Transactivation of the DNA-binding proteins nuclear factor-κB (NF-κB) and activator protein (AP)-1 by de novo oxyradical generation is a stereotypic redox-sensitive process during hypoxic stress of the liver. Systemic trauma is associated with splanchnic hypoxia-reoxygenation (H/R) followed by intraportal gram-negative bacteremia, which collectively have been implicated in posttraumatic liver dysfunction and multiple organ damage. We hypothesized that hypoxic stress of the liver before stimulation by Escherichia coli serotype O55:B5 (EC) amplifies oxyradical-mediated transactivation of NF-κB and AP-1 as well as cytokine production compared with noninfectious H/R or gram-negative sepsis without prior hypoxia. Livers from Sprague-Dawley rats underwent perfusion for 180 min with or without 0.5 h of hypoxia (perfusate Po2, 40 ± 5 mmHg) followed by reoxygenation and infection with 109 EC or 0.9% NaCl infusion. In H/R + EC livers, nuclear translocation of NF-κB and AP-1 was unexpectedly reduced in gel shift assays vs. normoxic EC controls, as were perfusate TNF-α and IL-1β levels. Preceding hypoxic stress paradoxically increased postbacteremic reduced-to-oxidized glutathione ratios plus nuclear localization of IκBα and phospho-IκBα, but not JunB/FosB profiles. Notably, xanthine oxidase inhibition increased transactivation as well as cytokine production in H/R + EC livers. Thus brief hypoxic stress of the liver before intraportal gram-negative bacteremia potently suppresses activation of canonical redox-sensitive transcription factors and production of inflammatory cytokines by mechanisms including xanthine oxidase-induced oxyradicals functioning in an anti-inflammatory signaling role. These results suggest a novel multifunctionality of oxyradicals in decoupling hepatic transcriptional activity and cytokine biosynthesis early in the posttraumatic milieu.
- liver-lung interactions
- complex systems
- transcription networks
- transcription factors
- gram-negative bacteremia
- posttraumatic sepsis
- transcription protein/DNA array
- gene regulation
- reactive oxygen species
- multiple organ failure
acute lung injury and multiple organ dysfunction are major life-threatening sequelae of severe trauma (24, 36). Early after systemic injury, reductions in O2 availability to the hepatosplanchnic region are particularly common (7, 35). Such hypoxic stress may enhance translocation of enteric gram-negative bacteria (31), culminating in portal bacteremia, microbial stimulation of hepatic Kupffer cells, and enhanced export of inflammatory cytokines by the liver, most notably TNF-α and IL-1β (8, 12, 36). Nevertheless, the pathogenesis of posttraumatic septic complications is poorly understood, from the standpoint of early posthypoxic transcriptional events in the liver and their modification by subsequent portal bacteremia.
Common to hypoxic stress of mammalian cells or their stimulation by gram-negative microbial products is the de novo generation of oxyradicals from mitochondrial and xanthine oxidase-derived origins (4, 6, 14, 28). Oxyradical-mediated signaling activates canonical oxidation-reduction (redox)-sensitive transcription factors, such as the NF-κB/c-Rel and activator protein (AP)-1 families of DNA-binding proteins, followed by their nuclear localization and binding to specific cis-acting motifs in the promoter regions of inflammatory cytokine genes (5, 15, 30). Oxyradical generation and signaling in the intact liver are well documented after experimental perturbations in the hepatic O2 supply, resulting in hypoxic stress (14, 28), or challenge by gram-negative stimuli (4, 33). Consequently, it has been assumed, but not proven, that combined or sequential hypoxic stress of the liver and intraportal bacteremia additively promote oxyradical-mediated transactivation of NF-κB and AP-1 complexes.
Complex biological systems, in general, share fundamental characteristics. A major feature is the potential for decoupling of regulatory systems by nonlinear behavior of independent elements as multiple stimuli are applied in a time- or phase-varying manner (13). Redox-sensitive hepatic transcriptional networks involving NF-κB and AP-1 multistep activation, nuclear translocation, and DNA binding surely fulfill such complexity criteria (1, 9, 15, 23, 26). Considering the diversity of integrated stress responses initiated by de novo oxyradical generation, host calibration of transactivation and subsequent cytokine biosynthesis may dynamically vary with the intensity or coupling frequency of oxidative stimuli (13). By extension, we reasoned that redox-sensitive transcriptional activation of NF-κB and AP-1 in the liver may diverge from predicted responses when oxyradical-mediated signaling from hypoxic stress and gram-negative bacterial stimulation are temporally linked, as may occur after traumatic injury (24, 31, 36).
We therefore performed these studies to test the hypothesis that brief and otherwise well-tolerated hypoxic stress of the liver, followed by reoxygenation and intraportal Escherichia coli bacteremia, amplifies transactivation of NF-κB and AP-1 by a xanthine oxidase-derived oxyradical signaling mechanism (14, 19). As a corollary, we postulated that hypoxia would modulate subsequent bacteremic transcriptional events and proportionately increase downstream cytokine production, as reflected by increases in TNF-α and IL-1β in hepatic venous effluent. Because intrahepatic oxyradical generation during combined hypoxic stress and bacterial stimulation may induce cysteine oxidation within the DNA-binding domains of NF-κB, Jun, and Fos (1, 20, 25, 26), we furthermore assessed transcriptional responses in relation to the concomitant equilibrium between reduced glutathione (GSH) and glutathione disulfide (GSSG), the most abundant and thoroughly characterized intracellular redox couple (9).
MATERIALS AND METHODS
Double-stranded consensus oligonucleotides were purchased from Promega (Madison, WI); antibodies to specific transcription factor subunits, IκBα and phospho-IκBα (Ser32), and human recombinant IκBα from Santa Cruz Biotechnology (Santa Cruz, CA); and [γ-32P]ATP from Amersham Biosciences (Piscataway, NJ). Molecular biology-grade reagents (Sigma Chemical, St. Louis, MO) were used for nuclear isolation, electrophoretic mobility shift assay (EMSA), and supershift assay procedures (19). The transcription protein/DNA array for NF-κB and AP-1 was obtained from Panomics (Redwood City, CA). Sodium taurocholate and cell culture-grade chemicals for liver perfusions were purchased from Sigma Chemical unless otherwise noted. Allopurinol (Allo; USP grade) was a gift from Burroughs-Wellcome (Research Triangle Park, NC).
Pathogen-free adult male Sprague-Dawley rats (Harlan, Indianapolis, IN) weighing 260–325 g were allowed free access to food and water before experiments. Studies were performed according to National Institutes of Health guidelines and were approved by the Animal Care Committee of Saint Louis University.
E. coli cultures.
Strain 12014 of E. coli (EC) serotype O55:B5 (American Type Culture Collection, Bethesda, MD) was maintained in trypticase soy broth and grown up over 18–24 h in fresh cultures (10, 19, 22). Organisms sedimented at 1,000 g for 10 min at 4°C were washed twice in sterile 0.9% NaCl (NS) and resuspended in NS to 1 × 109 colony-forming units (cfu) in 1 ml. Freshly prepared inocula were kept at 4°C until use. Confirmation of infused inocula by quantitative streak-plate cultures (37°C, 24 h) yielded monomorphic colonies averaging 9.2 ± 0.8 × 108 (SE) cfu/ml in all EC experimental groups.
Ex situ liver perfusion.
Perfusates were freshly prepared in pyrogen-free glassware as previously described (10, 19, 22) by addition of 10 mM glucose, 5% (vol/vol) sterile rat serum, sodium taurocholate (50 μM), 105 U of penicillin, and 100 mg of streptomycin to 1 liter of Krebs-Henseleit Ringer bicarbonate (297 ± 2 mosM final osmolarity, pH 7.4). After equilibration with 95% O2-5% CO2 gas, perfusate arterial Po2 values were 550–620 mmHg (IL-1306 blood gas analyzer, Instrumentation Laboratories, Lexington, MA). Perfusates were filtered (0.22 μm) and cultured on nutrient agar (37°C, 24 h) to confirm sterility before each experimental run and were again cultured after circuit priming (125 ml) before liver harvesting to confirm circuit sterility before each experiment. Perfusate endotoxin levels were ≤10 pg/ml by the Limulus assay (QCL-1000, Whittaker Bioproducts, Walkersville, MD).
After anesthesia was induced in rats with pentobarbital sodium (50 mg/kg ip), aseptic organ harvesting was performed (10, 19, 22). Livers were equilibrated for 30 min before the end of each experiment at time 0 to ensure steady-state conditions of hepatic O2 consumption (V̇o2) and portal venous pressure. Supplemental NaHCO3 was added as needed to maintain arterial pH at 7.36–7.44. Ex situ perfusions were performed in the recirculating mode in a temperature-controlled apparatus (37.0 ± 0.5°C) at a flow rate of 3.75 ml·min−1·g liver−1 (19, 22), which was maintained for the remainder of experiments.
Baseline portal (afferent) and hepatic venous (efferent) samples were obtained at time 0. To assess the influence of organ harvesting and 30 min of perfusion without hypoxia or infection on transactivation of hepatic DNA-binding proteins and to serve as reference for subsequently timed nuclear samples, three livers were immediately snap frozen in liquid N2 at time 0. Normoxic EC control livers (n = 10) were intraportally infected with 1 × 109 cfu of viable EC over 2–3 min, and normoxic NS control organs (n = 7) received isovolumetric NS. These control livers were then perfused under normoxic conditions and harvested at 180 min. In certain experiments, other EC and NS control livers were harvested to assess postbacteremic transactivation of NF-κB and AP-1 at 60 min (n = 3 each).
Sequential hypoxic stress followed by reoxygenation and gram-negative bacterial stimulation of other livers was induced by rapid flushing of the circuit reservoir with 95% N2-5% CO2 at a flow rate of 10 l/min for 5 min commencing at time 0 for 0.5 h. Resulting perfusate arterial Po2 values verified in each preparation averaged 42 ± 7 mmHg within 5 min, representing a constant-flow reduction in the hepatic O2 supply of > 90%. At 30 min, portal and hepatic venous perfusate samples were again obtained. The perfusion reservoir was then flushed with 95% O2-5% CO2 gas for reoxygenation, and livers were infected as in normoxic EC controls with 1 × 109 cfu of EC in the hypoxia-reoxygenation (H/R) + EC group (n = 6) or received NS in the H/R + NS group (n = 5). Normoxic perfusion was maintained for the remainder of experiments until 180 min of posthypoxic EC or NS exposure had elapsed. In certain experiments, additional posthypoxic EC- and NS-challenged livers were harvested 60 min later (n = 3 each) for time-matched comparisons of transcription factor activation with their respective normoxic controls.
We next evaluated the effects of inhibiting formation of xanthine oxidase-derived oxyradicals generated during combined hypoxic stress and E. coli bacteremic stimulation compared with E. coli infection under normoxic conditions on hepatic NF-κB and AP-1 transactivation and cytokine production. The xanthine oxidase inhibitor Allo was given to rats as 50 mg/kg by gavage 18 h before liver harvest, and an additional 3 mg/kg were given intravenously just before liver harvest (19). Allo was also added to initial perfusates to achieve a final concentration of 500 μM in these Allo + H/R + EC (n = 5), normoxic Allo + EC (n = 4), and Allo + H/R + NS (n = 4) livers.
For each liver, paired portal and hepatic venous perfusate samples were obtained at 0, 30, 60, 90, 120, 150, and 180 min for determination of pH, Po2, and Pco2; then fresh perfusate was isovolumetrically added. Sodium taurocholate (50 μM) was given every 30 min to replace bile acids. The hepatic V̇o2 (in μmol·min−1·g−1) was calculated as previously described (22). The number of circulating EC (in cfu/ml) in hepatic venous perfusates was determined at each time point by duplicate nutrient agar streak plates. Additional aliquots of venous perfusate at these time points were immediately placed on ice and stored at −70°C until analyses for TNF-α and IL-1β. Viability of the preparations was continually assessed by sequential determinations of aspartate aminotransferase (AST) levels (Sigma) and hepatic V̇o2 ≥1.0 μmol·min−1·g−1 after hypoxic stress. At 180 min, we also evaluated receptor-mediated glycogenolysis using glucagon (10−7 M final concentration) for analysis of incremental V̇o2 responses and histological examination of formalin submersion-fixed liver tissue samples after hematoxylin and eosin staining (10, 22).
Hepatic nuclear extraction.
Standardized liver sections were snap frozen in liquid N2 at time 0, 60 min after E. coli bacteremic stimulation, or at the conclusion of experimental runs and stored at −70°C. Nuclear extracts were prepared using a modification of the protocol described by Essani et al. (11). Briefly, 0.2–0.4 g of frozen liver was homogenized (EMI, Clinton, CT) in 1.5 ml of ice-cold buffer containing 10 mM HEPES (pH 7.9), 1.5 mM MgCl2, 10 mM KCl, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10 μg/ml each of the protease inhibitors pepstatin, aprotinin, and leupeptin, and 5 mM α-glycerophosphate. After a 10-min incubation on ice, homogenates were centrifuged for 10 min at 850 g at 4°C. Supernatants were aliquoted and stored at −70°C for protein studies. The pellet was resuspended in 0.75 ml of ice-cold buffer containing the above reagents with 0.1% Triton X-100 and reincubated for 10 min. Samples were then vortexed and recentrifuged as described above, the supernatant was again decanted, and the nuclear pellet was resuspended in 50 μl of an ice-cold buffer containing 20 mM HEPES (pH 7.9), 25% glycerol (vol/vol), 0.42 M NaCl, 1.5 mM MgCl2, 0.2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 mM α-glycerophosphate, and 10 μg/ml each of pepstatin, aprotinin, and leupeptin. After incubation on ice for 30 min, samples were centrifuged at 14,000 g for 30 min at 4°C. Supernatants containing the nuclear proteins were aliquoted and stored at −70°C. To avoid denaturation from repeated freeze-thaw cycles, each thawed aliquot was used only once.
EMSA and protein/DNA array assays for NF-κB and AP-1 complexes.
Nuclear protein concentrations were determined by the bicinchoninic acid assay (Pierce, Rockford, IL) with bovine serum albumin as the standard. Nuclear protein (10 μg) was incubated in binding buffer (10 mM Tris, pH 7.5, 1 mM EDTA, 5 mM MgCl2, 25 mM NaCl, 5% glycerol, 5% sucrose, and 0.01% Nonidet P-40) for 10 min at 37°C. Poly(dIdC) (3 μg; Pharmacia Biotech, Piscataway, NJ) was added to samples before addition of the oligonucleotide probe. Double-stranded consensus oligonucleotides for the transcription factors NF-κB (5′-AGTTGAGGGACTTTCCCAGGC-3′) and AP-1 (5′-CGCTTGATGAGTCAGCCGGAA-3′) (Promega) were end labeled with [γ-32P]ATP using T4 polynucleotide kinase, and ∼1 × 105 counts/min were added to reaction mixtures. For competition studies, 100-fold excess of unlabeled NF-κB or AP-1 oligonucleotide or noncompetitive mutant oligonucleotides [5′-AGTTGAGGCGACTTTCCCAGGC-3′ (NF-κB mutant) and 5′-CGCTGATATTGGCGGAA-3′ (AP-1 mutant)] was added before addition of the radiolabeled probe.
For supershift analyses of NF-κB, a 100-fold excess of antibodies cross-reactive to the rat p65 subunit of NF-κB was added to the reaction mixture. The AP-1 transcription factor complex consists of c-Jun, JunB, or JunD homodimers or Jun/Fos complexes binding to the palindromic sequence TGA/(G)TCA (30). Accordingly, 100-fold excess of antibodies cross-reactive to the Jun family elements c-Jun, JunB, JunD, and Fos gene family proteins c-Fos, FosB, and Fra-1 was used for supershift studies. After addition of 1 μl of 10× gel loading buffer to the reaction mixture, samples were run through a 4% acrylamide-bis gel in 1× running buffer (0.025 M Tris and 0.2 M glycine) at 250 V for 3–4 h in a 4°C room. Gels were vacuum-dried and exposed to X-ray film (Hyperfilm MD, Amersham) for 24–72 h at −70°C; then autoradiographs were developed. Individual bands were quantitated densitometrically over a linear range (Molecular Dynamics, Sunnyvale, CA), normalized to the signal of recombinant p50 (1 μl of a 0.135 gel shift units/μl solution, Promega), which was concomitantly loaded on each gel an as internal loading control, and averaged for subsequent analysis.
For independent confirmation of the results of gel shift assays, transcription factor protein/DNA array analyses were also performed on the same time-matched EC and H/R + EC nuclear extracts by incubation of the extracts with biotin-labeled DNA-binding oligonucleotides corresponding to NF-κB and AP-1 consensus-binding sequences according to the manufacturer's instructions (TranSignal Protein/DNA Array I, Panomics).
Western blot analyses for IκBα.
Whole liver cytoplasmic and nuclear extracts (300 μg) from at least three livers per group prepared as described above with added 1:100 phosphatase inhibitor cocktail I (catalog no. P-2850, Sigma) containing microcystin LR, cantharidin, and (−)-p-bromotetamisole were electrophoresed on 16 × 16 cm 12% SDS-polyacrylamide gels, transferred to semidry nitrocellulose membranes, and reacted with rabbit IgG polyclonal antibodies specific for IκBα or phospho-IκBα (Ser32), with human recombinant IκBα used as a positive control. Blots were reacted to goat anti-rabbit IgG horseradish peroxidase conjugate (200 μg/ml, 1:10,000 dilution), and proteins were detected by enhanced luminescence using Supersignal West Pico (Pierce). As controls, blots were reacted against a rabbit polyclonal IgG antibody to β-actin and a biotinylated avidin-horseradish peroxidase antibody (Bio-Rad, Hercules, CA). Densitometric evaluation of blots was performed as described above.
Liver tissue samples for GSH analysis were immediately homogenized in 5% trichloroacetic acid-0.01 N HCl at a weight-to-volume ratio of 1:10 as previously described (19, 22) and stored at −70°C until analyzed. Total glutathione, including GSH and GSSG, was analyzed by a modified enzymatic recycling procedure described by Tietze (34). Briefly, GSH oxidized by 5,5′-dithio-bis-2-nitrobenzoic acid forms GSSG and 5-thio-2-nitrobenzoic acid. GSSG is then reduced by NADPH and GSSG reductase back to GSH. The rate of 5-thio-2-nitrobenzoic acid production, representing original GSH-GSSG partitioning in samples, is measured spectrophotometrically at 412 nm. Standard curves were generated with known amounts of GSH to derive original sample concentrations of total glutathione. Results are expressed as the average of a minimum of three determinations.
TNF bioactivity in venous perfusates was quantitated using mycoplasma-free, actinomycin D-treated murine L929 cells (American Type Culture Collection) in duplicate over a fourfold dilution range on the basis of measurement of cytotoxicity at 550 nm (10, 19). Results were calibrated with standard curves on each plate with use of murine recombinant TNF-α (specific activity ≥5 × 107 U/mg; Genzyme, Cambridge, MA). Internal controls were spiked with recombinant murine TNF-α to assess recovery. No interference was observed when Allo was added over a proportional concentration range.
Venous perfusate samples were assessed in duplicate for IL-1β by an ELISA sensitive over 0–960 pg/ml protein (BioSource International, Camarillo, CA) with use of a monoclonal anti-IL-1β capture antibody with absorbance measured at 450 nm.
Serial changes in within-group variables were analyzed by a repeated-measure analysis of variance, with post hoc pairwise comparisons performed when appropriate by a Newman-Keuls test (22). Between-group comparisons of time- and group-specific variables were performed using a Kruskal-Wallis analysis, with pairwise comparisons made when appropriate using a two-tailed paired t-test. Significance was accepted at P < 0.05. Values are means ± SE.
Hypoxia suppresses postbacteremic NF-κB transactivation.
Gel shift assays specific for NF-κB performed on nuclear lysates from normoxic EC control livers exhibited a >75% mean increase in postbacteremic transactivation of DNA-binding activity composed of p50 and p65 subunits over 3 h compared with NS control organs, in which signals did not differ from time 0 levels. However, brief hypoxic stress preceding subsequent E. coli bacteremic stimulation strongly suppressed, rather than enhanced, NF-κB activation in H/R + EC livers (Fig. 1). Thus mean DNA binding of NF-κB was 45% less in H/R + EC livers than in time-matched normoxic EC control organs and remained similar to baseline unstimulated (time 0) livers (P < 0.05). This potent suppressive effect of combined hypoxic stress and E. coli stimulation on the activation and nuclear translocation of NF-κB in H/R + EC livers contrasted with the robust hypoxic stress-induced transactivation of NF-κB without sepsis in H/R + NS livers (Fig. 1). Therefore, two potent oxidative stimuli, each of which separately enhanced hepatic DNA binding of NF-κB in this model, reduced postbacteremic NF-κB transactivation when combined.
Hypoxic suppression of postbacteremic NF-κB transactivation is reversed by xanthine oxidase inhibition.
Xanthine oxidase-derived oxyradicals are abundantly generated during severe H/R stress of the liver without microbial exposure (14, 28). Hepatic oxyradical generation likewise follows gram-negative bacterial stimulation without preceding hypoxic stress (4, 33). Considering the anomalous reduction of NF-κB transactivation in H/R + EC livers, we reasoned that additive generation of oxyradicals from the combination of stimuli might be responsible. Accordingly, we examined the effects of the anti-inflammatory compound Allo on hepatic nuclear localization and DNA binding of NF-κB in H/R + EC livers. Figure 1 shows that xanthine oxidase inhibition in these Allo + H/R + EC organs substantially increased postbacteremic NF-κB transactivation, despite preceding hypoxic stress. Densitometric averaging of EMSA signals confirmed a ∼50% mean increase over time-matched values of H/R + EC livers. In agreement with the well-characterized anti-inflammatory effects of xanthine oxidase inhibition under conventional single-stimulus conditions of gram-negative bacterial infection in particular or of H/R (5), EMSA signals were reduced in Allo + EC or Allo + H/R + NS livers, respectively (Fig. 1). These results indicate a paradoxical reduction in proinflammatory NF-κB transactivation after the liver is exposed to dual oxidative stimuli and an increase, rather than the expected attenuation, of postbacteremic NF-κB EMSA signals when generation of xanthine oxidase-derived oxyradicals in H/R + EC livers is reduced by Allo pretreatment.
Sequential hypoxic stress and E. coli infection paradoxically increase the hepatic GSH-GSSG equilibrium.
Baseline total hepatic GSH + GSSG concentration in freshly harvested and perfused livers of nonfasted rats at time 0 was 4.56 ± 0.3 μmol/g, similar to previous findings (19, 22). Mean GSSG level (23.2 ± 0.007 nmol/g) in these surgical controls was increased compared with normoxic NS controls after 180 min of perfusion (Table 1), most likely from incomplete resolution of early surgical oxidative stress. Nevertheless, mean GSH-to-GSSG ratios in time 0 livers and normoxic NS controls were equivalent (196 vs. 218, respectively).
The effects of preceding hypoxic stress and intraportal E. coli infection on the hepatic GSH-GSSG equilibrium after 180 min of perfusion, with and without prior xanthine oxidase inhibition with Allo, are summarized in Table 1. The lowest GSH:GSSG ratio, indicative of the highest degree of oxidative stress, was found in normoxic EC control livers. By contrast, GSSG concentrations exhibited a 2.4-fold reduction in H/R + EC organs (P < 0.05 vs. normoxic EC controls). The corresponding group-specific mean GSH:GSSG ratio was likewise higher when brief hypoxic stress preceded E. coli infection than when infection occurred under sustained normoxic conditions (Table 1). Thus the mean GSSG values and corresponding GSH:GSSG ratios of H/R + EC livers approximated those of normoxic NS controls. After xanthine oxidase inhibition in Allo + H/R + EC livers, neither total GSH + GSSG levels nor hepatic GSSG concentrations differed from H/R + EC or normoxic EC control values (Table 1). Total GSH + GSSG levels, GSSG concentrations, or GSH:GSSG ratios also did not change over time after hypoxic stress and reoxygenation in NS livers, with or without corresponding Allo pretreatment (Table 1).
Sequential hypoxic stress and E. coli infection elevate cytoplasmic and nuclear IκBα concentrations.
We next investigated whether phosphorylation and degradation of the cytoplasmic NF-κB inhibitor IκBα preceded suppression of hepatic NF-κB nuclear transactivation in H/R + EC livers and whether hypoxia-related increases in nuclear IκBα were associated with reduced postbacteremic NF-κB binding to DNA. Representative Western blot results for cytoplasmic and nuclear IκBα and their respective phosphoactive forms in H/R + EC livers 60 min after bacterial stimulation are shown in Fig. 2 compared with time-matched normoxic EC controls.
Notably, mean cytoplasmic IκBα densitometric signals increased more than threefold and phospho-IκBα values increased more than twofold in H/R + EC vs. normoxic EC livers (Fig. 2). Moreover, nuclear IκBα densitometric signals rose strikingly in H/R + EC livers, while phospho-IκBα values increased a modest 2.8-fold, compared with their respective signals in normoxic EC control organs. Collectively, these results suggest that brief preceding hypoxia of the liver interferes with neither E. coli bacteremia-induced phosphorylation of IκBα in the cytoplasmic compartment nor with the nuclear transport of IκBα or its phosphoactive moiety, to account for the suppression of NF-κB transactivation in H/R + EC livers.
Hypoxic stress suppresses postbacteremic AP-1 activation.
We next sought to determine whether the anomalous hypoxic suppression of postbacteremic NF-κB transactivation in H/R + EC livers was generalizable to other redox-sensitive transcription factors. The EMSA results for the AP-1 complex in perfused rat livers are shown in Fig. 3. Supershift analysis for Jun family elements indicated a predominance of JunB subunits in all experimental groups, with lesser amounts of c-Jun (Fig. 3). The primary Fos gene family element in hepatic AP-1 complexes was FosB. There was no evidence of c-Fos subunit binding.
AP-1 transactivation after normoxic EC infection exhibited a 75% mean increase over densitometrically averaged baseline (time 0) signals. Hypoxic stress and reoxygenation alone without subsequent infection in NS + H/R organs also increased AP-1 transactivation by ∼50%. Nevertheless, prior hypoxic stress in H/R + EC livers suppressed postbacteremic AP-1 transactivation and binding to DNA consensus motifs in nuclear isolates to a level equivalent to that in normoxic NS controls. Xanthine oxidase inhibition in Allo + H/R + EC livers resulted in a 20% mean increase in AP-1 transactivation. Thus Allo pretreatment resulted in a >66% mean increase in AP-1 transactivation over baseline. By supershift analyses, we found no evidence to implicate an altered profile of Jun/Fos subunit activation or combinatorial binding in H/R + EC livers compared with time-matched EC controls.
These results for AP-1 and NF-κB were separately confirmed in the protein/DNA array analyses, where reduced transactivation of each transcription factor was found in nuclear lysates of H/R + EC livers compared with EC controls. By these arrays, signals for NF-κB were increased 8.2-fold and AP-1 by 16.7-fold at 180 min compared with time 0 values in normoxic EC controls. In contrast, similarly timed signals in H/R + EC livers were not different from baseline values, with NF-κB values being only 0.26-fold different from time 0 values and changes for AP-1 averaging 0.26-fold.
Hypoxic suppression of postbacteremic NF-κB and AP-1 transactivation is associated with reduced hepatic cytokine secretion.
Bioactive TNF-α concentrations in venous perfusates of H/R + EC livers were reduced at 180 min compared with time-matched normoxic EC controls (46 ± 9 vs. 181 ± 52 U/ml, P < 0.05; Fig. 4). As for NF-κB and AP-1 transactivation, reductions in postbacteremic TNF-α secretion by brief prior hypoxia were completely abrogated by xanthine oxidase inhibition in Allo + H/R + EC livers [216 ± 56 U/ml at 180 min, P = not significant (NS) vs. time-matched normoxic EC controls].
Peak hepatic secretion of immunoreactive IL-1β was likewise suppressed in H/R + EC livers compared with time-matched normoxic EC controls (265 ± 38 pg/ml at 180 min vs. 710 ± 190 pg/ml, P < 0.001; Fig. 5). IL-1β secretion did not however, rise with xanthine oxidase inhibition in Allo + H/R + EC organs at 180 min (359 ± 273 pg/ml) compared with time-matched H/R + EC values.
Hypoxic stress does not alter postbacteremic E. coli clearance or liver function.
Hypoxic suppression of postbacteremic NF-κB and AP-1 transactivation and cytokine production were not associated with changes in the kinetics of E. coli clearance from perfusates. After intraportal infection, peak colony-forming units per milliliter in perfusates occurred at 30 min in all E. coli-infected livers, regardless of preceding hypoxic stress or xanthine oxidase inhibition. By 120 min, complete clearance of bacterial inocula was observed in all groups (data not shown). Enzymatic assessment of liver function by release of AST into perfusates did not differ among E. coli-infected groups at any time. Peak AST values at 180 min were 330 ± 64 U/l in normoxic EC controls vs. 416 ± 57 U/l in H/R + EC livers and 465 ± 122 U/l in Allo + H/R + EC organs (P = NS). Similarly, there were no differences in AST values between NS controls and H/R + NS livers (353 ± 24 and 268 ± 50 U/l, respectively, P = NS). Additional assessments of liver function by comparison of group-specific biliary flow rates and examination of acinar histological integrity by light microscopy yielded similar results (not shown).
Hepatic V̇o2 and its stability over time represent established additional indexes by which to confirm the viability of perfused livers (10, 19, 22, 35). Hepatic V̇o2 values were similar at baseline in all preparations studied here (Fig. 6) and remained stable thereafter in normoxic EC and NS control livers. Reductions in hepatic O2 availability induced by constant-flow hypoxic stress in H/R + EC and H/R + NS livers and in their respective Allo-pretreated counterparts averaged 92 ± 2%. Hepatic V̇o2 decreased in parallel. However, utilization of O2 by the liver promptly returned to prehypoxic levels during the reoxygenation phase in all groups (Fig. 6), irrespective of Allo pretreatment. Hepatic oxidative responses to 10−7 M glucagon, defined as a >5% increase in V̇o2 (35), were also preserved in H/R groups compared with organs perfused under normoxic conditions. Thus a 23 ± 3% rise in V̇o2 followed glucagon stimulation in H/R + EC livers as well as in Allo + H/R + EC and H/R + NS preparations.
In these studies, we have defined a novel multiagent interaction whereby brief and otherwise well-tolerated hypoxic stress of the liver strongly suppresses E. coli-induced transactivation of redox-sensitive DNA-binding proteins as well as subsequent hepatic secretion of inflammatory cytokines. Remarkable features of this interaction, which parallels the posttraumatic environment, include its divergence from responses anticipated by two established oxidative stimuli arriving at the liver in succession. Thus knowledge generalized from less complex in vitro systems or from conventional single-stimulus experiments using only H/R or gram-negative bacterial stimulation would predict additive de novo oxyradical generation and, consequently, enhanced redox-sensitive transactivation and DNA binding. However, juxtaposing antecedent hypoxic stress with intraportal E. coli bacteremia led to directionally opposite effects on postbacteremic transactivation of NF-κB and AP-1 (Figs. 1 and 3). Corresponding to nonlinear behavior typifying complex adaptive systems (13), xanthine oxidase-derived oxyradicals as key signaling intermediates unexpectedly reduced activation of these canonical redox-sensitive transcription factors and hepatic secretion of TNF-α and IL-1β. Stated differently, oxidant generation that was insufficient to measurably alter the GSH:GSSG equilibrium nonetheless significantly reduced NF-κB and AP-1 transactivations, which are conventionally viewed as anti-inflammatory events (15, 30). Moreover, the characteristic anti-inflammatory effects of xanthine oxidase inhibition with Allo (5) were not observed in these studies. Rather, transactivation of NF-κB and AP-1, along with hepatic export of biologically active TNF-α, was increased in Allo + H/R + EC livers (Figs. 1, 3, and 4). Collectively, these results suggest that brief posttraumatic reductions in hepatic O2 availability may modulate subsequent postbacteremic transcriptional responses in the liver to paradoxically diminish cytokine-mediated inflammation.
Hypoxic stress of the intact liver has been diversely modeled. However, methods for inducing hypoxia, the duration of Po2 reductions, coinclusion of hypoxia with low-flow ischemia, variable provision of reoxygenation phases, association with other stimuli, and other important factors have differed considerably across studies (10, 14, 16, 32). Accordingly, data are sparse concerning specific thresholds for hypoxic modulation of postbacteremic transcriptional responses compared with normoxic profiles. Such information is highly relevant to intrahepatic events during the postresuscitation phase of traumatic injury. In this milieu, cardiopulmonary dysfunction limiting arterial oxygenation (24), stress hormone-induced increases in hepatic lobular V̇o2 (21), and other factors (32) culminate in hypoxic stress, despite quasi-normalization of hepatic blood flow (35). Our modeling of a constant-flow limitation of hepatic O2 availability followed by reoxygenation reasonably reflects such events, as does the temporal proximity of secondary hepatic stimulation by gut-derived Enterobacteriaciae, corresponding to bacterial translocation across a gut mucosal barrier previously impaired by trauma-related hypoxia (31). To our knowledge, this is the first report of such a modeling strategy performed in conjunction with assessment of the hepatic cellular redox state.
The NF-κB and AP-1 signaling pathways are among the most extensively examined redox-sensitive processes because of their involvement in numerous homeostatic and pathophysiological gene activation programs (15, 23, 30). NF-κB activation, in particular, has been linked to lethal organ injury in critically ill septic humans (36). The DNA-binding activity of NF-κB is conventionally attributed to nuclear translocation of heterodimeric p50 and p65 subunits after oxidant-induced phosphorylation of the cytosolic inhibitor IκBα (15). However, emerging data suggest a more intricate regulatory picture, in terms of interactions among NF-κB/RelA family subunits (17) and with respect to cytoplasmic-nuclear compartmentalization of IκBα and its intranuclear competition with NF-κB for binding to DNA consensus motifs (3, 29).
In this context, several explanations could potentially account for the reduced NF-κB activation in H/R + EC livers (Fig. 1). First, hypoxia-specific changes in the heterodimeric composition of NF-κB could reduce apparent postbacteremic NF-κB activation and DNA binding compared with normoxic EC controls, if a relatively constant p50 and p65 subunit dominancy was erroneously assumed. However, our supershift analysis of livers undergoing combined hypoxic stress and E. coli infection did not support this possibility. A second potential explanation centers on the nuclear interaction between NF-κB and IκBα regulating overall DNA binding. Conceivably, preceding hypoxia might selectively alter the profile of cytoplasmic vs. nuclear NF-κB in relation to total IκBα and phospho-IκBα abundance and degradation. Nevertheless, we found no significant differences in compartmentalization of either of these IκBα signals by Western analysis from H/R + EC livers compared with EC controls (Fig. 2). It may also be argued that hypoxic impairment of liver function in some manner limited subsequent EC-induced NF-κB transactivation. We consider this possibility unlikely, because all physiological, biochemical, microbiological, histological, and ultrastructural analysis employed here and in our previous studies (19, 22) failed to uncover any hypoxia-specific organ damage. In particular, the lack of increases in AST or GSSG levels, delayed clearance of circulating bacteria, or reoxygenation-phase decreases in hepatic V̇o2 in H/R + EC livers compared with EC controls support this view. Making this possibility further unlikely is the restored NF-κB transactivation and DNA binding in Allo + H/R + EC livers (Fig. 1). However, the lack of change in GSH:GSSG ratios in H/R + EC livers (Table 1) does not exclude the possible involvement of other cellular redox couples such as thioredoxin in the hypoxic suppression of postbacteremic NF-κB transactivation as well as TNF-α and IL-1β biosynthesis.
AP-1 is the designation for a redox-sensitive group of dimeric basic region leucine-zipper proteins that bind to the palindromic sequence 5′-TGAGTCA-3′ and control a broad array of biological processes (23, 30). Given AP-1's distinct regulation via the ERK, JNK, and other signaling pathways (30), cobinding within the TNF-α and IL-1β promoter regions (36), and evidence for differential activation compared with NF-κB (23), we sought to determine whether the hypoxic suppression of postbacteremic NF-κB was generalizable to AP-1. The generalizability of the phenomenon and the reversibility of the process by xanthine-oxidase inhibition were confirmed (Fig. 3). Equally important, we found no evidence for hypoxia-specific changes in the dominance of c-Fos and JunB subunit composition of AP-1 in H/R + EC rat livers (Fig. 3).
To what redox-sensitive mechanisms can hypoxic suppression of postbacteremic NF-κB and AP-1 transactivation in the liver be attributed? Abate et al. (1) reported that Fos and Jun DNA-binding activity in vitro were reduced by conformational changes in the local redox state of protein sulfhydryls. With respect to NF-κB, the necessity for Cys62 of the p50 subunit to remain in the reduced state for optimal DNA binding has been confirmed in vitro using chemical reducing agents as the oxidative stimuli (20, 25). More recently, Nishi et al. (26) refined this concept by demonstrating a spatial dependency of intranuclear reduction of the critical Cys62 of p50 mediated by redox factor (Ref)-1, a ubiquitous multifunctional protein with regulatory effects on NF-κB and AP-1. The small GTPase Rac-1 promotes de novo oxyradical generation in response to environmental stress and, in so doing, upregulates Ref-1 (2). However, Rac-1-mediated oxyradical generation is insensitive to xanthine oxidase inhibition (2). Although we therefore consider it unlikely that Rac-1 played a role in our system, we can neither confirm nor deny the possibility of involvement of Ref-1, which is targeted for additional investigation. On the basis of the above-mentioned considerations, we favor the explanation that low-level oxyradical generation during initial hypoxic stress summates with additional oxidant production during E. coli infection to induce conformational changes in the AP-1 and NF-κB DNA-binding domains. In support of this explanation, we recently potentiated these hypoxia-mediated reductions in E. coli-induced NF-κB and AP-1 transactivation by depleting hepatic GSH with diethyl maleate (18).
The pleiotropic cytokines TNF-α and IL-1β are beneficial and even indispensable for host defense during gram-negative bacteremic sepsis. However, when produced in excess, they mediate shock, liver dysfunction, and multiple organ failure (8, 12, 31, 36). The importance of NF-κB and AP-1 in regulating promoter activities of these cytokines was underscored and the biological significance of the hypoxic suppression of transactivation reported here was confirmed by the decline in TNF-α and IL-1β secretion in H/R + EC livers compared with normoxic EC organs (Figs. 4 and 5). Nevertheless, the data do not permit insight into the discrepancy between the full return of TNF-α values to control levels after xanthine oxidase inhibition in Allo + H/R + EC perfusates and the lack of restoration for IL-1β. We previously showed that Allo in similar doses fully reversed suppression of E. coli-induced hepatic IL-1β production by hypoxia after a primary intraportal infection (22). In view of the multiple redox-sensitive transcriptional and posttranscriptional regulatory mechanisms for IL-1β gene expression not investigated here, it is conceivable that Allo reversed the effects of antecedent hypoxic stress on IL-1β message accumulation or production of cell-associated IL-1β protein, even though the amount of secreted cytokine failed to return to EC control levels.
In summary, we have provided evidence in support of a more comprehensive model of redox-sensitive transcriptional regulation in the liver after stress using dual, context-specific physiological stimuli, rather than chemical reducing agents. Additional studies are indicated to define more fully the extent to which such regulatory strategies are utilized by the host to balance the beneficial and pathophysiological aspects of acute inflammation early after posttraumatic sepsis.
This work was supported by National Institute of General Medical Sciences Grant R01 GM-43513 (G. M. Matuschak).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 the American Physiological Society