Tumor Necrosis Factor α Induces the Adenovirus Early 3 Promoter by Activation of NF-κB

The early transcription unit 3 (E3) of human adenoviruses encodes proteins which appear to subvert host defense mechanisms. For example, the E3/19K protein inhibits the transport of major histocompatibility complex (MHC) class I molecules to the cell surface and thereby prevents cell lysis by cytotoxic T cells. Tumor necrosis factor α (TNF) stimulates expression of MHC molecules on the cell surface of normal cells but not of E3+ cells, rather, a further reduction of MHC expression is evident. This was attributed to the increased expression of E3/19K upon TNF treatment, an effect also observed for other E3 proteins. We investigated the mechanism of the TNF-mediated up-regulation of E3 products. We show that TNF stimulates expression of a luciferase reporter gene driven by the E3 promoter. Mutation of individual transcription factor binding sites within the E3 promoter reveals the importance of the NF-κB binding site κ2 for TNF inducibility. Electrophoretic mobility shift assays using antibodies directed against various members of the NF-κB family demonstrate that stimulation by TNF is mediated by the p50-p65 NF-κB complex. TNF inducibility does not depend on coexpression of E1A and can be observed during infection. Interestingly, the E3 promoter seems to be the only early promoter responsive to TNF and the only adenovirus promoter containing an NF-κB site. The implications of this regulatory mechanism for the adenovirus life cycle and its pathogenesis are discussed.

The early transcription unit 3 (E3) of human adenoviruses encodes proteins which appear to subvert host defense mechanisms. For example, the E3/19K protein inhibits the transport of major histocompatibility complex (MHC) class I molecules to the cell surface and thereby prevents cell lysis by cytotoxic T cells. Tumor necrosis factor ␣ (TNF) stimulates expression of MHC molecules on the cell surface of normal cells but not of E3 ؉ cells, rather, a further reduction of MHC expression is evident. This was attributed to the increased expression of E3/19K upon TNF treatment, an effect also observed for other E3 proteins. We investigated the mechanism of the TNF-mediated up-regulation of E3 products. We show that TNF stimulates expression of a luciferase reporter gene driven by the E3 promoter. Mutation of individual transcription factor binding sites within the E3 promoter reveals the importance of the NF-B binding site 2 for TNF inducibility. Electrophoretic mobility shift assays using antibodies directed against various members of the NF-B family demonstrate that stimulation by TNF is mediated by the p50-p65 NF-B complex. TNF inducibility does not depend on coexpression of E1A and can be observed during infection. Interestingly, the E3 promoter seems to be the only early promoter responsive to TNF and the only adenovirus promoter containing an NF-B site. The implications of this regulatory mechanism for the adenovirus life cycle and its pathogenesis are discussed.
Human adenoviruses (Ads) 1 cause acute infections of the respiratory and gastrointestinal tract as well as infections of the eye (1,2). A proportion of patients develop persistent infections which can last for months or years (2,3). A key role for regulating the interaction of the virus with its host and perhaps for viral persistence has been attributed to proteins encoded in the non-essential early transcription unit 3 (E3) of the virus (4 -6). Several E3 proteins, 14.7, 14.5, and 10.4, can protect cells from TNF-mediated lysis (5), while the most abun-dant protein E3/19K down-regulates expression of host class I major histocompatibility complex (MHC) antigens on the cell surface (7,8). This results in a profound reduction of cytotoxic T cell mediated lysis of virus-infected cells (6, 9 -12). Both mechanisms may prolong survival of infected cells in vivo, allowing efficient virus reproduction in the face of an ongoing immune response. In support of this hypothesis, an Ad5 mutant with an E3/19K deletion shows a much stronger immunopathology in vivo than wild type viruses (13). Furthermore, E3 transgene products expressed in transplanted pancreas tissue seem to prevent tissue rejection, leading to prolonged survival of the allogeneic transplant (14).
Efficient transcription of the E3 region as well as of the other early regions, E1B, E2, and E4, is dependent on expression of the immediate early gene product E1A which has a variety of functions (15,16). 130 base pairs immediately upstream of the E3 transcription start site are sufficient for both basal promoter activity and E1A induction of the E3 promoter in fibroblasts (17,18). DNase I footprinting analysis and electrophoretic mobility shift assays (EMSA) revealed four protein binding sites within the E3 promoter, namely those for NF-1, ATF, and AP-1, and the TATA box (19,20); the latter three sites are located within the 130-base pair segment. AP-1 and ATF are both responsible for the E1A response and can individually confer E1A inducibility, AP-1 being more efficient in this process (21,22). More recently, E3 promoter activity analyzed in lymphoid cell lines lead to the identification of two additional binding sites located between the AP-1 and NF-1 sites (23). These two sites show homology to the NF-B binding consensus (24) and are necessary for full transcriptional activity in lymphoid cells.
NF-B is composed of two subunits and is present in the cytoplasm in an inactive form bound to an inhibitory subunit termed IB. Various stimuli lead to the phosphorylation and degradation of IB. This allows the nuclear translocation of an active heterodimer which can bind DNA and activate transcription (for reviews, see Refs. 25 and 26). The heterodimer is composed of a p50-like and a p65-like subunit. Two p50 homologues, p50 and p52, are distinguished whereas the second subunit can either be p65 (also called RelA), RelB, or c-Rel.
We recently investigated the effect of TNF on cell surface expression of MHC molecules in E3 ϩ fibroblasts. Unlike the stimulation of MHC expression in E3 Ϫ cells, TNF treatment further reduced MHC expression on the cell surface of E3 ϩ cells. Concomitantly, an increased expression of E3/19K (27) as well as other E3 proteins was noted (28). This suggested that TNF stimulates E3/19K expression and thereby reduces MHC antigen levels on the cell surface. TNF is an inflammatory cytokine, produced primarily by macrophages but also by T cells (for reviews, see Refs. 29 -31). It exhibits several antiviral properties. High doses of TNF can inhibit replication of RNA and DNA viruses and many virus-infected cells become sensi-tive to TNF-mediated lysis (32, 33). In addition, TNF activates neutrophils and stimulates the killing activity of macrophages and NK cells, both of which are involved in the early antiviral response of the innate immune system (29). Moreover, TNF also supports the adaptive immune system by up-regulating class I major histocompatibility (MHC) antigens, which in turn increases the efficiency of killing by cytotoxic T cells (34).
Interestingly, Ad appears to induce TNF, IL-1, and IL-6 in the infected tissue in mice. Thus, the observed stimulation of E3 expression by TNF may be a feedback mechanism allowing efficient expression of immunosubversive E3 proteins which might be beneficial for survival of the virus in its host. Consistent with this interpretation, there is good correlation between high levels of TNF in the blood and severity of Adinduced disease in humans (35).
We have now investigated the mechanism by which TNF stimulates E3 expression. We found that TNF activates the E3 promoter by utilizing the cytosolic transcription factor NF-B. Two NF-B family members, p50/relA, strongly bind to the 2 site within the E3 promoter and therefore seem to mediate this effect. Interestingly, E3 appears to be the only early promoter sensitive to the presence of TNF. The implications of this response mechanism for the natural infection are discussed.

MATERIALS AND METHODS
Plasmid Constructs-The EcoRV C fragment, subcloned in pBluescript II KS Ϫ (Stratagene, Heidelberg, Germany) encompasses the entire E3 region of Ad2 including 273 base pairs upstream of the E3 transcription start site (27). The E3 promoter, contained within a 360base pair long SacI fragment, was cut out of the subcloned EcoRV C fragment utilizing the SacI site of the vector and the E3 SacI site at nucleotide ϩ29 and inserted into the SacI site in front of the luciferase reporter gene of the "pGL2 basic" vector (Promega, Heidelberg, Germany). The orientation of the inserted fragment was determined by sequencing using the Sequenase kit (U. S. Biochemicals, Bad Homburg, Germany). Polymerase chain reaction-mediated site-directed mutagenesis (36) of the E3 promoter using vector and mutant primers listed in Table I was carried out as described previously (37). Mutagenized double-stranded fragments were cleaved with SacI and inserted into pGL2 basic. Orientation and correct sequence of the resulting mutant E3 promoter constructs were confirmed by sequencing.
Cell Lines, Transient Transfections, and Enzymatic Assays-293 and HeLa cells were routinely grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 2 mM glutamine and antibiotics. Transient transfections were performed by the calcium-phosphate precipitation method according to a standard protocol (38). Briefly, a 6-cm dish of cells was transfected with 5 g of DNA of each luciferase construct together with 2 g of the Rous sarcoma virus promoter-␤-galactosidase internal control plasmid. Routinely, 200 l of calcium-phosphate-DNA precipitate was added to a 6-cm dish of cells at 25% confluency. The cells were washed two times with complete Dulbecco's modified Eagle's medium 4 h later (for 293 cells) or 24 h later (for HeLa cells) and further incubated in the same medium. For enzymatic assays, cells were harvested 48 h after the start of the transfection. TNF (PBH, Hannover, Germany) treatment (320 units/ml) was performed for 5 h at the end of the incubation period. Cells were scraped with a rubber policeman in phosphate-buffered saline, washed, and resuspended in sodium phosphate buffer (100 mM, pH 7.6), and lysed by three freeze-thaw cycles consisting of sequential incubations in liquid nitrogen and in a 37°C waterbath. Cell debris were pelleted by centrifugation in a microcentrifuge and the supernatants were analyzed for enzymatic activities. Luciferase, CAT, and ␤-galactosidase assays were performed as described previously (38 -40).
TNF Treatment, Nuclear Extracts, Electrophoretic Mobility Shift Assays-TNF treatment of the cells was carried out as described above. For the extraction of the nuclear proteins, cells from a 6-cm dish were lysed for 5 min on ice in 500 l of Nonidet P-40 lysis buffer (0.5% Nonidet P-40, 0.5 M sucrose, 15 mM Tris, pH 7.5, 60 mM KCl, 0.25 mM EDTA, 0.125 mM EGTA, 0.5 mM spermine, 0.125 mM spermidine, 1 mM dithiothreitol, and a mixture of protease inhibitors: 0.5 mM phenylmethylsulfonyl fluoride, 2 mM benzamidine, 5 g/ml aprotinin, 5 g/ml pepstatin, 5 g/ml leupeptin), the nuclei were pelleted by centrifugation at 600 ϫ g for 5 min at 4°C. After resuspending the nuclei in 105 l of NEB 0.14 (140 mM NaCl, 20 mM Hepes, pH 7.9, 25% glycerol, 1.5 mM MgCl 2 containing the same concentration of protease inhibitors as above), nuclear proteins were extracted by adding 75 l of NEB 0.7 (700 mM NaCl, 20 mM Hepes, pH 7.9, 25% glycerol, 1.5 mM MgCl 2 with protease inhibitors) and incubation for 20 min on ice. Soluble extracts were separated from non-soluble material by a 5-min centrifugation at 11,000 ϫ g (4°C). Protein concentrations were determined with the Bradford dye-binding assay using a Bio-Rad kit (Bio-Rad, Munich, Germany). The nuclear extracts were frozen in liquid nitrogen and stored at Ϫ80°C.
End-labeled double-stranded oligonucleotides (sequences are listed in Table I) were used as probes for EMSA. All complementary oligonucleotides were designed to have a 5Ј overhang for labeling by Klenow fragment in the presence of [␣-32 P]dCTP and cold dATP, dGTP, and dTTP. EMSA binding reactions were performed at room temperature for 15 min. The reaction mixture contained 5 g of nuclear extracts, 0.5 ng of the radiolabeled probe (20,000 cpm), and 2 g of poly(dI-dC) in 16 l of binding buffer (25 mM Hepes, pH 7.6, 60 mM KCl, 5% glycerol, 0.1 mM EDTA, 0.8 mM dithiothreitol, 2.5 mM MgCl 2 ). For EMSA using the NF-B probe, unspecific complexes were eliminated by further incubating the binding reactions with 3 mM chloroquine (24). For supershift analysis the binding mixture was preincubated with 1 l of specific antibody for 1 h on ice before adding the radiolabeled probe. The binding reactions were run on a native 5% polyacrylamide 0.5 ϫ TBE gel for 2 h at 150 V.
Virus Infections and Immunoprecipitations-Virus infections of HeLa cells and subsequent immunoprecipitations were performed essentially as described (27).

TNF Up-regulates Transcription Driven by the E3 Promoter-
We previously reported that TNF up-regulates expression of the E3/19K protein (27) as well as the other E3 proteins (28). The coordinated induction of all E3 proteins indicated a mechanism acting prior to translation. Indeed, the level of E3/19K mRNA is also increased (27). To examine if this increased synthesis is due to stimulation of the E3 promoter, we subcloned the Ad2 E3 promoter (Ϫ273, ϩ29) in front of the luciferase reporter gene (the resulting plasmid is termed pE3Luc). As efficient expression from the E3 promoter requires E1A, human 293 cells which constitutively express E1A (41) were used for the transfection assays. Transient transfection of pE3Luc into 293 cells gave rise to higher luciferase activity than the positive control plasmid expressing luciferase under the control of the SV40 promoter/enhancer (data not shown), showing that the E3 promoter is highly active in these cells. However, when the transiently transfected cells were treated with TNF, the luciferase activity measured was further increased by a factor of 3.3 (see top of Fig. 1). This level of induction is very similar to the increase of E3/19K protein synthesis that we reported previously (27). Therefore, it seems that the TNF-mediated up-regulation of E3 expression may be entirely due to activation of transcription driven by the E3 promoter.
The Nuclear Factor B Elements of the E3 Promoter Play a Key Role in the TNF-mediated Induction of the E3 Promoter-Previous analysis of the E3 promoter led to the identification of transcription factor binding sites for AP-1, ATF, and NF-1. Two additional sites showing homology to the NF-B consensus binding site were specifically occupied in lymphoid cells (18 -23, 42) (see schematic representation on Fig. 1, left). To examine which cis-acting element of the E3 promoter is required for its TNF inducibility, each transcription factor binding site was separately mutated by substitution of 3-6 nucleotides (for details, see Table I). The inability of the mutated sites to bind the corresponding nuclear factor was confirmed by EMSA (see below, Fig. 2A). Mutated E3 promoters were inserted in front of the luciferase gene and the resulting constructs were transfected into 293 cells. The luciferase activity was determined in the presence and absence of TNF (Fig. 1). Luciferase activity obtained with the wild type promoter construct without TNF treatment was set to 1. For each construct, the ratio between luciferase activities with and without TNF treatment is indicated on the right. As expected from previous studies (18 -22, 42), both AP-1 and ATF mutations have a drastic effect on the basal level and/or E1A driven expression of the reporter gene, while TNF stimulation appears only weakly affected. Mutation of the NF-1 binding site alters neither the TNF induction potential of the promoter, nor its basal activity. A more pronounced reduction of the TNF-mediated induction is exhibited by mutating the potential binding sites for NF-B, 1, and 2.
In particular, the mutation of 2 reduces inducibility by 90%. A double mutation of 2 and 1 abolishes TNF induction completely. Thus, it seems that the NF-B binding sites of the E3 promoter are primarily responsible for the induction of the E3 promoter by TNF. Apparently the 2 element plays the most critical role in this phenomenon. It is intriguing that the 2 site is strongly conserved among the known Ad E3 promoter sequences while 1 is not (Table II). Moreover, all 2 elements align well with the NF-B consensus binding site (24). In con-trast, the 1 sites generally do not fit the NF-B consensus except for 1 of Ad2 and Ad5 which contains only one mismatch to the NF-B consensus (Table II).
NF-B Proteins Bind to the Elements of the E3 Promoter after TNF Treatment of Fibroblasts-To analyze which proteins bind to the E3 promoter in the absence and presence of TNF, EMSA were performed. Nuclear extracts of non-induced and TNF-induced 293 cells were prepared and tested with the probes indicated below the autoradiographs in Fig. 2A. The specificity of the binding reactions was confirmed by competition with a 200 M excess of unlabeled wild type and mutant oligonucleotides. The mutant double-stranded oligonucleotides were identical to the primers used for generation of the E3 promoter constructs tested above (Fig. 1). While the wild type oligonucleotides efficiently competed with binding of the probes ( Fig. 2A, lanes 4, 9, 14, and 19), the mutant oligonucleotides were unable to do so ( Fig. 2A, lanes 5, 10, 15, and 20). This demonstrated that the introduced mutations disrupt the binding to the corresponding nuclear factors.
As can be seen, NF-1, AP-1, and ATF bind to the E3 promoter in the absence of TNF ( Fig. 2A, lanes 2, 12, and 17). In agreement with the results of the mutational analysis shown in Fig.  1, induction with TNF has no effect on proteins binding to the NF-1 and ATF probes ( Fig. 2A, lanes 3 and 18), whereas a slight increase is observed with the AP-1 probe (1.6-fold; Fig.  2A, lane 13). In contrast, a protein complex with 2 is only formed when extracts from TNF-treated 293 cells are used ( Fig.  2A, compare lanes 7 and 8). A complex of similar size but in much lower amounts can be detected using the 1 probe after TNF induction (data not shown). We therefore compared the affinities of 1 and 2 for the TNF-induced complex in competition assays using radiolabeled 2 (Fig. 2B). A 10 times higher concentration of cold 1 is necessary to give an equivalent competition to 2 (Fig. 2B, compare lanes 6 and 7). This showed that the affinity of the bound protein(s) for 2 is much greater than for 1.
The question arises which members of the NF-B family of transcription factors bind to the NF-B elements within the E3 promoter. Therefore, supershift experiments were performed using antibodies directed against individual NF-B subunits (Fig. 3). Antibodies against p50 and p65 decrease the migration of the complex whereas antibodies against RelB, c-Rel, and p52 (data not shown) seem to have no effect. Thus, we conclude that the TNF-activated transcription factor binding to the 2 and 1 elements of the E3 promoter in 293 cells is composed of the  Table I). Luciferase activity was normalized with the ␤-galactosidase activity measured from the same extracts. The values represented by the bars correspond to the average of seven independent experiments and the error bars are standard deviations. Luciferase activity of the wild type E3 promoter in the absence of TNF was set arbitrarily to 1. The numbers on the right represent the ratio between luciferase activities obtained with and without TNF treatment for each construct. NF-B proteins p50 and p65 (RelA). Co-transfection of the NF-B Subunits p65 or IB Alters TNF Induction of the E3 Promoter-To corroborate our conclusion that TNF induction of the E3 promoter is mediated by NF-B, we sought to overexpress the activating subunit of NF-B, p65. Overexpression of p65 would be expected to activate gene expression driven by the E3 promoter, if this were NF-B dependent. Therefore, pE3Luc was transfected with a vector expressing p65 under control of the cytomegalovirus promoter (pCMVp65, a kind gift from P. Baeuerle). As expected, cotransfection of pCMVp65 strongly activates expression of the pE3Luc construct. Furthermore, TNF treatment does not lead to a further increase of the measured luciferase activity (Fig. 4). The opposite effect is observed when the inhibitory subunit IB is overexpressed by co-transfection of pCMVIB (a kind gift from P. Baeuerle). Overexpressed IB abolishes TNF inducibility of the E3 promoter (Fig. 4). These two findings strongly confirm the involvement of NF-B in the induction of the E3 promoter by TNF.
TNF Does Not Activate the E2 and E4 Promoter-Considering the function of E3 proteins, which seem to counteract functions of the host immune system, it is interesting to know whether TNF inducibility is a specific feature of the E3 promoter. To answer this question, we first tried to find other NF-B binding sites in the Ad2 or Ad5 genome by searching with the NF-B consensus sequence. Apart from those in the E3 promoter, no other Ad promoter contained an NF-B site. Second, we tested the effect of TNF on two other early promoters, namely E2 and E4. E2CAT, E3CAT, and E4CAT plasmids (43; kindly provided by M. B. Mathews) were transfected into 293 cells and the CAT activity was measured in TNF-treated and untreated cells. As shown in Fig. 5, only the E3 promoter is up-regulated upon TNF treatment. These results, together with our previous data indicating that E1A is also not upregulated by TNF (27), provide strong evidence that within the Ad genome TNF-inducibility is a special feature of the E3 promoter.
TNF Up-regulation of E3 in the Absence of E1A-E1A proteins play a key role for the efficient expression of the Ad early genes (for reviews, see Refs. 15, 16, and 44). All the above described experiments were performed with 293 cells, a human embryonic kidney cell line transformed with the E1 region of Ad5 (41). These cells constitutively express E1A proteins which activate the E3 promoter. To test whether TNF induction is dependent on the presence of E1A, we performed transient transfection assays with pE3Luc in HeLa cells with and without co-transfection of a plasmid carrying the E1A region (Fig.  6). As expected, the activity of the E3 promoter in the absence of E1A is much lower than in 293 cells (30-fold lower, data not shown). Coexpression of E1A increases its activity by a factor of 7. In the presence of TNF, the activity of the E3 promoter is up-regulated by a factor of ϳ3, irrespective of the presence of E1A. Furthermore, the double mutant construct 21, like in 293 cells, cannot be activated by TNF in HeLa cells, while the AP-1 mutant, although having a lower basal expression, can be activated ϳ3-fold in the presence of TNF in these cells (Fig. 6). These results resemble those obtained with 293 cells except for the lower basal level of E3 expression which can be increased by E1A co-transfection. Thus, the 3-fold induction of the E3 promoter by TNF which is mediated by NF-B is observed in different human cell lines and this induction is independent of E1A.
All the above experiments were performed using cell transfections and in vitro assays. We next examined whether TNF inducibility of E3 expression can also be observed during virus infection. HeLa cells were infected with Ad2, Ad5, and dl312, a derivative of Ad5 in which the E1A gene is deleted (45). This latter virus allows us to examine whether TNF-mediated expression of E3 is possible during infection in the absence of E1A. HeLa cells were infected with a multiplicity of infection of ϳ3 for 14 h and incubated with or without TNF. Then, the most abundant E3 protein, E3/19K, was immunoprecipitated and analyzed by SDS-polyacrylamide gel electrophoresis (Fig. 7). In accordance with the observed 3-fold higher E3 promoter activity, TNF treatment increases the expression of E3/19K in Ad2 and Ad5 infected cells by a factor of 3 (average value from five independent experiments, see also Ref. 27). The induction is much higher (6 -20-fold in three different experiments) in cells infected with the E1A deletion mutant dl312. The level of E3/19K expression can reach that of Ad2 and Ad5 (ϩE1A; Fig.  7, compare lane 6 with lanes 1 and 3). Therefore, during infection, TNF can substitute for the E1A-mediated induction of the E3 promoter. DISCUSSION We previously reported that TNF amplifies the effects of the adenovirus E3/19K protein by increasing its expression at the protein and mRNA level (27). The subsequent observation that all E3 proteins tested seem to be up-regulated, suggested that TNF might activate the E3 promoter (28). Using E3 luciferase reporter constructs, we demonstrate that this is indeed the case (Fig. 1). By mutating individual transcription factor binding sites within the E3 promoter, it was shown that neither the NF-1 nor the ATF site contributed to TNF-mediated induction of the E3 promoter. Most critical for this effect was one of the two potential NF-B binding sites, termed 2. Mutating this site resulted in a 90% reduction of TNF inducibility, whereas mutations in 1 reduced the TNF induction only by 36% (Fig.  1). The reason for this differential influence appears to be a 10-fold lower affinity of the 1 site for the NF-B complex (Fig.  2B). Nevertheless, 1 contributes to the TNF-mediated induction of the E3 promoter since mutation of both 1 and 2 completely abrogated the ability of the promoter to respond to TNF. The same construct also showed a drastic reduction of constitutive promoter activity indicating that this portion of the promoter influences basal and/or E1A mediated activity. At present, we do not know whether additional factors bind to the NF-B site in the uninduced state or the low amount of NF-B detected after long exposure contributes to basal transcription. Confirming the importance of NF-B, overexpression of an inhibitor of NF-B, IB, essentially eliminated the stimulatory capacity of the E3 promoter for TNF (Fig. 4).
An even greater reduction of basal promoter activity was observed with the promoter mutated at the AP-1 site. This can be accounted for by the loss of E1A-mediated stimulation of E3 promoter activity (18,21,46). However, inducibility by TNF was also reduced by 30%. Consistent with these data, induction with TNF increases the signal seen with an AP-1 oligonucleotide ( Fig. 2A). This suggests that TNF treatment also influences the occupancy of the AP-1 site, presumably by induction of c-Fos and c-Jun, the constituents of AP-1 (47).
Having identified NF-B as a predominant regulator of TNFmediated induction of the E3 promoter, we next examined which members of the NF-B family of transcription factors p50, p52, RelA, RelB, or c-Rel are mediating the response. Several criteria suggest that p50/RelA activate the E3 promoter in response to TNF. First, overexpression of p65 (RelA) increases promoter activity 7-fold in the absence of TNF and TNF treatment does not further increase this factor (Fig. 4). Second, EMSA experiments in the presence of antibodies directed to p50 and p65 induced a supershift which was not observed with RelB or c-Rel specific antibodies (Fig. 3) nor with antibodies directed to p52 (data not shown).
We further investigated whether or not induction of the E3 promoter by TNF is independent of the E1A protein. Two types of experiments were performed: HeLa cells lacking E1A were transiently transfected with the E3 reporter construct in the presence and absence of TNF. Fig. 6 shows that the induction by TNF was basically identical to that observed in 293 cells which constitutively express E1A. Co-transfection of an E1A expression vector did not alter the inducibility of the E3 promoter by TNF, although basal expression is increased 7-fold. Again, no induction was seen when both NF-B sites were mutated, while mutating AP-1 did not affect TNF induction in HeLa cells. Moreover, TNF induction of E3/19K protein synthesis was observed upon infection with an Ad virus containing a deletion of the E1A genes (dl312). While basal expression of E3/19K was poor in the uninduced state, TNF treatment raised the level of E3/19K to that of Ad2, which expresses E1A. Thus, TNF can functionally replace E1A for efficient expression of E3 proteins. No evidence is available that TNF substitutes for E1A   (24)) are written in lower case. The sequences for Ad12, Ad35, Ad2, Ad5, Ad19, Ad40, and Ad41 have been published (52-58). in supporting the replication of dl312.
We believe that the regulation of the E3 promoter by TNF/ NF-B is biologically important for the Ad replication cycle in vivo and for Ad pathogenesis. Considering the known activities of E3 proteins, inhibition of antigen presentation by E3/19K and interference with TNF lysis by 14.5K, 10.4K, and 14.7K, it is intriguing that the sensitivity to TNF/NF-B appears to be restricted to the E3 promoter. Searching the Ad2 and Ad5 genome sequences did not reveal NF-B consensus sequences in other promoters than the E3 promoter (data not shown). Furthermore, examination of CAT reporter constructs driven by the early promoters E2, E3, and E4 indicated that only the E3 promoter is activated by TNF (Fig. 5). Strikingly, the high affinity binding site for NF-B, 2, is particularly well conserved among Ad subtypes and in all cases conforms to the consensus sequence for NF-B binding sites (Table II). This strongly suggests an important role for this element in the virus life cycle.
On the basis of occupancy of the E3 NF-B sites in lymphoid cells and the activity of CAT reporter constructs, Williams and co-workers (23) previously suggested that NF-B might drive E3 expression in lymphoid cells and thereby might support persistence of Ad in lymphoid tissue. Our data extend these studies by showing that in other cell types, such as fibroblasts and probably epithelial cells, TNF can activate NF-B, thereby promoting E3 expression and thus enhancing the functional activity of these viral proteins. When infection conditions were chosen which mimic the in vivo situation, that is, when cells were infected with a low multiplicity of infection, induction of E3/19K by TNF was about 10-fold (27). In light of earlier in vivo data showing that adenovirus infection induces TNF in infected lung tissue in mice (48), we suggest the existence of an autocrine feedback loop to insure efficient E3 expression. The following scenario can be considered: macrophages/monocytes first infiltrate the site of infection and produce the inflammatory cytokines TNF, IL-1, and IL-6. If E3 expression is not sufficiently high, TNF may be able to lyse infected cells. However, binding of TNF to the TNF receptor activates NF-B leading to increased transcription of MHC genes. The simultaneous activation of the E3 promoter allows efficient expression of E3 proteins and therefore protect infected cells from the lytic activity of TNF and the subsequently infiltrating cytotoxic T cells (6,48). Several lines of evidence support this model. First, TNF treatment of E3 ϩ target cells did not increase their lysis  (51). The amount of E3/19K precipitated was determined by PhosphorImager analysis. In this particular experiment the ratio of immunoprecipitated E3/19K from TNF-treated versus untreated cells was 2.5 for Ad2, 3.2 for Ad5, and 9.0 for dl312. Average numbers are given in the text. by cytotoxic T cells. 2 Second, lipopolysaccharide treatment of mice transgenic for the E3 region of Ad2 induces TNF-activated E3 transcription in several organs which were E3-negative in the absence of TNF. The induction in organs showing a basal activity of the E3 promoter ranged from 2-fold in spleen to 62-fold in the lung (49). These data indicate that the E3 promoter can be regulated by TNF in vivo. Moreover, there is evidence that the severity of an adenovirus infection (those with fatal outcome) correlates with high levels of TNF (and IL-6 and IL-8) in the sera of patients (35). Further work is necessary to clarify whether TNF causatively determines the outcome of Ad infections or whether the above correlation reflects an increased tissue damage triggered by adenovirus.
There is also evidence that viral membrane proteins such as E3/19K which accumulate in the endoplasmic reticulum may cause endoplasmic reticulum stress resulting in NF-B activation (50). This can be interpreted as a primitive response mechanism of cells to virus infection with the consequence of activating immunologically important genes. However, the significance of this phenomenon during infection remains to be investigated.