Mechanism of Inactivation of NF-κB by a Viral Homologue of IκBα

Activation of the nuclear factor κB plays a key role in viral pathogenesis, resulting in inflammation and modulation of the immune response. We have previously shown that A238L, an open reading frame from African swine fever virus (ASFV), encoding a protein with 40% homology to porcine IκBα exerts a potent anti-inflammatory effect in host macrophages, where it down-regulates NF-κB-dependent gene transcription and proinflammatory cytokine production. This paper reveals the mechanism of suppression of NF-κB activity by A238Lp. A238Lp is synthesized throughout infection as two molecular mass forms of 28 and 32 kDa, and vaccinia-mediated expression of A238L demonstrated that both proteins are produced from a single gene. Significantly, the higher 32-kDa form of A238L, but not the 28-kDa form, interacts directly with RelA, the 65-kDa subunit of NF-κB, indicating that the binding is dependent on a post-translational modification. Immunoprecipitation analysis shows the NF-κB p65-A238L p32 heterodimer is a separate complex from NF-κB-IκBα, and it resides in the cytoplasm. Moreover, we show that ASFV infection stimulates the NFκB signal transduction pathway, which results in the rapid degradation of endogenous IκBα, although both forms of A238Lp are resistant to stimulus-induced degradation. Using the proteasome inhibitor MG132, we show that when degradation of IκBα is inhibited, A238Lp binding to NF-κB p65 is reduced. The results suggest that the virus exploits its activation of the NF-κB pathway to enable its own IκB homologue to bind to NF-κB p65. Last, we show that synthesis of IκBα is increased during ASFV infection, indicating RelA-independent transcription of the IκBα gene.

The NF-B family of transcription factors is involved in regulation of the expression of numerous cellular genes involved in the immune response, inflammation and apoptosis (reviewed in Ref. 1). The NF-B/Rel family is composed of homodimeric and heterodimeric complexes of the Rel family of proteins, which include RelA (p65), NF-B1 (p50), NF-B2 (p52), c-Rel, and RelB (2). The heterodimer of p50 and p65 is the most common form of NF-B dimer. In resting cells, it is present in an inactive form bound to one of a family of IB proteins, ensuring that proinflammatory gene expression is turned off. The family of IBs, characterized by structural ankyrin repeat motifs that bind NFB, includes IB␣, IB␤, IB⑀, the C-terminal ends of NF-B precursors p100/p105, the Bcl3 protein, and the Drosophila protein Cactus (2)(3)(4). Upon cell stimulation, IB␣ is first phosphorylated by IB kinase, a large multisubunit complex (5)(6)(7); it is then ubiquitinated and degraded by the 26 S proteasome, allowing NF-B to translocate to the nucleus and bind target B sites (reviewed in Ref. 8). Activated NF-B is then responsible for the rapid induction of cellular anti-viral activity.
Viruses exploit diverse and complex strategies to counteract the host response to infection. Large DNA viruses encode proteins that are nonessential for viral replication but that can inhibit a range of cellular functions, including blocking of cytokines and their receptors (reviewed in Ref. 9), the cytotoxic immune defense (10), or proteins that regulate cellular signaling (11). A novel strategy has been elucidated for the porcine virus, African swine fever virus (ASFV). 1 ASFV is a large double-stranded DNA virus that infects macrophages, and virulent isolates cause a fatal hemorrhagic disease of pigs (12,13). In addition to hemorrhage, the pathology is characterized by lymphoid tissue destruction due to apoptosis (14). We found that secretion and transcription of proinflammatory cytokines were inhibited in ASFV-infected macrophages (15). The ASFV genome of 170 -180 kilobase pairs was sequenced (16) and revealed at least 150 open reading frames, one of which, designated A238L in the BA71V isolate, has ankyrin repeats similar to those in IB␣ (16). Interestingly, we found that expression of the A238L gene alone inhibited NF-B-dependent gene transcription and prevented NF-B binding to its cognate B target sequence (15).
In this study, we investigated the mechanism of inhibition of NF-B by A238Lp. We show here that there are two forms of A238Lp synthesized throughout infection. One is the 28-kDa protein predicted from the ORF; the second is a higher molecular mass 32-kDa form produced by post-translational modification. Significantly, the higher molecular mass 32-kDa form and not the 28-kDa form bound NF-B p65, thus demonstrating that the post-translational modification was required to inhibit the NF-B pathway. We also show that, in common with many viruses (reviewed in Ref. 17), infection with ASFV acti-vated the NF-B signal transduction pathway. Remarkably, ASFV exploits the resulting signal-induced degradation of IB␣ to enable its own IB homologue to bind NF-B p65. The result is a A238Lp-NF-B heterodimer that forms an inactive complex in the cytoplasm that is not subject to signal-induced degradation. Furthermore, ASFV infection was found to increase IB␣ synthesis, although NF-B p65 activity was inhibited, thus supporting previous studies showing RelA-independent control of IB␣ gene transcription (18 -20).

EXPERIMENTAL PROCEDURES
Cells and Viruses-Cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Vero cells were infected with the tissue culture adapted strain of ASFV, BA71V, at a multiplicity of infection of 5:1 for 1 h and then incubated in complete medium for the indicated times. In some experiments, cells were stimulated with 10 M phorbol 12-myristate 13-acetate (PMA) and 10 g/ml lipopolysaccharide (LPS). In experiments where proteasome inhibitor was used, MG132 (Calbiochem) was added at 25 M to the medium after the initial 1-h infection for the rest of the incubation period (7 h). Reporter genes were transfected into pig kidney IBRS2 cells or into the mouse macrophage cell line, RAW 264.7 (ECACC 91062702). Vaccinia virus (MVA-T7) expressing T7 RNA polymerase was grown in baby hamster kidney cells (21).
Antibodies and Plasmids-Peptides corresponding to the N and C terminus of A238Lp (MEHMFPEREIENLFVKWIKKHIRNGNLTLF and VFHRWFKKKPKIIITGCKNNVYEKLPEQNP, respectively) were conjugated to maleimide-activated KLH and OVA and used to raise antibodies in rabbits. Two antibodies raised to the N-terminal peptide were characterized as cross-reacting with A238Lp and termed N1 and N2. One antiserum raised to the C-terminal peptide cross-reacted with A238Lp and was termed C1. Rabbit anti-NFB p65 antiserum against an N-terminal peptide (SC109) was used for Western blot analysis and, where indicated, for immunoprecipitation, and a goat antibody against a C-terminal peptide of NF-B (SC-372-G) was used in separate immunoprecipitations as indicated (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). IB␣ was detected by Western blot with a monoclonal antibody (SC-1643) and by immunoprecipitation by rabbit polyclonal antibody (SC 203). Antibody recognizing ERP60 was raised in rabbits as described previously (22). Normal rabbit serum was from Sigma. Anti-ASFV vp30 was from Dan Rock (USDA, Plum Island Animal Disease Center). A238L in pcDNA3 was as described previously (15). NF-B p65 and IkB␣ cDNAs in RcCMV was from Heike Pahl (University of Freiberg).
Vaccinia Virus/T7 RNA Polymerase-mediated Expression of A238L-BSC40 cells were infected for 1 h at 37°C with the MVA-T7 strain of vaccinia (21) expressing bacteriophage T7 RNA polymerase. Cells were washed in serum-free medium and transfected with pA238L or pcDNA3 alone (Invitrogen) using SuperFect (Quiagen). Cells were lysed in immunoprecipitation buffer 24 h post-transfection.
Metabolic Labeling and Immunoprecipitation-Cells (2 ϫ 10 6 ) were preincubated with methionine-and cysteine-free Eagle's medium for 30 min at 37°C. They were pulse-labeled for 60 min using 35 S-labeled Promix (Amersham Pharmacia Biotech) in methionine-and cysteinefree medium. Cells were washed twice in phosphate-buffered saline and lysed in immunoprecipitation buffer (10 mM Tris, pH 7.8, 0.15 M NaCl, 1% Brij 35, 10 mM iodoacetamide, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 1 g/ml each of leupeptin, pepstatin, chymostatin, and antipain (Roche Molecular Biochemicals). In phosphate labeling experiments, cells were washed in phosphate-free medium and incubated for a further 5 h in phosphate-free medium containing [ 32 P]orthophosphate (Amersham Pharmacia Biotech) and lysed and immunoprecipitated as described above. In subcellular localization experiments, nuclear and cytoplasmic fractions were obtained by standard procedures. Lysates or subfractions were precleared with protein A (Sigma) and immunoprecipitated with antibodies immobilized on protein A-or protein G-Sepharose (Sigma). After an overnight incubation, proteins were separated by SDS-PAGE and visualized by autoradiography. In some experiments, half the lysate was reprecipitated with a second antibody after incubation in 1% deoxycholate or 1% SDS for 30 min at room temperature.
Western Blot Analysis-Cells were infected with ASFV for time periods described in the individual experiments. In some experiments, cells were incubated for 1 h with cycloheximide (100 g/ml) with or without PMA/LPS (10 M and 10 g/ml, respectively) before lysis in immunoprecipitation buffer. Lysates were freeze-thawed and sonicated. Protein concentrations were determined by the Pierce BCA protein assay. Proteins were separated by SDS-PAGE on 10% gels and transferred to nitrocellulose membranes (Protran BA 85, Schleicher and Schuell). Filters were blocked in 10% dried milk and incubated with the primary antibody in 5% dried milk 10% goat serum, 0.05% Tween 20 for 1 h. Immunoreactive proteins were detected with a horseradish peroxidase-conjugated donkey anti-rabbit secondary antibody using ECL (Amersham Pharmacia Biotech).
Transfections and Reporter Gene Assays-Transfections were carried out in the pig kidney cell line IBRS2 or in the mouse macrophage cell line RAW264.7 using Transfectam (Promega). The RANTES promoter contains sequences from Ϫ952 to ϩ53 of the human RANTES gene cloned into the HinDIII and XbaI site of the reporter vector pCAT basic (Promega). Sequences between Ϫ680 and ϩ11 of the murine TNF␣ promoter were polymerase chain reaction-amplified from Balb/c genomic DNA and cloned between the HindIII and XbaI sites of the reporter vector pCAT basic. The human IL-8 and mutant IL-8 promoter constructs have been described previously (23). Transactivation studies with p65 and inhibition studies with IB␣ were performed by co-transfection of the cDNAs for human NF-B p65 or IB␣ into cells with the promoter-reporter constructs. Cell lysates were prepared from transfected cells after 18 h and assayed for luciferase activity (Promega) or for chloramphenicol acetyltransferase activity using a CAT enzymelinked immunosorbent assay kit (Roche Molecular Biochemicals) according to the manufacturer's instructions.

RESULTS
There Are Two Forms of A238L-The characteristics of the viral IB homologue, A238Lp, were followed during infection by raising antibody to peptides representing the N and C termini of the protein encoded by the open reading frame. One antibody, termed N2, recognized denatured protein and was used in Western blotting experiments to follow the time course of A238Lp expression during infection (Fig. 1A). A238Lp was first detected 2 h after infection with levels rising over 16 h. Interestingly, the antibody recognized two forms of A238Lp throughout the infectious cycle. One form migrated at 28 kDa, the molecular mass anticipated from translation of the A238L reading frame. The second form migrated at 32 kDa, suggesting post-translational modification of the protein or alternative splicing of the gene. When the A238L gene was expressed transiently from a T7 promoter using a recombinant vaccinia virus expressing T7 RNA polymerase (MVA-T7), two forms of the protein were again identified at 32 and 28 kDa by Western blot analysis (Fig. 1C, lane 3). These bands were not seen in uninfected cells (lane 1) or in cells transfected with vector alone (lane 2). Since both forms of the protein were expressed from a single cDNA, we concluded that the 32-kDa form was a posttranslational modification of the faster migrating 28-kDa form. Moreover, since the two forms occur in transfected cells in the absence of the rest of the ASFV genome, the modification is either the result of an endogenous host cell process or, more unlikely, caused by a vaccinia virus-encoded protein.
Two further antibodies were identified that immunoprecipitated A238Lp from infected cell lysates. Surprisingly, one antibody (C1) raised against the C terminus of A238Lp immunoprecipitated only the 28-kDa form of the protein, while a second antibody (N1) raised against the N terminus recognized only the 32-kDa form (Fig. 1B). Both antibodies failed to recognize A238Lp by Western blotting, suggesting that they recognize conformational epitopes. Importantly for this study, the antibodies could be used to distinguish between the two forms of A238Lp.
The 32-kDa Form of A238L Coprecipitates with the 65-kDa Subunit of NF-B-Given the homology between A238Lp and IB␣, the potential association of the A238L gene product with NFB was analyzed. Control cells or cells infected with ASFV for 8 h prior to the experiment were pulse-labeled for 60 min, lysed in detergent, and immunoprecipitated with antibody specific for a peptide representing either the N terminus or the C terminus of the 65-kDa subunit of NF-B ( Fig. 2A). The NF-B p65 N-terminal peptide antibody immunoprecipitated a 65-kDa protein from uninfected cells and a second protein at 37-kDa shown below to be IB␣ ( Fig. 2A, lane 3) (a nonspecific band is seen in all lanes at 50 kDa). Significantly, when the immunoprecipitation was repeated for infected cells, a third protein was seen at 32 kDa ( Fig. 2A, lane 4). The 32-kDa protein comigrated with the higher molecular weight form of A238Lp immunoprecipitated from the cell lysates with N1 anti-A238Lp antibody ( Fig. 2A, lane 1). A 28-kDa protein was not seen in the NF-B immunoprecipitate, even though substantial levels of the 28-kDa form of A238Lp could be recovered from the lysates using the p28-specific antibody, C1 ( Fig. 2A, lane 2). An antibody recognizing the C terminus of p65 immunoprecipitated only two proteins, NF-B p65 and IB␣, at 37 kDa from both uninfected ( Fig. 2A, lane 5) and infected cells ( Fig. 2A, lane 6).
The proteins at 65, 37, and 32 kDa were identified as NF-B p65, IB␣, and A238L p32 by immunoprecipitation followed by Western blot analysis (Fig. 2B). Vero cell lysates that had been immunoprecipitated with either anti-IB␣ antibody (lane 1) or anti-p65 antibody (lane 2) or an irrelevant control antibody of anti-IB kinase antibody (lane 3) or the anti-A238Lp antibody N2 (lane 4) were transferred onto nitrocellulose. The upper half of the membrane was probed with an anti-p65 antibody, while the lower half was probed with either a monoclonal antibody to IB␣ or the anti-A238Lp antibody, N2. Western blot analysis confirmed that the 37-kDa protein was IB␣, which co-precipitated a 65-kDa protein detected with anti-NF-B p65 antibody (lane 1). In a reciprocal manner, immunoprecipitation with NF-B p65 antiserum (lane 2) followed by Western blot analysis with NF-B p65 antibody (upper panel) and IB␣ antibody (middle panel) also identified the 65-kDa/37-kDa complex as NF-B p65 and IB␣. When immunoprecipitates of p65 were generated from ASFV-infected cells and analyzed by Western blot using antibody N2, which detects both forms of A238Lp, a 32-kDa band was seen (lane 2, bottom panel), thus identifying the 32-kDa protein as A238L p32. A control immunoprecipitation with anti-IB kinase antibody (Fig. 2B, lane 3) demonstrated that the Western blotting was specific. When immunoprecipitates of A238Lp were generated and transferred to nitrocellulose and the membrane was probed with the N2 anti-A238Lp antiserum, both A238L p28 and A238L p32 were seen (lane 4, bottom panel); however, Western blotting of the A238L immunocomplex with anti-p65 did not detect NF-B p65 (lane 4, top panel). Interestingly, therefore, this result suggests that either N2 did not recognize A238Lp complexed to p65 or that the antibody displaced p65 from A238L p32. The interaction of p65 with A238L p32 was established further in the next experiment.
A sequential reimmunoprecipitation approach was taken to confirm the identify of the 32-kDa protein associated with NF-B p65 in infected cells (Fig. 2C). Lysates from an equal number of infected cells were immunoprecipitated with either N1 anti-A238Lp antiserum (lane 1) or anti-p65 antiserum (lane 2). The complex of three proteins (p65, IB␣, and p32) in the anti-p65 immunoprecipitate were dissociated at room temperature with 1% deoxycholate and reprecipitated with the N1 antibody specific for the 32-kDa form of A238Lp (Fig. 2C, lane 4). A single band at 32 kDa was detected, confirming the identity of the viral protein precipitating with p65 as the higher molecular weight form of A238Lp. A control using normal rabbit serum to reimmunoprecipitate the complex is shown (lane 5).
Surprisingly, a comparison of lanes 1 and 2 of Fig. 2C showed that more A238L p32 was recovered from lysates with anti-p65 antibody (lane 2) than when the same lysate was immunoprecipitated with excess N1 antibody specific for A238L p32 (lane 1). The results suggested that most of the A238Lp in cells was bound to NF-B p65 and that the epitope on A238Lp recognized by the anti-N-terminal peptide antibody was masked by NF-B p65. This was consistent with two observations: first, that increased levels of A238L p32 could be recovered from the complex with p65 if they were first dissociated in deoxycholate and then reprecipitated (comparing Fig. 2C, lanes 1 and 4); second, p65 is not detected by Western blot of an N2 immunoprecipitate (Fig. 2B, lane 4).
An experiment was performed to investigate whether a trimolecular complex existed between p65, IB␣, and A238L p32 (Fig. 2D). Lysates from Vero cells infected for 8 h with ASFV were immunoprecipitated with an excess of anti-IB␣ antiserum, revealing proteins at 37 kDa (IB␣), 65 kDa (p65), and 70 kDa (perhaps other Rel proteins) but not at 32 kDa (Fig. 2D,  lane 1). This result indicated IB␣ bound p65 separately from A238L p32. The supernatant from this IB␣-depleted lysate FIG. 1. The A238L gene is expressed as two proteins throughout ASFV infection. A, time course of A238Lp expression. Vero cells were infected with ASFV for the indicated period and then lysed, and protein was separated by SDS-PAGE and blotted to membrane filters. Expression of A238L protein was analyzed by Western blot using the N2 anti-A238L N-terminal peptide antibody. Two forms of A238Lp were detected throughout infection at 28 and 32 kDa. B, differential recognition of A238L p28 and p32 by conformation-dependent antibodies N1 and C1 in ASFV-infected cells. Vero cells, either uninfected (lanes 1 and 2) or infected with ASFV for 8 h (lanes 3-6), were pulse-labeled with [ 35 S]cysteine/methionine for 1 h, lysed, and immunoprecipitated with either C1, the anti-A238L C-terminal peptide antibody (lanes 1 and 5); N1, the anti-A238L N-terminal peptide antiserum (lanes 2 and 6); or the respective preimmune serum (lanes 3 and 4). At 8 h postinfection, N1 antibody recognized the 32-kDa protein A238L p32, while C1 recognized the 28-kDa protein A238L p28. C, vaccinia virus-mediated expression of A238L cDNA. BSC40 cells infected with vaccinia virus expressing T7 RNA polymerase (MVA-T7) were used to express the A238L gene under the control of a T7 promoter. Lysates from uninfected cells (lane 1) or from vaccinia-infected cells transfected with pcDNA3 alone (lane 2) or A238L in pcDNA3 (lane 3) were blotted onto membrane filters, and proteins were detected with N2 anti-A238L antiserum. In A238L-transfected cells, two bands were detected at 28 and 32 kDa not seen in uninfected cells or cells transfected with pcDNA3 alone. was then immunoprecipitated with NF-B p65 antiserum and showed two major proteins at 65 and 32 kDa and a reduced amount of IB␣ at 37 kDa (Fig. 2D, lane 2). NF-B p65-IB␣ complexes were much reduced but not completely absent from this sample, probably due to low affinity of the IB␣ antibody (as seen in Fig. 2B, lane 1). These results indicate that A238L p32 bound to NF-B p65 that was not complexed with IB␣. In infected cells, therefore, NF-B p65 exists as two pools, either complexed to IB␣ or to A238L p32 (Fig. 2D, lane 3).
Cytoplasmic Localization of the NF-B p65-A238L p32 Complex-The localization of the NF-B p65-A238L p32 complex was determined in infected cells after pulse labeling and fractionation into cytoplasm and nuclei. Nuclear or cytoplasmic fractions were immunoprecipitated with N1 or N2 anti-A238Lp or with anti-p65 antibodies. Both forms of A238L, p28 and p32, were found in the cytoplasm (Fig. 3, N1 and N2), as was the complex of NF-B p65 with A238L p32 (Fig. 3, ␣ p65). Interestingly, NF-B p65 was only found in the cytoplasm of infected cells.
IB␣ Is Degraded during ASFV Infection, and A238Lp Is Resistant to Cell Activation-Upon infection of cells, many viruses stimulate the signal transduction pathway leading to IB␣ degradation (17). In resting cells, NF-B is held in an inactive complex with IB␣. Upon cell stimulation, phosphorylation of IB␣ by an IB kinase complex provides a signal for ubiquitination and degradation of IB␣ by cytosolic proteasomes (8,24). The released NF-B dimer enters the nucleus to increase antiviral and proinflammatory gene transcription. To investigate the possible activation of the NF-B signal transduction pathway by ASFV, steady state levels of IB␣ were analyzed in the presence and absence of the proteasome inhib-itor, MG132 (Fig. 4A, middle panel). Lysates from equal numbers of uninfected cells, either untreated (Fig. 4A, lane 1) or treated with MG132 (Fig. 4A, lane 2) were transferred to membranes along with lysates from ASFV-infected cells either untreated (Fig. 4A, lane 3) or treated with MG132 (Fig. 4A, lane  4). Steady state IB␣ levels were lower in infected cells com- pared with uninfected cells and were increased toward normal by proteasome inhibition. These results show that ASFV infection itself stimulates IB␣ degradation by proteasomes. Significantly, therefore, the signal transduction pathway leading to IB kinase activation and resulting in IB phosphorylation is not blocked during ASFV infection, since signal-induced phosphorylation of IB is a prerequisite for its degradation.
Synthetic rates of IB␣ were measured in infected cells by metabolic labeling and immunoprecipitation with anti-IB␣ antiserum (Fig. 4A, top panel). IB␣ synthesis decreased in cells after proteasome inhibition (lane 2), showing that the synthesis of IB␣ is dependent on its own degradation (25).

Interestingly, the results show that synthesis of IB␣ was increased in ASFV-infected cells (lane 3) compared with uninfected cells (lane 1), and this increase could be blocked by proteasome inhibition (lane 4).
Although ASFV inhibits p65 activity, other members of the Rel family have been shown to up-regulate the IB␣ promoter activity (18). IB␣ is also transcriptionally regulated by transforming growth factor-␤ (19), a cytokine that has been shown to be up-regulated in ASFV infection (15). The results demonstrate a complex regulation of the IB␣ gene, which is under further investigation. The high synthetic rate of IB␣ together with its low steady-state level in infected cells indicates that ASFV infection provides a potent signal for IB␣ degradation.
One explanation for the observed suppression of NF-B gene transcription by A238L protein could be the formation of a complex with A238Lp where the viral protein is resistant to proteasomal degradation. The relative stabilities of IB␣ and the two forms of A238Lp in response to cell stimulation were therefore tested. Vero cells, either uninfected or infected with ASFV for 8 h, were incubated with cycloheximide for 1 h either in the presence or absence of PMA/LPS (Fig. 4B). Levels of both IB␣ and A238Lp were detected by Western blotting with an anti-IB␣ monoclonal antibody (top panel) and the N2 antibody that recognizes both forms of A238Lp (middle panel). In control experiments in uninfected cells, cycloheximide treatment showed that the half-life of IB␣ was about 1 h (Fig. 4B, lanes  1 and 2), and IB␣ was completely degraded after stimulation with PMA/LPS (Fig. 4B, lane 3) as expected (26). In infected cells, steady state levels of IB␣ were lower than those in uninfected cells (Fig. 4B, lane 4), and IB␣ was undetectable after the cycloheximide chase in the presence or absence of PMA/LPS (Fig. 4B, lanes 5 and 6) degradation. Both forms of A238Lp were stable during cell activation with PMA/LPS, indicating resistance to stimulus-induced degradation (Fig. 4B,  middle panel, lanes 4 -6). A control showing equal lane loading and transfer to nitrocellulose is shown with an abundant resident ER protein, ERP60 (Fig. 4B, bottom panel).
A238Lp Replaces IB␣ bound to NF-B p65-The previous results point to the conclusion that the virus exploits signalinduced degradation of IB␣ to expose NF-B for binding to the IB homologue. A238Lp then forms a stable complex with NFB in the cytoplasm, preventing the transcription factor from moving into the nucleus and activating gene expression. In the next experiment, the requirement of proteasome activity for the replacement of IB␣ with the viral A238L protein was tested by incubating cells with the proteasome inhibitor MG132 (Fig. 5). Cells were pulse-labeled, and lysates were immunoprecipitated with anti-p65 antibody (Fig. 5A, IP). Significantly, incubation of cells with MG132 decreased the levels of A238L p32 bound to p65. The immunoprecipitate also showed an apparent reduction in the amount of IB␣ associated with p65 in MG132-treated cells, confirming the result in Fig. 4A, where immunoprecipitation with IB␣ antibody detected decreased IB␣ synthesis after proteasomal inhibition. The levels of IB␣ bound to p65 were analyzed by Western blot of the complex precipitated with anti-p65 using an IB␣ monoclonal antibody (Fig. 5A, WB). The Western blot gave the predicted result that there was more IB␣ associated with p65 in MG132-treated cells.
To investigate whether MG132 had an inhibitory effect on A238Lp synthesis, metabolically labeled cell lysates were immunoprecipitated with the N1 anti-A238Lp antibody (Fig. 5B). In fact, levels of A238L p32 were slightly increased in the presence of MG132. It has been shown above (Fig. 2C) that the N1 epitope is masked in the p65-A238Lp complex. The result shown in Fig. 5B suggests that there was more free A238Lp available in MG132-treated cells, supporting the idea that inhibition of IB␣ degradation blocks formation of the A238Lp-p65 complex. Levels of another viral protein, vp30, an early viral gene product, were unchanged in the presence of MG132 (Fig. 5B). Taken together therefore, these results show that proteolytic removal of IB␣ from p65 was required for the binding of A238Lp to p65.
Analysis of the Post-translational Modification on A238L p32-Throughout these experiments, we have shown that only the 32-kDa form of A238L bound to NF-B p65. Clearly, the post-translational modification providing the extra 4 kDa was important for A238Lp to inhibit NF-B p65. Given the functional and sequence homology between IB␣ and A238Lp, we investigated the possibilities of phosphorylation, ubiquitination, and sumoylation (27) to explain the increased molecular weight seen for A238L p32.
For phospholabeling analysis, cells were incubated in phosphate-free medium containing [ 32 P]orthophosphate for 5 h, and cell lysates were immunoprecipitated with either antibody N1 or antibody to vp30, the major phosphoprotein of ASFV as a positive control (Fig. 6). Antibody N1 did not detect any proteins labeled with [ 32  Uninfected and ASFV-infected cell lysates, either immunoprecipitated with NF-B p65 or the whole cell lysate, were separated by SDS-PAGE, blotted to nitrocellulose, and analyzed with anti-ubiquitin and anti-SUMO antibodies. Although many ubiquitinated and SUMO-modified proteins were detected in whole cell lysates, there was no specific band seen for A238Lp in infected cells (data not shown). Also, Western blotting of an NF-B p65 immunoprecipitate that coprecipitated A238Lp did not detect any ubiquitinated or SUMO-containing protein at 32 kDa (data not shown). These experiments indicated that the modification of A238L p32 is not due to either ubiquitin, SUMO, or phosphate addition to A238L p28.

A238Lp Inhibits NF-B-dependent Cytokine Gene Expression in a Macrophage-like Cell
Line-A238Lp was first identified as a putative IB␣ homologue because it inhibited an NF-B-dependent reporter construct and NF-B binding to DNA in a porcine kidney cell line (15). However, since the primary site of ASFV replication in vivo is the macrophage (14), a functional analysis of A238Lp expression was carried out in a macrophage cell line, RAW 264.7, using macrophage-specific cytokine promoter-reporter constructs co-transfected with A238Lp. This would more closely represent the effect on gene expression of A238Lp binding to p65 found in the infected host cell in vivo. TNF␣, IL-8, and RANTES genes are rapidly up-regulated through the NF-B signal transduction pathway in activated macrophages. The upstream promoters of these genes cloned upstream of reporter cDNAs were transfected into RAW 267 cells. A 2-3-fold increase from each of the IL-8, RANTES, and TNF␣ promoters was recorded when RAW cells were stimulated with PMA/LPS (Fig. 7A). Since macrophage cell lines are inherently difficult to transfect, with only 5-10% of cells ex-

FIG. 5. A238Lp replaces IB␣ bound to NF-B p65.
A, proteasome inhibition decreases the binding of A238Lp to NFB p65. Metabolically labeled Vero cell lysates, 8 hpi with ASFV, either untreated (Ϫ) or after the addition of the proteasome inhibitor MG132 (ϩ) were immunoprecipitated with anti-NF-B p65 antiserum. There was less A238Lp bound to NF-B p65 when IB␣ degradation was inhibited. Western blot analysis of IB␣ showed total amounts of IB␣ co-precipitating with p65 was greater when degradation was inhibited with MG132 (WBϩ). However, there was less metabolically labeled IB␣ (IPϩ), since synthesis of IB␣ is decreased in the presence of proteasome inhibitor (see Fig. 4). B, proteasome inhibition does not alter synthesis of viral proteins vp30 and A238Lp. Immunoprecipitation of early viral protein vp30 (anti-vp30) and A238L p32 (N1) showed that viral protein synthesis was not inhibited in the presence of MG132. In fact, N1 detected slightly more A238L p32 after proteasome inhibition, since the N1 epitope is revealed in free A238Lp that is not bound to p65.
pressing the transgene, reporter levels were lower than normally expected from other cell lines. When cells were co-transfected with the A238Lp gene, basal and PMA/LPS-stimulated activities were inhibited from IL-8, RANTES, and TNF␣ promoters (Fig. 7A). The experiment shown is a representative of the results obtained in three separate experiments. A plasmid encoding NF-B p65 co-expressed with the IL-8 promoter transactivated gene expression, but when the A238L gene was also expressed, transactivation was inhibited. This experiment supports the findings that A238Lp influences p65 directly rather than stepping upstream of NF-B activation. As a control, an IL-8 promoter mutated at the NF-B binding site showed no activity, indicating that the increases were totally dependent on the NF-B binding to its cognate site. In a second control of specificity of NF-B inhibition, AP1-dependent transcription was unaffected by A238L (Fig. 7A). These results show that the viral IB homologue specifically inhibited several promoters containing NF-B sites driving the expression of macrophage-specific proinflammatory genes.
The previous experiments demonstrated that in infected cells the endogenous IB␣ associated with the p65 was replaced by the viral protein A238Lp (Fig. 5). It was also shown that the viral homologue was resistant to activation-induced degradation, suggesting that binding to p65 was long lived (Fig.  4B). Taken together, the results suggested that sustained binding of A238Lp to p65 would lead to prolonged inactivation of the protein. The reversibility of inactivation of NF-B by either IB␣ or A238Lp was therefore compared using the NF-B-dependent reporter assay (Fig. 7B). In the control, stimulation of cells for 120 min led to a 2-3-fold increase in IL-8 promoter activity. Activity was then measured in cells coexpressing the IL-8 promoter plasmid with genes encoding either IB␣ or A238Lp after stimulation with PMA/LPS. Up to 30 min poststimulation, both IB␣ and A238Lp inhibited promoter activity. However, luciferase activity began to recover between 60 and 120 min post-stimulation in cells transfected with IB␣ but not in cells expressing the A238L protein. This experiment demonstrated that, in contrast to the stimulus-induced degradation of IB␣, the effect of A238Lp on NF-B-dependent transcription did not wane over 2 h. DISCUSSION This study has probed the molecular mechanism for the inhibition of the pro-inflammatory response of macrophages during infection by African swine fever virus. We show that the ASFV-encoded protein, A238Lp, with homology to IB␣, binds FIG. 7. ASFV IB␣ homologue, A238Lp, inhibits NF-B-dependent gene expression. A, RAW 264 cells, a macrophage-like cell line, were transfected with the following constructs: an IL-8 promoter-luciferase reporter construct (IL8); part of the IL-2 promoter containing an AP1 site (AP1); a RANTES promoter-CAT reporter (RANTES); a TNF␣ promoter-CAT reporter (TNF). Where indicated, cells were co-transfected with a cDNAs for A238L and NF-B p65 and/or stimulated with PMA/LPS for 4 h. Luciferase activity measured as relative light units (RLU) and CAT activity measured by CAT enzyme-linked immunosorbent assay is presented as percentage over background. Mutant is an IL-8 promoter construct that does not bind NF-B. B, inhibition of NF-B-dependent gene expression by A238Lp. Porcine IBRS2 cells were transfected with IL-8 promoter-luciferase reporter gene alone (IL8) or cotransfected with genes for IB␣ (IL8 ϩ IB␣) or A238L (IL8 ϩ A238L). Cells were stimulated with PMA/LPS for 0, 30, 60, or 120 min, and luciferase activity was measured (RLU). Increased luciferase activity demonstrated that inhibition with IB␣ was diminished after 60 min, whereas A238L inhibition was sustained over 2 h. directly to the p65 subunit of NFB. Immunoprecipitation and Western blotting studies showed that infected cells synthesize two forms of A238Lp: a smaller form that migrated at 28 kDa, a size predicted from the A238L reading frame, and a larger form that migrated at 32 kDa. We assume that the larger form is post-translationally modified, because transfection of the cDNA for A238L into cells results in expression of two proteins, but we do not know the nature of the modification. Significantly, the post-translational modification activated A238Lp, since only the 32-kDa form of the viral protein bound to NF-B. Furthermore, the virus activates the NF-B signal transduction pathway, and this is exploited by the virus to replace the endogenous IB␣ with its own IB homologue.
During activation of cells, IB is targeted for degradation, allowing transport of the NF-B into the nucleus (28,29). The simplest model for the action of the viral IB homologue would be for the viral protein to bind NF-B but be resistant to signal induced degradation. Essentially, the protein would act as a dominant negative inhibitor of NF-B by retaining the protein in the cytoplasm. Several lines of evidence support this model. First, both forms of A238Lp were stable in cells activated by PMA and LPS, conditions that stimulated the rapid degradation of IB␣. The viral protein was not therefore subject to stimulus-induced degradation. Interestingly, since both forms of A238Lp were stable, resistance to degradation is an inherent property of the viral protein and does not result from the post-translational modification that promotes binding of A238Lp to NF-B p65. Second, the cytoplasmic localization of the A238L-p65 complex suggests that A238Lp prevents translocation of p65 into the nucleus in a manner analogous to IB␣. Functional evidence for compromised nuclear translocation of NF-B by A238Lp comes from gel shift experiments, demonstrating that in cells expressing the protein, levels of NF-B DNA binding activity are reduced following stimulation (15). Unlike IB␣, which can also inhibit NF-B activity by enhancing export of nuclear NF-B (30,31), our experiments suggest that A238Lp does not act in the nucleus.
This study showed that, in common with many viruses, ASFV infection activated the NF-B signal transduction pathway. Interestingly, when degradation of IB␣ was blocked by MG132, greatly reduced levels of A238Lp were bound to NF-B. We think that the virus exploits the signal-induced degradation of IB␣ to expose NF-B for binding to the viral IB homologue. Immunoprecipitation of infected cell lysates with anti-IB␣ antiserum failed to detect a trimolecular complex between IB␣, p65, and A238Lp, supporting the idea that a separate pool of NF-B p65 exists bound as a heterodimer with A238Lp. It also suggests that A238Lp binds to the same domain of RelA as IB␣, but this question needs to be addressed using deletion or point mutations of p65. It is also of interest that an anti-C-terminal p65 antibody did not precipitate the complex, suggesting that it might have displaced A238L p32. It has been established that IB binds to NF-B through ankyrin repeat domains in the central region together with a PEST domain in the C terminus of the molecule (32,33). The central region of A238Lp shows the highest homology to the ankyrin repeats 3-6 in IB␣, and it may bind through this domain to p65.
Interestingly, although this study has shown an increased degradation of IB␣ during ASFV infection, it has also demonstrated an increased synthesis of IB␣ in the presence of NF-B p65 inhibition by A238Lp. An increase in IB␣ expression following a decrease in NF-B binding has been seen in other systems, for example, after transforming growth factor-␤ treatment of B cell lymphomas (19). ASFV infection has been previously shown to increase transforming growth factor-␤ syn-thesis (15), so this cytokine may account for IB␣ transcriptional activation. Alternatively, during infection, there may be increases in other transcriptional activators of IB␣, such as heat shock proteins (20) or other Rel family members (18) that are not inhibited by A238Lp. Given the complexity of this signal transduction pathway, these possibilities are being investigated further.
One of the unique outcomes of this study is that we were unable to detect binding of the 28-kDa form of A238Lp to NF-B. The viral protein, therefore, has homology with IB␣, yet binding to NF-B requires post-translational modification by the host cell. Furthermore, it appears that the post-translational modification acts as a molecular switch whereby the virus can inhibit two signal transduction pathways with one protein. It has been shown recently, using A238Lp as bait in a yeast two-hybrid screen of a pig macrophage library, that the 28-kDa molecular mass form of A238Lp interacts with cyclophilin and the small subunit of calcineurin (protein phosphatase 2B) (34), thus inhibiting the nuclear factor of activated T cells pathway. A238Lp is therefore a remarkable viral protein with two forms affecting different signal transduction pathways, both of which have the potential to play an important role in immune evasion. The interaction between calcineurin and the NF-B pathway has yet to be defined, but interestingly, a recent report showed that calcineurin activity suppresses NF-B activation in macrophages (35).
Unfortunately, we have been unable to determine the nature of the modification. Given the structural and functional homology with IB␣, we investigated whether modifications known for IB␣ were found on A238Lp. The presence of two lysines at positions 19 and 20 in the predicted amino acid sequence of A238Lp were similar in position to two lysine residues at amino acids 21 and 22 that are targets for ubiquitination in IB␣ (36,37). Western blot analysis with anti-ubiquitin antibodies did not, however, detect the 32-kDa form of A238L in lysates from infected cells, suggesting that A238Lp is not ubiquitinated. These results were consistent with the lack of degradation of A238Lp in cells stimulated by PMA and LPS. Phosphorylation of neighboring serine residues at positions 32 and 36 in IB␣ by IB kinase acts as a signal for ubiquitin ligase recognition (38). Notably, these serines are absent from analogous positions in A238Lp. This raised the possibility that A238Lp was modified by SUMO-1, a reaction that occurs at lysine 21 of IB␣ without phosphorylation of Ser 32 and Ser 36 (27). We were however unable to demonstrate a SUMO-1 modification to A238Lp. Analysis of the sequence of A238Lp predicts three possible consensus sites for phosphorylation: a casein kinase II site, a protein kinase C site, and one tyrosine phosphorylation site. Even so, phosphate labeling experiments and immunoprecipitation with isoform-specific antibodies followed by phosphatase digestion showed that there was no phosphate modification. Although we do not know the precise modification, it is possible to predict the mechanism of activation of A238Lp. Antibodies raised against the extreme C and N terminus of A238Lp were able to differentiate between the two forms of the protein. At first, this was difficult to reconcile with the prediction that anti-peptide antibodies should be conformation-independent and bind both forms of the protein. The selective binding of C-terminal specific antibody to the smaller 28-kDa form could be explained if the epitope were blocked by the post-translational modification. However, the selective binding of the Nterminal specific antibody to the 32-kDa form suggests that this epitope is masked in the 28-kDa protein. We favor a model where post-translational modification of A238Lp produces a large conformational change that exposes the N terminus of the protein for binding to NF-B. This would explain why the antibody raised against this region of the protein recognizes only A238L p32 not bound to p65.
It has been well established that activation of NF-B protects cells from apoptosis (39,40). One would therefore expect ASFV-infected cells that inhibit NF-B to be more sensitive to activation-induced cell death (41). This would not be an advantage to the virus that needs to replicate and spread. However, the ASFV genome encodes two proteins with anti-apoptotic function, a Bcl2 homologue (42) and an Iap homologue (43), which may counteract the effect of NF-B inhibition.