Chimeric anti-TNF-α monoclonal antibody cA2 binds recombinant transmembrane TNF-α and activates immune effector functions

CHIMERIC ANTI-TNF-α MONOCLONAL ANTIBODY cA2 BINDS RECOMBINANT TRANSMEMBRANE TNF-α AND ACTIVATES IMMUNE EFFECTOR FUNCTIONS Bernard J. Scallon,

Maria Arevalo Moore, Han Trinh, John Ghrayeb

David M. Knight,

Results of clinical trials have indicated that cA2, a neutralizing mouse/human IgG1 chimeric anti-human TNF-α monoclonal antibody, may have therapeutic benefit for rheumatoid arthritis patients. Arthritic joints contain, in addition to elevated levels of soluble TNF-α, high numbers of CD41 T cells and macrophages, cells known to express transmembrane TNF-α upon activation. For that reason, we sought to determine if cA2 binds to transmembrane TNF-α and what effects such binding may have on TNF-α-expressing cells. A cell line expressing a cell-surface, mutant form of transmembrane TNF-α was prepared for these studies. Analysis of these TNF1 cells by flow cytometry, direct binding, and competitive binding assays showed that cA2 binds to the transmembrane form of TNF-α with high avidity. Binding of the IgG1 isotype of cA2, but not an IgG4 version of cA2, resulted in efficient killing of the TNF1 cells by both antibody-dependent cellular toxicity and complementdependent cytotoxicity effector mechanisms. These findings indicate that, in addition to blocking soluble TNF-α activity, cA2 can bind to transmembrane TNF-α in vitro and suggest that cA2 binding may lead to lysis of TNF-α-expressing cells in vivo.

Abbreviations: TNF-α, tumour necrosis factor-α; ADCC, antibody-dependent cellular cytotoxicity; CDC, complement-dependent cytotoxicity; LDH, lactate dehydrogenase; PBL, peripheral blood lymphocytes; RA, rheumatoid arthritis; cA2 G1, IgG1 isotype of cA2; cA2 G4, IgG4 isotype of cA2; Ab, antibody

It is now clear that excessive levels of tumour necrosis factor-α (TNF-α) can exacerbate the types of inflammatory reactions that occur in numerous disease states, including rheumatoid arthritis (RA). A central role for TNF-α in this disease is apparently due, at least in part, to its ability to induce other proinflammatory mediators such as interleukin 1,1 interleukin 6,2 and granulocyte-macrophage colony-stimulating factor,3,4 to induce the chemoattractant peptide interleukin 8,5 and to upregulate adhesion molecules that play a role in recruitment of inflammatory leukocytes (reviewed in 6–9). The prediction that TNF-α may

From the Department of Molecular Biology, Centocor, Inc. Malvern, PA USA 19355 Correspondence to: Dr Bernard Scallon, Dept. of Molecular Biology, Centocor, Inc., 200 Great Valley Parkway, Malvern, PA 19355, USA Received 26 July 1994; revised and accepted for publication 30 September 1994 ©1995 Academic Press Limited 1043-4666/95/03025119 $08.00/0 KEY WORDS: antibody-dependent cellular cytotoxicity/anti-TNFa/complement-dependent cytotoxicity/transmembrane TNF-α CYTOKINE, Vol. 7, No. 3 (April), 1995: pp 251–259

therefore be an attractive therapeutic target for certain diseases such as RA1 has been borne out in animal studies using inhibitors of TNF-α activity. Neutralizing anti-TNF antibodies and recombinant TNF receptors have been shown to improve the pathological state in animal models of RA,10–12 sepsis,13–17 and experimental autoimmune encephalomyelitis,18 as well as in human rheumatoid arthritis patients.19,20 Soluble TNF-α is a homotrimer of 17 kDa subunits secreted mainly by macrophages and activated T cells. The mature 17 kDa polypeptide is derived by proteolytic release from the carboxyl-terminus of a 26 kDa transmembrane form of TNF-α that is oriented with its amino-terminus inside the cell.21 Little is known about the function of this cell-bound form of TNF-α. One possibility is that a transmembrane form may serve to ready this cytokine for immediate release by proteolysis. But recent reports indicate transmembrane TNF-α does have activities similar to that seen with soluble TNF-α.21 Cell-to-cell contact between NIH-3T3 or chinese hamster ovary cells expressing a cell-surface version of TNF-α and the TNF-sensitive murine L929 cells or virally infected C3HA cells resulted in the TNF-sensitive cells being killed.22 In addition, transmembrane TNF-expressed on activated T cells has been reported to be part of the cell-to-cell contact-dependent signal that induces immunoglobulin production by B cells.23 Transmembrane TNF-α expressed on cloned T cells 251

252 / Scallon et al.

CYTOKINE, Vol. 7, No. 3 (April 1995: 251–259)

in activation of immune effector functions, a recombinant version of transmembrane TNF-α was expressed on the surface of SP2/O murine myeloma cells. We report here that cA2 binds the transmembrane form of TNF-α with high avidity and induces efficient killing of the TNF-α-expressing cells in vitro when coincubated in the presence of either human peripheral blood lymphocytes (PBL) or rabbit complement. An IgG4 version of cA2 was only weakly active in triggering ADCC and did not activate complement. These results suggest that the IgG1 isotype of cA2 may have therapeutic benefit, not only by its ability to inhibit soluble TNF-α activity, but also by inducing lysis of cells that express TNF-α.

infected with human immunodeficiency virus-1 appears to be responsible for polyclonal B cell activation.24 Another member of the TNF family, the CD40 ligand, is also expressed on activated T cells and has been shown to activate B cells efficiently.25,26 CD41 T cells and macrophages accumulate in the synovium of RA patients27,28 and apparently are responsible for the elevated levels of TNF-α seen in inflamed joints.29 Because anti-TNF-α antibodies may bind transmembrane TNF-α on the surface of these cells in addition to soluble TNF-α, it is important to determine what effects such binding may have on TNF-α-expressing cells. Indeed, activation of effector functions such as antibody-dependent cellular cytotoxicity (ADCC) or complement-dependent cytotoxicity (CDC) may result in depletion of TNF-α-expressing cells which could have therapeutic benefit in diseases involving high levels of TNF-α. We previously reported the construction of a mouse/human chimeric antibody, cA2 (IgG1 isotype), and demonstrated that it binds and efficiently neutralizes human TNF-α activity.30,31 cA2 has since shown encouraging results in both open and randomized placebo-controlled clinical trials with RA patients.19,20 To determine if cA2 binds the transmembrane form of TNF-α and if such binding would result

A

ATG

RESULTS TNF-α structure A modified genomic copy of the human TNF-α gene was constructed and placed downstream of a cytomegalovirus (CMV) promoter in plasmid pB936 (Fig. 1). The encoded transmembrane TNF-α differs from the wild-type transmembrane form by lacking amino acids Ala21 to Pro12 and by having an Ala in place

P

TGA

wild-type TNFα gene

mutant TNFα gene in pB936

CMV

gpt

B +1

+12

wild-type mutant Figure 1. Structure of the modified human TNF-α gene. Genomic DNA was used as template in polymerase chain reactions (PCR) designed to introduce amino acid deletions near the proteolytic cleavage site (P) in wild-type TNF-α. (A) Schematic illustration of the intron-exon structure of the wild-type gene and the mutant TNF-a gene in pB936. Boxes represent exon coding sequences and horizontal lines intron sequences. The stippled boxes represent 5′ or 3′ non-coding sequence, white boxes represent the pro-TNF-α sequence that includes the cytoplamic and transmembrane domains, and black boxes represent the cytoplasmic portion of TNF-α that is released by proteolysis to form soluble TNF-α. The positions of the translation initiation and stop codons are shown. The promoter/enhancer from CMV and the gpt gene used to select for resistance to mycophenolic acid are shown. The triangles between the two genes mark the regions that were deleted to make the mutant TNF-α gene. (B) Comparison of DNA sequence and amino acid sequence (single-letter code) of the wild-type and mutant TNF-α genes in the region flanking the site of proteolytic cleavage (P). Pairs of diagonal lines in the sequences mark the positions of introns. The positions of amino acids 1 and 12 in soluble TNF-α are shown. The Ala residue that was substituted for Val16 is circled.

of Val16 (Fig. 1). Deletion of amino acids Val1 to Pro12 was reported to result in a transmembrane form of TNF-α that was resistant to the proteolytic cleavage that otherwise releases mature soluble TNF-α from the cell surface.22 The Val16 to Ala substitution was the result of engineering in an SphI restriction enzyme site used to join the two halves of the TNF-α gene that were cloned separately (see Materials and Methods).

TNF-a expression assays Murine SP2/O myeloma cells were transfected with pB936 DNA and cells expressing the plasmid vector gpt gene were selected using mycophenolic acid. Resistant colonies were first assayed for TNF-α expression by Western blot analysis of whole cell extracts. Detection using polyclonal goat anti-human TNF-α and alkaline phosphate-conjugated rabbit anti-goat IgG showed the expected 25 kDa band (approximately 1 kDa smaller than wild-type TNF-α) in some of the cell lines (data not shown). To test whether the TNF-α was expressed on the cell surface, live cells were incubated with murine anti-human TNF-α mAb 6401.1, known to bind transmembrane TNF-α,23 and bound mAb detected using fluorescein isothiocyanate (FITC)-conjugated antimouse IgG and flow cytometry. The results showed that cells which tested positive for TNF-α in the Western blot assay, but not control SP2/O cells, did indeed bind mAb 6401.1 (see Fig. 2A and B; Fig. 2B shows mAb 6401.1 binding to the TNF1 SP2/O subcloned cell line, T72-18, see below). SP2/O cells transfected with vector alone were not prepared and tested in these assays. An isotype control Ab did not bind the TNF1 cells (data not shown). The negative controls showed that the Abs were not binding to Fc receptors on the SP2/O cells. These results indicated that the epitope recognized by mAb 6401.1 was preserved in this recombinant TNF-α molecule. No differences in the appearance or growth characteristics between the TNF1 SP2/O cells and nontransfected SP2/O cells were apparent. Flow cytometry analysis initially revealed heterogeneity in the TNF1 parent cell population, with approximately 40% of the cells staining brightly and most of the remaining 60% staining near background levels (data not shown). Subcloning of one parent cell line and analyses by flow cytometry identified a subclone (T72) that showed 80% of the cells staining brightly, but for reasons that are not known, the fluorescent profile of this subclone was later observed to be identical to that seen with the parental cell line (data not shown). A second subcloning of the T72 subcloned cell line was performed in which homogeneity of cell populations was assessed by incubating with cA2, peroxidase-conjugated anti-human Fc, and the substrate diaminobenzidine. Microscopic visualization of the percentage of brown staining cells enabled identification

Relative cell number

Antibody-induced lysis of TNF-α-expressing cells / 253

100

A

SP2/O

B

TNF+

C

SP2/O

D

TNF+

E

TNF+

F

TNF+

101

102

103

104 100

101

102

103

104

Relative fluorescence Figure 2. Flow cytometric analysis of non-transfected SP2/O cells and TNF1 T72-18 SP2/O cells. Cells of the type indicated were incubated with either no Ab (shaded) or a primary Ab (non-shaded), washed, and incubated with FITC-conjugated anti-mouse or anti-human immunoglobulin where appropriate. Fluorescence was detected on a FACScan (Becton Dickinson) flow cytometer. Non-shaded histograms in panels A and C are hidden by the shaded histograms. Primary Abs were: (A and B) mouse anti-human TNF-α mAb 6401.1, (C) cA2 G1, (D) mouse/human chimeric IgG1 control Ab, (E) cA2 G1, (F) cA2 G4.

of a subclone (T72-18) that was nearly homogenous for anti-TNF-α Ab binding (Fig. 2B, E and F). All assays reported here were performed on this T72-18 cell line.

cA2 binding to transmembrane TNF-a Two isotypes of the chimeric anti-TNF-α mAb, cA2, were tested for their ability to bind the recombinant transmembrane TNF-α by flow cytometry analysis. The G1 (Fig. 2C) and G4 (not shown) isotypes of cA2 did not bind to non-transfected SP2/O cells but did not bind to TNF1 T72-18 cells (Fig. 2E and F), showing the same fluorescent profile seen with mAb 6401.1 (Fig. 2B). An irrelevant chimeric IgG1 mAb did not bind T72-18 cells (Fig. 2D), but the Fab and F(ab′)2 fragments of cA2 G1 did bind to T72-18 cells confirming that binding was not Fc-mediated (data not shown). A competition experiment confirmed that binding of cA2 to the cells was human TNF-α-specific binding. Increasing concentrations of soluble human TNF-α, but not soluble murine TNF-α, blocked binding of iodinated cA2 G1 to T72-18 cells (Fig. 3).

254 / Scallon et al.

CYTOKINE, Vol. 7, No. 3 (April 1995: 251–259)

Scatchard analyses showed that the T72-18 cells bound an average of between 25 000 cA2 F(ab′)2 fragments and 35 000 cA2 G1 molecules per cell (Fig. 4B). This small difference may be due to partial loss of activity of the F(ab′)2 fragment after its preparation by protease treatment. Whether 35 000 cA2 molecules binding per cell can be interpreted as 35 000 TNF molecules expressed per cell depends on whether the TNFα molecules are monomeric, dimeric or trimeric and how many cA2 molecules can simultaneously bind each TNF-α molecule. This information is currently not known.

7000 6000

cpm bound

5000 4000 3000 2000 1000 0 10–8 10–7 10–6 10–5 10–4 10–3 10–2 10–1 10–0 101 102 Conc. of TNF competitor (nM)

Antibody-dependent cellular cytotoxicity

Figure 3. Inhibition of cA2 binding to T72-18 cells by soluble TNF-α. 125

I-labelled cA2 G1 at 0.2 nM was incubated with T72-18 cells in the presence of varying concentrations of (j) soluble human TNF-α, (h) murine TNF-a, or (s) no competitor. Incubations were at room temperature for 4 h in Iscove’s media with 5% FCS. Counts bound were determined using a gamma counter.

Affinity measurements Binding assays using T72-18 cells and different forms of cA2 were performed to (a) measure the avidity between cA2 G1 and transmembrane TNF-α, and (b) analyse the effects of differences in the Fc domain on avidity to compare with results of subsequent immune effector function assays (see below). cA2 G1, cA2 G4, and the F(ab′)2 and Fab fragments of cA2 G1 were labelled with 125I-NaI and incubated with T72-18 cells at varying concentrations. Each cA2 tracer showed dose-dependent, saturable binding to the T72-18 cells (Fig. 4A). Scatchard analyses of the binding curves revealed that cA2 G1, cA2 G4, the cA2 G1 F(ab′)2 fragment, and the cA2 G1 Fab fragment had Kd values of 46 pM, 92 pM, 98 pM, and 2.44 nM, respectively (Fig. 4B).

8000 6000 4000 2000

0

Complement-dependent cytotoxicity Different forms of cA2 were allowed to bind T7218 cells and then assayed for their ability to trigger complement-dependent lysis of the target cells. Lysis was assessed by measuring the amount of lactate dehyrogenase (LDH) activity released into the cell supernatant. The G1 isotype of cA2, but not the G4 isotype or the cA2 G1 fragments, triggered efficient killing of 800 000

A Bound (Ab/cell)/Free (nM)

Specific cpm bound

10 000

The ability of cA2 bound to T72-18 cells to activate effector functions was first tested in an ADCC assay. T72-18 cells or non-transfected SP2/O target cells were labelled with 111Indium-Cl (111In-Cl), incubated with varying amounts of Ab, and mixed with human PBL effector cells. Cell lysis was quantitated by measuring counts in the cell supernatant. The G1 isotype of cA2 triggered efficient killing of the target cells in a dosedependent manner with killing being detectable at doses as low as 5 ng/ml cA2 G1 (Fig. 5). In contrast, the G4 isotype of cA2 triggered significantly lower levels of killing even when present at 4 µg/ml. Control cells showed only a very low level of killing in the presence of cA2 G1.

1

600 000

B

Kd(pM) cA2 G1: y = 776010 – 21.8x cA2 G4: y = 351960 – 10.9x F(ab')2: y = 249420 – 10.2x Fab: y = 14880 – 0.41x

46 92 98 2440

400 000

200 000

2 3 4 5 6 7 8 0 Ab concentration (nM) Figure 4. Binding analyses of different forms of cA2 to T72-18 cells.

10 000 20 000 30 000 40 000 Ab molecules bound/cell

(A) (j) cA2 G1, (h) cA2 G4, (n) cA2 F(ab′)2 fragment, or (d) cA2 Fab fragment were radiolabelled using 125I-NaI and incubated with T72-18 cells at varying concentrations for 18 h at 48C. Counts bound to control SP2/O cells were subtracted from counts bound to T72-18 cells to derive specific counts bound. (B) Scatchard analysis of specific binding.

Antibody-induced lysis of TNF-α-expressing cells / 255

T72-18 cells (Fig. 6). The F(ab′)2 fragment of cA2 did not activate complement, consistent with the role of the Fc domain of the heavy chain in activating complement.31 cA2 G1 did not trigger killing of control SP2/O cells.

Percent cytotoxicity

35

25

DISCUSSION

15

5

–5

0

1 10 100 Ab concentration (ng/ml) Figure 5. ADCC in the presence of cA2.

1000

10 000

Human PBL effector cells were mixed with 111In-labelled T72-18 target cells at a ratio of 10:1 in the presence of varying amounts of (j) cA2 G1, (h) cA2 G4, (n) cA2 G1 F(ab′)2 fragment, or (d) a control IgG1. The mixture was incubated at 378C for 18 h. Counts released into cell supernatant were measured to quantitate lysis of target cells. Results shown were derived by averaging the results from using PBL cells from two different donors. Results using non-transfected SP2/O cells as target cells are shown for cA2 G1 only (s), but were negative for all samples. Assays were performed in quadruplicate. Data are representative of two independent experiments using PBL cells from three different donors.

80

Percent cytotoxicity

70

TNF+

SP2/O

60 50 40 30 20 10 0 –10 control cA2 lgG1 F(ab')2

cA2 G1

cA2 G4

control cA2 lgG1 F(ab')2

cA2 G1

Figure 6. Complement-dependent cytotoxicity activated by cellbound cA2. T72-18 cells or non-transfected SP2/O cells were pre-incubated with 5 mg/ml of the indicated Ab at room temperature for 3 h. Rabbit complement was added to a final concentration of 10% and the mixture incubated at room temperature for 20 min. The extent of cell lysis was measured by assaying levels of LDH released into the cell supernatant in a colorimetric assay. Absorbance from cells incubated without Ab and in the presence of rabbit complement was subtracted to determine specific lysis as described in the Materials and Methods section. Assays were performed in quadruplicate. Standard deviations (bars not shown) were less than 5% of the mean for each test reagent. Pretreating rabbit complement with a 568C incubation for 30 min resulted in LDH levels equal to samples not treated with complement (data not shown). Data shown are representative of three independent experiments. An experiment using 111In-labelled T72-18 and SP2/O cells to measure cell cytotoxicity gave similar results.

Because some cytokines and antigens can be either soluble or membrane-bound, it is important to understand the effects that potentially therapeutic monoclonal antibodies can have upon binding to either form. Indeed, the ability of an Ab to neutralize a soluble cytokine that exacerbates a disease and also bind to its transmembrane form on cells that produce it may be advantageous if secretion from the producer cells is repressed or the producer cells are lysed. We report here that the chimeric monoclonal antibody, cA2, binds the cell-surface, transmembrane form of human TNF-α in addition to the soluble form of TNF-α, and mediates both antibody-dependent (ADCC) and complement-dependent (CDC) cytotoxicity on a cell line expressing abundant transmembrane TNF-α. The cA2 that has recently been successfully tested in clinical trials of rheumatoid arthritis patients19,20 has a human IgG1 constant domain. Human IgG1 antibodies, when complexed with antigen, can efficiently bind and activate Fc receptor-bearing cells and may activate the complement cascade. Such effector systems can result in clearance of antigen from circulation by the reticuloendothelial system and/or killing of cells with surface-bound antigen. In contrast, the human IgG4 isotype binds relatively weakly to the Fc receptors and has very little complement-activating activity. The results reported here are in keeping with these established isotype behaviours. The IgG1 isotype of cA2 was much more active than the IgG4 isotype in inducing ADCC, and only the IgG1 isotype showed activity in CDC assays. Cell binding measurements indicated that these differences were not due to differences in their avidity for transmembrane TNF-α. The dimeric forms of cA2 G1 (cA2 G1 and its F(ab′)2 fragment) had 25- to 50-fold greater avidity for transmembrane TNF-α than the monomeric cA2 G1 Fab fragment. This is most likely due to both arms of dimer cA2 molecules binding TNF-α on the cell surface. Such bivalent interactions would be expected to result in slower dissocation of cA2 from the cell than for monovalent interactions. Interestingly, the fact that cA2 G4 had as much avidity as cA2 G1 for transmembrane TNF-α suggests that the IgG4 hinge, believed to limit segmental flexibility,33–35 permits bivalent binding of cA2 G4 to two TNF-α molecules. This is in contrast to the IgG4 isotype of an anti-intercellular adhesion molecule Ab that showed significantly less avidity for cell-surface antigen compared to the IgG1 isotype of the same Ab.36

256 / Scallon et al.

The difference between the bivalent and monovalent forms of cA2 in their avidities for transmembrane TNF-α contrasts with their relative affinities for soluble TNF-α. When cA2 G1 or its fragments were captured on an EIA plate coated with goat anti-human IgG F(ab′)2 fragment and incubated with varying concentrations of soluble 125I-labelled TNF-α, identical affinities were observed (our unpublished data). The result was consistent with previous data suggesting that a single soluble TNF-α molecule binds only one arm of a cA2 molecule (our unpublished data) and, therefore, that cA2 valency does not affect affinity for soluble TNF-α. Comparisons of isotype effects with two different anti-TNF-α mAbs in sepsis models were recently reported.37 Their human IgG4 anti-TNF-α was more effective than an IgG1 isotype of the same Ab at preventing pyrexia in rabbits given human TNF-α. Likewise, mice challenged with LPS were protected better by a murine IgG1 (functionally analogous to a human IgG4) anti-TNF-α than a murine IgG2a (functionally analogous to a human IgG1) isotype of that same Ab. As Suitters et al.37 pointed out, these differences demonstrate that choice of isotype can have important consequences in vivo for a given Ab and disease. However, a particular isotype of a given Ab may be more efficacious for one disease state whereas a different isotype of that same Ab may be more efficacious in a different disease state. The choice of an IgG1 isotype for an anti-TNF-α Ab may be beneficial for rheumatoid arthritis, a chronic and mostly localized disease, whereas an IgG4 isotype may be more beneficial for sepsis, an acute and systemic inflammatory reaction. Despite the expression of abundant TNF-α on the SP2/O cells, numerous attempts to demonstrate that the mutant transmembrane TNF-α was functional in cytotoxicity assays (e.g. using murine L929 cells) or in adhesion molecule induction assays (e.g. coincubating with endothelial cells and measuring E-selectin expression) were not successful. In addition, membranes prepared from the TNF1 T72-18 cells did not show significantly more killing of TNF-sensitive cell lines than membranes from non-transfected SP2/O cells. Perez et al.22 reported that a mutant TNF-α molecule that lacked amino acids 1–12 was expressed on the cell surface and had cytolytic activity. Our mutant TNF-α, in addition to lacking amino acids 1–12, was lacking Ala21 and had a Val to Ala substitution at position 16. The latter change may result in a reduction in TNF-α activity.38 Interestingly, however, the TNF1 cells did bind to soluble TNF receptor/IgG fusion proteins, indicating that the receptor binding site was intact (our unpublished data). Both CD41 T cells and macrophages are present at elevated levels in arthritic joints.27,28 Because these two cell types express both soluble and transmembrane forms of TNF-α, it is conceivable that the beneficial

CYTOKINE, Vol. 7, No. 3 (April 1995: 251–259)

effect of cA2 (IgG1) in RA patients may be the result of multiple mechanisms including (a) blocking the proinflammatory activities of soluble TNF-α, and (b) binding to surface TNF-α on CD41 T cells and macrophages, and triggering their subsequent destruction by some effector mechanism. Although the in vitro experimental conditions described here using murine myeloma cells expressing mutant human TNF-α are clearly different from conditions found in inflamed joints of RA patients, the demonstration that cA2mediated lysis of TNF-α-expressing cells is possible raises the intriguing notion that such activity contributes to the therapeutic efficacy of cA2. Such a mechanism might explain the prolonged beneficial effect of cA2 noted in clinical trials.19,20

MATERIALS AND METHODS Materials Oligonucleotides were purchased from Integrated DNA Technologies, Inc. (Coralville, IA). PCR amplification kits were from Perkin-Elmer (Norwalk, CT) and DNA sequencing kits from U.S. Biochemical Corporation (Cleveland, OH). Restriction enzymes were from New England Biolabs (Beverly, MA). Klenow enzyme and T4 DNA ligase were from Boehringer Mannheim (Indianopolis, IN). FITC-conjugated goat F(ab′)2 fragments against total human Ig or human IgG F(ab′)2 fragments were obtained from Tago, Inc (Burlingame, CA) and Jackson ImmunoResearch (West Grove, PA), respectively. Recombinant human TNF-α, goat anti-huTNF-α, and mouse anti-huTNF-α mAb 6401.1 (an unconjugated form of product number FAB-210) were obtained from R&D Systems (Minneapolis, MN). Recombinant murine TNF-α was from Genzyme Corporation (Cambridge, MA). Alkaline phosphatase-conjugated rabbit anti-goat IgG was from Jackson ImmunoResearch. Polymorphprep™ was from NYCOMED PHARMA AS (Norway). Complement from 3–4 week old rabbits was from Pel-Freez Clinical Systems (Brown Deer, WI). A lactate dehydrogenase assay kit (CytoTox 96) was from Promega Corporation (Madison, WI). 125I-labelled human TNF-α, 125I-NaI, and 111In-Cl were purchased from DuPont Company, NEN (Boston, MA). Iodobeads were from Pierce (Rockford, IL) and Iodogen and tropolone were from Sigma Chemical Company (St. Louis, MO). A mouse/human IgG1 chimeric monoclonal Ab, cSF25, was used as a negative control in various experiments.

PCR amplification and expression of mutant TNF-a gene Human genomic DNA isolated from HL-60 cells was used as template in PCR reactions. The whole TNF-a coding sequence was first amplified using the 5′ oligonucleotide 5′-CACATACTGACCCACGGCTT-3′ (corresponding to nucleotides 247 to 228 relative to the translation initiation site) and the 3′ oligonucleotide 5′-TCCCCCGGGGGTAATAAAGGGATTGGG-3′ (corresponding

Antibody-induced lysis of TNF-α-expressing cells / 257

to 40 to 60 nucleotides downstream of the translation stop codon).39 Total amplified products generated in this reaction were used as template in subsequent PCR reactions to amplify either the 5′ half or 3′ half of the TNF-a gene separately. The 5′ oligonucleotide 5′-CAGCatgAGCACTGAAAGCATG-3′ (translation initiation site in lower case letters) and the 3′ oligonucleotide 5′-AGCATGCGCGACTGCCTGGGCCAGAGGGCT-3′ (corresponding to amino acids Ser26 to Ala21 and Val13 to His15; SphI site underlined) were used to generate an altered version of the 5′ half of the gene. The 5′ oligonucleotide 5′-AGcatGCGTAGGTAAGAGCTCTGA-3′ (His15 codon in lower case letters; SphI site underlined; intron 3 sequences in bold) and the 3′ oligonucleotide 5′-CGCGTCGACAAGGTTGGATGTTCGTCCT-3′ (corresponding to 3 to 22 nucleotides downstream of the translation stop codon; SalI site underlined) were used to amplify the 3′ half of the TNF gene. Both PCR fragments were blunt-end ligated into pUC19 that had been digested with SalI and blunt-ended with the Klenow fragment of DNA polymerase. The 5′ half thus reformed a SalI site at its 5′ end. DNA sequence analysis revealed an unintentional 3 base deletion in the 5′ half of the gene that resulted in deletion of the codon for Ala21. This alteration was not repaired. Sequence analysis also showed that the first 11 base pairs were missing from the 5′ end of cloned 3′ half fragments. This was repaired by annealing two complementary oligonucleotides (5′-CTGTAGGTAAGTAGG-3′ and 5′-CCTACTTACCTACAGCATG-3′) designed to form an SphI sticky end and a blunt end at the 5′ and 3′ ends, respectively, of the doublestranded fragment. This fragment was cloned between the SphI site in the pUC19 plasmid and the StuI site 137 bp into intron 3 (note resulting deletion of intron 3 sequences depicted in Fig. 1). The two halves of the TNF gene were ligated together at their SphI sites to form pB908. The resulting insert could be removed from the pUC19 vector with SalI. To express the pB908 insert in SP2/O myeloma cells, a vector previously used for expressing chimeric immunoglobulins in mouse myeloma cells30 was modified. The entire coding sequence for the human IgG1 constant regions was replaced by the promoter/enhancer of the major-immediate early gene of CMV.40 An immunoglobulin transcription termination sequence was joined to the promoter/enhancer with a unique SalI restriction site to be used to insert genes of interest. Sequences coding for resistance to ampillicin and mycophenolic acid were retained in this expression vector. The SalI fragment in pB908 was inserted into the expression vector SalI site downstream of the CMV promoter to form plasmid pB936. This plasmid was linearized using BamHI prior to transfecting SP2/O cells by electroporation as previously described.30 Transfected cells were selected by incubation in 0.5 µg/ml mycophenolic acid, 50 µg/ml xanthine, and 2.5 µg/ml hypoxanthine.

Flow cytometry 96-well plates were seeded with 2.5 3 105 TNF1 T72-18 or TNF2 SP2/O cells/well in duplicate and incubated with 5 mg/ml Ab for 18 h at 48C in media (Iscove’s, 5% fetal calf serum (FCS)). Cells were washed four times with cold media and incubated with 10 mg/ml FITC-conjugated F(ab′)2 frag-

ments of goat anti-human Ig or anti-human F(ab′)2 antibody for 2 h at 48C in media. Cells were washed twice with media and twice with PBS before resuspending in cold Isotone. Samples were analysed on a FACScan (Becton Dickinson) flow cytometer using LYSYS II software. 5000 events were monitored for each sample.

Cell binding with iodinated antibody 100 µg each of cA2 G1, cA2 G4, and the F(ab′)2 and Fab fragments of cA2 G1 were iodinated using Iodobeads or Iodogen to specific activities of 5–6 µCi/µg. These iodinated proteins were shown by WEHI cytotoxicity assays to be as effective as their non-labelled counterparts in blocking TNF cytotoxicity (data not shown). 96-well plates were seeded with 1.5 3 105 T72-18 or TNF2 SP2/O cells/well and incubated with varying concentrations (0.01–5 nM) of labelled Ab for 18 h at 48C in media (Iscove’s, 5% FCS). Cells were washed four times with media and removed with a cotton tip for determination of cpm bound in a gamma-counter. Counts bound to control SP2/O cells were considered background binding and were subtracted from counted values to derive specific counts bound. Data were analysed by the method of Scatchard.41 For a competitive binding experiment, 1 3 105 T72-18 cells were incubated with 0.2 nM of 125I-labelled cA2 G1 and varying concentrations of unlabelled human TNF-α, murine TNF-α, or no competitor in media at room temperature for 4 h. No pre-incubations were done. Cells were washed four times with media and counts bound determined using a gamma-counter.

Antibody-dependent cellular cytotoxicity Human peripheral blood lymphocytes (effector cells) were isolated using Polymorphprep™ as described by the manufacturer. T72-18 cells or control SP2/O cells (target cells) were labelled with 111In-Cl by a modification of a described method.42 Briefly, 7.5 3 107 cells were resuspended in 5 mls of labelling buffer (5 mM glucose, 25 mM HEPES in Hank’s balanced salt solution with Ca11/Mg11). Tropolone was added to a final concentration of 0.4 mM. 500 µCi of 111In-Cl was added and the mixture incubated at room temperature for 10 min. Cells were washed three times in labelling buffer and resuspended in RPMI media with 10% FCS. 96-well round bottom plates were seeded with 1 3 104 labelled target cells per well and incubated with varying concentrations of Ab (1–4000 ng/ml) for 30 min at room temperature. PBL effector cells were added to have effector cell:target cell ratios of 10:1. The cell mixture was incubated at 378C for 18 h with each sample being tested in quadruplicates. Cells were pelleted by centrifugation and cell supernatant counted in a gammacounter. Maximum counts released was determined by lysing 111 In-labelled cells in 2.5% Triton X. Spontaneous release was determined from mixtures of effector and target cells which did not contain Ab. Percent cytotoxicity was calculated using the formula: % cytotoxicity 5 [sample release 2 spontaneous release]/ [maximum release 2 spontaneous release] 3 100.

Complement-dependent cytotoxicity 96-well plates were seeded with 2.0 3 106 T72-18 or TNF2 SP2/O cells and incubated with 5 µg/ml Ab for 3 h at

258 / Scallon et al.

room temperature in media (Iscove’s, 5% FCS). Complement from young rabbits was added to a final concentration of 10% and the mixture incubated at room temperature for 20 min. Cells and cell debris were pelleted by centrifugation and 50 µl of the supernatant removed for a colorimetric assay for LDH activity using the CytoTox 96 kit as described by the manufacturer. Optical absorbance at 490 nm was measured. Each reagent was tested in quadruplicates. Background values were considered to be that obtained by incubating cells with rabbit complement but no Ab. Optical absorbance values taken from cells taken through two freeze-thaw cycles was taken as representing 100% lysis. Percent cytotoxicity was calculated using the formula: % cytotoxicity 5 [sample OD value 2 background value]/[100% lysis value 2 background value] 3 100.

Acknowledgements We thank Jill Giles-Komar and Cheryl Kinney for their help in performing flow cytometry experiments, Steve McCarthy for iodinating cA2 samples, and Dr Marc Feldmann for his helpful suggestions on the manuscript.

REFERENCES 1. Brennan FM, Chantry D, Jackson A, Maini R, Feldmann M (1989) Inhibitory effect of TNF-a antibodies on synovial cell interleukin-1 production in rheumatoid arthritis. The Lancet ii:244–247. 2. Hirano T, Matsuda T, Turner M, Miyasaka N, Buchan G, Tang B, Sato K, Shimizu M, Maini R, Feldmann M, Kishimoto T (1988) Excessive production of interleukin 6/B cell stimulatory factor-2 in rheumatoid arthritis. Eur J Immunol 18:1797–1801. 3. Alvaro-Gracia JM, Zvaifler NJ, Brown CB, Kaushansky K, Firestein GS (1991) Cytokines in chronic inflammatory arthritis. VI. Analysis of the synovial cells involved in granulocyte-macrophage colony-stimulating factor production and gene expression in rheumatoid arthritis and its regulation by Il-1 and tumor necrosis factoralpha. J Immunol 146:3365–3371. 4. Haworth C, Brennan FM, Chantry D, Turner M, Maini RN, Feldmann M (1991) Expression of granulocyte-macrophage colonystimulating factor (GM-CSF) in rheumatoid arthritis: regulation by tumor necrosis factor-a. Eur J Immunol 21:2575–2579. 5. Brennan FM, Zachariae COC, Chantry D, Larsen CG, Turner M, Maini RN, Matsushima K, Feldmann M (1990) Detection of interleukin-8 biological activity in synovial fluids from patients with rheumatoid arthritis and production of IL-8 mRNA by isolated synovial cells. Eur J Immunol 20:2141–2144. 6. Aggarwal BB, Vilcek J (1992) Tumour Necrosis Factors: Structure, Function, and Mechanism of Action. Marcel Dekker, Inc., New York. 7. Beutler B (1992) Tumor Necrosis Factors: The Molecules and Their Emerging Role in Medicine. Raven Press, New York.. 8. Fiers W (1991) Tumor necrosis factor: characterization at the molecular, cellular and in vivo level. FEBS Letters 285:199–212. 9. Beutler B, Cerami A (1989) The biology of cachectin/TNF— a primary mediator of the host response. Ann Rev Immunol 7:625–655. 10. Williams RO, Mason LJ, Feldmann M, Maini RN (1994) Synergy between anti-CD4 and anti-tumor necrosis factor in the amelioration of established collagen-induced arthritis. Proc Natl Acad Sci USA 91:2762–2766. 11. Wooley PH, Dutcher J, Widmer MB, Gillis S (1993) Influence of a recombinant human soluble tumor necrosis factor

CYTOKINE, Vol. 7, No. 3 (April 1995: 251–259) receptor Fc fusion protein on type II collagen-induced arthritis in mice. J Immunol 151:6602–6607. 12. Williams RO, Feldmann M, Maini RN (1992) Anti-tumor necrosis factor ameliorates joint disease in murine collagen-induced arthritis. Proc Natl Acad Sci 89:9784–9788. 13. Mohler KM, Torrance DS, Smith CA, Goodwin RG, Stremler KE, Fung VP, Madani H, Widmer MB (1993) Soluble tumor necrosis factor (TNF) receptors are effective therapeutic agents in lethal endotoxemia and function simultaneously as both TNF carriers and TNF antagonists. J Immunol 151:1548–1561. 14. Ashkenazi A, Marsters SA, Capon DJ, Chamow SM, Figari IS, Pennica D, Goeddel DV, Palladino MA, Smith DH (1991) Protection against endotoxic shock by a tumor necrosis factor receptor immunoadhesin. Proc Natl Acad Sci 88:10535–10539. 15. Lesslauer W, Tabuchi H, Gentz R, Brockhaus M, Schlaeger DJ, Grau G, Piguet PF, Pointaire P, Vassalli P, Loetscher H (1991) Recombinant soluble tumor necrosis factor receptor proteins protect mice from lipopolysaccharide-induced lethality. Eur J Immunol 21:2883–2886. 16. Silva AT, Bayston KF, Cohen J (1990) Prophylactic and therapeutic effects of a monoclonal antibody to tumor necrosis factor-alpha in experimental gram negative shock. J Infect Dis 162:421–427. 17. Tracey KJ, Fong Y, Hesse DG, Manogue KR, Lee AT, Kuo GC, Lowry SF, Cerami A (1987) Anti-cachectin/TNF monoclonal antibodies prevent septic shock during lethal bacteremia. Nature 330:662–664. 18. Baker D, Butler D, Scallon BJ, O’Neill JK, Turk JL, Feldmann M (1994) Control of established experimental allergic encephalomyelitis by inhibition of tumour necrosis factor (TNF) activity within the central nervous system using monoclonal antibodies and TNF receptor-immunoglobulin fusion proteins. Eur J Immunol 24: 2040–2048. 19. Elliott MJ, Maini RN, Feldmann M, Long-Fox A, Charles P, Katsikis P, Brennan FM, Walker J, Bijl H, Ghrayeb J, Woody J (1993) Treatment of rheumatoid arthritis with chimeric monoclonal antibodies to tumor necrosis factor α. Arthr Rheum 36:1681–1690. 20. Elliott MJ, Maini RN, Feldmann M, Kalden JR, Antoni C, Smollen JS, Leeb B, Breedveld FC, Macfarlane JD, Bijl H, Woody JN (1994) Randomised double-blind comparison of chimeric monoclonal antibody to tumour necrosis factor α versus placebo in rheumatoid arthritis. The Lancet 344: 1105–1110. 21. Kriegler M, Perez C, DeFay K, Albert I, Lu SD (1988) A novel form of TNF/cachectin is a cell surface cytotoxic transmembrane protein: ramifications for the complex physiology of TNF. Cell 53:45–53. 22. Perez C, Albert I, DeFay K, Zachariades N, Gooding L, Kriegler M (1990) A nonsecretable cell surface mutant of tumor necrosis factor (TNF) kills by cell-to-cell contact. Cell 63:251–258. 23. Aversa G, Punnonen J, de Vries JE (1993) The 26-kD transmembrane form of tumor necrosis factor α on activated CD41 T cell clones provides a costimulatory signal for human B cell activation. J Exp Med 177:1575–1585. 24. Macchia D, Almerigogna F, Parronchi P, Ravina A, Maggi E, Romagnani S (1993) Membrane tumor necrosis factor-α is involved in the polyclonal B-cell activation induced by HIV-infected human T cells. Nature 363:464–466. 25. Noelle RJ, Roy M, Shepherd DM, Stamenkovic I, Ledbetter JA, Aruffo A (1992) A 39-kDa protein on activated helper T cells binds CD40 and transduces the signal for cognate activation of B cells. Proc Natl Acad Sci 89:6550–6554. 26. Lane P, Traunecker A, Hubele S, Inui S, Lanzavecchia A, Gray D (1992) Activated human T cells express a ligand for the human B cell associated antigen CD40 which participates in T cell-dependent activation of B lymphocytes. Eur J Immunol 22:2573–2578. 27. Janossy G, Panayi G, Duke O, Bofill M, Poulter LW, Goldstein G (1981) Rheumatoid arthritis: a disease of T lymphocyte/macrophage immunoregulation. Lancet ii:839–841. 28. Klareskog L, Forsum U, Scheynius A, Kabelitz D, Wigzell

Antibody-induced lysis of TNF-α-expressing cells / 259 H (1982) Evidence in support of a self perpetuating HLA-DR dependent delayed type cell reaction in rheumatoid arthritis. Proc Natl Acad Sci USA 72:3632–3636. 29. Chu CQ, Field M, Feldmann M, Maini RN (1991) Localization of tumor necrosis factor a in synovial tissues and at the cartilage-pannus junction in patients with rheumatoid arthritis. Arthr Rheum 34:1125–1132. 30. Knight DM, Trinh H, Le J, Siegel S, Shealy D, McDonough M, Scallon B, Arevalo Moore M, Vilcek J, Daddona P, Ghrayeb J (1993) Construction and initial characterization of a mouse-human chimeric anti-TNF antibody. Mol Immunol 30:1443–1453. 31. Siegel SA, Shealy DJ, Nakada MT, Le J, Woulfe DS, Probert L, Kollias G, Ghrayeb J, Vilček J, Daddona PE (1995) The mouse/human chimeric monoclonal antibody cA2 neutralizes TNF in vitro and protects transgenic mice from cachexia and TNF lethality in vivo. Cytokine 7: 15–25. 32. Duncan AR, Winter G (1988) The binding site for C1q on IgG. Nature 332:738–740. 33. Oi VT, Vuong TM, Hardy R, Reidler J, Dangl J, Herzenberg LA, Stryer L (1984) Correlation between segmental flexibility and effector function of antibodies. Nature 307:136–140. 34. Dangl JL, Wensel TG, Morrison SL, Stryer L, Herzenberg LA, Oi VT (1988) Segmental flexibility and complement fixation of genetically engineered chimeric human, rabbit and mouse antibodies. EMBO J 7:1989–1994. 35. Horgan C, Brown K, Pincus SH (1993) Studies on antigen binding by intact and hinge-deleted chimeric antibodies. J Immunol 150:5400–5407.

36. Morelock MM, Rothlein R, Bright SM, Robinson MK, Graham ET, Sabo JP, Owens R, King DJ, Norris SH, Scher DS, Wright JL, Adair JR (1994) Isotype choice for chimeric antibodies affects binding properties. J Biol Chem 269:13048–13055. 37. Suitters AJ, Foulkes R, Opal SM, Palardy JE, Emtage JS, Rolfe M, Stephens S, Morgan A, Holt AR, Chaplin LC, Shaw NE, Nesbitt AM, Bodmer MW (1994) Differential effect of isotype on efficacy of anti-tumor necrosis factor a chimeric antibodies in experimental septic shock. J Exp Med 179:849–856. 38. Yamagishi J, Kawashima H, Matsuo N, Ohue M, Yamayoshi M, Fukui T, Kotani H, Furuta R, Nakano K, Yamada M (1990) Mutational analysis of structure-activity relationships in human tumour necrosis factor-alpha. Prot Eng 3: 713–719. 39. Marmenout A, Fransen L, Tavernier J, Van d HJ, Tizard R, Kawashima E, Shaw A, Johnson MJ, Semon D, Muller R, Ruysschaert M-R, Van Vliet A, Fiers W (1985) Molecular cloning and expression of human tumor necrosis factor and comparison with mouse tumour necrosis factor. Eur J Biochem 152:515–519. 40. Foecking MK, Hofstetter H (1985) Powerful and versatile enhancer-promoter unit for mammalian expression vectors. Gene 45:101–105. 41. Scatchard G (1949) The attractions of proteins for small molecules and ions. Ann NY Acac Sci 51:660–672. 42. Tonnesen MG, Anderson DC, Springer TA, Knedler A, Avdi N, Henson PM (1989) Adherence of neutrophils to cultured human microvascular endothelial cells. J Clin Invest 83:637–646.