Journal of Immunological Methods, 130 (1990) 177-185
Highly sensitive enzyme immunoassay for human lymphotoxin (tumor necrosis factor fl) in serum Giinther R. Adolf and Herbert R. Lamche Ernst Boehringer - Institut j~" Arzneimittelforschung~ Bender & Co GmbH, Department of Cell Biology, Vienna, Austria
(Received 22 January 1990, revisedreceived28 February 1990, accepted 5 March 1990)
We have developed a rapid, simple and highly sensitive 'sandwich' enzyme immunoassay (ELISA) for the detection and quantification of human lymphotoxin (ffi tumor necrosis factor fl) in serum. The assay, performed in microtiter plates, employs two monoclonal murine antibodies able to neutralize the cytotoxic activity of lymphotoxin. In a one-step procedure, antibody LTX-21 (IgG2b) coated on to the solid phase captures antigen present in the sample; subsequently antibody LTX-22 (IgG1), covalently coupled to horseradish peroxidase, labels the bound antigen. The assay is able to detect lymphotoxin spiked into human serum in concentrations as low as 7 p g / m l , whereas human tumor necrosis factor a does not cross-react even at 107-fold higher concentrations. Only biologically active protein is recognized by the antibodies, since inactivation of iymphotoxin measured by bioassay results in a parallel decrease in immunoreactivity. Natural, glycosylated lymphotoxin shows the same reactivity as recombinant, unglycosylated protein. The assay will be useful for the quantification of endogenous human lymphotoxin in serum, other body fluids, and culture supernatants of human cells, and can also be used to monitor levels of recombinant human lymphotoxin in animal studies and clinical trials. Key words:
Lymphotoxin; Monoclonalantibody; Immunoassay; Tumor necrosis factor ,8
Introduction Lymphotoxin, a slightly acidic glycoprotein consisting of 171 amino acids, is mainly a product of activated lymphocytes and was originally discovered by its ability to lyse tumor cells in culture (for reviews, see Evans, 1982; Devlin et al., 1984;
Correspondence to: G.R. Adolf, Department of Cell Biology, Bender& Co GmbH, Dr. Boehringer-Gasse5-11, A-1121 Vienna, Austria. Abbreviations: ELISA, enzyme-linked immunosorbent assay; HRPO, horseradish peroxidase; IFN, interferon; MoAb, monoclonal antibody; PBS, phosphate-buffered saline; TNF, tumor necrosis factor.
Paul and Ruddle, 1988). Lymphotoxin is structurally related to tumor necrosis factor a ( T N F - a ) with which it shares 28% amino acid sequence identity (Gray et al., 1984; Pennica et al., 1984) and both proteins bind to the same cell surface receptor. Since recombinant T N F - a and specific antibodies to this protein became available, a large number of studies have used these reagents to study the wide spectrum of biological activities of this protein and its physiological and pathophysiological role (reviewed by Beutler and Cerami, 1989). In comparison, much less is known conceming the biology of lymphotoxin. In a variety of experimental systems, lymphotoxin has been shown to share many of the biological effects of
0022-1759/90/$03.50 © 1990 ElsevierScience Publishers B.V. (Biomedical Division)
178 TNF-a, most notably its ability to cause hemorrhagic necrosis of transplantable tumors in mice; the designation 'TNF-fl' was therefore suggested to emphasize the structural and functional relationship of the two cytokines (Gray et al., 1984). A few reports, however, have demonstrated quantitative and even qualitative differences between some of the effects of lymphotoxin and TNF-a (e.g., Locksley et al., 1987; Oster et al., 1987). Several immunoassay systems for TNF-a, based on highly specific antisera or monoclonal antibodies (MoAbs), have recently been developed. A number of studies have since demonstrated increased levels of T N F ~ in serum, CSF or other body fluids in various pathological conditions, including septic shock, bacterial infections, and AIDS, and have thus provided essential contributions to our understanding of the pathophysiological role of TNF-a (reviewed by Beutler, 1988). In contrast, virtually nothing is known of the in vivo production of lymphotoxin under physiological or pathophysiological conditions, the main reason being a lack of adequate assay systems. Biological assays, such as the widely used cytotoxicity assay on murine L929 or L-M cells, are liable to nonspecific interference and moreover do not distinguish between TNF-a and lymphotoxin. A further important problem is the very low concentration of the cytokines in serum. The lymphotoxin immunoassays described to date show sensitivities of less than 400 pg/ml (Bringman and Aggarwal, 1987), 100-200 pg/ml (Tada et al., 1989a), 50 p g / m l (Tada et al., 1989b), or 10 pg/ml (Meager et al., 1987). However, these values were obtained in buffer solutions and experience shows that when assay systems developed for solutions with relatively low protein concentration (e.g., cell culture supernatants) are used for measurements in serum there is usually a dramatic (up to ten-fold) loss in sensitivity. We have therefore attempted to develop a highly sensitive assay for determination of lymphotoxin levels in serum. An enzyme immunoassay (ELISA) format was preferred since this system has considerable advantages in terms of safety and convenience over radioimmunoassays or immunoradiometric assays. To this end, we have first developed a panel of murine MoAbs, and by extensive testing defined a pair of antibod-
ies that could be used as capture and detector antibodies, respectively, in a highly sensitive onestep, solid-phase 'sandwich' ELISA. The characteristics and some applications of this assay are reported here.
Materials and methods
Cytokines Recombinant (E. coli-derived) human lymphotoxin was produced at Bender, Vienna, under licence from Genentech (San Francisco, CA), and was kindly provided by Dr. G. Bodo. Compared with the amino acid sequence deduced from cloned cDNA, this recombinant protein lacks the 23 N terminal amino acids, and thus consists of 149 amino acids including the initiator methionine which is present in essentially all molecules (M r = 16,490). This protein corresponds to the Mr= 20,000 lymphotoxin originally isolated from a human lymphoblastoid cell line (Aggarwal et al., 1984, 1985), but lacks N-linked glycosylation. Recombinant (E. coli-derived) human TNF-a, interferon (IFN)-a2c, and IFN-~, were produced at Bender, Vienna, Austria. All recombinant cytokines were > 99% pure. Natural human lymphotoxin was partially purified from supernatants of cell line RPMI 1788 as described previously (Adolf and Fogy, 1984).
Development of monocional antibodies The murine MoAbs LTX-21 and LTX-22 described in this paper are derived from a series of cell fusion experiments using spleen cells of female BALB/c mice immunized with recombinant human lymphotoxin. Details of the immunization, cell fusion and screening procedures as well as characteristics of the antibodies obtained will be given elsewhere (Voigt and Adolf, in preparation). In brief, mice were immunized 3-4 times by intraperitoneal injection of a lymphotoxin solution emulsified with Freund's adjuvant. A further injection of lymphotoxin without adjuvant was given 4 days before fusion. Fusion of spleen cells with P3X63Ag8.653 mouse myeloma cells (Kt~hler and Milstein, 1975; Kearney et al., 1979) was performed using 50% polyethylene glycol 4000 and 5% dimethyl sulfoxide. Following the selection of
179 hybridoma cells with H A T medium, the cultures (typically 1500 wells seeded from one spleen) were first screened by an ELISA technique. Plates coated with rabbit antibodies to lymphotoxin were incubated with lymphotoxin and hybridoma supernatants; murine antibodies captured were detected with rabbit antibodies to murine immunoglobulins coupled to horseradish peroxidase (HRPO). Positive cultures were tested for their ability to inhibit the cytotoxic activity of lymphotoxin towards mouse L-M cells essentially as described (Kramer and Carver, 1986; Kramer et al., 1986); supernatants were incubated with lymphotoxin for 90 rain at 37 ° C before application to the cells. Positive cultures were subcloned by limiting dilution. MoAbs were produced by inoculation of hybridoma cells into pristaneprimed B A L B / c mice and purified from ascitic fluids by ammonium sulfate precipitation followed by protein A-scpharose affinity chromatography, dialysed against phosphate-buffered saline (PBS) and stored at - 7 0 ° C . The isotypes and subisotypes of the antibodies were determined using commercially available ELISA kits (Bio-Rad and Zymed). Antiserum to recombinant lymphotoxin raised in New Zealand White rabbits was kindly provided by Dr. F.J. Schneider.
ELISA techniques MoAbs were covalently coupled to H R P O (Boehringer Mannheim) using the periodate method as described by Wilson and Nakane (1978). In a typical experiment, H R P O (1 m g / m l water, 1 ml) was oxidized with N a l O 4 (0.1 M, 0.2 ml) for 30 min at room temperature and dialysed against 1 mM sodium acetate buffer p H 4.4 overnight at 4 ° C . The pH of the H R P O solution was then adjusted to 9.0-9.5 with 0.2 M sodium carbonate buffer pH 9.5 (about 0.2 ml), and a solution of purified antibody (2 m g / m l in 0.01 M sodium carbonate buffer pH 9.5, 1 ml) was added immediately. The solution was stirred for 2 h at room temperature; 0.1 ml of freshly prepared sodium borohydride solution (4 m g / m l water) was then added and the reaction mixture further incubated for 2 h at 40C. The antibody-peroxidase conjugate was finally precipitated with ammonium sulfate (50% saturation) and redissolved in 2 ml PBS containing 10 mg bovine serum albumin as a
stabilizing agent. This stock solution was stored in aliquots at - 70 o C. Antibodies were adsorbed to the wells of enzyme immunoassay plates (Nunc, Denmark) in 0.05 M sodium carbonate buffer pH 9.6 (5 # g / m l , 0.1 ml/well) for 1 h at 3 7 ° C or overnight at 4 ° C . The wells were washed once with wash buffer (PBS containing 0.05% Tween 20; 0.2 ml/well), blocked with assay buffer (PBS containing bovine serum albumin at 5 m g / m l and 0.05% Tween 20; 0.2 ml/well) for 1 h at room temperature, and washed with wash buffer. The wells were then filled with dilution medium (pooled normal human serum; 0.1 ml/well); samples, standards and controls were added and, if appropriate, diluted in serial two-fold steps directly in the plate. 0.05 ml of a solution of peroxidase-labelled antibody LTX-22 appropriately diluted in assay buffer were applied to each well; the plates were then incubated for 4 h at room temperature. The appropriate concentration of the peroxidase conjugate, resulting in an absorbance of about 2.0 for the highest standard concentration and a background absorbance below 0.1, was determined in preliminary titrations. The wells were finally washed three times with wash buffer. Freshly prepared substrate solution was added (o-phenylenediamine, 3 m g / m l , and sodium perborate, 1 m g / m l in 0.067 M potassium citrate pH 5.0, 0.1 ml/well) and the plates incubated for 20 rain at room temperature in the dark. The enzyme reaction was stopped by the addition of 0.1 ml 2 M sulfuric acid. The absorbance in the wells was measured at 492 nm in a microtiter plate reader; the absorbance at 690 nm was automatically subtracted. The data were transferred on-line to a Hewlett Packard Vectra computer and analysed using the Titercalc software package (log-logit transformation).
Monocional antibodies to human lymphotoxin Sera from female B A L B / c mice immunized with recombinant human lymphotoxin showed high antibody titers as determined by ELISA and high neutraliTirtg titers in a standard cytotoxicity assay using L-M cells. Supernatants of more than
ten hybridoma cell lines derived from three cell fusion experiments with mouse spleen cells and the non-secretor myeloma cell line, P3X63Ag8.653, gave positive results in an ELISA-based primary screening. Eight of these supernatants were able to neutralize the cytotoxic activity of lymphotoxin towards L-M cells in standard cytotoxicity assays. All neutraliTirlg antibodies were purified from ascitic fluids and characterized in terms of their neutraliTing potency towards lymphotoxin and TNF-a, their epitope specificities, suitability for detection of lymphotoxin on Western blots and other parameters; detailed results of these experiments will be given elsewhere (Voigt and Adolf, in preparation). One of the antibodies (LTX-21) was used to study lymphotoxin production of activated leukocytes by immunofluorescence (Andersson et al., 1989).
Development and characteristics of an ELISA for lymphotoxin Two-site solid-phase ELISAs were performed for the selection of a pair of antibodies resulting in optimal sensitivity, using MoAbs coupled to horseradish peroxidase (HRPO) for the detection of antigen captured by polyclonal rabbit antibodies or MoAbs. Only neutrali7Jng antibodies were considered, since we expected that these - due to their specificity for the receptor-binding site of the protein - would recognize only biologically active protein. After extensive testing of all binary antibody combinations, MoAbs LTX-21 and LTX-22 were selected for further study. Both antibodies were able to neutralize recombinant lymphotoxin as well as natural, glycosylated lymphotoxin purified from supernatants of RPMI 1788 cells, but had no inhibitory activity for human TNF-a (Voigt and Adolf, in preparation). In ELISA procedures using isotype- and subisotype-specific antisera, LTX-21 was determined to be of the IgG2b type, whereas LTX-22 was identified as IgG1. Using an assay configuration consisting of LTX-21 as the capture antibody and LTX-22 coupled to HRPO for the detection of bound antigen, the coating antibody concentration, incubation temperature and time of the antigen-antibody reaction, as well as the conjugate concentration were optimized. The resulting assay protocol is given in the materials and methods section and a typical
! ~ I 0 2
Fig. 1. Standard curve of the lymphotoxin serum ELISA. Recombinant lyrnphotoxin was diluted in serial two-fold steps in pooled normal h u m a n serum; symbols indicate the m e a n s of four parallel titrations. The bar indicates the background absorption plus three standard deviations (n - 30); the minimal detectable dose in this particular experiment was 4 p g / m l .
standard curve is shown in Fig. 1. The useful assay range was observed to extend from 15 "pg/ml to 1000 p g / m l and serum samples could be applied without external dilution. The limit of detection for recombinant lymphotoxin spiked into normal human serum, defined as the analyte concentration resulting in an absorption significantly higher than the absorption of the dilution medium (mean plus three standard deviations) was determined to be 7 p g / m l (mean of three independent assays). To determine the reproducibility of the assay, artificial samples were generated by spiking recombinant lymphotoxin at three dose levels into pooled normal human serum; nominal concentrations of 750, 200 and 40 pg/ml, respectively, were prepared. Depending on the lymphotoxin concentration in the sample, the intra-assay coefficients of variation varied between 7.3% and 8.5%, and the interassay coefficients of variation between 8.7% and 11.2% (Table I). Human TNF-a is structurally related to lymphotoxin (28% sequence identity, with many of
TABLE I PRECISION O F T H E LYMPHOTOXIN ELISA Artificial samples were generated by spiking recombinant human lymphotoxin into pooled normal human serum. These samples were tested in five independent assays (three plates per assay, four repficates per plate). Two standard curves were run on each plate. Nominal concentrations were 40 p g / m l ('low'), 200 p g / m l (' medium'), and 750 p g / m l (' high'), respectively.
(n Coefficient of variation ($) Intra-assay precision 'low' level sample 'medium' level sample 'high' level sample Interassay precision 'low' level sample 'medium' level sample 'high' level sample
8.5 8.2 7.3 10.6 11.2 8.7
the differences due to conservative changes (Pennica et al., 1984)) and binds to the same cell membrane receptor. Nevertheless, recombinant human T N F - a did not generate a significant signal even at the highest concentration tested (0.5 mg/ml), and neither did recombinant human IFN-a2c at concentrations up to 1 m g / m l . Unexpectedly, recombinant human IFN-y, which is unrelated to lymphotoxin in its primary structure, showed a very low but reproducible cross-reactivity of 0.000125. Since serum levels of IFN-y are well below 1 ng/rrd, this cross-reactivity is irrelevant for all practical purposes. Thermal inactivation of lymphotoxin was used to investigate the correlation of the ELISA with a widely used cytotoxicity bioassay for lymphotoxin (Kramer and Carver, 1986; K.ramer et al., 1986). Samples of lymphotoxin dissolved in phosphatebuffered saline (0.1 m g / m l ) were heated to 80 ° C for 5, 10, 20, 40 and 80 min, respectively, and thereafter immediately cooled in an ice bath. The samples, together with the untreated control, were then diluted to lie within the assay range and tested in the ELISA as well as in the bioassay. Both assay systems showed that the activity of lymphotoxin was reduced with a half-life of approximately 7 rain and the correlation coefficient was calculated as 0.996 (Fig. 2). The stability of artificial samples (lymphotoxin spiked into pooled normal human serum at three
ELISA (log ng/ml) Fig. 2. Correlation of ELISA and bioassay. Recombinant lymphotoxin was denatured at 80 ° C for various time periods; samples were tested in the ELISA as well as in the cytotoxicity bioassay on L-M cells. The correlation coefficient was 0.996.
_0 . 2 i.a.
e,, 0Ul .1m