the two pieces produced by a single cleavage of the molecule, but must reflect different processing of the larger form. Interestingly, similar observations have been made in studies of IL-8. This molecule, when released by endothelial cells, has an extended amino terminus (Gimbrone et al., 1989) compared to that produced by monocytes (Carveth et al., 1989).

In normal tissues, this protein is expressed in the cells and ECM of connective tissue and in other tissues of mesenchymal origin such as bone and tendon. It is not present in muscle fibers, endothelial cells and bone marrow. In general, this distribution coincides with the distribution of the mRNA (Martins-Green and Bissell, 1990), suggesting that this protein is biologically active in the tissues that produce it. However, 9E3 was also detected in two tissues that do not express the mRNA at detectable levels, the smooth muscle layer of arteries (intima layer) and the epidermis (Martins-Green and Bissell, 1990). In the intima of arteries the protein appears to be localized in the ECM between the cells, suggesting that it diffuses into this layer after production by the fibroblasts of the surrounding connective tissue. Thus the 9E3 protein produced by one cell type could be interacting with a different cell type.

The staining of the epidermis is more difficult to explain. The western blot analysis performed to test for crossreactivity between 9E3 and keratins was negative. Although suggestive, this negative result is not definitive because the keratins are denatured for such analysis, leaving the possibility that crossreactivity could occur between one or more keratins in their in situ forms. The large amount of keratins present in the stratum corneum of the skin and the high charge carried by these molecules (Moll et al., 1982) also raises the possibility that the 9E3 antibody could be binding nonspecifically to keratins, but if that were the case we probably would have seen similar staining with the rabbit IgG control. Fortunately, a similar uncertainty does not arise for other tissues because none of the other cells and tissues that stain for the protein contain keratins. In summary, the 9E3 protein is essentially found in cells and tissues that express the mRNA.

The fact that some cells stain more than others both in normal tissues and in the granulation tissue of wounds is intriguing. One possibility is that the cells that are staining more intensely are macrophages. In normal tissues this is unlikely because the number of macrophages present at any one time in the tissue is very small compared with the number of cells staining intensely for 9E3. In the granulation tissue this is more plausible but we did not detect any cells staining for peroxidases. Moreover, previous work has shown that at this stage of wound healing (10 days after wounding) there are very few macrophages in the granulation tissue (for review, see Clark and Henson, 1988). Alternatively, these cells could be special wound fibroblasts or normal fibroblasts that are in a specific phase of the cell cycle. Support for the latter possibility comes from the observation that the expression of the 9E3 gene is induced during the G₀ to G₁ transition or early in G₁ and that the mRNA subsides during the S-phase (Martins-Green et al., 1991). That is, the cells that stain more intensely could be cells in the early phase of the cell cycle. This appears to be a plausible explanation for the numbers of cells that we see staining more intensely in normal tissues because these newly hatched birds are growing very rapidly. These results are consistent with the observations of Barker and Hanafusa (1990), who showed that expression of the 9E3 mRNA also correlates with increases in cellular phosphorylation and morphological changes in cultured CEFs, phenomena known to occur during early stages of the cell cycle.

We had shown previously that RSV-induced tumors in vivo do not contain the mRNA for 9E3 (Martins-Green and Bissell, 1990). However, it was still possible that the protein could have a long half-life and be present in the absence of the mRNA. The absence of protein in healthy tumor tissue indicates that this protein is not involved in maintenance of the tumor. The small number of cells in the necrotic areas of tumors that show peroxidase staining could represent macrophages. This is consistent with previous findings that 9E3 may also be involved in the early stages of wound healing (Martins-Green and Bissell, 1990).

The results presented here, in combination with previous observations (Martins-Green and Bissell, 1990; Martins-Green et al., 1991), indicate that the 9E3 protein may have multiple functions. Expression of the gene is stimulated by wounding, serum and specific growth factors when cells leave the resting stage and enter the cell cycle (Martins-Green et al., 1991); the mRNA accumulates during G₁ and declines in S-phase. The protein is made as a larger molecule (14 kDa) that is secreted and binds to cell surfaces and ECM, and can be processed into smaller forms either simultaneously or subsequently upon appropriate stimulus. The small forms could act as autocrine or paracrine factors to stimulate (or perhaps inhibit) growth of cells during normal tissue remodeling. Additionally, upon wounding, one or more of the smaller molecules could be chemotactic for neutrophils, macrophages or fibroblasts and at later stages of the healing process may be involved in neovascularization.

We thank Prof. J. Oppenheim, NCI, for providing the hr IL-8, G. Parry and C. Streuli for advice in the immunoprecipitation techniques, C. Schmidhauser for advice in transfection procedures and A. Czernik for helpful discussions. The excellent sections were prepared by I. Chen. This work was supported by a National Research Service Award postdoctoral fellowship to M. M-G., by NIH grant EYO 6473 to L.M.H., and by the Office of Health and Environmental Research of the U.S. Department of Energy, under contract no. DE-AC03-76SF00098 to M.J.B.