Mechanism for Biphasic Rel A· NF-κB1 Nuclear Translocation in Tumor Necrosis Factor α-stimulated Hepatocytes

The proinflammatory cytokine, tumor necrosis factor α (TNFα), is a potent activator of angiotensinogen gene transcription in hepatocytes by activation of latent nuclear factor-κB (NF-κB) DNA binding activity. In this study, we examine the kinetics of TNFα-activated translocation of the 65-kDa (Rel A) and 50-kDa (NF-κB1) NF-κB subunits mediated by inhibitor (IκB) proteolysis in HepG2 hepatoblastoma cells. HepG2 cells express the IκB members IκBα, IκBβ, and IκBγ. In response to TNFα, Rel A·NF-κB1 translocation and DNA binding activity follows a biphasic profile, with an “early” induction (15-30 min), followed by a nadir to control levels at 60 min, and a “late” induction (>120 min). The early phase of Rel A·NF-κB1 translocation depends on simultaneous proteolysis of both IκBα and IκBβ isoforms; IκBγ is inert to TNFα treatment. The 60-min nadir is due to a rapid IκBα resynthesis that reassociates with Rel A and completely inhibits its DNA binding activity; the 60-min nadir is not observed when IκBα resynthesis is prevented by cycloheximide treatment. By contrast, selective inhibition of IκBβ proteolysis by pretreatment of HepG2 cells with the peptide aldehyde N-acetyl-Leu-Leu-norleucinal completely blocks the late phase of Rel A·NF-κB1 translocation. These studies indicate the presence of inducible and constitutive cytoplasmic NF-κB pools in hepatocytes. TNFα induces a coordinated proteolysis and resynthesis of IκB isoforms to produce dynamic changes in NF-κB nuclear abundance.

Multicellular organisms have evolved mechanisms for the coordinate expression of inducible genes through ligand-dependent receptors. Ligand binding to high affinity receptors located on the plasma membrane generate second messenger signals that can influence the activity or abundance of transcription factors through post-translational modifications including signal-induced phosphorylation and/or proteolysis. Hormone-activated gene transcription plays an important role in many homeostatic processes, including the cytokine cascade for lymphocyte expansion (1), and the change in expression of liver genes in response to systemic inflammation known as the hepatic acute-phase response (APR 1 ; reviewed in Refs. [2][3][4]. The APR is the consequence of inducible transcriptional activation of hepatic genes required for blood pressure regulation, such as angiotensinogen (2,5), and those involved in macrophage opsonization and wound repair (6) through the effects of macrophage-derived interleukins-1, interleukin-6, and tumor necrosis factor ␣ (TNF␣) (6). Hepatocyte-specific transactivators modified during the APR include AP-1 (7), signal transducers and activators (8), nuclear factor-interleukin 6 (9), and nuclear-factor-B (NF-B) (10). The angiotensinogen gene is transcriptionally activated during the APR by the effect of a single regulatory element, the acute-phase response element (APRE) (11,12). The APRE is a target for intracellular signaling initiated by the liganded TNF␣ type I receptor that activates latent DNA binding activity of the potent NF-B transcription factor family (3,12,13).
NF-B is a family of homo-and heterodimeric proteins related by an NH 2 -terminal ϳ300 amino acid Rel homology domain including the proteolytic processed NF-B1 and NF-B2 subunits, as well as the Rel A (p65), c-Rel, and Rel B subunits (reviewed in Ref. 10). Dimerization of various NF-B subunits produce complexes with various intrinsic DNA-binding specificities (14), transactivation potentials (10,(15)(16)(17)(18), and subcellular localization (19). For example, NF-B1 homodimers are constitutively nuclear and bind DNA avidly, but lack significant transcriptional activity; by contrast, Rel A⅐NF-B1, Rel A⅐c-Rel, and Rel A⅐NF-B2 heterodimers are cytoplasmic and exhibit various degrees of transcriptional activator properties (reviewed in Refs. 10,20). UV cross-linking (11), gel mobility shift assays with subunit-specific NF-B antibodies (12), and transient overexpression assays (12,13) indicate that Rel A⅐NF-B1 heterodimers are the major species of hormone-inducible NF-B subunits that bind the APRE in hepatocytes.
The Rel A⅐NF-B1 complex is sequestered in a latent cytoplasmic form by association with various inhibitor (IB) proteins, including IB␣ (pp40/MAD-3) (21)(22)(23), IB␤ (24), IB␥ (the COOH-terminal product encoded by translation of the alternative splicing of the p105 NF-B1 mRNA precursor (25)), and p105 itself (26,27) that associate with Rel A through a protein interactive domain homologous to erythrocyte ankyrin. Dissociation of Rel A from IB is prerequisite for Rel A nuclear translocation (23, 28 -30); current evidence favors a two-step dissociation that first requires inducible NH 2 -terminal phosphorylation (IB␣ is phosphorylated at serine residues 32 and 36) followed by proteolysis through the 26 S proteasome (30 -32).
The observations that distinct IB family members are expressed, and perhaps regulated, in a tissue-restricted fashion (24,31) prompted us to investigate the kinetics of latent Rel A:NF-B1 activation in hepatocytes. We report the unanticipated findings that TNF␣ produces a biphasic Rel A⅐NF-B1 nuclear translocation, with an "early" peak at 15-30 min, return to control (1 h), and a later peak (Ͼ2 h induction). In hepatocytes, the IB family members ␣, ␤, and ␥, but not the NF-B1 p105 precursor are expressed. Early Rel A⅐NF-B1 translocation is due to simultaneous proteolysis of both IB␣ and IB␤. By 1 h, IB␣ is rapidly synthesized and reassociates with Rel A; this reassociation (due to "overshoot" synthesis) results in complete inhibition of Rel A⅐NF-B1 binding even in the continued absence of IB␤. As IB␣ levels fall after 1 h, Rel A⅐NF-B1 binding is again detectable in the nucleus. Inhibition of IB␤ proteolysis by the peptide aldehyde N-acetyl-Leu-Leunorleucinal completely prevents the second phase of Rel A⅐NF-B1 binding, demonstrating the requirement of IB␤ for prolonged Rel A⅐NF-B1 nuclear action.

Cell Culture and Treatment
The human hepatoblastoma cell-line HepG2 was obtained from ATCC (Rockville, MD) and grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc., Grand Island, NY) supplemented with 10% (v/v) heat-inactivated fetal bovine serum, 2 mM L-glutamine, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, and antibiotics (penicillin/streptomycin/fungizone) in a humidified atmosphere of 5% CO 2 . Recombinant human TNF␣ (rTNF␣, Genentech) was added to a final concentration of 30 ng/ml in culture medium and cells were incubated for the indicated time periods at 37°C. For pretreatments, calpain inhibitor I (200 M, CalBiochem, San Diego, CA) or cycloheximide (50 g/ml, Sigma) were added in medium 1 h or 30 min prior to TNF␣ stimulation, respectively.

Electrophoretic Mobility Shift Assays (EMSAs)
EMSAs were performed as described previously with minor modifications (12,13). Nuclear extracts (10 g) were incubated with 40,000 cpm of 32 P-labeled APRE WT duplex oligonucleotide probe and 2 g of poly(dA-dT) in a buffer containing 8% glycerol, 100 mM NaCl, 5 mM MgCl 2 , 5 mM dithiothreitol, and 0.1 mg/ml phenylmethylsulfonyl fluoride in a final volume of 20 l, for 15 min at room temperature. The complexes were fractionated on 6% native polyacrylamide gels run in 1 ϫ TBE buffer (25 mM Tris, 25 mM boric acid, and 0.5 mM EDTA), dried, and exposed to Kodak X-AR film at Ϫ70°C. Competition was performed by the addition of 100-fold molar excess nonradioactive double-stranded oligonucleotide competitor at the time of addition of radioactive probe. The sequences of the APRE double-stranded oligonucleotides are as shown below. Antibody supershift assays were performed by adding to the binding reaction 1 l of affinity-purified polyclonal antibodies and incubating for 1 h on ice. All of the antibodies used in these assays were obtained commercially (Santa Cruz Biotech, Santa Cruz, CA). For the NF-B-DNA binding inhibition assays, the nuclear extracts were mixed with the indicated amounts of bacterially-expressed full-length IB␣ protein in binding reaction. Inactivated IB␣ was prepared by boiling the recombinant IB␣ protein for 30 min in phosphate-buffered saline.

Expression of Polyhistidine-tagged IB␣
The full-length cDNA encoding human IB␣ was subcloned as an EcoRI fragment into the pRSETB expression plasmid under the control of the T7 promoter (InVitrogen, San Diego, CA). The plasmid was transformed into Escherichia coli BL21(DE3)pLysS and the 47-kDa protein was induced with the addition of 1 mM isopropyl-1-thio-␤-Dgalactopyranoside (final concentration) during logarithmic growth and purified under native conditions on a nickel-agarose column (35). The protein was Ͼ90% pure as judged by SDS-PAGE and Coomassie Blue staining.

Biphasic Induction of Rel A⅐NF-B1 DNA-binding in Hepa-
tocyte Nuclei-Exposure of cultured HepG2 human hepatocytes to 20 ng/ml rTNF␣ for 6 h results in an induction of NF-B DNA binding activity and transcriptional activity (12,13). To resolve the induction kinetics of various NF-B family members in the rTNF␣ response, we examined a 6-h time course of APRE DNA binding activity using extracts of sucrose cushionpurified nuclei using EMSAs. In EMSAs performed under conditions that resolve the individual heterodimeric NF-B species binding to the radiolabeled APRE, four nucleoprotein complexes could be resolved (C1-C4, Fig. 1A). The C1 complex was weakly and variable inducible at 15 min. By contrast, the C2 complex was strongly inducible in a biphasic manner with the first peak occurring at 15 min (a 16-fold induction relative to control), declining by 30 min, and was undetectable at 60 min (the early binding phase). At 120, the C2 complex reappeared (4.1-fold relative to control) and persisted as long as 360 min (the "late" binding phase). Binding specificity of the complexes was demonstrated using site-specific competitors of the APRE in the EMSA (Fig. 1B). Complexes C1 and C2 both competed with homologous APRE WT but not APRE M2 or APRE M6 oligonucleotides, indicating sequence-specific recognition of the NF-B contact points on the APRE (5).
Gel mobility supershift assays using subunit-specific NF-B antibodies was used to demonstrate the composition of the strongly inducible APRE-binding C2 complex (Fig. 1C). Addition of Rel A, but not preimmune antibody resulted in the selective diminution of the C2 complex with the simultaneous appearance of a supershifted band. Similarly, addition of NF-B1 antibody also diminished the intensity of the C2 nucleoprotein complex. Taken together, these data indicate that the DNA binding activity of the C2 is biphasic upon rTNF␣ treatment, C2 binds to the APRE with NF-B binding specificity and is composed of the Rel A⅐NF-B1 heterodimer.
Nuclear Translocation of Rel A and NF-B1 Parallel the Biphasic Changes in DNA Binding of the C2 Complex-HepG2 cells fractionated into cytoplasmic and highly purified nuclear extracts (by sucrose cushion centrifugation) were assayed in Western immunoblots for changes in relative abundance of Rel A and NF-B1. The Rel A antibody recognized a single ϳ65-kDa antigen (Fig. 2, arrow) that could be specifically blocked by preadsorbtion using recombinant Rel A protein (not shown). In unstimulated cells, the majority of Rel A was located in the cytoplasmic fraction. By 15 min of rTNF␣ treatment, a slight depletion in the cytoplasmic fraction was noted with a concomitant 3.7-fold increase in nuclear Rel A. At 60 min, nuclear Rel A has diminished to levels approximating control followed by a second increase in Rel A abundance at 120 and 360 min. Changes in nuclear NF-B1 abundance, detected by a specific polyclonal antibody as a 50-kDa band, paralleled those observed for nuclear Rel A (Fig. 2), and for DNA binding of the C2 complex (Fig. 1). These data indicate that rTNF␣ controls cytoplasmic:nuclear positioning of Rel A and NF-B1 in hepatocytes in a biphasic pattern.
Dynamic Expression and Differential Regulation of IB Isoforms in Response to rTNF␣ Treatment-Cytoplasmic extracts of control and rTNF␣-treated HepG2 cells were assayed for the expression and relative changes in IB abundance using antibodies that recognized specific epitopes of IB␣, IB␤, and IB␥ as determined by the appropriate molecular weight and ability of peptide preadsorption to compete for the immunostaining (Fig. 3A). In control cells, 37-kDa IB␣ was abundantly detected, as was 46-kDa IB␤ and 70-kDa IB␥ (Fig. 3B). With rTNF␣ treatment, both IB␣ and IB␤, but not IB␥, disappeared within 15 min of treatment. Abundance of IB␣ returned to a 2-fold greater than control levels at 60 min producing an "overshoot" in its synthesis; by 120 min, IB␣ returned to control levels. In contrast, although 46-kDa IB␤ disappeared simultaneously with IB␣ after TNF␣ treatment, no resynthesis of IB␤ was observed. These data indicate the abundance of IB␣ and ␤ is regulated by rTNF␣ treatment, whereas the abundance of IB␥ is not. Moreover the robust IB␣ resynthesis at 1 h corresponds to the "nadir" of Rel A⅐NF-B1 DNA binding activity and nuclear abundance (cf. Figs  Autoradiogram of competition EMSA using 10 g of 15-min stimulated HepG2 nuclear extract binding to radiolabeled APRE WT in the absence (Ϫ) or presence of 100-fold molar excess of APRE site mutations ("Experimental Procedures"). Location of complexes is located at left. Complexes C1 and C2 compete with wild type APRE, but not mutants Overshoot in Inhibiting Nuclear Rel A Abundance-We noted that the disappearance of nuclear Rel A occurred simultaneously with enhanced IB␣ abundance, indicating that enhanced IB␣ synthesis (overshoot) and reassociation with Rel A may underlie the phenomenon of the 60-min nadir. To directly measure association of Rel A with IB␣, we performed a two-step immunoprecipitation/Western immunoblot using cytoplasmic extracts from HepG2 hepatocytes. In this assay, Rel A complexes are captured and washed under nondenaturing conditions on protein A-agarose beads. After the immune complexes are eluted in denaturing SDS-PAGE loading buffer, IB␣ association is measured by Western immunoblot with IB␣ antibodies (Fig. 4A; Rel A antibody is used as a control for recovery from the immunoprecipitate). Controls demonstrating specificity of the Rel A⅐IB␣ association include: 1) use of preadsorbed Rel A antibody as the primary antibody. As shown in Fig. 4A (lane 2), preadsorbed Rel A antibody does not bring down Rel A nor IB␣; and 2) use of preadsorbed IB␣ antibody in the Western immunoblot to confirm that the 37-kDa antigen is IB␣ (Fig. 4A, lane 1).
In control cytoplasmic extracts, Rel A is associated with IB␣ (Fig. 4A, lane 3), but this association is lost upon 15 min of rTNF␣ treatment (a time coincident with its proteolysis, Fig.  3B). At 30 min (and later times), IB␣ is reassociated with Rel A; we note at 60 min that more IB␣ is associated with Rel A than at other time points, indicating that there is strong association of Rel A with IB␣ (during the nadir in Rel A⅐NF-B1 binding).
The late phase of Rel A induction occurs in the continued presence of IB␣. To exclude the possibility that Rel A is modified in a way that prevents inhibition of its DNA binding activity upon IB␣ binding, purified recombinant human IB␣ (rhIB␣) was added to nuclear extracts prepared from either the early or the late phases of NF-B activation (Fig. 4B). Heat-inactivated IB␣ was used as a control for nonspecific salt or detergent effects in the EMSA. Addition of 200 ng of rhIB␣ completely inhibited DNA binding activity of the C1 and C2 complexes, the latter representing the Rel A:NF-B1 heterodimer, whereas the nonspecific C4 complex was unaffected. These data demonstrate that Rel A⅐NF-B1 complex associated with IB␣ is unable to bind DNA at any phase of its induction.
Our data indicated that enhanced IB␣ synthesis results in inhibition of Rel A⅐NF-B1 binding at 60 min of rTNF␣ stimulation. To confirm this model, EMSA was used to determine the pattern of Rel A⅐NF-B1 binding after IB␣ resynthesis was inhibited using the protein synthesis inhibitor cycloheximide. In the presence of 50 g/ml cycloheximide, the biphasic Rel A⅐NF-B1 binding pattern was abolished and converted into a single monotonic profile as shown in EMSA (Fig. 5). In this same experiment, cytoplasmic IB␣ disappeared at 15 min and was undetectable at 1 h of rTNF␣ treatment assayed by Western immunoblot (data not shown), indicating requirement of new protein synthesis for the nadir in Rel A⅐NF-B1 binding. These data indicate: 1) IB␣ reassociated with the Rel A⅐NF-B1 complex at 60 min; 2) IB␣ is capable of inhibiting DNAbinding of Rel A⅐NF-B1; and 3) IB␣ appearance at 60 min requires new protein synthesis. We conclude that the enhanced IB␣ resynthesis at 60 min (IB␣ overshoot) underlies the biphasic pattern of Rel A⅐NF-B1 translocation, by producing the nadir in C2 binding. IB␤ Proteolysis Is Required for the Late Phase of Rel A⅐NF-B1 Translocation-Previous work has indicated that chymotrypsin-like enzyme(s) mediate IB proteolysis (36), prompting us to examine whether inhibitors could selectively interfere with IB␤ proteolysis so that we could determine its role in biphasic NF-B activation in hepatocytes. The effect of pretreatment with the peptide aldehyde N-acetyl-Leu-Leu-norleucinal (calpain inhibitor I) was determined using Western immunoblot assays of cytoplasmic extracts (Fig. 6). In response to rTNF␣, IB␣ proteolysis was clearly evident at 15 and 30 min. By contrast, IB␤ abundance was similar, or exceeded, control values from 15 to 360 min (compare with Fig. 3B). IB␥ was unaffected throughout the time course of the experiment. We conclude that calpain inhibitor I preferentially blocks IB␤, but not IB␣ proteolysis in hepatocellular cells.
Under conditions where IB␤ proteolysis was preferentially inhibited, we next determined the kinetics of Rel A⅐NF-B1 translocation. The effect of calpain inhibitor I pretreatment was to block completely the second late phase of Rel A⅐NF-B1 induction (Fig. 7A). The early appearance of C2 at 15 and 30 min was attenuated, but not abolished, indicating the contribution of IB␤ proteolysis to the early phase of translocation. Western immunoblots for nuclear Rel A and NF-B1 abundance confirmed that Rel A and NF-B1 were induced corre- sponding to the early but not the late phase of nuclear translocation (compare Fig. 7B (360 min) with Fig. 2 (360 min)). These data indicate calpain inhibitor I markedly affects the biphasic pattern of Rel A⅐NF-B1 induction, converting is induction to a monophasic pattern. These data provide direct evidence for the dynamic effect of IB␣ and IB␤ proteolysis and resynthesis during the biphasic Rel A⅐NF-B1 induction. DISCUSSION An important mechanism in the transcriptional activation of the angiotensinogen gene during the APR is the TNF␣-mediated activation of latent NF-B subunits that bind to the APRE (3, 12, 13). In this study, we have identified mechanistically distinct phases of Rel A⅐NF-B1 nuclear translocation that depend on dynamic changes in individual IB subunit abundance. In this report, we describe the unanticipated and novel observation that rTNF␣ produces a biphasic pattern of Rel A⅐NF-B1 binding. Previous studies have only identified a monophasic induction pattern using 70Z/3 pre-B lymphocytes (24), Jurkat T-cells (38), and U937 macrophages (36) of varying duration. In hepatocytes, the biphasic pattern consists of an Autoradiograms of Western immunoblots using HepG2 cytoplasmic extracts from unstimulated cells probed with preimmune rabbit serum, anti-IB␣, IB␤, and IB␥ primary antibodies, or the same antibodies preadsorbed with respective peptides (PreAd-IB␣, PreAd-IB␤, and PreAd-IB␥). The molecular weight (in kDa) of the protein standards is indicated at the left. IB␣ staining produces a single specific 37-kDa band; IB␤ appears as a 46-kDa band (small arrow); IB␥ is identified as a 70-kDa band (arrow). No NF-B1 precursor is recognized by IB␥ antibody (this antibody detects epitope shared by both isoforms). B, changes in steady-state IB levels in response to rTNF␣ treatment. Western immunoblots from HepG2 cytoplasmic extracts taken from cells treated for indicated times with rTNF␣ (top) and probed with anti-IB␣, IB␤, and IB␥ primary antibodies (left). The abundance of IB␣ is rapidly diminished at 15 and 30 min, followed by enhanced levels (2.1-fold relative to control values) at 60 min, and return to control levels after 120 min. IB␤ staining is detectable in unstimulated cytoplasm and vanishes after 15 min of rTNF␣ treatment. IB␥ abundance is not affected by rTNF␣. FIG. 4. A, rapid reassociation of IB␣ with Rel A during rTNF␣ treatment. Western immunoblot of immunoprecipitates from control and rTNF␣-treated HepG2 cytoplasmic lysates. Antibody used in immunoprecipitation (IP), indicated in top panel, includes Rel A COOHterminal antibody (raised to amino acids 434 -551), or the same antibody preadsorbed with recombinant Rel A (Pre Ad-Rel A). Antibodies used in Western immunoblot (IB) includes Rel A and either IB␣ or IB␣ preadsorbed with recombinant IB␣ (PreAd-IB␣). The locations of Rel A, IgG, and IB␣ are indicated on left. IB␣ staining is dependent on the use of both anti-Rel A in the immunoprecipitation and anti-IB␣ in the immunoblot. IB␣ association with Rel A is lost at 15 min and rapidly returns at 30 and 60 min. Note the maximal amount of IB␣ associated with Rel A occurs at 60 min, the nadir of both Rel A⅐NF-B1 DNA-binding (Fig. 1A) and nuclear abundance (Fig. 2). B, association of IB␣ inhibits Rel A⅐NF-B1 binding taken from early and late phases of induction. Autoradiogram of EMSA using nuclear extracts from rTNF␣treated HepG2 for 15 min (Early Induction) and 120 min (Late Induction) binding radiolabeled APRE WT. Homogenous recombinant polyhistidine-tagged human IB␣ was added to extracts in the indicated amount for 15 min prior to fractionation. Rel A⅐NF-B1 (C2) binding is completely inhibited with 200 ng of rhIB␣. C1 binding (Rel A⅐c-Rel) is weakly affected. Heat inactivated rhIB␣ (Inact) has no effect. early peak (15-30 min), return to control levels (60 min nadir), and a late peak (Ͼ120 min). The early peak is the consequence of simultaneous proteolysis of both the IB␣ and IB␤ subunits (without alterations in IB␥ abundance). Although IB␤ resynthesis is not detectable during the 360-min study, IB␣ resynthesis, by contrast, is rapid. As a result of the overshoot of IB␣ synthesis, IB␣ appears at greater than control levels at 60 min and is apparently responsible for the nadir in Rel A⅐NF-B1 binding. The use of calpain inhibitor I, a protease inhibitor that selectively blocks IB␤ proteolysis, allows us to demonstrate that the second phase of RelA⅐NF-B1 translocation is solely dependent on signal-induced hydrolysis of IB␤.
IB Forms Expressed in Hepatocytes Include Regulated IBs ␣ and ␤ as well as Constitutive IB␥-In non-B lymphocytes, NF-B proteins are sequestered in an inactivated form in the cytoplasm by binding various members of the inhibitory IB protein family. IB␣, -␤, -␥ and NF-B1 p105 are all candidate regulators of Rel A translocation because other studies have shown direct association with Rel A⅐NF-B1 in cellulo and their ability to inhibit RelA⅐NF-B1 DNA binding activity in vitro (23)(24)(25). The pattern of IB expression and modification by signal transduction systems has not been systematically studied in liver cells. Relative levels of mRNA encoding the IB isoforms have been shown to be expressed in a tissue-restricted fashion. For example, IB␤ mRNA is expressed highly in testes (where no IB␣ mRNA is detectable) and IB␣ is expressed more abundantly in spleen and lung than IB␤ (24). Although we have not measured transcript levels, we observe that IB␣ protein appears to be more abundant than IB␤ in hepatocytes. IB␥ is a 70-kDa translation product of a unique mRNA that represents either an alternative splice product or a cryptic promoter from the nf-kb1 gene (25), and is reported to have the most highly restricted pattern of expression previously detected only in mouse lymphoid cell lines (25). Although we have not assayed for expression of the unique IB␥ transcript, we can identify IB␥ expression on the basis of antibody specificity and its appropriate 70,000 molecular weight in hepatocellular cells. In marked contrast to the behavior of IB␣ and IB␤, the steady state abundance of IB␥ appears not to be regulated by rTNF␣ treatment. The presence of constitutive IB␥ may account for the presence of residual Rel A in the cytoplasm of rTNF␣-treated cells at times (15 min) when nearly complete IB␣ and IB␤ proteolysis have occurred, and underscores the concept that there may be pools of Rel A in complex with IB proteins whose abundance are themselves differentially responsive to distinct second messenger pathways. Finally, it is important to emphasize that our observations are made on a cell population and hence represent a statistical average of individual cellular responses to rTNF␣. Based on this experimental design, we therefore cannot differentiate between the following interpretations: 1) two subsets of cells are present in the HepG2 culture, one group expresses IB's ␣ and ␤ and a second group expresses IB␥ only, with only the former population being rTNF␣-responsive; versus 2) a homogenous cell population is present in the HepG2 culture that expresses all three IB isoforms. Although this cell population is rTNF␣responsive, the abundance of individual IB isoforms are controlled by different intracellular signaling pathways. Additional studies, at individual cell resolution, will be required to differentiate between these possibilities.
Mechanisms of TNF␣-induced Changes in IB Abundance-TNF␣ activates Rel A translocation by signal-induced modifications of the IB inhibitor. TNF␣-activated intracellular signals induces phosphorylation of IB␣ at NH 2 -terminal serine residues (amino acids 32 and 36 (28,30)), by a ubiquitinationdependent kinase, a process that subsequently targets IB␣ for proteolysis (31,32). Phospho-IB␣ migrates at a distinct position on SDS-PAGE gels (ϳ3 kDa larger), allowing for its identification (23,28,30,32). Although we have not directly demonstrated inducible phosphorylation of IB in HepG2 cells, indirect evidence for IB␣ phosphorylation is seen in Western immunoblots of cytoplasmic extracts from rTNF␣-treated cells where the slower phospho-IB␣ migrating species is seen with longer exposures (data not shown). Other studies have shown that phospho-IB␣ is itself rapidly polyubiquitinated (Ub n ) at lysine residues 21 and 22 (31) and subsequently proteolyzed through the 26 S proteasome pathway (29,32). The 26 S proteasome, a 700-kDa protease complex, is known to degrade Ub n -conjugated proteins in an ATP-dependent fashion, but the protease(s) involved and their specificity are incompletely characterized (37).
In epithelial cells and lymphocytes, serine protease inhibitors (L-1-tosylamido-2-phenylethyl chloromethyl ketone) and the peptide aldehyde (calpain inhibitor I) are effective in blocking IB␣ proteolysis at concentrations that interfere with proteasome activity (31,37). Based on this inhibitor sensitivity profile, the enzyme(s) in the 26 S proteasome complex mediating IB␣ hydrolysis has been characterized as chymotrypsinlike. In this regard, we are surprised to find in hepatocytes, that calpain inhibitor I blocks IB␤ proteolysis relatively selectively. These data indicate a potential involvement of celltype specific factors in the inducible proteolysis of IB␣.
Overshoot of IB␣ Synthesis: Evidence for Exaggerated Rel A⅐IB␣ Autoregulatory Pathway in Hepatocytes-Upon Rel A translocation into the nucleus, the synthesis of IB␣ is activated (38,39). Newly resynthesized IB␣ reassociates with Rel A, inactivating its nuclear location and transcriptional effect resulting in an autoregulatory feedback pathway. The important role of IB␣ in terminating Rel A⅐NF-B1 activation is illustrated in the persistent NF-B activation upon rTNF␣ stimulation in fibroblasts cultured from mice with homozygous deletion of ib␣ (40). The surprising observation in our studies is that, although IB␣ is involved in terminating NF-B action, it does so only transiently by overshoot resynthesis. That IB␣ resynthesis results in a transient inhibition of Rel A⅐NF-B1 DNA binding activity is supported by the following observations: 1) increased IB␣ abundance by Western immunoblot (Fig. 3) is coincident with inhibition of Rel A⅐NF-B1 binding at 60 min (Fig. 1A). 2) Increased IB␣ is associated with Rel A by coimmunoprecipitation assays (Fig. 4A). 3) Rel A⅐NF-B1 DNA binding activity from both early and late phases of induction is completely inhibited by rhIB␣ (Fig. 4B). 4) Rel A⅐NF-B1 inhibition at 60 min is dependent on new protein synthesis (Fig. 5). At present, we are unable to explain why the increased IB␣ levels at 60 min return to control levels at 120 min; this phenomenon is probably the result of accelerated turnover of IB␣ in the presence of TNF␣ (as for LPS activation in 70/Z3 cells (23)).
Late Phase Translocation of Rel A⅐NF-B1 Is Dependent on IB␤ Proteolysis-In Jurkat T-lymphocytes (24) and in mouse embryo fibroblasts (4), rTNF␣ produces only IB␣, but not IB␤ proteolysis, associated with a transient monotonic induction of Rel A⅐NF-B1 binding. In hepatocytes, by contrast, rTNF␣ produces a biphasic, prolonged induction of Rel A⅐NF-B1. A requirement for IB␤ proteolysis in producing the late phase of Rel A⅐NF-B1 translocation is seen under conditions where IB␤ proteolysis is abolished. Earlier reports in other cell lines have shown that ib␤ is not subject to the Rel A-induced autoregulatory pathway, and once proteolyzed, is not resynthesized during the time course of these experiments (24); our data are consistent with this phenomenon. IB␤ proteolysis apparently allows for a prolonged nuclear response to the actions of TNF␣ in a cell restricted pattern. Based on their similar domain organizations and similar rapid proteolytic response to TNF␣, it might be expected that IB␣ and IB␤ would be proteolyzed through similar pathways. However, several features in HepG2 cells indicate that the mechanism for signal-induced proteolysis of IB␤ is distinct from that used for IB␣. Phosphorylated IB␣ isoforms migrate more slowly that nonphosphorylated forms (28,30), allowing their identification by Western immunoblot. In calpain inhibitor I-treated cells, rTNF␣-inducible slower migrating species of IB␣ corresponding to phosphorylated forms can be detected upon longer exposure. No such forms are seen with IB␤ (Fig. 6, middle panel), even under conditions when complete inhibition of proteolysis is observed. Second, and more importantly, proteolysis of IB␣ is insensitive to 200 M calpain inhibitor I whereas proteolysis of IB␤ is completely blocked by this concentration. These data probably indicate that rTNF␣ activates IB proteolysis through two distinct pathways.
In summary, our studies have focused on the mechanisms for the novel observation of a biphasic induction of Rel A⅐NF-B1 binding in hepatocytes. Rel A⅐NF-B1 is associated with three IB isoforms that are differentially regulated by rTNF␣, producing pools of constitutive and inducible NF-B complexes. We provide evidence for how dynamic changes in IB␣ abundance influences the two distinct phases of Rel A⅐NF-B1 translocation. Moreover, we report that separate proteolytic pathways are involved in IB␣ and IB␤ degradation in hepatocytes producing unanticipated complexity in signal-induced regulation of IB abundance.