Portal de Periódicos da CAPES

Insertion of Foreign T Cell Epitopes in Human Tumor Necrosis Factor α with Minimal Effect on Protein Structure and Biological Activity

2004; Elsevier BV; Volume: 279; Issue: 32 Linguagem: Inglês

10.1074/jbc.m403072200

1083-351X

F. S. NIELSEN, Jørgen Sauer, Johan Bäcklund, Bjørn G. Voldborg, Klaus Gregorius, Søren Mouritsen, Tomas Bratt,

Immunotherapy and Immune Responses

To create a human therapeutic vaccine able to circumvent self-tolerance against tumor necrosis factor (TNF) α, foreign T helper epitopes were inserted into human TNFα, with minimal effect on the native three-dimensional structure. TNFα variants were screened for solubility, structural stability, biological activity, and after immunization, for eliciting inhibitory antibodies. The longest and most flexible loop in TNFα, also designated loop 3, is the only region that is not involved in intra- or intermolecular interactions and therefore constitute an attractive insertion site. However, the extension of the flexible loop by epitope insertions destabilized the TNFα molecule. Therefore, two cysteines were introduced to form a stabilizing disulfide bond between loops 2 and 3. In a second design approach, three TNFα monomers were linked by two T cell epitopes and expressed as a single chain TNFα trimer. TNFα variants that were expressed as soluble proteins also had a conserved tertiary structure, as determined by circular dichroism. The biological activity of the TNFα variants was of the same magnitude as human TNFα in cellular assays. Introduction of three separate single-point mutations (D143N, A145R, or Y87S) diminished the cytotoxicity of the mutated variants 50-800-fold compared with native TNFα. Antisera from mice immunized with the different TNFα variants were able to cross-react with native human TNFα and to inhibit TNFα signaling via the TNF receptors in vitro, suggesting that the structural binding epitopes of native human TNFα and thus the native conformation were conserved in the constructed vaccine candidates. To create a human therapeutic vaccine able to circumvent self-tolerance against tumor necrosis factor (TNF) α, foreign T helper epitopes were inserted into human TNFα, with minimal effect on the native three-dimensional structure. TNFα variants were screened for solubility, structural stability, biological activity, and after immunization, for eliciting inhibitory antibodies. The longest and most flexible loop in TNFα, also designated loop 3, is the only region that is not involved in intra- or intermolecular interactions and therefore constitute an attractive insertion site. However, the extension of the flexible loop by epitope insertions destabilized the TNFα molecule. Therefore, two cysteines were introduced to form a stabilizing disulfide bond between loops 2 and 3. In a second design approach, three TNFα monomers were linked by two T cell epitopes and expressed as a single chain TNFα trimer. TNFα variants that were expressed as soluble proteins also had a conserved tertiary structure, as determined by circular dichroism. The biological activity of the TNFα variants was of the same magnitude as human TNFα in cellular assays. Introduction of three separate single-point mutations (D143N, A145R, or Y87S) diminished the cytotoxicity of the mutated variants 50-800-fold compared with native TNFα. Antisera from mice immunized with the different TNFα variants were able to cross-react with native human TNFα and to inhibit TNFα signaling via the TNF receptors in vitro, suggesting that the structural binding epitopes of native human TNFα and thus the native conformation were conserved in the constructed vaccine candidates. Tumor necrosis factor α (TNFα) 1The abbreviations used are: TNF, tumor necrosis factor; TNFRI, tumor necrosis factor RI; TNFRII, tumor necrosis factor RII; MALDITOF, matrix-assisted laser desorption ionization time-of-flight; ELISA, enzyme-linked immunosorbent assay. is a pleiotropic cytokine, which plays a central role in immune regulation and as an inflammatory and host defense mediator (1Feldman M. Taylor P. Paleolog E. Brennan F.M. Maini R.N. Transplant. Proc. 1998; 30: 4126-4127Crossref PubMed Scopus (103) Google Scholar). The multiple biological effects of TNFα are mediated both through binding of membrane-bound and soluble TNFα to two cell surface TNF receptors TNFRI and TNFRII, each displaying diverse kinetics of activation upon binding of TNFα (2Vandenabeele P. Declercq W. Beyaert R. Fiers W. Trends Cell Biol. 1995; 5: 392-399Abstract Full Text PDF PubMed Scopus (739) Google Scholar). The successful use of TNFα antagonists in relieving pain and symptoms in patients suffering from rheumatoid arthritis and Crohn's disease has confirmed the important role of TNFα in inflammatory diseases. At present, the anti-TNFα monoclonal antibodies Infliximab (3Lochs H. Adler G. Beglinger C. Duchmann R. Emmrich J. Ewe K. Gangl A. Gasche C. Hahn E. Hoffmann P. Kaskas B. Malchow H. Pohl C. Raedler A. Renner E. Scholmerich J. Schreiber S. Stange E. Tilg H. Vogelsang H. Weigert N. Zeitz M. Z. Gastroenterol. 1999; 37: 509-512PubMed Google Scholar) and Adalimumab (4Machold K.P. Smolen J.S. Expert Opin. Biol. Ther. 2003; 3: 351-360PubMed Google Scholar), as well as the soluble TNF receptor Etanercept (5Yung R.L. Curr. Opin. Investig. Drugs. 2001; 2: 216-221PubMed Google Scholar), have been approved for clinical use as TNFα antagonists. In addition, several alternative anti-TNFα treatments, such as intracellular kinase inhibitors, TNFα synthesis inhibitors, TNFα membrane-release inhibitors, as well as other soluble TNFα blockers are under development (6Girolomoni G. Pastore S. Albanesi C. Cavani A. Curr. Opin. Investig. Drugs. 2002; 3: 1590-1595PubMed Google Scholar). Another potential strategy to target overexpression of TNFα is the induction of antibody responses against TNFα in the patient through immunization or active immunotherapy. However, one obstacle in the development of such a vaccine is that immunological T cell tolerance against TNFα needs to be circumvented to induce a sufficiently strong immune response with the capacity to neutralize endogenous TNFα. By the use of the AutoVac™ technology, where a foreign immunodominant T helper cell epitope is inserted into the native self-protein, T cell tolerance to the self-protein can be circumvented (7Dalum I. Butler D.M. Jensen M.R. Hindersson P. Steinaa L. Waterston A.M. Grell S.N. Feldmann M. Elsner H.I. Mouritsen S. Nat. Biotechnol. 1999; 17: 666-669Crossref PubMed Scopus (131) Google Scholar). Furthermore, the activation of hetero-reactive T cells leads to an induction of T cell-dependent polyclonal antibody production, which has the ability to cross-react with and neutralize the endogenous target protein. Most importantly, immunization with a modified self-protein, containing a foreign and immunodominant T cell epitope, allows for the immune response to be driven by hetero-reactive, rather than auto-reactive, T cells and re-immunization is a prerequisite to maintain high levels of neutralizing antibodies. Clearly, the development of a therapeutic and safe vaccine against TNFα would be preferable to the use of monoclonal anti-TNFα antibodies or engineered TNF receptors with respect to the amount of protein needed for each treatment and the convenience of the patients. The aim of the current investigation was to develop a modified human TNFα protein that, upon immunization, would induce the activation of T cells specific for the modified protein and the generation of an antibody response that could cross-react and neutralize native human TNFα. However, to be able to induce a polyclonal B cell response with high cross-reactivity to the wild type protein, it was of pivotal importance to insert the foreign immunodominant T cell epitope into the native TNFα sequence in such a way that the structural identity and, consequentially, conformational B cell epitopes between the modified protein and wild-type human TNFα were maintained. Introduction of foreign epitopes in a trimeric protein, like TNFα, without interfering with the native conformation is not trivial. Active TNFα consists of three homologous monomers that self-associate into a symmetrical non-covalent trimer with a molecular weight of 52,000 (8Jones E.Y. Stuart D.I. Walker N.P. Nature. 1989; 338: 225-228Crossref PubMed Scopus (482) Google Scholar). The individual 157-amino acid monomer consists of anti-parallel β-sheets organized into a jellyroll-like structure (9Narhi L.O. Philo J.S. Li T. Zhang M. Samal B. Arakawa T. Biochemistry. 1996; 35: 11447-11453Crossref PubMed Scopus (35) Google Scholar) with an intra-molecular disulfide bond between residues Cys69 and Cys101. By circular dichroism, it has been shown that the trimeric structure unfolds at 60 °C where α-helical structures are introduced in the monomer (9Narhi L.O. Philo J.S. Li T. Zhang M. Samal B. Arakawa T. Biochemistry. 1996; 35: 11447-11453Crossref PubMed Scopus (35) Google Scholar). Furthermore, more than half the amino acids in TNFα have been investigated through different point mutations by various researchers (10Van Ostade X. Vandenabeele P. Tavernier J. Fiers W. Eur. J. Biochem. 1994; 220: 771-779Crossref PubMed Scopus (49) Google Scholar). Most of the point mutations resulted in decreased stability, solubility, and/or trimer formation leading to low expression yields and the formation of inclusion bodies when expressed in Escherichia coli. The subsequent refolding required for the formation of trimeric proteins is at best a cumbersome process, mainly because of the hydrophobic interactions that stabilize a correctly folded molecule. To avoid the generation of inclusion bodies and to maintain the biophysical characteristics of native human TNFα, we have investigated several different strategies. These include the identification of a flexible loop where the foreign T cell epitope could be inserted with negligible effect on the intra- or intermolecular interactions of the trimer protein, and expression of a single chain TNFα trimer with the T cell epitopes inserted as linkers between the monomers. Construction of TNFα AutoVac™ Variants from Synthetic Genes—Synthetic genes encoding either a TNFα monomer or a single chain TNFα trimer were designed by codon optimization of the TNFα gene for E. coli expression, according to the codon usage data base. Codons with a frequency in E. coli of less than 10% were excluded. The genes were obtained as sequenced clones from GeneArt GMBH, in the case of the TNFα monomer as a single clone, and in the case of the trimer, as three colony PCR products each representing one of the three embedded TNFα genes. Codon usage was modified to create three TNFα monomers with identical amino acid sequences but with distinct cDNA sequences, for the single chain TNFα trimer (TNF_T) gene. This was done to circumvent PCR recognition problems upon assembly of the monomers. The three PCR products were cloned using the TOPO system (Invitrogen), into the pCR2.1-TOPO vector. These clones were then used as templates in three separate PCR to generate fragments that were assembled by SOE PCR (11Horton R.M. Hunt H.D. Ho S.N. Pullen J.K. Pease L.R. Gene (Amst.). 1989; 77: 61-68Crossref PubMed Scopus (2639) Google Scholar). Briefly, two fragments were assembled to generate a dimeric TNFα gene, which subsequently was used as template to assemble the final single chain TNFα trimeric gene. The assembled gene was cloned into the pET28b+ vector and transformed into competent HMS174 E. coli cells. A sequence verified clone was isolated and used as template for the generation of the single chain TNFα trimer constructs. The TNFα monomeric cDNA was moved from the pCR-Script vector to the pET28b+ vector, and used as template for the generation of flexible loop constructs. Insertion of Epitopes—The artificial T cell Pan DR epitope named PADRE (sequence AKFVAAWTLKAAA) (12Alexander J. Sidney J. Southwood S. Ruppert J. Oseroff C. Maewal A. Snoke K. Serra H.M. Kubo R.T. Sette A. Immunity. 1994; 1: 751-761Abstract Full Text PDF PubMed Scopus (445) Google Scholar) was inserted into the flexible loop variants and the tetanus toxoid epitopes P2 and P30 (sequence QYIKANSKFIGITEL, and sequence FNNFTVSFWLRVPKVSASHLE) into the single chain TNFα trimer by the SOE PCR technique, to generate TNFα AutoVac™ variants. All variants were sequence verified and transformed into the E. coli strain HMS174(DE3). TNFα AutoVac™ Expression—All proteins were expressed in HMS174 cells in defined media (13Dubendorff J.W. Studier F.W. J. Mol. Biol. 1991; 219: 45-59Crossref PubMed Scopus (349) Google Scholar). Different combinations of temperature (37 and 25 °C) and isopropyl-1-thio-β-d-galactopyranoside were tested for optimal expression. Purification of TNFα Variants—Purification of the flexible loop TNFα variants was based on a combination of Bio-Gel hydroxyapatite HTP (Bio-Rad) and Q-Sepharose anion exchange chromatography (Amersham Biosciences) (14Paquet A. Levesque A. Page M. J. Chromatogr. A. 1994; 667: 125-130Crossref PubMed Scopus (7) Google Scholar) with minor modifications. The single chain TNFα trimer variants were purified to apparent homogeneity by monoclonal antibody affinity chromatography. The monoclonal antibody Infliximab (3Lochs H. Adler G. Beglinger C. Duchmann R. Emmrich J. Ewe K. Gangl A. Gasche C. Hahn E. Hoffmann P. Kaskas B. Malchow H. Pohl C. Raedler A. Renner E. Scholmerich J. Schreiber S. Stange E. Tilg H. Vogelsang H. Weigert N. Zeitz M. Z. Gastroenterol. 1999; 37: 509-512PubMed Google Scholar) was coupled to CNBr-activated Sepharose CL-4B (Amersham Biosciences). After capture, the single chain TNFα trimers were eluted with 100 mm glycine, 500 mm NaCl, pH 11. Fractions were collected, pooled, and transferred directly to dialysis against 20 mm Tris-HCl, 150 mm NaCl, pH 7.0. MALDI-TOF Mass Spectrometry—Mass spectrometric analysis was performed using Autoflex equipment (BRUKER Daltonik, Germany). Sinapinic acid was used as matrix with samples of TNFα variants previously desalted and concentrated using ZipTipHPL (MilliPore, Denmark). Trypsin digestion (Promega) of TNFα variants were performed in 20 mm Tris-HCl, 150 mm NaCl, pH 8.0, using sequence grade trypsin by incubation for 20 h at room temperature. Peptides were mixed directly (1:1) with a saturated solution of α-cyano-4-Hydroxycinnamic acid in 33% acetonitrile (v/v), 0.1% trifluoroacetic acid (v/v) followed by application onto an AnchorChip target plate. The analysis was carried out in linear mode. Circular Dichroism Analysis—Circular dichroism measurements were carried out on a JASCO J-715 spectropolarimeter instrument (Gross-Umstadt, Germany) with quartz cells of 1-mm path length. Proteins were examined at a concentration of 0.5-1.0 mg/ml in 20 mm Tris-HCl, 150 mm NaCl, pH 7.5. The thermal denaturation was followed by scans from 280 to 200 nm at temperatures ranging from 25 to 85 °C. N-terminal Sequencing—N-terminal sequencing of variants were routinely done on a model 476A Protein Sequencer (Applied Biosystems). ProSorb cartridges (PerkinElmer Life Sciences) were used for desalting samples prior to the analysis. Measurement of Direct Binding to TNFRI and TNFRII—Polystyrene microtiter plates (Maxisorp, Nunc A/S, Denmark) were coated overnight with 50 ng/ml TNFR (I or II; R&D Systems), respectively, in a coating buffer (0.1 m Na2CO3, pH 9.5). After blocking with an ELISA buffer (10 mg/ml bovine serum albumin and 0.0005% phenol red in phosphate-buffered saline, pH 7.2), the plates were washed with a washing buffer (phosphate-buffered saline containing 0.5 m NaCl and 1% Triton X-100, pH 7.5). The investigated TNFα variant was applied to the microtiter plate in serial dilutions in the ELISA buffer and incubated for 1 h at 20 °C. The plates were subsequently washed and bound TNFα variants were detected by applying biotinylated goat anti-human TNFα (BAF210, R&D Systems), diluted to 0.5 μg/ml in the ELISA buffer, and incubated for 1 h. After washing, the plates were incubated for 1 h with streptavidin-conjugated horseradish peroxidase (RPN1231V, Amersham Biosciences) diluted 1:1000 in the ELISA buffer, washed, and developed with o-phenylenediamine dihydrochloride (Sigma P-8287) (1 mg/ml) dissolved in citrate phosphate buffer, pH 5, and H2O2 (0.03%). The plates were read on an Elx 808 ELISA reader (Bio-Tek Instruments, Burlington, VT) at 490 nm, using 620 nm as reference. Quantitative TNFα ELISA—7.0 μg/ml rabbit anti-goat antibodies were coated overnight in coating buffer. Goat anti-human TNFα in the ELISA buffer was incubated for 1 h and subsequently washed. TNFα variants and human TNFα (R&D Systems) as standard were applied to the microtiter plate in serial dilutions in the ELISA buffer and incubated for 1 h at 20 °C. After washing, detection was made as described for the receptor ELISA. The concentrations of the variants were calculated by interpolation to the TNFα standard fitted to a four-parameter plot using KC4 software (Bio-Tec Instruments). Immunizations—Groups of 10 Balb/C female mice (Taconic M&B, Denmark) were immunized subcutaneously with 50 μg of protein emulsified in Complete Freund's adjuvant (Sigma). Two, 6, and 10 weeks later, mice were boosted with 25 μg of protein in incomplete Freund's adjuvant (Sigma). One week after each boost (i.e. 3, 7, and 11 weeks after the first immunization), blood samples were collected and serum was prepared. Sera were stored at -20 °C until analyzed. All mice were between 7 and 12 weeks of age, at the start of the experiments. All experimental procedures had been approved by the Danish Animal Experimentation Inspectorate. Cell Line—The KYM-1D4 (KD4) cell line derived from human rhabdomyosarcoma cells was kindly provided from A. Meager (Division of Immunology, National Institute of Biological Standards and Controls). Culturing of cells was essentially performed as described previously (15Meager A. J. Immunol. Methods. 1991; 144: 141-143Crossref PubMed Scopus (34) Google Scholar). Cell Culture Assays—For neutralization assays, KD4 cells were cultured in triplicates in microtiter plates at 2 × 104 cells/well in the presence of a constant concentration of in-house produced human wild type TNFα (100 pg/ml a concentration inducing at least half-maximum cellular death) and titrated amounts of mouse sera. In addition, pooled control serum from mice, which had received 3-4 immunizations with an irrelevant protein (KLH) in Complete Freund's adjuvant/incomplete Freund's adjuvant, was used as a negative control in the neutralization assay. For cytotoxicity assays, cells were cultured as above in the presence of titrated amounts of human TNFα or TNFα AutoVac™ variants. After 48 h of incubation, cell survival was determined by MTS, inner salt (CellTiter 96® AQueous 1 Solution Cell Proliferation Assay, Promega). In the neutralization assay, relative cell viability was calculated by dividing the mean optical density value from wells with cells cultured in the presence of TNFα and sera with the mean optical density value from wells where cells had been cultured in the absence of both protein and sera. The IC50 value, corresponding to the amount of sera resulting in 50% inhibition of TNF-mediated cell cytotoxicity, was determined by interpolating the OD50 value onto the x axis (Y50 = (Ymax - Ymin)/2 + Ymin, where Ymax is the relative cell viability of cells cultured in the absence of sera and TNFα (i.e. 100%) and Ymin is the maximum relative cell viability observed for cells cultured in the presence of TNFα and titrated amounts of sera from control immunized mice). In the cytotoxicity assay, relative cell viability was calculated by dividing the mean optical density value of cells cultured in the presence of TNFα variants with the mean optical density value of cells cultured in the absence of TNFα. An ED50 value, corresponding to the concentration of TNFα inducing half-maximum cytotoxicity, was calculated as described for the IC50 value (i.e. Y50 = (Ymax - Ymin)/2 + Ymin, where Ymax is the relative cell viability of cells cultured in the absence of TNFα (i.e. 100%) and Ymin is the minimal OD value obtained from cells cultured in the presence of TNFα variants). Design of Flexible Loop Variants—The purpose of this work was to insert a foreign T cell epitope into the native TNFα molecule to create stable and soluble TNFα AutoVac™ variants. The crystal structure of TNFα bound to its receptor has not been published. However, from the structure of the TNFα family member TRAIL bound to the DR5 receptor, it is expected that the three major receptor interacting sites of the TNFα molecules are located along the edge of TNFα (16Cha S.S. Sung B.J. Kim Y.A. Song Y.L. Kim H.J. Kim S. Lee M.S. Oh B.H. J. Biol. Chem. 2000; 275: 31171-31177Abstract Full Text Full Text PDF PubMed Scopus (113) Google Scholar). To create TNFα AutoVac™ variants that could induce a polyclonal antibody response that was directed to receptor binding epitopes and thereby able to inhibit TNFα activity, the T cell epitope insertions were not performed in the proposed receptor binding regions. Prior to this work, we constructed numerous variants where, e.g. parts of β-strands of TNFα had been substituted with foreign T cell epitopes. However, this strategy resulted in expression as inclusion bodies in E. coli. This is in contrast to native human TNFα that is readily expressed as a soluble protein in E. coli (17Van Ostade X. Tavernier J. Fiers W. Protein Eng. 1994; 7: 5-22Crossref PubMed Scopus (88) Google Scholar). Consequently, we used another approach to find permissive sites for epitope insertions within regions of TNFα that would not affect the conformation of the core TNFα molecule. This was done by identification of flexible regions from crystal structures of TNFα, together with information from superimposition of human and murine TNFα structures and sequence alignment of TNFα from different species. In this way loop 3 (Arg103-Lys112) was identified as a potential site for insertion of epitopes. The PADRE epitope was inserted at several positions in flexible loop 3 by direct insertion and/or with deletion of one or more amino acids in TNFα (Fig. 1). The resulting variants were analyzed for soluble expression in E. coli. Interestingly, only variants where the T cell epitope had been inserted C-terminal (e.g. TNF4, TNF7, and TNF8) and not N-terminal of Gly108 were expressed as soluble proteins. The stability of proteins can be increased by insertion of cysteines that form disulfide bonds (18Matsumura M. Becktel W.J. Levitt M. Matthews B.W. Proc. Natl. Acad. Sci. U. S. A. 1989; 86: 6562-6566Crossref PubMed Scopus (256) Google Scholar). We therefore designed another TNFα variant, TNF4A, containing an extra disulfide bond, to examine if this would have a beneficial effect on the variant stability. Stabilization of the TNF4A variant was done by introduction of a new disulfide bond opposite the native disulfide bond, connecting the flexible loop 3 to the non-flexible loop 2 (Lys65-Val74). Thereby, the inserted PADRE peptide became positioned between two disulfide bonds both connecting loops 3 to 2 (Fig. 2). To achieve this the crystal structure of TNFα was analyzed, and Gln67 and Ala111 were selected for substitution with cysteines. The distance between the Cα atoms of these residues is 5.2 Å, approximately the same distance as between the residues harboring the native disulfide bond (Cys69 and Cys101; 5.7 Å). As for TNF4 (without the extra disulfide bond), TNF4A was also successfully expressed as a soluble protein in E. coli (Fig. 1). A schematic view of the structure of the flexible loop variants with and without the introduced disulfide bond is seen in Fig. 3, A and B. The highest expression levels (200 mg/liter) of soluble flexible loop variants, as determined by ELISA, were observed when cultured at 25 °C, both pre- and post-induction.Fig. 3Design of TNFα AutoVac™ variants. Two different strategies were used for design of TNFα variants. In the flexible loop variants, the PADRE epitope was inserted into the TNFα sequence with or without deletions (A). A stabilizing disulfide bridge was introduced (B), and in the single chain TNFα trimer approach the epitopes were inserted between the monomers (C). Blue, TNFα wild-type sequence; red, PADRE; black lines, disulfide bond; purple, P2; and pink, P30.View Large Image Figure ViewerDownload Hi-res image Download (PPT) Design of Single Chain TNFα Trimer—As an alternative approach to produce stable TNFα variants, single chain TNFα trimer variants were designed. cDNA of three TNFα monomers were linked by glycine linkers and two tetanus toxoid-derived T cell epitopes, P2 and P30, were placed in the linker regions. From the crystal structure of native trimeric TNFα it was perceived that the C-terminal Leu157 is located at the entrance of the interface region of the three monomers, where they make important salt bridges to Lys11 in the successive monomer (8Jones E.Y. Stuart D.I. Walker N.P. Nature. 1989; 338: 225-228Crossref PubMed Scopus (482) Google Scholar). The distance between the Cα atom of residue Leu157 to the Cα of Arg6 of the next monomer (the first five residues are not determined in the crystal structure) is only 12-15 Å. To minimize the disruptive forces on Leu157, glycine linkers (GGG) of approximately 10 Å length were introduced between the epitopes and the TNFα monomers. A schematic overview of the structure is seen in Fig. 3C. Optimal growth conditions were found to be the same as for the flexible loop variants, i.e. 25 °C, both pre- and postinduction. However, protein expression levels were lower for the single chain TNFα trimers with an approximate expression of 20 mg/liters. Characterization of Purified Variants—After purification, the N-terminal sequences were determined and found to be identical to the TNFα sequence. The molecular masses, as determined by MALDI-TOF, correlated with the theoretical values within 1 or 2 daltons. Reduced SDS-PAGE analysis revealed that the purity was about 90% for all the variants. Under certain expression conditions, 5-10% of the flexible loop constructs were degraded. The single chain TNFα trimer (TNF_T) was expressed as a protein with a molecular weight corresponding to a trimer, but breakdown products corresponding to dimeric and monomeric bands were also detected in Western blots (data not shown). Non-reduced Coomassie-stained SDS-PAGE demonstrated that the flexible loop variants migrated as monomeric TNFα. In the same analysis, the variant containing the introduced cysteines (TNF4A) also migrated as a monomer, indicating that the intramolecular disulfide bonds were correctly formed. Flexible loop variants were tested for solubility at high concentrations and were readily soluble at concentrations ranging from 0.75 to 50 mg/ml in 150 mm sodium phosphate buffer, pH 7.0, without significant precipitation. Far-UV spectra equivalent to human TNFα were obtained by circular dichroism analyses. All examined variants had a characteristic minimum at 220 nm, indicating a predominantly β-stranded protein structure (data not shown). Stability of the TNFα variants was analyzed by near-UV circular dichroism during heat-induced denaturation. Denaturation temperatures were above 40 °C for the single chain TNFα trimer, ∼50 °C for the monomer variants, and ∼65 °C for the variant that was stabilized by introduction of a new disulfide bond (see Fig. 4). As expected, native human TNFα denatured at 60 °C (9Narhi L.O. Philo J.S. Li T. Zhang M. Samal B. Arakawa T. Biochemistry. 1996; 35: 11447-11453Crossref PubMed Scopus (35) Google Scholar). The correct formation of the two disulfide bonds in TNF4A was confirmed by MALDI-TOF analyses on non-reduced and reduced trypsin-digested samples. Under non-reduced conditions a peak of 4,955.18 daltons was identified. This peak consists of a tripeptide held together by the two introduced cysteines and the two naturally occurring cysteines (Fig. 5A). After reduction of the trypsin-digested TNF4A, the tripeptide complex disappeared and a new peak of 2,583.81 daltons appeared. This peak has a mass analogous to the Ala119-Lys141 peptide representing the longest cysteine containing peptide (Fig. 5B). The second cysteine containing peptide Gly66-Arg82 was detected as a weaker peak at 1779.28 daltons. The third peptide was too short to be identified in this assay. Further evidence of the stabilizing effect of the introduced disulfide bond was done by a comparison of reduced and non-reduced SDS-PAGE that included purified TNF4A with partial cleavage in the PADRE region. Two degradation products, resulting from PADRE cleavage, were seen in reduced gels at 12 and 6 kDa, below the 18-kDa intact monomer band. Because TNF4A migrates as a single monomeric band at 18 kDa in non-reduced SDS-PAGE, the introduced disulfide bond must link the two degradation fragments together (data not shown). All variants eluted analogous to the human TNFα in size exclusion chromatography analyses corresponding to a theoretical TNFα trimer weight of 52,000 for the variants (results not shown). Biological Activity of TNFα Variants and Design of Partially Inactivated Variants—TNFα has toxic activity when injected into animals (19Asher A. Mule J.J. Reichert C.M. Shiloni E. Rosenberg S.A. J. Immunol. 1987; 138: 963-974PubMed Google Scholar) and humans (20Yang S.C. Grimm E.A. Parkinson D.R. Carinhas J. Fry K.D. Mendiguren-Rodriguez A. Licciardello J. Owen-Schaub L.B. Hong W.K. Roth J.A. Cancer Res. 1991; 51: 3669-3676PubMed Google Scholar). Interestingly, both the flexible loop and single chain TNFα trimer variants exhibited cytotoxicity comparable with the wild type TNFα in a cell based assay (Fig. 6A). TNF4A was even 3-5-fold more toxic than human TNFα. The cytotoxicity of the TNFα variants is suggestive of a native-like conformation. A strong toxicity is, however, not desirable in a vaccine. Consequently, to reduce the cytotoxicity of the TNFα AutoVac™ molecules, three different point mutations known to significantly reduce the toxicity of TNFα were intro