Noncleavable Transmembrane Mouse Tumor Necrosis Factor-α (TNFα) Mediates Effects Distinct from Those of Wild-type TNFα in Vitro and in Vivo *

Tumor necrosis factor-α (TNFα) exists in two biologically active forms, a 26-kDa transmembrane form and a proteolytically cleaved and secreted form. We sequentially inactivated all three known cleavage sites of mouse TNFα by mutating the corresponding DNA sequences. A murine T cell hybridoma transfected with the nonsecretable mutant TNFα efficiently lysed L929 target cells in a cell contact-dependent manner and induced expression of vascular cell adhesion molecule-1 on mouse endothelioma cells. A genomic mouse TNFα clone encoding this mutant was subsequently introduced as a transgene into TNFα−/−lymphotoxin-α−/− mice. The 3′ AU-rich regulatory elements of the TNF locus were maintained in the transgene to assure adequate gene regulation. Transmembrane TNFα transgenic mice were fully protected from endotoxic shock, and no TNFα bioactivity was detectable in the serum after stimulation with lipopolysaccharide. Activated CD4 T cells from these animals, however, lysed L929 cells in a cell contact-dependent way. After administration of lipopolysaccharide, transmembrane TNFα transgenic mice produced significantly higher levels of interleukin-12 than wild-type mice or TNF-deficient mice. This indicates that transmembrane TNFα may greatly affect the course of a cellular immune responses in vivo and exerts quantitatively and qualitatively distinct functions from secreted TNFα in vitro and in vivo.

Tumor necrosis factor-␣ (TNF␣) exists in two biologically active forms, a 26-kDa transmembrane form and a proteolytically cleaved and secreted form. We sequentially inactivated all three known cleavage sites of mouse TNF␣ by mutating the corresponding DNA sequences. A murine T cell hybridoma transfected with the nonsecretable mutant TNF␣ efficiently lysed L929 target cells in a cell contact-dependent manner and induced expression of vascular cell adhesion molecule-1 on mouse endothelioma cells. A genomic mouse TNF␣ clone encoding this mutant was subsequently introduced as a transgene into TNF␣ ؊/؊ lymphotoxin-␣ ؊/؊ mice. The 3 AU-rich regulatory elements of the TNF locus were maintained in the transgene to assure adequate gene regulation. Transmembrane TNF␣ transgenic mice were fully protected from endotoxic shock, and no TNF␣ bioactivity was detectable in the serum after stimulation with lipopolysaccharide. Activated CD4 T cells from these animals, however, lysed L929 cells in a cell contact-dependent way. After administration of lipopolysaccharide, transmembrane TNF␣ transgenic mice produced significantly higher levels of interleukin-12 than wild-type mice or TNF-deficient mice.

This indicates that transmembrane TNF␣ may greatly affect the course of a cellular immune responses in vivo and exerts quantitatively and qualitatively distinct functions from secreted TNF␣ in vitro and in vivo.
Tumor necrosis factor-␣ (TNF␣) 1 is a pleiotropic cytokine produced by a wide variety of cell types of mostly hematopoietic, but also of nonhematopoietic, origin (for review, see Ref. 1). TNF␣ is instrumental in the immune elimination of various infectious agents such as Candida albicans (2), Listeria monocytogenes (3), or mycobacteria (4) and exerts potent proinflammatory effects, e.g. by inducing the expression of adhesion molecules such as VCAM-1, intercellular adhesion molecule 1 (ICAM-1), or E-selectin on endothelial cells and other cell types (5,6). Aberrant production of TNF␣, however, has been also implicated in the pathogenesis of various diseases, such as rheumatoid arthritis, insulin-dependent diabetes-mellitus, sialoadenitis, and inflammatory bowel disease, in particular Crohn's disease (7)(8)(9)(10)(11).
TNF␣ mediates its effects by binding to either TNF receptor 1 (TNFR1) or TNFR2. As revealed by mice deficient for either TNFR1 (12) or TNFR2 (13), these two receptors can mediate distinct effects (14). TNF␣ is synthesized as a 26-kDa precursor that is also expressed as a type II transmembrane molecule. The 26-kDa transmembrane molecule can be cleaved by membrane bound metalloprotease(s), including the TNF␣ converting enzyme (TACE) (ADAM-17) into 17-kDa secreted monomers that form biologically active homotrimers (15)(16)(17)(18)(19). By deleting the DNA sequence encoding the first 12 amino acids of the processed human 17-kDa TNF␣ monomer, Kriegler et al. (19) generated a nonsecretable 26-kDa transmembrane TNF␣ ⌬1-12 mutant (tm TNF␣). This human tm TNF␣ mutant is capable of lysing TNF␣-sensitive target cells (20). Using transfectants overexpressing this human transmembrane TNF␣, Grell et al. (21) were able to demonstrate in a series of elegant in vitro experiments that human transmembrane TNF␣ is signaling mainly through TNFR2. Indications for distinct roles of transmembrane and secreted TNF␣ in vivo have been recently obtained in transgenic mice and by the use of inhibitors of TACE inhibitors. Mice overexpressing murine tm TNF␣ are prone to developing arthritis (22) and concanavalin A-induced inflammatory liver disease (23). The inhibition of proteolytic TNF␣ cleavage by TACE inhibitors, however, protected mice completely from endotoxic shock (24). These results clearly indicate that the pleiotropic effects of TNF␣ in vivo may be further subdivided into effects mediated primarily by secreted, trimeric TNF␣, and tm TNF␣.
To further dissect the functions exerted by secreted and transmembrane TNF␣ and to assess the possible use of TACE inhibitors in modulating the potentially deleterious effects of secreted TNF␣ while maintaining the tm TNF␣-mediated functions, we generated tm TNF␣ tg mice. To this end, the three known proteolytic cleavage sites of mouse TNF␣ were first inactivated by mutating the corresponding DNA sequences. The functionality of these mutants was first tested by transfecting NIH-3T3 fibroblasts and a mouse T cell hybridoma. Identical mutations were subsequently introduced into a genomic clone of mouse TNF␣. The resulting transmembrane TNF␣ transgene is controlled by the mouse TNF␣ promotor, and the 3Ј AU-rich elements of the genomic TNF␣ locus are preserved to assure adequate gene regulation in vivo (25,26). In contrast to wt TNF mice, these tm TNF␣ tg mice are pro-* This research was supported by Grants 31-43495.95 and 31-53Ј961.98 from the Swiss National Science Foundation and by grants from the Swiss Diabetes Foundation and the Canton of Bern. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed. tected from LPS ϩ D-galactosamine-induced mortality, and no TNF␣ bioactivity is found in the serum of these challenged mice. Interestingly, in the absence of secreted TNF␣, in vivo stimulation of tm TNF␣ tg mice with LPS (100 g/mouse) leads to significantly increased IL-12 serum levels when compared with wt TNF mice, thus demonstrating quantitatively distinct effects mediated by transmembrane TNF␣ in vivo in the absence and presence of secreted TNF␣.

EXPERIMENTAL PROCEDURES
Generation of a cDNA Encoding a Noncleavable Mouse TNF␣ Mutant-Primers used to construct the various TNF␣ mutants were purchased from Microsynth (Balgad, Switzerland), and their sequences were as follows: to generate ⌬[Leu Ϫ2 to Leu ϩ1 ], ttgagaagatgatctggtctgggccataga; to generate ⌬[Leu Ϫ12 to Leu Ϫ10 ], ggccatagaactgatgccatttgggaactt; to mutate Lys ϩ11 to Glu ϩ11 , gggctacaggctcgtcactcg; for PCRmediated 5Ј and 3Ј linker addition contributing the XhoI and NotI cloning sites for subsequent cloning, gtcgactcgaggatccacatctccctccag and taccgcggccgctgcagagtgggggctggg.
To construct the mouse TNF␣ cDNA mutants, the 1108-base pair EcoRI fragment of the TNF␣ cDNA (kindly provided by Genentech Inc, South San Francisco, CA) was subcloned into the EcoRI site of M13mp18. DNA sequences encoding amino acids Leu Ϫ2 to Leu ϩ1 , Leu Ϫ12 to Leu Ϫ10 , and Leu Ϫ2 to Leu ϩ1 /Leu Ϫ12 to Leu Ϫ10 were deleted using the U-DNA mutagenesis kit from Roche Molecular Biochemicals, and the resulting EcoRI fragments were isolated and cloned into the EcoRI site of vector pBluescript KSϩ. Subsequently, 5Ј and 3Ј untranslated regions were trimmed off by PCR-mediated linker addition using Deep Vent polymerase (New England Biolabs). The resulting TNF␣ cDNA genes wt TNF␣, L1 TNF␣ (⌬[Leu Ϫ2 to Leu ϩ1 ]), L2 TNF␣ (⌬[Leu Ϫ12 to Leu Ϫ10 ]), and L3 TNF␣ (⌬[Leu Ϫ2 to Leu ϩ1 /Leu Ϫ12 to Leu Ϫ10 ]) were cloned into the XhoI/NotI site of the eukaryotic expression vector BCMGS-Neo (27) under the control of the constitutive cytomegalovirus promoter. Using the U-DNA Mutagenesis Phasmid set (Roche Molecular Biochemicals), an additional TNF␣ mutant L6 was constructed by mutating Lys ϩ11 of mutant L3 to Glu ϩ11 . Sequences of TNF␣ cDNA genes were verified by the dideoxy chain termination method using Sequenase® 2.0 (United States Biochemical Corp.) according to the manufacturer's instructions for sequencing double strand DNA.
DNA Transfection-Transfection of NIH-3T3 and By155.16 cells with various TNF␣ expression constructs was accomplished by electroporation. Typically, log phase NIH-3T3 cells were intensively washed in sucrose buffer, and 5 ϫ 10 6 cells in 0.8 ml of ice-cold sucrose buffer were electroporated with 5 g of plasmid in a 0.4-cm-gap cuvette (Bio-Rad) at 1000 V and 1 uF, applying four pulses separated by 1-min intervals on ice. Electroporated cells remained for an additional 10 min on ice before being transferred into supplemented Iscove's modified Dulbecco's medium. Three days later, the selection of transfectants was performed using G418 sulfate at 750 g/ml. Resistant colonies were expanded in supplemented Iscove's modified Dulbecco's medium in the presence of G418 sulfate.
Electroporation of By155.16 cells was performed as described above except that only three pulses were applied. One day after electroporation, 500 g/ml G418 sulfate was added to the medium to select for transfectants. TNF␣-expressing By155.16 cells were finally enriched by sorting brightly fluorescent cells on a FACS Vantage using serum from mice transgenic for soluble TNFR p55-human Fc␥3 fusion protein (9) as a primary reagent, and FITC-conjugated F(abЈ) 2 rabbit anti-human IgG as a secondary reagent (Dako, Copenhagen, Denmark) for detection.
TNF␣ Bioassay-Biologically active TNF␣ in the supernatant was detected using the TNF-sensitive L929 fibroblast bioassay as described (29). As a specificity control, a neutralizing polyclonal rabbit anti-mouse TNF␣ antibody (IP-400; Genzyme) was added to the L929-containing wells together with cell culture supernatants or TNF␣ producing cells.
Isolation of TNF␣ by Ammonium Sulfate Precipitation-Where required, purification of TNF␣ from the medium was accomplished by ammonium sulfate precipitation (30). Total proteins in culture supernatants were precipitated at 85% saturation of ammonium sulfate, and the protein pellet was extracted twice with 40% saturated ammonium sulfate and 50 mM Tris, pH 7.8. Finally, TNF␣ in the remaining protein pellet was redissolved in 50 mM Tris, pH 7.8, and used for Western blot analysis. Using this purification scheme we were able to quantitatively isolate TNF from the medium and to reduce more than 80% of bovine serum albumin added to supplement the culture medium.
Western Blot Analysis-Protein samples were concentrated by chloroform/methanol extraction (31) and electrophoretically separated on 12% polyacrylamide gels under reducing conditions. Electroblotting to 0.2 m nitrocellulose membranes (Schleicher & Schuell, Inc.) was carried out using a Trans-Blot electrophoretic blotting system (Idea Scientific Co.). After blocking the filters overnight in 3% bovine serum albumin, the immunoblots were developed with rabbit anti-murine TNF␣ antibody (Genzyme IP-400) and horseradish peroxidase-conjugated goat anti-rabbit immunoglobulin (Dako). 3,3Ј-Diaminobenzidine was used as the chromogenic substrate (Sigma).
An Asp 718 /SalI fragment of a murine genomic clone of LT␣/TNF␣ locus (34) was subcloned into pBluescript KSϩ via the flanking Asp 718 and SalI sites. This plasmid, pTNF A/S, was used as template for the generation of mutated fragments A, B, and C by PCR amplification using primer pairs muTNF1/muTNF2, muTNF3/muTNF4, and muTNF5/muTNF6, respectively. Prior to further use, PCR fragments were treated with mung bean nuclease and purified by gel excision and phenol extraction. Due to their complementary ends, fragments B and C were then annealed and linked in a conventional PCR containing primers muTNF3 and muTNF6. Fragment A was subsequently linked by PCR to the fragment BC using the primers muTNF1 and muTNF6. The resulting NarI/AvaI PCR fragment carrying the deletions and point mutation was finally used to replace the NarI/AvaI wild-type fragment in pTNF A/S, thus yielding pmutgenTNF12. Reintroduction of a 1.5kilobase pair AvaI fragment from pTNF A/S into pmutgenTNF12 at the respective site delivered pmutgenTNF13, a full-size mutant TNF␣ gene, which should lead to uncleavable TNF␣ expression. All mutations were verified by DNA sequencing using Sequenase® 2.0 (United States Biochemical Corp.) according to the manufacturer's instructions.
Generation of tm TNF␣ Transgenic Mice-A 4.842-kilobase Asp 718 / SalI fragment from pmutgenTNF13 containing the mutated genomic clone of TNF␣ was isolated and used to generate transgenic mice. The microinjection of about 1-2 pl buffer into the male pronucleus of TNF␣/ LT␣ double deficient C57Bl/6 ϫ 129 mice (TNF Ϫ/Ϫ mice) (35) was performed by standard procedures with a final DNA concentration of 2 ng/l. Detection of transgenic mice was carried out by PCR on mouse tail DNA using the primer pair muTNFtg1 (aagttcccaaatggcatcag) and muTNFtg2 (ctacgacgtgggctacag), allowing the amplification of the transgene only.
Administration of LPS ϩ D-Galactosamine-B6 ϫ 129 mice, TNF Ϫ/Ϫ B6 ϫ 129 mice, and tm TNF␣ tg TNF Ϫ/Ϫ B6 ϫ 129 mice were challenged with intraperitoneal administration of 10 g of LPS ϩ 20 mg of Dgalactosamine. Mice were observed for 72 h. Mice were sacrificed when they became moribund or at the end of the observation period, i.e. 72 h after LPS ϩ D-galactosamine administration.
In Situ Hybridization-A 1108-base pair cDNA fragment of the murine TNF␣ gene (positions 1-1108; obtained from Genentech Inc., San Francisco, CA) was subcloned into pGEM-2. After linearization of the plasmid, sense and antisense RNA probes were prepared using the appropriate RNA-polymerase as described previously (36). In situ hybridizations of paraformaldehyde-fixed cryostat sections were performed as described previously (36).
Generation of CD4 T Cell Blasts from ex Vivo Isolated Splenocytes-After osmotic lysis of erythrocytes, splenocytes were incubated with 0.5-1 g of biotinylated anti-CD8␣ and anti-B220 monoclonal antibody per 10 6 cells for 15 min on ice. Avidin magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany) were used as second step reagents for the negative selection of CD4 T cells. The CD4 T cell-enriched negative fraction (90% purity) was resuspended at 2 ϫ 10 6 cells/ml in RPMI medium ϩ 10% fetal calf serum in the presence of 3 ϫ 10 6 cells/ml of irradiated antigen-presenting cells (2000 rad) and was incubated with 2 g/ml concanavalin A for 48 h. After 2 days, the cells were washed and incubated for additional 72 h in the presence of 100 units/ml IL-2. The CD4 T cell blasts were then purified on a Ficoll gradient, and 2 ϫ 10 6 cells/ml were incubated for 4 h in the presence of 50 units/ml IL-2, 20 g/ml phorbol 12-myristate 13-acetate, and 200 g/ml ionomycin and subsequently co-cultured with L929 cells.
Determination of IL-12 and TNF␣ Production upon Stimulation in Vivo with LPS-Mice were sacrificed 3 and 6 h after intraperitoneal administration of 100 g of LPS, and serum was collected for subsequent determination of IL-12 p70 levels. IL-12 p70 was detected by enzyme-linked immunosorbent assay in 96-well microtiter plates using the anti-IL-12 p70 monoclonal antibody 9A5 as a coating antibody and biotinylated and IL-12 p40 monoclonal antibody C17.8 as a detecting antibody. The standard curve was constructed using recombinant mouse IL-12 p70 (all reagents from Pharmingen). For statistical analysis and calculation of P values, the Student t test was performed using InStat version 2.03 software (GraphPad Inc). For the detection of TNF␣ in the serum, mice were sacrificed 1.5 h after administration of 100 g of LPS, intraperitoneal and the serum was collected for subsequent assessment of bioactive TNF␣ using L929 indicator cells.

Generation of a Mouse TNF␣ cDNA Encoding Nonsecretable
TNF␣-To assess the differential proteolytic cleavage of mouse TNF␣, the three known proteolytic cleavage sites of mouse TNF␣ (17,18) were sequentially deleted in a mouse TNF␣ cDNA clone. To this end, four different mutants of mouse TNF␣ cDNA were created, with inactivation of the cleavage sites at positions ϩ1 (TNF␣ mutant L1); Ϫ10 (TNF␣ mutant L2); ϩ1 and Ϫ10 (TNF␣ mutant L3) and ϩ1, Ϫ10, and ϩ11 (TNF␣ mutant L6) (Fig. 1). Mutant cDNAs were subsequently cloned into the BCMGS-Neo expression vector and transfected into NIH-3T3 fibroblasts or the T cell hybridoma By155.16 (28). As shown in Fig. 2A, in the L3 TNF␣ mutant, the inactivation of the two proteolytic cleavage site at Ϫ10 and at ϩ1 by the deletion of three amino acids each leads to the expected reduction in the molecular weight of the TNF␣ precursor compared to nonmutated tmTNF␣. When transfected with wt TNF␣ construct, NIH3T3 cells secrete three protein species with a molecular mass of around 17 kDa, as revealed by Western blot analysis (Fig. 2B, lane 2). Sequential deletions of the known cleavage sites display altered Western blot patterns. In particular, deletion of the cleavage site at ϩ1 results in the loss of one band and the appearance of a higher mobility band (Fig. 2B,  lane 3). On the other hand, deleting the cleavage site at Ϫ10 results in the disappearance of the slowest mobility band only (Fig. 2B, lane 4). However, when cleavage sites at ϩ1 and Ϫ10 were deleted, only one band was visible on Western blots (Fig.  2B, lane 5). This TNF␣ species still exerts biological activity in a L929 bioassay. 2 Hence, we inactivated the third potential cleavage site of mouse TNF␣ at position ϩ11 by a Lys 11 3 Glu substitution (18). By155.16 T cells transfected with the resulting mutant L6 do not secrete TNF␣ into the supernatant as shown in a Western blot analysis (Fig. 2C, lane 4) and in a L929 2 M. Imboden and S. Trachsel, unpublished observations. bioassay (Fig. 3A). However, these L6 transfectants were unequivocally positive when tested for surface TNF␣ by FACS analysis using soluble TNFR-p55 hc␥3 fusion protein (9) for detection. 2 Therefore, to assess the functionality of cell surfaceexpressed tmTNF␣, TNF␣-sensitive L929 indicator cells were co-cultured with the L6-transfected By155.16 cells. Co-cultures of L929 cells with wt TNF␣-transfected By155.16, or vectoronly (BCMGS-Neo)-transfected By155.16 cells as positive and negative controls, respectively. As shown in Fig. 3B, L6 tmTNF␣ transfectants lyse L929 cells as efficiently as wt TNF␣ transfectants in a cell contact-dependent manner, whereas the supernatant of L6 overexpressing By155.16 transfectants, even upon 20-fold concentration by ammonium sulfate precipitation, is unable to lyse L929 cells in contrast to the supernatant of the wt TNF␣ transfectants that exerts potent cytotoxic activity in the L929 bioassay (Fig. 3A). The cell contact-dependent lysis of L929 cells is inhibited by neutralizing anti-TNF␣ antibodies in a dose-dependent manner (Fig. 3B).
Taken together, these results clearly indicate that the L6 TNF␣ mutant is functional and cannot be cleaved into secreted, bioactive, or inactive TNF␣ in transfected By155.16 cells.
tm TNF␣ tg Mice Are Protected from Endotoxic Shock-To further elucidate the distinct functions exerted in vivo by tm TNF␣ and secreted TNF␣ and hence to assess the potential use of TACE inhibitors to modulate TNF␣-mediated effects in vivo, we introduced the three mutations of TNF␣ mutant L6 also in the coding sequence of a genomic clone of mouse TNF␣. The resulting transgene is thus under the control of the mouse TNF␣ promotor, and the AU-rich elements at the 3Ј end are maintained. The resulting tm TNF␣ construct was microinjected into the male pronucleus of TNF␣ Ϫ/Ϫ LT␣ Ϫ/Ϫ mice (35). The resulting transgenic mice (tmTNF␣ tg mice) still show an impaired splenic architecture and lack peripheral lymph nodes, as do their nontransgenic littermates. 3 TNF␣ Ϫ/Ϫ LT␣ Ϫ/Ϫ (TNF Ϫ/Ϫ ) mice challenged with LPS ϩ D-galactosamine are fully protected from the LPS-induced mortality observed in B6 ϫ 129 mice (wt TNF mice) (35). Hence, we determined first whether tmTNF␣ tg mice are also protected from this LPS ϩ D-galactosamine-induced lethality. As shown in Fig. 5A, TNFϩ/ϩ mice died within 6 h upon administration of LPS ϩ D-galactosamine, whereas tmTNF␣ tg and TNF Ϫ/Ϫ mice alike were protected from death throughout the entire observation period of 72 h. No bioactive TNF␣ was detected in the serum of tmTNF␣ tg or wt TNF mice 1.5 h post-LPS ϩ D-galactosamine administration, whereas high concentrations of bioactive TNF␣ were detected in the serum of wt TNF mice, as shown in a TNF bioassay (Fig. 5B). The absence of TNF␣ in the serum of LPS ϩ D-galactosamine challenged tmTNF␣ tg mice, however, cannot be attributed to a defective transcription of the transgene because in the spleen of these animals, high frequencies of tmTNF␣ mRNA expressing cells were found (Fig. 5C), whereas splenic tissue sections of untreated wt TNF and tmTNF␣ tg mice showed no cells, or only a few, expressing TNF␣ mRNA at low levels. 4 To assess the functionality of the tmTNF␣ transgene, CD4-positive splenic T cells from tmTNF␣ tg, wt TNF, and TNF Ϫ/Ϫ mice were isolated and activated ex vivo with phorbol 12-myristate 13-acetate and ionomycin. As shown in Fig. 6, only activated CD4 T cells from wt TNF and tmTNF␣ tg mice lysed the L929 target cells upon co-culture. The cytotoxic activity, however, is generally slightly lower in activated CD4 T cells from tmTNF␣ tg than from wt TNF donor mice. No cytotoxic activity is detected in the supernatant of activated tmTNF␣ tg CD4 T cells, whereas L929 cells are lysed by the supernatant of activated splenic CD4 T cells from wt TNF mice. 4 tm TNF␣ tg Mice Produce More IL-12 Than wt TNF Mice upon Activation with LPS in Vivo-Mycobacteria-or LPS-induced production of IL-12 has previously been shown to be TNF-dependent (37). To determine the effect of transmembrane TNF␣ on the production of IL-12 p70 in vivo, TNF Ϫ/Ϫ , wt TNF, and tm TNF␣ tg mice were challenged with 100 g of LPS alone. 3 and 6 h later, mice were sacrificed to collect serum for subsequent determination of IL-12 p70 levels. As shown in Fig.  7, 3 h following LPS administration, IL-12 p70 serum levels were significantly increased in wt TNF mice over TNF Ϫ/Ϫ mice. In tm TNF␣ tg mice, however, IL-12 p70 serum levels significantly (p ϭ 0.014) exceeded even those seen in wt TNF mice. At 6 h after LPS administration IL-12 p70 serum levels are 3  slightly reduced, maximum IL-12 serum levels, however, were still observed in tm TNF␣ tg mice also at this time point. DISCUSSION TNF␣, like the majority of its family members, exists as a transmembrane molecule that can be cleaved and secreted as a homotrimer. Although the existence of tm TNF␣ has been documented soon after characterization of TNF␣ (38,39), and although experimental evidence for distinct functions of human transmembrane and secreted TNF␣ has been provided (19,20), the functional significance of transmembrane TNF␣ has been neglected for several years. Recent progress in the understand-ing of the signaling and the possible TNFR2-mediated functions of tm TNF␣ (21), together with the characterization and cloning of the metalloprotease(s) responsible for cleaving TNF␣ (15,16,40), rekindled the interest in the biological relevance of tm TNF␣. The identification of specific inhibitors of the TNF␣ processing proteases (24,41), furthermore, might offer the opportunity to selectively modulate the pleiotropic effects of TNF␣ through specifically blocking secreted TNF␣-mediated effects, such as endotoxic shock induction (24), while maintaining tm TNF␣-mediated functions.
Indications for an involvement of transmembrane TNF␣ in inflammatory reactions in vivo have been recently obtained in transgenic mice, overexpressing a mouse ⌬1-12 TNF␣-human ␤-globin hybrid transgene (22,42). These mice developed signs of inflammatory liver diseases upon experimental induction with concanavalin A (23) and were susceptible to the development of arthritis (22). For the generation of this transmembrane TNF␣ transgene, the 3Ј end of the TNF␣ locus containing the 3Ј AU-rich element was replaced by the regulatory 3Ј sequence of human ␤-globin. As has been recently demonstrated by the same group, this modification of the TNF␣ locus can induce itself immunopathological disorders (26). Furthermore, in these transmembrane TNF␣ tg ⌬1-12 mutant mice, a minor portion of the 26-kDa TNF␣ is still cleaved to yield a biological inactive, secreted TNF␣ molecule (22). Therefore, we decided to make a different construct for the generation of mice transgenic for a nonsecretable homologous tm TNF␣ by the deletion of all known proteolytic cleavage sites of mouse TNF␣. Sequential deletion of the proteolytic cleavage sites in the mouse TNF␣ cDNA leads to the disappearance of distinct bands in the Western blot analysis of TNF␣ secreted from transfected NIH-3T3 (Fig. 2B) as predicted from the po- sitions of the mutated sites (Fig. 1B). Western blot analysis of the T cell hybridoma By155.16, transfected with the same constructs, however, always revealed secreted TNF␣ molecules of a uniform size (Fig. 2C). This intriguing finding may indicate that either the cellular microenvironment controls the accessibility of the different cleavage sites for the membrane bound metalloproteases or, alternatively, that the 26-kDa TNF␣ precursor is processed by different proteases expressed in a cell type-specific manner. The distinct amino acid composition forming the three potential cleavage sites may indicate that the 26-kDa TNF␣ may be processed by different proteases with distinct cleavage specificities.
The successful use of TNF neutralizing antibodies in the treatment of patients with rheumatoid arthritis (7) or with fistulizing Crohn's disease (8) demonstrates that modulating the activity of TNF␣ represents a promising strategy to treat inflammatory disorders. The initial observation that due to the absence of secreted TNF␣ following the administration of TACE inhibitors, mice are protected from endotoxic shock (24), together with the increasing evidence that transmembrane and secreted TNF␣ may exert quantitatively and/or qualitatively distinct functions, prompted us to generate a tm TNF␣ tg mouse as a tool to determine distinct effects mediated by transmembrane and secreted TNF␣ in vivo. Such an animal model may turn out to be indispensable in assessing the potential of TACE inhibitors as a more selective treatment of inflammatory disorders. To ensure appropriate transcriptional and posttranscriptional regulation, the transgene was placed under the control of a mouse TNF␣ promotor, and the regulatory 3Ј AUrich element of the TNF locus were maintained. The functionality of the tm TNF␣ tg is clearly demonstrated by the cell contact-mediated cytotoxicity of CD4 T cells from tm TNF␣ tg mice, but not from TNF Ϫ/Ϫ mice against L929 target cells (Fig.  6). The observed protection of tm TNF␣ tg mice from endotoxic shock, together with the induced expression of TNF␣ mRNA in the spleen after LPS treatment in tm TNF␣ tg mice and wt TNF mice alike (Fig. 5C), indicates that the tm TNF␣ transgene is inducible, although it does not yield biologically active TNF␣ in the circulation. The histopathological analysis of the liver at the end of the observation period generally revealed more severe damage of the hepatocytes in tm TNF␣ tg mice than in TNF Ϫ/Ϫ mice. These alterations, however, were always much less severe than in LPS ϩ D-galactosamine-treated wt TNF mice. 4 Similar observations have previously been made in experimentally induced hepatitis in TACE inhibitor GM-6001treated mice (43). The observed enhanced production of IL-12 upon LPS stimulation in tm TNF␣ tg mice when compared with wt TNF mice is intriguing, and the molecular and cellular mechanisms involved have to be defined yet. It is tempting to speculate, however, that secreted TNF␣ is able to down-regulate production of IL-12, either directly through signaling via TNFRs, presumably through TNFR1, on the IL-12 producing cells (mainly phagocytic cells), or indirectly, through attenuating the activity of T cells. In support of the latter notion are observations by Cope et al. (44), who showed that T cells from TNF␣-deficient mice and from mice treated with neutralizing anti-TNF antibodies exert enhanced proliferative and functional activity (44). Hence, secreted TNF␣ may exert immunomodulatory properties and may indeed attenuate T cell activity. Because tm TNF␣ tg mice lack secreted TNF␣ (but in contrast to TNF Ϫ/Ϫ mice), TNFR activation is still possible through tm TNF␣, T cells of tm TNF␣ tg mice may represent potent inducers of IL-12 production in phagocytic cells, e.g. through the production of elevated amounts of interferon-␥ and enhanced expression of CD154. Indications for quantitatively and qualitatively distinct responses mediated by tm TNF␣ and secreted TNF␣ have previously been reported using in vitro systems: increased production of prothrombotic tissue factor is observed in co-cultures of human umbilical vein endothelial cells with tm TNF␣-transfected Chinese hamster ovary cells when compared with the addition of recombinant soluble TNF␣ (21), and tm TNF␣ is able to lyse target cells that are resistant to cytolysis mediated by secreted TNF␣ (21,39) In conclusion, the tm TNF␣ tg mice described in this report represent a useful tool to further dissect the pleiotropic functions exerted by secreted and transmembrane TNF␣. This has been clearly demonstrated by the striking early IL-12 production in tm TNF␣ tg mice upon LPS administration that exceeds the response seen wt TNF and TNF Ϫ/Ϫ mice. Due to the central role of IL-12 in determining the functional differentiation of naïve T cells (45), this tm TNF␣-mediated effect may have far reaching consequences on the outcome of a cellular immune responses in vivo. The definition of the distinct functions exerted by the two molecular forms of TNF␣ will be instrumental in assessing the potential of TACE inhibitors as a therapeutic agent. Furthermore, the tm TNF␣ tg mice should allow us to address more specifically the hypothesis that secreted TNF␣ exerts some of the anti-inflammatory properties described for TNF␣ in vivo and in vitro (2,46,47).