Human Anti-TNF Antibodies

1. Field

This disclosure relates generally to monoclonal antibodies and specifically with human antibodies that bind to human tumor necrosis factor (TNF.alpha.).

2. Prior Art

TNF.alpha. is a pluripotent and pleiotropic cytokine. It is produced principally by activated macrophages, however its synthesis and secretion have also been observed using granulocytes, tonsil B cells, B cell lines, natural killer cells, T cell lines, primary chronic malignant B cell isolates, and peripheral blood T cells.

TNF.alpha. can also be expressed on cell surfaces, apparently in two forms. One is a 26 kd molecular weight integral type 2 transmembrane protein on monocytes, T cells and some other cells. The other form is the secreted 17 kd product which is bound to its receptor.

Among the many activities of secreted TNF.alpha. are thymocyte growth factor, B cell growth and maturation factor, production in vivo of hemorrhagic necrosis, weight loss, cardiovascular collapse and multiple organ failure. Naturally, these latter activities are the source of the clinical interest in TNF.alpha..

During septic shock, as well as inflammatory diseases, synthesis and secretion of TNF.alpha., IL-1, IL-6 and IL-8 have been documented. Hence the immune systems of some individuals are exposed chronically to these cytokines. Indeed, low affinity antibodies to TNF.alpha. have been reported (A. Fomsgaard et al "Auto-antibodies to Tumor Necrosis Factor .alpha. in Healthy Humans and Patients with Inflammatory Diseases and Gram-Negative Bacterial Infections." Scand. J. Immunol. 30:219-23, 1989; and, K. Bendtzen et al "Auto-antibodies to IL-1.alpha. and TNF.alpha. in Normal Individuals and Infectious and Immunoinflammatory Disorders." Prog. Leukocyte. Biol. 10B:447-52, 1990). These anti-TNF.alpha. autoantibodies may, however, not be specific (H. -G. Leusch et al "Failure to Demonstrate TNF.alpha. Specific Autoantibodies in Human Sera by ELISA and Western Blot." J. Immunol. Meth. 139:145-147, 1991).

One peculiar feature of human serum, as well as sera from other animals, is its content of natural and so-called polyreactive antibodies. These are usually IgM antibodies which bind to various autoantigens with low affinity (A. B. Hartman et al "Organ Reactive Autoantibodies from Non-Immunized Adult Balb/c Mice are Polyreactive and Express Non-Biased Vh Gene Usage." Molec. Immunol. 26:359-70, 1989; and, P. Casali et al "CD5+ B Lymphocytes, Polyreactive Antibodies and the Human B cell Repertoire." Immunol. Today. 10:364-8, 1989). Hence the autoantibody-like reactivity to human TNF.alpha. might be expected to be low affinity and probably cross-reactive with several other antigens.

Some high affinity neutralizing antibodies to IL-1.alpha. have been reported in normal sera (N. Mae et al "Identification of High-Affinity Anti-IL-1.alpha. Autoantibodies in Normal Human Serum as an Interfering Substance in a Sensitive Enzyme-Linked Immunosorbent Assay for IL-1.alpha.." Lymphokine Cytokine and Research 10:(1) 61-68, 1991) or patient (H. Suzuki et al "Demonstration of Neutralizing Autoantibodies Against IL-1.alpha. in Sera from Patients with Rheumatoid Arthritis." J. Immunol. 145:2140-6, 1990).

Despite these considerations, we are unaware of the disclosure of any monoclonal human antibodies specifically binding to TNF.alpha. even though it is thought such antibodies may have significant clinical value. Thus, there has remained a need for monospecific monoclonal antibodies to TNF.alpha..

We have made monoclonal human antibodies which bind to both human and mouse TNF.alpha.. The antibodies bind to recombinant human TNF.alpha. (rhTNF.alpha.) with a titer comparable to three high affinity neutralizing mouse mAbs, when tested by ELISA. The antibodies most fully characterized are of the IgM isotype although we also prepared antibodies of the IgG isotype. By competition binding experiments, the antibody appears to bind to epitopes on rhTNF.alpha. distinct from those bound by the neutralizing mouse mAbs so far described. Specificity analyses indicate that the human IgM autoantibody binds to both human and mouse recombinant TNF.alpha., but not to other antigens commonly recognized by polyreactive natural IgM autoantibodies. The high level of amino acid identity between the human and mouse TNF.alpha. molecules suggest that the antibody is monospecific for a given epitope shared by these two forms of TNF.alpha..

The B5 antibody also binds to cell surface TNF.alpha. (cs TNF.alpha.) on human T cells, B cells, monocytes, a variety of lymphoid and monocyte lineage cell lines of human origin, as well as astrocytomas, a breast carcinoma, and a melanoma. The antibody also binds to chimpanzee lymphocyte and mouse T lymphoma cell line csTNF.alpha.. Binding of the antibody to csTNF.alpha. is specific since it can be inhibited by TNF.alpha. but not by TNF.beta., a neutralizing mouse anti-TNF.alpha. mAb, nor by a recombinant form of the extracellular domain of the p55 TNF receptor (TNFR). The B5 autoantibody can inhibit LPS induced TNF.alpha. secretion by cells of the human monocyte-like cell line THP-1.

Several monoclonal mouse anti-human TNF.alpha. antibodies have been described in the literature. None, however, also bind to mouse TNF.alpha..

The specificity, the autoantibody nature, the binding to cell surface TNF.alpha. and the ability to inhibit TNF.alpha. secretion make B5 a novel mAb.

Characterization of the antibodies and how to make them are described below .

FIGS. 1A and 1B show, in graph format, a comparison of solid phase ELISA format binding to rhTNF.alpha. of the B5 (human) and A10G10 (murine) monoclonal antibodies. ELISA plates were coated with various concentrations of TNF and titrated doses of mAb were then allowed to bind. Shown are the binding curves for each antibody for the various TNF coating concentrations.

FIGS. 2A and 2B show, in graph format, the lack of competition for binding to TNF.alpha. between mouse mAbs and B5 mAb. FIG. 2A shows the binding of three mouse anti-TNF mAbs and the control C7F7 anti-rFVIII mAb binding to solid phase rhTNF.alpha.. FIG. 2B shows the lack of inhibition of B5 binding to plate bound TNF when the mouse monoclonals are first allowed to bind to TNF plates and B5 antibody is added subsequently.

FIG. 3 shows, in bar graph format, the binding of human IgM anti-TNF mAbs to rhTNF.alpha. captured and presented as a complex by the combination of plate bound mouse mAbs A10G10, B6 and A6. ELISA plates were precoated with the three mouse mAbs and then incubated with rhTNF.alpha.. Plates were washed and 20 ug/ml of the indicated human IgM mAbs were then allowed to bind. Solid bars show the binding of the human IgM mAbs to the three mouse mAbs which had been incubated with TNF, the hatched bars show binding of the IgM mAbs when the attached mouse mAbs had not been exposed to TNF.

FIGS. 4A-4F show, in multiple graph format, results of an analysis of the binding specificity of several monoclonal antibodies. Plates where precoated with either recombinant human TNF.alpha.(.box-solid.), recombinant human lymphotoxin (.diamond-solid.), human insulin (.quadrature.), porcine thyroglobulin (.tangle-solidup.), BSA (.largecircle.), ssDNA (.box-solid.), dsDNA () or human IgG Fc fragments (.DELTA.). Mouse mAb A10G10 is shown in panel A. Human IgM mAbs B5, 7T1, H5, 1A6B5F and F2.2.34 are shown in panels B, C, D, E and F, respectively. Antibody binding was assessed by ELISA.

FIG. 5 shows, in graph format, binding of B5 to recombinant mouse TNF.alpha.. Plastic plates were precoated with a neutralizing monoclonal hamster anti-mouse TNF.alpha. antibody at 8 ug/ml (squares), 4 ug/ml (triangles) and 2 ug/ml (circles). Recombinant mouse TNF.alpha. was then added at 2 ug/ml (filled symbols) or was not added (open symbols). Human mAb B5 was then allowed to bind at the concentrations indicated. Binding was then assessed by ELISA using anti-human IgM antibody.

FIG. 6 shows, in graph format, a comparison of B5 mAb (triangles) and mAb A10G10 (circles) binding to soluble rhTNF.alpha.. Antibodies were bound to plastic plates precoated with anti-human or anti-mouse antibody. Biotinylated TNF was then incubated with the antibodies. Binding of soluble TNF.alpha. was detected by enzyme-avidin conjugates.

FIG. 7 shows, in graph format, that captured B5 mAb binds soluble TNF.alpha. and weakly presents it to A10G10 mAb. B5 mAb anti-TNF.alpha. or 6F11 (human anti-LPS IgM) as a control, were allowed to bind to plates precoated with anti-human IgM. Soluble TNF.alpha. was then allowed to bind to the complexed human mAbs. Mouse mAb A10G10 was added and its binding to TNF complexed to B5 mAb was detected by enzyme linked anti-mouse IgG antibody.

FIG. 8 shows, in photograph format of Western blots, the binding of several human IgM antibodies to mouse TNF.alpha. and binding of the human B5 mAb to human TNF.alpha.. Recombinant mouse TNF.alpha. (lanes A-G) and rhuTNF.alpha. (lanes H and I) were electrophoresed under reducing conditions and transferred to nitrocellulose. Mouse TNF.alpha. was blotted with the following monoclonal antibodies: 7T1 (lane A), B5 (lane B), 1A6B5F (lane C), 6F11 (lane D), H5 (lane E), A8 (lane F), and no primary antibody (lane G). Human TNF.alpha. was electrophoresed in lanes H and I. Lane H was then blotted with B5 mAb and lane I with 6F11 mAb. Lanes A-F, H and I were then exposed to biotinylated anti-human IgM. Lane F was exposed to biotinylated anti-human IgG, since A8 is an IgG antibody. All lanes were then exposed to the developing reagent avidin coupled horse radish peroxidase. Molecular weight standards, ranging in molecular weight from 211 kd to 15.4 kd, were run in parallel and their positions are indicated.

FIG. 9 shows, in graph format, the neutralization of rhTNF.alpha. by A10G10 mouse mAb and lack of neutralization by human mAbs. WEHI 164 cells were incubated with a cytotoxic dose of rhTNF.alpha. in the presence of titrated concentrations of mAb. Viability was subsequently assessed.

FIGS. 10A-10H show, in histogram format, the fluorescence staining profiles of two cell lines stained with human IgM anti-TNF.alpha. mAbs. 8B9 cells (FIG. 10A, 10C, 10E, 10G) and THP-1 cells (FIG. 10B, 10D, 10F, 10H) were stained with no antibodies (FIGS. 10A, 10B), with FL-F(ab)'.sub.2 anti-human IgM (FIGS. 10C, 10D), B5 IgM anti-TNF.alpha.+FL-anti-IgM (FIGS. 10E, 10F) and 6F11 anti-LPS+FL-anti-IgM(FIGS. 10G, 10H). Fluorescence intensity channel numbers, in arbitrary units are plotted against the cells per channel on the ordinate. For each sample 5000 cells were accumulated. The percentages of cells falling within the indicated markers, scored as fluorescence positive, are given.

FIGS. 11A-11B show, in graph format, the detection of cell surface expression of TNF.alpha. on THP-1 and U937 cells with the B5 anti-TNF.alpha. mAb, and increase in expression with LPS and PMA. THP-1 (FIG. 11A) and U937 (FIG. 11B) cells were incubated 3 hours with medium (open circles), LPS (filled circles) or LPS+PMA (filled triangles).

FIGS. 12A-12D show, in graph format, the shift in staining intensity when B5 anti-TNF.alpha. IgM mAb binds to cells being stained with F1-anti-IgM antibody. CD19 positive splenocytes are shown. These were stained with phycoerythrin conjugated anti-CD19 and only positive cells were further analyzed for fluorescein conjugated antibody staining. FIG. 12A shows C19+ splenocytes not stained with FL-anti-IgM. FIG. 12B shows staining of these cells with B5+FL-anti-IgM, FIG. 12C shows staining with FL-anti-hIgM alone, and FIG. 12D shows staining with control 6F11 anti-LPS IgM+FL-anti-IgM. The percentages of cells within the indicated markers are given, indicating the percentage of cells staining positively with the fluorescein conjugated antibody. The median channel numbers for the positive populations are also given. These numbers reflect the staining intensity, measured in arbitrary units, for the florescence positive populations.

Reagents

Bayer A. G., Wuppertal, Germany provided rhTNF.alpha.. The rmTNF.alpha. and rhLT were purchased from Genzyme. Human IgG Fc fragments were purchased from Chemicon. Insulin was purchased from Novo Nordisk Labs and all the other antigens used in ELISAs were purchased from Sigma. The Staph. aureus Cowan strain was purchased from Calbiochem (San Diego, Calif.). The anti-human IgD-Dextran conjugate was obtained from a private source. Phorbol myristic acid, mouse IgG.sub.1, staphylococcal enterotoxin B (SEB) and phytohemagglutinin (PHA) were purchased from Sigma. E. coli LPS was obtained from a private source. The different fetal bovine sera (FBS) were purchased from Hyclone.

The cell lines mentioned in Table 2 were all purchased from the American Type Culture Collection (ATCC), except for the 8B9 EBV transformed human B cell line which was obtained from Genetic Systems Corporation. The American Type Culture Collection (ATCC) is located at 12301 Parklawn Drive, Rockville, Md. 20852.

TNF was biotinylated using standard techniques; briefly, N-hydroxysuccinimidyl ester of biotin was added to TNF dissolved in 50 mM NaHCO.sub.3, pH 8.5 for 15 min, quenched with NH.sub.4 C1 then dialyzed to remove unreacted biotin.

The mouse A10G10 anti-TNF.alpha. IgG.sub.1 mAb was generated in collaboration with Chiron Corporation and has an ATCC designation number HB 9736, identified as hybridoma cell line 2-2-3E3.

The A6 and B6 mouse IgG.sub.1 mAb were generated from mice hyperimmunized in our laboratory. All three mouse mAbs neutralize TNF cytotoxicity and have been described in Galloway et al "Monoclonal anti-Tumor Necrosis Factor (TNF) Antibodies Protect Mouse and Human Cells from TNF cytotoxicity." J. Immunol. Meth. 140:37-43, (1991) which is incorporated herein by reference. These mAbs were purified by affinity chromatography.

The polyreactive IgM mAbs 1A6B5F and F2.2.34 were produced and characterized by Kasaian et al "Identification and Analysis of a Novel Human Surface CD5- B Lymphocyte Subset Producing Natural Antibodies." J. Immunol. 148:2690-702 (1992). The 7T1 human IgM mAb was produced and provided in ascites by a private source.

The 6F11-E4 (6F11) EBV transformed B cell lymphoblastoid line having ATCC designation number CRL 1869, produces a human anti-Fisher type 2 Pseudomonas LPS specific IgM antibody and was purchased from Genetic Systems Corporation. The monoclonal antibody from this cell line was produced in our laboratory. It serves as an isotype matched control mAb for the human anti-rhTNF.alpha. mAbs. The C7F7 mAb is a mouse IgG.sub.1 anti-hFVIII developed in collaboration with Genentech Inc. and is used as a isotype matched control mAb for the mouse anti-rhTNF.alpha. mAbs.

Goat anti-mouse IgG and biotinylated goat anti-human IgG were purchased from Jackson Labs. Biotinylated goat anti-mouse IgG and biotinylated mouse anti-human IgM were purchased from Zymed. Avidin coupled HRP and avidin coupled alkaline phosphatase were purchased from Zymed.

Phycoerythrin conjugated anti-CD3 and anti-CD19 antibodies were purchased from Dakopatts. Phycoerythrin conjugated anti-LeuM3 was purchased from Becton Dickinson. Fluorescein (FL) conjugated F(ab)'.sub.2 anti-human IgM, FL-F(ab)'.sub.2 anti-human IgG and FL-F(ab)'.sub.2 anti-mouse IgG antibodies were purchased from Cappel.

ELISAs

Antigens or capture antibodies (anti-immunoglobulin antibodies) were coated to plastic plates in carbonate/bicarbonate buffer, or PBS containing 20 ug/ml BSA, overnight at 4.degree. C. or 3 hrs at 37.degree. C. Secondary incubations were carried out overnight at 4.degree. C. or at room temperature for a period of 2 hrs or less. Secondary antibodies were biotinylated and their binding was revealed using avidin coupled HRP and avidin coupled alkaline phosphatase.

Hybridoma Production

The human IgM mAbs were produced by fusion with the mouse P3X63Ag8.653 non-secreting myeloma. Peripheral blood mononuclear cells from a CMV positive donor were separated by centrifugation on Ficoll, treated with L-leucyl leucine methyl ester, incubated in vitro with antigen and subsequently transformed with EBV. Transformants were distributed at limiting concentrations and cells producing antibody binding to TNF were fused and subsequently subcloned. The B5 hybridoma was subcloned a minimum of 5 times and was deposited with ATCC on Mar. 24, 1993, as deposit CRL 11306. The H5 and 7T1 mAbs were produced by fusion of human tonsillar cells immunized in vitro. Monoclonal human IgM antibodies were affinity purified by standard techniques for use in subsequent experiments.

Cytotoxicity Assay

To assess the TNF neutralizing ability of various mAbs, the assay described by Galloway et al (cited above) was used with the following minor modifications. Briefly, 20 pg/ml TNF were incubated overnight with 60,000 WEHI 164 cells and the test mAb. Viable cells were then detected by crystal violet staining and reading optical density at 570 nm.

Western Blotting

Recombinant huTNF.alpha. (100 ug/ml plus 100 ug/ml BSA) and recombinant mTNF.alpha. (5 ug/ml with 100 ug/ml BSA) were electrophoresed in the presence of .beta.mercaptoethanol and SDS on 12% polyacrylamide gels. Proteins were then electro-transferred to nitrocellulose which was then blocked with BSA. Test mAbs were allowed to bind and were subsequently detected with biotinylated anti-immunoglobulin reagents. Streptavidin-HRP was then added followed by substrate.

Fluorescence Analyses

One million cells were stained with optimal concentrations of primary antibody, usually 2.5-40 ug/ml at 4.degree. C. for 1/2 hour in PBS containing 1% FBS and 0.02% sodium azide. Optimal concentrations of fluorescent secondary antibodies were added, after two cell washes, for a similar time in similar buffer. After washing, cells were fixed with 2% paraformaldehyde solution. Cell fluorescence was then analyzed on a FACSCAN (name of instrument).

Inhibition of LPS stimulation of TNF.alpha. Secretion

One million THP-1 cells/ml were incubated 4 hrs with 1 ug/ml E. coli LPS in the presence or absence of 40 ug/ml human IgM antibodies. Supernatants were harvested, centrifuged, filtered and assayed for TNF.alpha. cytotoxicity in the WEHI 164 assay mentioned above. Supernatants were titrated and viability was plotted against supernatant dilution. These curves were compared to a standard curve using rhuTNF.alpha. to determine the actual concentrations of TNF.alpha. produced by the cells.

The monoclonal human IgM antibody B5 binds to solid phase recombinant human TNF (rhTNF.alpha.). Several hybridomas secreting monoclonal anti-rhTNF.alpha. antibodies have been established in our laboratory. An endpoint titer analysis was performed comparing a panel of 6 human IgM mAbs and 3 human IgG mAbs to three high affinity neutralizing mouse mAbs, A10G10, A6 and B6. ELISA plates were coated with 2 ug/ml rhTNF.alpha.. The indicated mAbs were added in titrated concentrations and binding was assessed spectrophotometrically. The minimum mAb concentrations yielding detectable rhTNF.alpha. binding are shown. B5 and F12 (F80-1B9-F12) were two of the best human IgM mAbs by this criterion, showing endpoint titers in the subnanogram/ml range. Table 1 presents the data below.

It should be noted that the ranges and endpoint titers were similar for the IgM anti-TNF.alpha. mAbs and the IgG anti-TNF.alpha. mAbs.

FIG. 1 presents a more extensive comparison of the human B5 and mouse A10G10 mAbs. Binding of both mAbs was concentration dependent regardless of TNF coating concentration. The B5 mAb bound slightly better than A10G10 with high TNF coating concentrations. As the TNF coating concentration was reduced, however, the binding of B5 decreased more rapidly than that of A10G10. This is consistent with B5 having a lower affinity than A10G10 for rhTNF.alpha.. These data show that the B5 mAb binds to solid-phase rhTNF.alpha..

B5 mAb binds to a different epitope on rhTNF.alpha. than those bound by three mouse anti-TNF mAbs. Competitive binding experiments have shown that A10G10 and B6 recognize similar epitopes on rhTNF.alpha. whereas A6 recognizes a different epitope (data not shown). To examine the epitope binding specificity of B5, competitive binding experiments were performed using the mouse mAbs and B5.

The mouse mAbs were added at different concentrations to ELISA plates previously coated with TNF.alpha.. An optimum concentration of B5 mAb was then added and binding was subsequently detected with biotinylated anti-human IgM. If the mouse mAbs recognize the same epitope as B5 mAb, they should inhibit B5 mAb binding in a concentration dependent manner.

As shown in FIG. 2A, binding of the mouse mAbs to plate bound rhTNF.alpha. is concentration dependent. FIG. 2B shows that none of the mouse mAbs interfered with rhTNF.alpha. binding by a fixed amount of B5 mAb, even at concentrations of the mouse mAbs significantly in excess of those required for maximal binding to the plate. These data suggest that B5 recognizes an epitope on rhTNF.alpha. different from those recognized by A10G10, A6 and B6.

To support this finding, rhTNF.alpha. was added to ELISA plates previously coated with the combination of A10G10, B6 plus A6 mAbs. B5 mAb was then added to test whether it could bind to rhTNF.alpha. complexed to, or captured by, the mouse mAbs.

FIG. 3 shows that B5 and all the other human IgM mAbs, except 7T1, bound to rhTNF.alpha. complexed to mouse mAbs. Binding of the human mAbs was not seen in the absence of rhTNF.alpha., demonstrating specificity for some epitope of rhTNF.alpha.. The failure of 7T1 mAb to bind to complexed TNF may be simply due to a low affinity. These results support the conclusion that the human IgM mAbs B5, F12, A1, B6 and D6 and the three mouse mAbs recognize different epitopes on rhTNF.alpha..

B5 mAb is not polyreactive. Since B5 mAb is a human IgM which binds to human TNF.alpha., and therefore has properties which define it as an autoantibody, it was important to determine the quality of this mAb and assess its polyreactivity. We chose a panel of human and non-human antigens typically used to define polyreactivity. Binding of these antigens by B5 mAb, A10G10, two control polyreactive human IgM mAbs 1A6B5F and F2.2.34 and two other human IgM anti-TNF mAbs was compared. The results have been normalized for each antibody to allow direct comparison.

FIG. 4 presents the data from one of four similar experiments. The mouse mAb A10G10 binds specifically to rhTNF.alpha. and none of the other antigens. In contrast, the polyreactive mAb 1A6B5F binds to virtually all of the antigens tested. The same was true for the other polyreactive mAb F2.2.34, although binding to BSA and TNF was much stronger than that seen with the other antigens. The B5 mAb showed specificity for rhTNF.alpha.. No binding by B5 mAb to recombinant human lymphotoxin (rhTNF.beta.) nor to any of the other antigens tested was observed. These data provide evidence that the B5 mAb is not polyreactive.

In contrast, the 7T1 and H5 human IgM mAbs bind to human Fc fragments indicating a rheumatoid factor nature. These two antibodies also bind to insulin and 7T1 binds BSA as well. The control polyreactive mAbs appear to define two classes of polyreactivity; one being very broad in specificity and the other being more restricted in the antigens recognized. The 7T1 and H5 mAbs belong to the more restricted class of polyreactive mAbs. The F12 anti-TNF mAb binds to human TNF.alpha. but only marginally to other antigens.

B5 mAb binds to recombinant mouse TNF.alpha.. During the course of analyzing the specificity of the B5 mAb, we noticed that it also bound to mouse TNF.alpha.. To demonstrate this, we first captured mouse TNF.alpha. with a neutralizing hamster monoclonal antibody and then allowed B5 to bind to this complex. FIG. 5 shows the results of this kind of experiment. The binding of B5 was dependent on both the concentration of B5 present, and on the concentration of hamster antibody used to coat the plates. No binding was observed when mouse TNF.alpha. was not added, indicating the specificity of B5 binding in this system. Other experiments not shown revealed binding to mouse TNF.alpha. by the F12 mAb.

B5 mAb binds to soluble rhTNF.alpha. with detectable but low affinity. Next, we assessed the mAb's ability to bind to soluble rhTNF.alpha.. ELISA plates were coated with anti-human IgM and B5 was then added. The ability of the bound B5 mAb to capture biotinylated rhTNF.alpha. was then determined.

FIG. 6 compares the abilities of A10G10 and B5 to bind soluble TNF.alpha. under these conditions. Although both mAbs bind soluble rhTNF.alpha., about 300-fold higher concentration of B5 mAb is required for binding equivalent .degree. 8 to that of A10G10. Furthermore, binding of soluble TNF.alpha. to immobilized B5 did not saturate with the concentrations of B5 tested. These results are consistent with a low affinity binding of rhTNF.alpha. by B5 mAb. Indeed, attempts to measure the binding constant of B5 mAb revealed an affinity too low to calculate by conventional methods (data not shown).

Soluble rhTNF.alpha. binding by B5 was also demonstrated by coating plates with anti-IgM, capturing B5 and then adding unmodified soluble rhTNF.alpha.. A10G10 was added next and its binding to this B5-complexed form of rhTNF.alpha. was detected with biotinylated anti-mouse IgG. FIG. 7 compares the abilities of B5 and a control human IgM, 6F11, to capture and present soluble rhTNF.alpha. to A10G10. Although some non-specific binding was seen with the control mAb, B5 mAb bound approximately four- to eight-fold more rhTNF.alpha. in this experiment. These data are consistent with a low binding constant of B5 and add further support for the concept that B5 mAb and A10G10 mAb recognize different epitopes on rhTNF.alpha..

B5 mAb recognizes rhTNF.alpha. in Western blots. FIG. 8 shows the results of an experiment using western blotting to demonstrate B5 binding to denatured TNF.alpha.. The images have been enhanced for clarity. In lanes A-G, binding to mouse TNF.alpha. was examined and in lanes H and I binding to human TNF.alpha. was examined. The 6F11 antibody did not bind to either TNF.alpha. species and so provides a specificity control. All the human IgM mAbs, 7T1, H5, 1A6B5F and B5 bind to mouse TNF.alpha.. Furthermore, the B5 antibody also binds to human TNF.alpha., under these conditions. These results suggest that B5 may recognize a linear epitope of rhTNF.alpha..

B5 mAb does not neutralize the cytotoxicity of rhTNF.alpha.. The TNF sensitive cell line WEHI 164 was used to assess the ability of B5 mAb to neutralize TNF.alpha. cytotoxicity. FIG. 9 shows that A10G10 clearly neutralizes rhTNF.alpha. in a dose dependent manner as previously demonstrated by Galloway et al (cited above). At no concentration of B5, however, was any neutralization of rhTNF.alpha. observed. The same is true for the three other human IgM anti-TNF.alpha. mAbs B6, F12 and 7T1 which were tested. These data add further support to the idea that B5 and A10G10 bind different epitopes of TNF and are consistent with the ability of B5 mAb to bind soluble rhTHF.alpha. weakly.

The B5 mAb anti-rTNF.alpha. binds to the surface of several different cell lines. Since the B5 mAb binds specifically to rTNF.alpha., several cell lines were chosen to test whether or not the mAb would bind to their surfaces. FIG. 10 shows the results of a typical experiment using two cell lines. The EBV transformed human B lymphoblastoid cell line 8B9 and the human monocyte cell line THP-1 were stained with either B5 anti-TNF.alpha. or the 6F11 anti-Pseudomonas LPS mAbs and then fluorescent anti-human IgM F(ab)'.sub.2 fragments.

The 8B9 cells were stained well with the B5 mAb whereas no significant binding to the cell surface was seen with the control 6F11 mAb. B5 staining was also observed with THP-1 cells. However, fewer cells in this population were stained and the observed staining was somewhat dimmer than that seen for the 8B9 cells. Nevertheless, nearly 1/3 of the cells in the THP-1 population expressed cell surface TNF.alpha. (csTNF.alpha.), as detected with the B5 mAb. It is unclear whether this level of staining reflects some regulation of csTNF.alpha. expression or whether it is due to clonal variation within the cell line.

The concentration dependence of B5 binding to cell surfaces was examined more closely with the THP-1 monocyte and U937 histiocyte cell lines. These cells were stained with titrated amounts of B5 antibody after incubation with either no stimulus, LPS or LPS+PMA for 3 hrs. The results are shown in FIG. 11. In all cases, B5 binding to cells was dose dependent. Interestingly, more binding was observed for both cell lines when they were preincubated with LPS or LPS+PMA. This was especially apparent for the U937 cell line. This increase is consistent with the known ability of these agents to induce TNF secretion by monocyte cell lines. Upon stimulation, B5 binding to the cells was apparent, even at several hundred nanograms/ml of antibody.

Table 2 shows the results of two experiments in which the binding of B5 anti-TNF.alpha. mAb was surveyed. Cells were stained with the indicated primary antibodies and fluorescein labeled anti-human IgM (.mu.-specific) secondary antibody. The percentages of cells staining positively are shown as determined on a FACSCAN instrument.

A variety of cell lines were tested including those of human B and T lymphocyte, breast carcinoma, astrocytoma, glioblastoma, monocyte, histiocyte, melanoma, and monoblast origin. A mouse T cell lymphoma was tested as well. Of the 15 lines tested, only the breast carcinoma U118MG showed no binding by B5. The others exhibited a range in the percentages of cells within each population which expressed csTNF.alpha. from a low of around 8% for the A375 melanoma to over 90% for EBV transformed B cells. The class matched 6F11 anti-LPS mAb failed to stain any of these cell lines. This and the negative cell line indicate that the B5 staining seen was specific and not the result of a general affinity for all cells.

Lack of Neutralizing Mouse anti-TNF.alpha. mAb Binding to cs TNF.alpha.

ELISA experiments have shown the TNF specificity of the B5 mAb and demonstrated its binding to an epitope on TNF.alpha. different from that bound by the neutralizing mouse mAb A10G10. We next examined whether or not the epitope recognized by A10G10 mAb was expressed on the surface of cells to which B5 binds.

Table 3 presents data from five experiments addressing this issue using the U937 and THP-1 cell lines. Cells were stained with the indicated primary antibodies and fluorescein labeled anti-mouse IgG (.gamma.-specific) or anti-human IgM (.mu.-specific) secondary antibodies. The asterisk (*) indicates that F(ab)'.sub.2 fragments of A10G10 mAb were used. The percentages of cells staining positively are shown as determined on a FACSCAN instrument. Not determined is signified by nd.

In all five experiments the B5 mAb bound to each cell line. On the other hand, A10G10 mAb did not bind, to a significant degree, in four of the experiments. In one of the five experiments, a small amount of binding by A10G10 to the U937 cells was observed. Taken together, these data suggest that TNF.alpha. is on the surface of these cell lines, but the epitope recognized by A10G10 is only rarely available for binding by mAbs in the absence of exogenous stimulation.

LPS Induction of cell surface TNF.alpha. Expression

LPS is a commonly used agent to induce TNF.alpha. secretion by human monocytes. We incubated THP-1 and U937 cells with LPS to examine whether or not csTNF.alpha. expression can be increased. Table 4 shows the results of three experiments. Stimulation was performed by 3 or 4 hour incubation with 100 ng/ml LPS. Cells were stained with the indicated primary antibodies and fluorescein labeled anti-mouse IgG (.gamma.-specific) or anti-human IgM (.mu.-specific) secondary antibodies. The asterisk (*) indicates that F(ab)'.sub.2 fragments of A10G10 mAb were used. The percentages of cells staining positively are given as determined on a FACSCAN. Not determined is signified by nd.

In all three experiments, LPS increased the amount of B5 binding to THP-1 cells. This was true also for U937 cells in two of the three experiments. In contrast to noninduced cells, LPS stimulation led to the ability to be stained by the A10G10 mAb for both the THP-1 and the U937 lines. Nevertheless, the percentages of cells in either line expressing TNF.alpha. epitopes recognized by A10G10 were small, in comparison to those percentages seen with the B5 mAb. These data suggest that csTNF.alpha. can be increased by incubation with LPS and that this increase correlates with the acquisition of TNF.alpha. epitopes recognized by neutralizing antibodies.

Influence of Factors other than LPS on csTNF.alpha. Expression

During the course of our experiments, some of our cell lines lost some spontaneous csTNF.alpha. expression. To examine the influence of fetal bovine serum (FBS) on csTNF.alpha. expression, THP-1 cells were cultivated four days in the different lots of fetal bovine sera and analyzed for cell surface TNF.alpha. expression. Table 5 shows typical results. Shown are the percentages of cells staining positively with the indicated primary and fluorescent secondary staining antibodies. The endotoxin concentrations, in Limulus amebocyte lysate units, for FBS lots 1079, 1087, 2081 and 1026 are 0.125, 0.250, 0.060 and 0.750, respectively. Analyses were performed with a FACSCAN instrument.

The FBS lot had a large influence on csTNF.alpha. expression by THP-1 cells. The difference in expression varied by about a factor of four depending on the particular FBS batch used. Comparison of the endotoxin levels in these different lots revealed no direct correlation with csTNF.alpha. levels. These data suggest that factors other than LPS can influence expression of csTNF.alpha..

Specificity of B5 mAb Binding to csTNF.alpha.

Table 6 presents data which confirm the specificity of B5 mAb binding to the THP-1 cells. B5 mAb at 10 ug/ml was incubated with the indicated concentrations of inhibitors prior to exposure to LPS stimulated THP-1 cells. Its binding was detected with fluorescein conjugated F(ab)'.sub.2 anti-human IgM antibody. LT is recombinant human lymphotoxin, ECD55 is the recombinant extracellular TNF.alpha. binding domain of the p55 TNF receptor and A10G10 is the neutralizing mouse IgG.sub.1 anti-TNF.alpha. mAb. Analyses were performed with a FACSCAN instrument.

Preincubation of the B5 IgM mAb with TNF.alpha. inhibited its subsequent cell surface binding, in a dose dependent manner, whereas preincubation with lymphotoxin did not, except for a small effect at the highest concentration. The lack of complete inhibition with the high doses of TNF.alpha. is consistent with the previously documented low affinity of this mAb for soluble TNF.alpha.. Interestingly, preincubation of B5 mAb with A10G10 and subsequent addition of both did not decrease B5 binding. These data suggest that neutralizing A10G10 does not compete for the same epitope on TNF.alpha. to which B5 mAb binds.

B5 Binds to csTNF.alpha. on Fresh Human Spleen Cells

The previous sections establish B5 binding to csTNF.alpha. on several different cell lines. To determine whether or not B5 binds to untransformed cells, experiments were performed with human splenocytes.

To analyze B cell expression of csTNF.alpha. by B5, we used unconjugated B5 IgM since direct fluoresceination or biotinylation of this antibody was very inefficient or interfered with its TNF.alpha. binding ability. Fluorescent F(ab)'.sub.2 fragments of anti-human IgM antibody were used to detect B5 binding. Since many normal B cells already express sIgM as an antigen receptor, it was not always possible to detect csTNF.alpha. as an increase in the percentage of sIgM+ cells. We could, however, detect csTNF.alpha. by measuring the increase in staining intensity with the fluorescent anti-IgM when cells are incubated with the B5 mAb compared to incubation with either control 6F11 IgM mAb or no antibody at all.

FIG. 12 demonstrates this shift in fluorescence intensity seen when the B5 mAb binds to B cells. FIG. 12A shows the fluorescence histogram of cells stained with anti-IgM antibody alone. FIG. 12B shows a histogram of these same cells when first reacted with B5 mAb anti-TNF.alpha. and subsequently restained with florescent anti-IgM antibody. The most useful statistic to measure this shift is the median channel of fluorescence intensity, or simply median channel. The median channel numbers are presented in the following tables when B cells are examined.

Table 7 presents the data from two experiments using splenic biopsy material. The expression of csTNF.alpha. on monocytes, T cells and B cells was examined by two color immunofluorescence analysis using phycoerythrin conjugated anti-LeuM3, anti-CD3 and anti-CD19, respectively, in conjunction with fluorescein conjugated anti-human IgM. Human splenocytes received one day after biopsy were analyzed for expression of cell surface staining with the indicated monoclonal antibodies. Small lymphocytes were gated by forward and side scatter properties and then analyzed. T cells, B cells and monocytes were stained with phycoerythrin conjugated anti-CD3, anti-CD19 and anti-LeuM3 antibodies, respectively. Two color analyses were then performed on these populations using fluorescein labeled F(ab)'.sub.2 anti-human IgM and the indicated IgM mAbs. Underlined values represent those which show significant increases in the percentage of positively stained cells or show greater than twice the fluorescence intensity of the appropriate control population. Analyses were performed with a FACSCAN instrument.

In both experiments, monocytes constituted less than 5% of the total splenocyte populations. Of these, a significant fraction in both experiments were stained with the anti-TNF.alpha. B5 mAb. On the other hand, these cells were not stained with the control 6F11 human IgM mAb. These results suggest that some splenic monocytes express csTNF.alpha..

CD3+ T cells showed variable expression of csTNF.alpha.. While the percentages of csTNF.alpha. positive T cells varied in these experiments, the staining with the B5 mAb was much weaker than that seen for B cells and monocytes. The median fluorescence intensity for T cell csTNF.alpha. was not even twice that seen for the background controls. These results suggest that a variable proportion of splenic T cells express small amounts of csTNF.alpha..

Analysis of B cell csTNF.alpha. expression revealed quite strong csTNF.alpha. expression. As seen in spleen Z, the percentage of IgM+ B cells increased after incubation with B5 mAb. Furthermore the staining intensity of the entire B cell population approximately tripled. No increase in staining was seen with the 6F11 control antibody, indicating the specificity of the B5 staining on B cells.

The polyreactive mAb 7T1 and H5 were included in these analyses. In addition to binding to TNF.alpha., these antibodies react with several other antigens. Hence the specificity of their cell surface binding is unknown. We include them for comparison not only since they do bind to TNF, but also since little data on binding of polyreactive mAbs to unfixed cells is available. These antibodies do appear to react with T cells and B cells but they react with monocyte surfaces far better. In addition to significant increases in the percentages of B and T cells staining with these antibodies, the majority of monocytes in both experiments were stained.

These data suggest that the B5 anti-TNF.alpha. mAb can react with splenic lymphocytes of the B and T lineages as well as being able to recognize and bind to splenic monocytes.

B5 Binding to csTNF.alpha. on Cultured Human Spleen Cells

Spleen cells from one individual examined in Table 7 were cultivated in vitro for 3 days with various stimuli and were then analyzed for B5 mAb binding. Results are shown in Table 8. Cultivation of these cells resulted in loss of monocytes so data for Leu-M3+ cells are not presented. The cells were stained for CD3 or CD19 with phycoerythrin conjugated antibodies to allow two color analyses with fluorescein conjugated F(ab)'.sub.2 anti-human IgM and the indicated human IgM mAbs. All cells were analyzed when no activator was included in culture but only large activated cells were analyzed from cultures which included activators. Underlined values represent those which show significant increases in the percentage of positively stained cells or show greater than twice the fluorescence intensity of the appropriate control population. Analyses were performed with a FACSCAN instrument.

The cells cultured in medium were 55% CD19+ (B cells) and 22% CD3+ (T Cells). Of the CD19+ cells, 85% were sIgM+ with a median channel intensity of 54. Staining with the B5 mAb increased this intensity to median channel 294--nearly six fold higher. This increase was not seen with the polyreactive or control IgM mAbs. Increases in percentages of CD3+ T cells which bound B5 mAb were also seen, although the intensity of staining was low. Despite the fact that anti-IgM alone revealed some T cell staining, addition of 6F11 to these T cells did not result in increased anti-IgM staining, showing the specificity of B5 staining and suggesting the B5 mAb is not binding to the IgM receptors expressed on activated T cells. These receptors are presumably already occupied and account for the background staining observed with the anti-IgM secondary antibody.

Stimulation with the superantigen Staphylococcal Enterotoxin B (SEB), which activates both T and B cells, resulted in about 24% of the T cells binding the secondary anti-human IgM antibody. However, about 66% of the SEB activated T cells bound B5 anti-TNF.alpha. mAb. No increase in sIgM+ T cells was seen with the 6F11 control mAb. These data indicate induction of csTNF.alpha. expression when T cells are activated.

B cells activated by either anti-IgD-dextran or Staphylococcus aureus Cowan Strain I (SAC), both potent B cell mitogens, demonstrated binding by B5 anti-TNF.alpha. mAb. The higher B5 staining fluorescence intensity seen after SAC induction suggests a higher B cell surface level of TNF.alpha. expression than seen on anti-IgD activated B cells, or B cells cultured in medium alone. These data suggest that both activated human B cells and T cells express csTNF.alpha.epitopes recognized by the B5 mAb.

Binding of B5 mAb to Human and Chimpanzee Peripheral Blood Lymphocytes

To extend the finding of human splenic lymphocyte expression of csTNF.alpha., peripheral blood lymphocytes of human and chimpanzee origin were examined. Table 9 shows the results obtained with blood from two chimpanzees and one human. The chimpanzee blood was received one day after it was drawn whereas the human blood was fresh. The delay in receipt of the blood appeared to result in loss of monocytes from the chimpanzee blood. Peripheral blood mononuclear cells were prepared by separation on Ficoll and stained with PE derivatized anti-CD3, CD19 or LeuM3. For the chimpanzees 171 and 203, less than 2% and 0.6% of cells were LeuM3+, respectively. Some 20.2% of the human cells were LeuM3+. T cells comprised 62% and 54% of the chimpanzee lymphocytes and 68% of the human lymphocytes. B cell percentages were 2.8 and 5.4 for the chimpanzees and 16.4% for the human. Cells were incubated with the indicated IgM primary antibodies and subsequently stained with the fluorescein conjugated F(ab)'.sub.2 anti-human IgM reagent. Analyses were performed with a FACSCAN instrument. Underlined values represent those which show significant increases in the percentage of positively stained cells or show greater than twice the fluorescence intensity of the appropriate control population.

In contrast to the previous results with human spleen, the fresh peripheral human monocytes did not express csTNF.alpha. as seen by the B5 mAb. A significant fraction of these cells did, however, bind the polyreactive mAbs 7T1 and H5.

The fresh human T cells did not express surface IgM whereas the chimpanzee T cells drawn one day previously did. T cells from both species, however, expressed modest amounts of csTNF.alpha. detected by the B5 mAb. This anti-TNF.alpha. staining was very weak, however, and suggests only low levels of csTNF.alpha. were present. T cells from neither species were recognized by polyreactive 7T1 or H5.

In contrast to the T cells, peripheral blood B cells from both chimpanzees and the human displayed high levels of csTNF.alpha. seen by B5 mAb. This expression was much more intense than that seen with the T cells. These results suggest that normal human peripheral blood monocytes do not express csTNF.alpha. whereas some T lymphocytes and most B lymphocytes from both species do express this cell surface cytokine.

B5 anti-TNF.alpha. mAb inhibits LPS induced secretion of TNF.alpha. by THP-1 Cells

To examine whether or not the binding of B5 mAb to csTNF.alpha. had any functional significance, we stimulated the THP-1 human monocyte cell line with LPS in the presence of B5 or other human IgM mAbs. We assayed secretion of biologically active TNF.alpha. by measuring cytotoxic activity of the supernatants on the TNF.alpha. sensitive WEHI 164 cell line. The results of two of four such experiments are given in Table 10. THP-1 cells were stimulated for 4 hours with 100 ng/ml E. coli LPS in the presence of 40 ug/ml of the indicated TNF non-neutralizing human IgM mAbs. Supernatants from these incubations were then tested for cytotoxicity against the TNF.alpha. sensitive WEHI 164 cell line. All supernatant cytotoxicity was concentration dependent and was neutralized by A10G10 anti-TNF.alpha. mAb, indicating cytotoxicity was due to TNF.alpha.. Concentrations of secreted TNF.alpha. were determined by comparison to a standard curve.

Stimulated THP-1 cells did secrete active TNF.alpha. and all of this cytotoxic activity was inhibited by including A10G10 in the cytotoxicity assay (data not shown). Previous experiments including B5 mAb in the cytotoxicity assay have shown that B5 does not neutralize TNF.alpha. (FIG. 9). Table 10 shows that coculture of the THP-1 cells with B5 mAb inhibits LPS induced TNF.alpha. secretion. These data suggest that B5 mAb interaction with csTNF.alpha. can inhibit LPS induced TNF secretion.

An additional feature of the antibodies of this disclosure is that they can be used to mediate inhibition of mitogen-induced proliferation of human lymphoid cells as shown in Table 11 below.

Human splenocytes were cultured for 3 days with the indicated mitogens and antibodies. 3HTdR was added on the final day for 6 hours, then cells were harvested and thymidine incorporation was assessed.

The mAbs of this Application have several useful features.

First, the human monoclonal anti-TNF antibodies of this disclosure could be used to detect and/or measure TNF in vitro using conventional immunoassay techniques. For example, they could be used is a diagnostic manner to assess expression of cell surface TNF on human and murine cells, and perhaps cells from other species.

Second, by binding to cells expressing surface TNF, the antibody may initiate killing of these cells via complement fixation. This might prove useful for depleting cells expressing surface TNF. For example, cells from a patient might be removed, treated with the antibody and complement, or a similar reagent leading to cytotoxicity of cells bound by the antibody, and the remaining cells might be reintroduced back into the donor. This might be useful in depleting patients of peripheral B cell leukemia cells, or other leukemic cells expressing surface TNF.

Third, the antibody might be introduced into patients as a therapeutic agent to help kill or remove cells expressing surface TNF. We have shown many activated cells, including some carcinoma cell lines express surface TNF. These types of cells, as well as others expressing surface TNF, could serve as appropriate targets for treatment by this antibody invention.

Fourth, the antibody might be introduced into patients where TNF production contributes to the disease process or state. The aim here would be to inhibit TNF production. Since we have shown that binding of the antibody can inhibit TNF secretion by some cells, this avenue of therapy may be beneficial.

Fifth, the antibody might be introduced into patients to slow or inhibit the growth of cells expressing surface TNF. We have shown that mitogen activated human cells can express surface TNF that is bound by the B5 antibody. (See Table 11.) Again, a specific application might be in cancer or leukemia therapy.

A primary advantage of this invention is that it comprises a human anti-TNF antibody and, as such, it is expected to be far less immunogenic than antibody from any other species. The properties which distinguish the invention from any putative and ill-defined natural anti-TNF antibody are its specificity and binding capacity. Unlike the B5mAb invention described in this disclosure, no other human antibodies in the literature have been proven to be TNF-specific.

To our knowledge, this is the first report of a monoclonal human autoantibody specific for human and mouse TNF.alpha.. It is unclear whether or not the CMV seropositive donor origin of B5 mAb is significant. The antibody is clearly different from the mouse mAbs we have generated to TNF.alpha., all of which are neutralizing, as shown previously by Galloway et al (cited above).

Three lines of evidence suggest that B5 mAb recognizes an epitope different from those recognized by the mouse mAbs described. First, there is no competition between the human and mouse mAbs for binding to plates coated with TNF. Second, TNF bound by the human mAb can be recognized by the mouse mAbs, and vice versa. Finally, B5 mAb does not neutralize rhTNF.alpha. whereas the mouse mAbs do. One might argue that TNF.alpha. is a trimer and, as such, TNF.alpha. bound to neutralizing mouse mAbs attached to plates can still present an identical epitope to be recognized by mAb B5. The lack of competition between the mouse mAbs and mAb B5 for plate bound TNF.alpha. is a strong argument against this possibility. The competition data in combination with the lack of neutralizing activity of B5 mAb support the interpretation of distinct epitope recognition by the mouse and human mAbs. The biological effects of TNF.alpha., especially its ability to promote Ig secretion, may preclude the generation of a high affinity neutralizing human anti-TNF.alpha. autoantibody by the techniques used. This ability may also explain the different epitope specificities of B5 mAb and the three neutralizing mouse mAbs.

The base of the bell shaped trimeric TNF.alpha. molecule, which contains the amino terminus apposed to the carboxy terminus, is the region of the molecule which binds to TNF receptors (M. J. Eck et al "The Structure of Tumor Necrosis Factor-.alpha. at 2.6.ANG. Resolution, Implications for Receptor Binding." J. Biol. Chem. 264:17595-605, 1989; and A. Corti et al "Antigenic Regions of Tumor Necrosis Factor Alpha and Their Topographic Relationships with Structural/Functional Domains." Molec. Immunol. 29:471-9, 1992). Since the mouse mAbs used in this report neutralize TNF.alpha., and have been found to block binding of TNF.alpha. to its receptors, it is likely that an epitope in the base of the trimer is recognized by these antibodies. From the data presented in this report, one might speculate that the B5 mAb sees a region of the TNF.alpha. molecule closer to the "top" of the trimer.

The weak binding of B5 mAb to soluble TNF.alpha. is consistent with a low binding constant of the mAb for the ligand. Nevertheless, the valency of this IgM mAb can outweigh this shortcoming so that B5 can bind to solid phase TNF.alpha. as well as, or better than, the high affinity neutralizing mouse anti-TNF.alpha. mAbs tested. Apparently, multipoint binding allows the mAb B5 to adhere strongly to TNF.alpha. when a sufficient antigen density is available.

Although B5 appears to bind with low affinity, we show that it binds specifically to TNF.alpha. and fails to bind to any of the other antigens tested. This contrasts with the observed binding of two other control polyreactive mAbs. Hence, B5 appears to be monospecific and is not polyreactive. B5 seems to bind specifically to an epitope, most likely a linear epitope, shared by mouse and human TNF.alpha.. These properties classify B5 as an autoantibody and distinguish it from other mAbs so far described.

The human B5 autoantibody binds to surface TNF.alpha. on a broad range of human cell lines and lymphoid cells. It is not surprising that it recognizes chimpanzee TNF.alpha. as there is no difference in the amino acid sequences of TNF.alpha. from chimpanzee and human. We have also shown that B5 recognizes mouse TNF.alpha. which is about 80% identical to human TNF.alpha. (D. Pennica et al "Cloning and expression in Escherichia coli of the cDNA for Murine Tumor Necrosis Factor", Proc. Natl. Acad. Sci. USA 82:6060-4, 1985). Hence it is not surprising that B5 recognizes mouse csTNF.alpha..

Others have certainly described TNF production by human B cells (M. Jaatela, "Biology of Disease. Biologic Activities and Mechanisms of Action of Tumor Necrosis Factor-a/Cachectin", Lab. Invest. 64:724-42, 1991; and Smeland et al "Interleukin 4 Induces Selective Production of Interleukin 6 from Normal Human B Lymphocytes", J. Exp. Med. 170:1463-68, 1989), T cells (S. -S. J. Sung et al "Production of Tumor Necrosis Factor/Cachectin by Human T Cell Lines and Peripheral Blood T Lymphocytes Stimulated by Phorbol Myristate Acetate and Anti-CD3 Antibody", J. Exp. Med. 167:937-, 1988), monocytes (Beutler et al "The Biology of Cachectin TNF-.alpha.: Primary Mediator of the Host Response", Ann. Rev. Immunol. 7:625-55, 1989), B cell lines (S. -S. J. Sung et al "Production of Tumor Necrosis Factor/Cachetin by Human T Cell Lines and Peripheral Blood T Lymphocytes Stimulated by Phorbol Myristate Acetate and Anti-CD3 Antibody", J. Exp. Med. 167:937-, 1988; and G. J. Jochems et al "Heterogeneity in Both Cytokine Production and Responsiveness of a Panel of Monoclonal Human Epstein-Barr Virus-Transformed B-Cell Lines", Hum. Antibod. Hybridomas 2:57-64, 1991), astrocytes (A. P. Lieberman et al "Production of Tumor necrosis Factor and other Cytokines by Astrocytes Stimulated with Lipopolysaccharide or a Neurotropic Virus", Proc. Natl. Acad. Sci. USA, 86:6348-52, 1989; and I. Y. Chung et al "Tumor Necrosis Factor Alpha Production by Astrocytes: Induction by Lipopolysaccharide, IFN-gamma, and IL-1 beta", J. Immunol. 144:2999-3007, 1990; and K. Selmaj et al "Identification of Lymphotoxin and Tumor necrosis Factor in Multiple Sclerosis Lesions", J. Clin. Invest. 87:949-54, 1991) as well as some TNF resistant cell lines (B. Y. Rubin et al "Nonhematopoietic Cells Selected for Resistance to Tumor Necrosis Factor Produce Tumor Necrosis Factor", J. Exp. Med. 164:1350-5, 1986). We extend these findings to include at least one metastatic breast carcinoma, DU4475, a melanoma, A375, and the U373 astrocytoma/glioblastoma. We also demonstrate csTNF.alpha. expression on human splenic lymphoid cells. This is somewhat surprising since previous demonstration of csTNF.alpha. by others tended to employ activated cells.

Although we examined small lymphocytes, as determined by light scatter properties, it is possible many of these cells were partially activated or at a stage of differentiation where they could express this cell surface molecule. The smaller percentages of T lymphocytes and monocytes from human peripheral blood expressing csTNF.alpha. is consistent with the resting phenotype of these cells. In any case, the breadth of csTNF.alpha. expression suggests it has an important role in the surface of many cells.

Others have shown that TNF.alpha. can exist as both an integral transmembrane protein and as a mature protein bound to its receptor on cell surfaces (B. Luettig et al "Evidence for the Existence of Two Forms of Membrane Tumor Necrosis Factor: an Integral Protein and a Molecule Attached to its Receptor", J. Immunol. 143:4034-38, 1989). Several observations suggest that the B5 mAb recognizes the integral transmembrane protein. B5 binding was increased when cells were activated with LPS or PMA. Both agents, but especially PMA, down regulate TNF receptor expression on a variety of cell types (A. H. Ding et al "Macrophages Rapidly Internalize their Tumor Necrosis Factor Receptors in Response to Bacterial Lipopolysaccharide", J. Biol. Chem. 264:3924-9, 1989; and B. A. Aggarwal et al "Effect of Phorbol Esters on Down-Regulation and Redistribution of Cell Surface Receptors for Tumor Necrosis Factor .alpha.", J. Biol. Chem. 262:16450-5, 1987).

B5 binds to unstimulated cell lines whereas cell lines normally need to be induced to secrete TNF. Hence, unstimulated cell lines would be expected to display little receptor bound TNF. We showed that B5 binding to cell surfaces was inhibited by preincubation with TNF.alpha., but not A10G10 anti-TNF.alpha. mAb. This demonstrates the specificity of the B5 antibody.

TNF.beta. binds to the same receptors as TNF.alpha. and so might compete off some receptor bound TNF.alpha. on cell surfaces. The data in Table 6 with high doses of TNF.beta. suggest that this did occur, and was detected by a decrease in B5 staining. For these reasons, it is likely that B5 recognizes the 26 kd transmembrane form of TNF.alpha. and possibly receptor bound TNF.

One puzzling result of these studies is that B5 mAb binds to csTNF.alpha. in many situations in which A10G10 binding is either absent or less than that seen with B5. It is clear that these two antibodies see non-overlapping epitopes. Since A10G10 neutralizes TNF.alpha. cytotoxicity and prevents TNF.alpha. binding to its receptor, this mouse mAb probably binds to TNF.alpha. near the receptor binding domain.

Others have shown that mAbs which bind the amino terminal 15 or so amino acids block TNF.alpha. cytotoxicity (S. H. Socher et al "Antibodies Against Amino Acids 1-15 of Tumor Necrosis Factor Block Its Binding to Cell-Surface Receptor" Proc. Natl. Acad. Sci. USA 84:8829-33, 1987). Hence, it is possible that A10G10 binds to some of the N-terminal amino acids which are most membrane proximal on the transmembrane form of TNF.alpha.. This region may not be accessible to A10G10 for binding, although the TNF molecule itself is present and can be recognized by B5 mAb.

Western blotting experiments suggest that A10G10 does not recognize TNF.alpha. monomers and probably recognizes a conformational epitope (data not shown). If transmembrane TNF.alpha. is primarily monomeric, epitopes recognized by A10G10 may not be present. Additional experiments may help to decide between these and other possibilities.

Interestingly, we did observe A10G10 cell surface binding when cells were activated with LPS. This induction causes secretion of the biologically active TNF.alpha. trimer which can then bind to remaining TNF receptors. Since trimeric TNF.alpha. is multivalent, it may bind to some receptors in a way which allows one or even two remaining receptor binding domains to remain free. It may be this form of csTNF.alpha. which is recognized by A10G10. Indeed, others have shown that incubating unactivated paraformaldehyde-fixed human monocytes with TNF.alpha. results in TNF.alpha. binding its receptors and renders these monocytes cytotoxic. Furthermore, this cytotoxicity is abolished by neutralizing anti-TNF antibodies (A Nii et al "The Incubation of Human Blood Monocytes with Tumor Necrosis Factor-Alpha Leads to Lysis of Tumor Necrosis Factor-Sensitive but not Resistant Tumor Cells", Lymphokine Res. 9:113-24, 1990).

One model which explains much of the data is that transmembrane TNF.alpha. monomers are recognized by B5 mAb. We have shown soluble monomer recognition by B5. Cell surface TNF.alpha. monomers might exhibit an overall conformation different from that of trimeric TNF. They may still expose TNF receptor binding domains and so be capable of mediating cytotoxicity through cell contact. Cells expressing many monomers could thus cause TNF receptor cross-linking on target cells. An activation signal could cause polymerization of TNF monomers in the cell membrane, leading to a conformational change which, in turn, might expose a proteolytic cleavage site leading to release of mature, biologically active trimeric TNF.alpha.. Release could be followed by interaction with TNF receptors and allow A10G10 binding, as suggested above. B5 apparently binds to membrane distal TNF domains and, by so doing, may interfere with either csTNF.alpha. polymerization, a subsequent conformational change, or both. B5 probably does not bind to the proteolytic cleavage site since it does bind to the mature trimeric molecule. This model would explain the cell surface staining results and also explain the observed inhibition of TNF secretion after LPS activation of THP-1 cells. It should be noted that this model allows for a role of the cytoplasmic domain in csTNF.alpha. polymerization. This is only a working model and, as such, is admittedly hypothetical.

The invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description. All changes coming within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.