Potential conflict of interest: Nothing to report.
During the last 3 days of fetal development in the rodent, a burst of hepatocyte proliferation results in a tripling of liver size. Despite stimulation of mitogenesis via multiple signaling pathways, including some that are considered stress response pathways, little apoptosis accompanies this cell growth. Given the accepted role of nuclear factor κB (NF-κB) in preventing hepatocellular apoptosis during proliferation in mid-development, we predicted that NF-κB would be functional during the period of rapid growth during late gestation in the rat. NF-κB binding in electrophoretic mobility shift assays was low in embryonic day (E) 19 liver nuclear extracts relative to adult liver nuclear extracts. An additional band that was present in E19 liver was purified and identified as nucleolin. Tumor necrosis factor α (TNF-α) administration to E19 embryos in utero produced minimal induction of NF-κB p50 homodimers and p50/p65 heterodimers, yet baseline apoptosis was not affected. Although p65 was present in E19 hepatocyte cytoplasm in amounts comparable to adult liver, we observed little translocation of p65 to the liver nuclei following TNF-α administration. Additionally, expression of several NF-κB–responsive genes remained minimally induced in E19 liver following TNF-α treatment. In conclusion, although the NF-κB components are present in late-gestation fetal liver, NF-κB as a transcription factor is relatively inactive and unresponsive to TNF-α. Given this finding and the high level of proliferation in late-gestation fetal liver, we predict that alternative antiapoptotic mechanisms are active during this period of rapid hepatic growth. (HEPATOLOGY 2005;42:326–334.)
Nuclear factor κB (NF-κB) is a mammalian transcription factor that regulates genes involved in cell growth, inflammation, and apoptosis. The most intensively studied form of NF-κB is a heterodimer consisting of p65 (RelA) and p50 subunits. In the absence of stimulation, most NF-κB remains in the cytoplasm, bound to its inhibitor, IκB.1 During activation of the canonical NF-κB signaling pathway, IκB is phosphorylated, ubiquitinated, and degraded, resulting in exposure of the NF-κB nuclear localization signal. This allows NF-κB to move to the nucleus and bind DNA. The phosphorylation of IκB is mediated by the IκB kinase (IKK) complex, containing IKK-α, IKK-β, and IKK-γ.2 However, there are variations on the canonical pathway. There are several forms of NF-κB, including homodimers and heterodimers of p65, p50/p105, c-rel, RelB, and p52/p100. Homodimers of p50 and p52 are considered inhibitory because they contain DNA-binding motifs without a transcription-activating domain.1 There is also evidence that the sequence of the NF-κB binding site as well as surrounding DNA sequence determines NF-κB family specificity and affects which coactivators will form productive interactions with the dimer.3 Finally, phosphorylation and acetylation of NF-κB and histones surrounding the NF-κB target gene can also enhance transcription.4 Thus, regulation of NF-κB can occur at several levels, is tissue type– and cell type–dependent, and can be stimulus-specific.
The importance of NF-κB in liver development is evidenced by various knockout mice missing important NF-κB pathway members. Many of these mice die mid-gestation from massive hepatocyte apoptosis. RelA−/− mice die on embryonic day (E) 15-16,5 NEMO/IKK-γ–deficient mice die on E12.5-13,6 and IKK-β–deficient mice die on E13-14.5.7, 8
One activator of NF-κB, tumor necrosis factor α (TNF-α), can also induce caspase activation and c-Jun N-terminal kinase (JNK) activity. Thus, TNF-α can elicit both pro- and anti-inflammatory mechanisms (JNK and caspase activation) and pro- and anti-apoptotic pathways (caspase and NF-κB activation). Consequently, the biological consequence of TNF-α exposure depends on the balance of these pathways.9 Adult hepatocytes are normally resistant to TNF-α–induced cytotoxicity. However, in the presence of a selective NF-κB inhibitor or in the setting of transcriptional or translational arrest, TNF-α induces hepatocyte apoptosis.10–13 Further evidence for the importance of NF-κB in preventing TNF-α–induced apoptosis is seen in RelA-deficient mice. When crossed with mice deficient in TNF-α or TNF-α receptor 1, the hepatocellular apoptosis associated with the RelA deficiency resolves.14–16 Similarly, hepatocellular apoptosis associated with IKK-β deficiency can be rescued by crossing with mice deficient in TNF-α receptor 1.8
In late-gestation fetal liver, signaling via the JNK pathway is constitutively active.17 Thus, it might be expected that there would be several antiapoptotic mechanisms in place to protect cells against the proapoptotic effect of JNK activation. Given the accepted role of NF-κB in preventing hepatocyte apoptosis following TNF-α exposure and in mid-gestation liver development, we predicted that NF-κB would be important in late-gestation liver growth. The present studies were originally undertaken to define the mechanism through which NF-κB was regulated in this model of hepatocyte proliferation. This led to the unexpected conclusion that NF-κB–independent antiapoptotic mechanisms may support the proliferation of late-gestation fetal hepatocytes.
Adult male Sprague-Dawley rats (125-175 g), timed-pregnant female rats, and postnatal day (PND) 1, 7, 21, 28 rat pups were anesthetized with sodium pentobarbital (50 mg/kg, intraperitoneal injection) and exsanguinated prior to removal of liver. Where noted, recombinant rat TNF-α (PeproTech, Rocky Hill, NJ) or phosphate-buffered saline was administered via intraperitoneal injection. To inject fetuses, timed-pregnant Sprague-Dawley rats (Charles River Breeding Laboratory, Wilmington, MA) were anesthetized, and an incision was made in the abdomen to expose the uterus. Fetuses were injected intraperitoneally through the uterine wall. The pregnant rat was sutured, and at the designated time point the fetuses were delivered via cesarean section. Two-thirds partial hepatectomies were performed on adult male rats (125-150 g) under isoflurane anesthesia. In all cases, livers were flash-frozen in liquid nitrogen and stored at −70°C until use. All procedures involving animals were performed in accordance with the guidelines of Rhode Island Hospital's Institutional Animal Care and Use Committee in compliance with guidelines set by the National Institutes of Health.
Preparation of Nuclear and Postnuclear Liver Extracts.
Nuclear and postnuclear extracts were prepared as previously described.18 Both the soluble and nuclear fractions were separated into aliquots and frozen at −70°C until use. The protein concentration of the soluble and nuclear components was determined using the bicinchoninic acid method (Pierce Chemical Co., Rockford, IL).
Electrophoretic Mobility Shift Assay.
Oligonucleotide target sequences were radio-labeled with [γ32P]adenosine triphosphate by T4 kinase. Binding reactions (10 μL) containing nuclear extracts (5 μg protein) and 5 μg of poly(dI-dC)poly(dI-dC) in binding buffer (10 mmol/L Tris-HCl [pH 7.5]; 50 mmol/L NaCl; 1 mmol/L MgCl2; 0.5 mmol/L EDTA; 0.5 mmol/L DTT; 4% glycerol) were incubated for 10 minutes at room temperature. The 32P-labeled oligonucleotide (5.25 nmol/L, 2.5 × 105 cpm) was then added to the reaction mixture and incubated for 45 minutes at room temperature. The sense strand oligonucleotide used was the NF-κB sequence (5′-agttgaggggactttcccaggc-3′) of the immunoglobulin κ promoter. Binding specificity was determined via competition experiments using a molar excess of unlabeled oligonucleotide before addition of the radiolabeled probe. Where indicated, antibody against p50, p52, p65, c-Rel, RelB, or nucleolin (Santa Cruz Biotechnology, Santa Cruz, CA; catalog numbers sc-114x, sc-298x, sc-372x, sc-70x, sc226x, and sc-13057, respectively) was included in the reaction mixture. Following electrophoresis, gels were dried and analyzed via autoradiography and phosphorescence detection.
E19 liver nuclear extract (1 mg) was incubated for a series of 1-hour incubations with fresh nucleolin antibody (5 μg) (Santa Cruz Biotechnology; catalog number sc-13057) immobilized on Protein A Sepharose (Amersham Biosciences, Piscataway, NJ). An aliquot was removed after each hour of immunodepletion for analysis via electrophoretic mobility shift assay (EMSA).
Fifty micrograms of protein were separated using a 10% polyacrylamide gel and transferred to a polyvinylidene difluoride membrane for immunoblotting. Primary antibodies to p50 (sc-114) and p65 (sc-372) were obtained from Santa Cruz Biotechnology. Detection employed an enhanced chemiluminescent method (Amersham Biosciences).
Cryosections (7 μm) were fixed in methanol and blocked in 1% normal donkey serum. They were incubated with primary antibodies for p50 (sc-114) or p65 (sc-372) for 60 minutes at room temperature. Detection employed fluorescein-conjugated donkey anti-rabbit antibodies (Pierce Chemical Co.). Omission of primary antibody was used as a negative control in each experiment.
Detection of Apoptosis.
Apoptosis was detected in 8-μm cryosections from frozen liver by terminal deoxy-nucleoltidyl transferase dUTP nick end-labeling (TUNEL) staining using ApopTag kit (Intergen, Purchase, NY). Tissue was counterstained with propidium iodide. Each tissue section was examined for TUNEL-positive cells and the apoptotic index was determined by the ratio of the number of TUNEL-positive cells to the total number of cells per ×60 microscopic field (3-4 animals per time point, 3-12 microscope fields per animal). Alternatively, apoptosis was detected using a poly(ADP-ribose)polymerase (PARP) cleavage detection kit (Calbiochem, La Jolla, CA). Fifty micrograms of nuclear extract was separated on 7.5% SDS-PAGE and immunodetection using an antibody for PARP was performed as described in the section on Western immunoblotting.
Ribonuclease Protection Assay.
To isolate total RNA, frozen liver was homogenized in guanidine thiocyanate and subjected to cesium chloride density centrifugation.19 Multiprobe template sets (BD RiboQuant Multi-Probe template sets; BD Biosciences, San Diego, CA) were custom-ordered, and ribonuclease protection assays were performed according to the manufacturer's instructions. Briefly, 32P-labeled antisense RNA probes (1.5 × 106 cpm) were hybridized in solution overnight in excess to target RNA (40 μg total RNA/treatment) in a total reaction volume of 10 μL. The free probe and other single-stranded RNA were digested with ribonucleases A and T1. The remaining ribonuclease-protected probes were precipitated, dissolved in 5 μL of sample buffer (BD Biosciences), and resolved on a denaturing polyacrylamide gel followed by autoradiography for 15 minutes to 3 days at −70°C.
Gel-Pro Analyzer software (version 3.0; Media Cybernetics, Newburyport, MA) was used to analyze the densitometry of bands on the autoradiograms produced by ribonuclease protection assays (RPAs), EMSAs, and Western blots. RPA densitometry results were corrected for the expression of GAPDH.
Results from apoptosis indices were analyzed using an unpaired t test. Results from ribonuclease protection assays were first analyzed via two-way ANOVA. When treatment and/or age effects occurred, a Bonferroni post hoc test was used. Two-sided significance testing was used, and a P value of less than .05 was considered significant.
Apoptosis in E19 Liver.
Cells in the rapidly proliferating E19 liver were smaller and more densely packed than in adult liver (Fig. 1). Apoptosis, as quantified by the number of TUNEL-positive cells versus the total nuclei per field, was slightly higher (P < .001) in fetal liver (1.45% ± 0.17% [mean ± SD]) than in adult liver (0.44% ± 0.08%).
NF-κB Activity in Fetal Rat Liver.
NF-κB activity in liver nuclear extracts was detected as the retarded migration of a radio-labeled consensus NF-κB oligonucleotide in an EMSA. In an ontogeny series of E19, E21, PND1, PND7, PND21, PND28, and adult liver nuclear extracts, two upper bands were barely detectable in E19 extracts but gradually became more abundant with advancing developmental age (Fig. 2A). Conversely, a lower band was prominent in E19 liver nuclear extract and became less abundant with advancing developmental age.
Because of the known induction of NF-κB by partial hepatectomy (PH), nuclear extracts from adult livers obtained 30 minutes after a PH were used to enhance the EMSA signal for identification of the individual bands. The specificity of binding to the consensus sequence was tested by adding an excess of the unlabeled oligonucleotide to the reaction before adding the radiolabeled probe (Fig. 2B). The three bands produced by PH nuclear extract were displaced using a 40-fold excess of unlabeled NF-κB target sequence oligonucleotide.
To further identify the proteins responsible for retarding the migration of the consensus NF-κB oligonucleotide on the gels, antibodies to specific NF-κB subunits were added to each reaction before electrophoresis. P65/p50 and p50/p50 dimers were identified in PH liver nuclear extracts by the presence of a supershift of the upper two bands on an EMSA (Fig. 2C). In addition, an unidentified lower band was present.
Identification of the Lower Band in NF-κB EMSAs.
Given the near absence of the identifiable forms of NF-κB in E19 nuclear extracts and the presence of the lower band on the EMSA of fetal nuclear extracts, we sought to determine whether the factor producing the lower band on the EMSAs could play a role in NF-κB–dependent gene regulation in late-gestation fetal liver.
E19 nuclear extracts were purified via heparin chromatography followed by DNA-affinity chromatography using the immunoglobulin κ promoter target sequence. The fractions collected from the columns were assayed via EMSA for NF-κB (data not shown). Fractions were run on SDS-PAGE, and prominent bands from the fractions with the highest EMSA binding activity were excised and submitted to matrix-assisted laser desorption ionization and tandem time-of-flight mass spectrometry (Laboratory for Proteomic Mass Spectrometry; University of Massachusetts Medical School, Shrewsbury, MA). The results were submitted to a sequencing database (the ProteinProspector program at the University of California–San Francisco Mass Spectrometry Facility was used to search the NCBInr.42803 database). The proteins identified were nucleolin and degraded products of nucleolin.
To verify that nucleolin was contributing to the lower band on the NF-κB EMSA, E19 nuclear extracts were incubated with the NF-κB probe and nucleolin antibody and analyzed for a supershift on EMSA. A supershift was not observed (data not shown). As an alternative approach, E19 nuclear exptracts were immunodepleted using an immobilized nucleolin antibody, and the unbound extract was analyzed via EMSA for NF-κB. Immunodepletion using nucleolin antibody reduced the amount of binding of the lower band, while control immunodepletions did not (Fig. 3).
NF-κB Cellular Localization.
Given the low NF-κB activity in E19 nuclear extracts, we determined whether there were differences in expression and/or cellular localization of NF-κB protein. Analysis of E19 and adult liver postnuclear fractions showed similar levels of p65 and p50 proteins as determined via Western immunoblotting (Fig. 4). Immunofluorescent staining of cryosections derived from livers obtained at various developmental ages showed the presence of p65 in the cytoplasm of both fetal and adult liver cells (Fig. 5). However, while p65 nuclear localization was present in adult liver, it was distinctly absent from most fetal and PND1 liver nuclei. The proportion of nuclei that excluded p65 from the nucleus increased from E13 to E14, a period when the hematopoietic cells, which make up early fetal liver, are being replaced by hepatocytes.20
Effects of TNF-α Administration on NF-κB.
To determine whether NF-κB could be induced in E19 liver, fetuses were given TNF-α in utero via intraperitoneal injection. Adult rats were used as a positive control. In both fetal and adult rat liver nuclear extracts, a peak of NF-κB activity, as assayed via EMSA, was observed 30 minutes after TNF-α administration (25 μg/kg) and continued for as long as 4 hours (time-course data from 0-4 hours not shown). NF-κB activity was consistently lower in fetal TNF-α–treated samples than in their adult counterparts at all timepoints. At 30 minutes after administration, p65/p50 binding following TNF-α treatment (25 μg/kg) was induced 10-fold higher in adult liver than in fetal liver (Fig. 6A). An increase in TNF-α to 715 μg/kg did not increase the NF-κB activity in fetal liver nuclear extracts. In fetal liver, binding of the inhibitory p50/p50 homodimer increased more than the transcriptionally active p65/p50 heterodimer. On the other hand, p65/p50 was induced more than p50/p50 in adult samples following 25 μg/kg TNF-α.
Western immunoblotting showed a marked increase in the nuclear content of p65 in adult animals following TNF-α, compared with a modest increase in fetuses (Fig. 6B). Immunofluorescent staining for p65 showed strong nuclear staining in the adult following TNF-α treatment, while no change was seen in fetal liver at the 25 μg/kg dose. A low level of nuclear localization of p65 occurred in E19 liver after 715 μg/kg TNF-α (Fig. 6).
Apoptosis Following TNF-α Treatment.
We have shown previously that TNF-α induces JNK activation in E19 fetal liver.17 Given the ability of TNF-α to induce apoptosis in the liver when the NF-κB pathway is blocked,10–13 we hypothesized that fetal hepatocytes, in which TNF-α did not induce NF-κB, might undergo apoptosis when exposed to TNF-α. However, PARP cleavage did not occur by 4 hours (Fig. 7). The effectiveness of TNF-α administration was confirmed by EMSAs, which showed continued low-level NF-κB induction at 4 hours after administration (data not shown). Apoptosis as measured by TUNEL staining did not increase in E19 liver by 15 hours after administration (data not shown).
NF-κB–Responsive Gene Expression Following TNF-α Treatment.
Ribonuclease protection assays were performed to examine the expression of NF-κB–responsive genes in the liver following TNF-α treatment. Several pro- and antiapoptotic genes were examined based on previous studies that showed NF-κB responsiveness in hepatocytes (bfl1,21 bak,21 iNOS22) or other cells (bcl-2,23 bcl-x(l/s),24, 25 FasL,26 TRAF127). Several genes were selected because their expression might provide insight into alternative antiapoptotic pathways (bcl-w, Fas28) or the regulation of the NF-κB pathway (RIP, IKK-β, IκBα, TNF-α9). The expression of KC, a TNF-α–responsive chemokine in the liver,29 was also evaluated as a positive control for TNF-α administration.
E19 and adult animals were treated with TNF-α. Livers were harvested after 1 hour. To verify the delivery of TNF-α, EMSAs for NF-κB were performed on nuclear extracts. TNF-α treatment induced NF-κB binding in both fetal and adult samples—although NF-κB activity was much higher in treated adults than in treated fetuses.
Fas expression was induced by TNF-α in adult but not fetal liver. TNF-α also increased expression of several genes in both fetal and adult liver, including IκBα, bak, bfl1, RIP, and KC (two-way ANOVA). However, the magnitude of the induction of most of these genes was much lower in fetal liver than in adult liver. Indeed, a comparison of control fetal liver to TNF-α–treated fetal liver using a Bonferroni post hoc test showed IκBα, a well-established NF-κB target gene, to be the only gene significantly induced by TNF-α in the fetus (Fig. 8).
Differences in baseline gene expression in fetal versus adult liver were also seen. Of note, IκBα, which was TNF-α–inducible, showed baseline expression that was no higher in the fetus than the adult. The proapoptotic genes bak and bcl-xs and the antiapoptotic gene bcl-xl were expressed at higher levels in fetal liver compared with adult liver, while antiapoptotic genes bcl-2 and bcl-w were decreased in fetal liver. No fetal/adult differences or TNF-α effects were seen for iNOS, TRAF1, FasL, TNF-α, or IKK-β.
Although mitogenesis in late gestation fetal liver is high relative to adult liver, apoptosis in fetal liver was only slightly increased. Given that in vivo apoptosis is detected primarily in proliferating tissues,30 a small increase in apoptosis in the rapidly proliferating fetal liver is not surprising. However, we had anticipated high activity of antiapoptotic signaling pathways. Given the constitutive activity of the JNK pathway in E19 fetal liver,17 we hypothesized that NF-κB would also be active. However, there was minimal NF-κB activity in late-gestation fetal liver. Additionally, NF-κB could only be minimally activated in response to in vivo administration of TNF-α to E19 fetuses, whereas a rapid and significant NF-κB response was induced in adult liver by TNF-α.
The supposition that NF-κB is important for survival in proliferating hepatocytes comes from observations from at least two models. NF-κB is rapidly induced during liver regeneration that follows PH,31 and transgenic mice lacking NF-κB pathway members die mid-gestation as a result of massive hepatocyte apoptosis.5–8 However, there are issues with each of these models that must be considered.
The observations that NF-κB is rapidly induced following PH and that inhibition of NF-κB by gliotoxin or the super-repressor IκBα results in extensive hepatocyte apoptosis after PH13, 32–34 led to the conclusion that hepatocyte survival after PH requires NF-κB. However, hepatocyte-specific expression of the super-repressor IκBα, which is mutated such that its degradation is inhibited, does not affect hepatocyte survival during liver regeneration.12 The current model of liver regeneration indicates that TNF-α induces NF-κB in Kupffer cells rather than hepatocytes leading to increased Kupffer cell expression of IL-6 and TNF-α. This in turn induces STAT3 in neighboring hepatocytes.35
RelA−/− mice die of massive hepatocellular apoptosis mid-gestation.5 However, these mice, as well as other transgenic mice lacking expression of important NF-κB family members, lack expression of the target gene in all cells and tissues. Thus it is not certain that hepatocyte expression of NF-κB is required for normal liver development. Also uncertain is whether NF-κB is active in mid-gestation liver. Although the presence of p65 has been shown in rat fetal liver during mid-gestation,12 the activity state of NF-κB has not been demonstrated. In fact, mice that are transgenic for a reporter gene under the control of an NF-κB–dependent promoter do not show NF-κB–mediated hepatic expression during mouse development.36 Our own data indicate that in fetal liver, expression of p65 in whole homogenates is not reflective of NF-κB activity. Both p50 and p65 were as abundant in late-gestation fetal liver cytoplasm as in adult, yet the levels of p65 found in nuclear extracts were extremely low in fetal liver compared with adult liver. Thus it remains to be shown which cell type requires NF-κB for the maintenance of hepatocyte viability during mid-gestation.
We found that exposure to an activating signal, TNF-α, did not induce NF-κB translocation to the nucleus. Therefore, although the NF-κB subunits were present in the cytoplasm, they were not activated under circumstances that would be expected to result in NF-κB activation. Moreover, TNF-α treatment resulted in a modest increase in p50 homodimer binding in fetal liver nuclear extracts.
Given our data on NF-κB activity state and the role of p50 homodimer as a competitive inhibitor of transcriptionally active NF-κB complexes,1 we hypothesized that NF-κB–responsive genes would not be induced in response to TNF-α in the fetus. In fact, several genes were induced in both fetal and adult liver by TNF-α treatment. However, only IκBα, a negative regulator of NF-κB, was significantly induced in the fetus. This result leaves open the question of what prevents TNF-α–induced apoptosis in the fetal liver. Although baseline gene expression of bcl family members was clearly different in fetal liver from adult liver, the trend was for increased expression of proapoptotic factors and decreased expression of antiapoptotic factors in fetal liver.
In late-gestation fetal liver, the JNK pathway is constitutively active and can be further activated by TNF-α administration in vivo.17 Although most models suggest that these circumstances would require an antiapoptotic mechanism to counter proapoptotic effects of JNK activation, several studies in liver-derived cells,37, 38 as well as other tissues,39 suggest that JNK may be antiapoptotic under some circumstances. Further characterization of the role of the JNK pathway in late-gestation liver development may be relevant to the present findings.
The absence of identifiable NF-κB in fetal liver along with the prominence of a lower band in the NF-κB EMSAs compelled us to identify the latter. DNA-affinity chromatography combined with matrix-assisted laser desorption ionization and tandem time-of-flight mass spectrometry identified the protein as nucleolin and degraded forms of nucleolin. Nucleolin is a ubiquitous nonhistone nucleolar phosphoprotein that is highly expressed in actively dividing cells. In nonproliferating cells, degraded forms of various molecular size are present due to autodegradation,40–43 which explains why several lower molecular weight forms were present in nuclear extracts. The functional significance of nucleolin binding to an NF-κB consensus sequence is not known. However, nucleolin is one of several “matrix attachment region” DNA-binding proteins known to provide a link between DNA and nuclear matrix scaffolding.44 Matrix attachment regions relieve negative superhelical strain and are common at the boundaries of transcription units, where they may delimit the ends of an active chromatin domain.45 Thus, nucleolin may interact with DNA and other DNA-binding proteins, such as NF-κB, to regulate transcription. Although our data indicate that nucleolin binds to our NF-κB binding site, we have no evidence to show that nucleolin binds to NF-κB target sequences in vivo. We also found that part of the required sequence for nucleolin binding was separate from the canonical NF-κB binding sequence and included a portion of flanking DNA in the oligonucleotide. Thus the functional significance of nucleolin binding to our NF-κB oligonucleotide has yet to be determined.
In conclusion, the unexpected results from these experiments leave us with several important questions. What are the antiapoptotic pathways that are active in fetal liver? Is nucleolin binding to a canonical NF-κB sequence functionally significant? What signals prevent abundant NF-κB from being activated in a fetus? Notwithstanding these issues, our data suggest an alternative to NF-κB–mediated antiapoptosis in the developing liver, thereby raising the possibility of an alternative pathway that might be involved in the physiology and pathophysiology of the hepatic response to injury and hepatocarcinogenesis.
We thank Gretchen Halpert, Virginia Hovanesian, and Theresa Bienieki for their technical assistance and Jennifer Sanders, Joan Boylan, and Anand Padmanabhan for their knowledgeable advice and discussions.