|Home | About | Journals | Submit | Contact Us | Français|
Type 2 diabetes is characterized by the inability of insulin to suppress glucose production in the liver and kidney. Insulin inhibits glucose production by indirect and direct mechanisms. The latter result in transcriptional suppression of key gluconeogenetic and glycogenolytic enzymes, phosphoenolpyruvate carboxykinase (Pepck) and glucose-6-phosphatase (G6p). The transcription factors required for this effect are incompletely characterized. We report that in glucogenetic kidney epithelial cells, Pepck and G6p expression are induced by dexamethasone (dex) and cAMP, but fail to be inhibited by insulin. The inability to respond to insulin is associated with reduced expression of the forkhead transcription factor Foxo1, a substrate of the Akt kinase that is inhibited by insulin through phosphorylation. Transduction of kidney cells with recombinant adenovirus encoding Foxo1 results in insulin inhibition of dex/cAMP–induced G6p expression. Moreover, expression of dominant negative Foxo1 mutant results in partial inhibition of dex/cAMP–induced G6p and Pepck expression in primary cultures of mouse hepatocyes and kidney LLC-PK1-FBPase+ cells. These findings are consistent with the possibility that Foxo1 is involved in insulin regulation of glucose production by mediating the ability of insulin to decrease the glucocorticoid/cAMP response of G6p.
Type 2 diabetes mellitus results from impaired insulin action and inadequate insulin secretion (1). A key abnormality in the pathogenesis of diabetes is insulin’s failure to restrain endogenous glucose production, resulting in increased blood glucose levels (2). Evidence from clamped dogs (2) and genetically engineered mice indicates that insulin action on glucose production includes indirect and direct effects (3, 4). For example, hepatocytes lacking insulin receptors lack the ability to suppress glucose output in response to insulin and display increased Pepck and G6p levels (5, 6). The primary sites of glucose production are the liver and kidney (2, 7). In vivo studies indicate that during prolonged fasting or diabetes, renal glucose production can account for up to 25% of total endogenous glucose production (8–13), although other studies suggest a much lower contribution (14, 15). Renal glucose production, like hepatic glucose production, is suppressed by insulin in vivo (10, 16–18). Insulin’s ability to reduce glucose production is preserved in cultured hepatocytes, but not in cultured kidney epithelial cells, the site of renal glucose production. The latter display hormonal (19, 20) and pH-regulated gluconeogenesis (20–29), but have never been shown to be sensitive to insulin inhibition.
Insulin controls glucose production by inhibiting expression of two rate-limiting enzymes in gluconeogenesis and glycogenolysis, Pepck and G6p (30, 31). The signaling pathways required for these effects are incompletely understood, but are thought to require activation of the lipid kinase phosphatidylinositol 3 (PI 3-kinase) (32–35). Among the PI-dependent kinases, Akt has been implicated as a mediator of insulin’s inhibition of a reporter gene driven by a Pepck promoter spanning the putative insulin response sequence (IRS) (36). However, the use of different dominant negative Akt mutants has yielded controversial results (34, 37). The forkhead transcription factor Foxo1 (previously known as Fkhr) (38) is phosphorylated in an insulin-responsive manner by PIP3-dependent kinases, such as Akt and Sgk (39–50). Phosphorylation leads to nuclear exclusion and inhibition of Foxo1-dependent transcription (39–42, 44–55). Studies in hepatoma cells suggest that Foxo1 and its closely related isoform Foxo3 possess the ability to regulate transcription of reporter genes containing insulin response elements from the G6p and Pepck promoters in an insulin-dependent manner (56, 57). However, it is unclear whether the endogenous genes can be regulated in a similar manner and whether Foxo proteins are the physiologic mediators of insulin action on Pepck and G6p. To address this question, we characterized the hormonal regulation of Pepck and G6p expression in LLC-PK1-FBPase+ kidney epithelial cells. In this study we show that Pepck and G6p in these cells are refractory to insulin inhibition. This refractoriness is associated with low levels of Foxo1 expression. Expression of Foxo1 by adenovirus-mediated gene transfer confers insulin inhibition onto the dex/cAMP cocktail–induced (see Methods) rise in G6p. Moreover, a dominant negative Foxo1 lacking the transactivation domain partially inhibits dex/cAMP cocktail–induced G6p and Pepck expression both in LLC-PK1-FBPase+ cells and in primary cultures of mouse hepatocytes. We suggest that Foxo1 plays an important role in insulin control of G6p expression and that dominant negative Foxo1 mutants provide a useful reagent to inhibit glucogenesis in experimental systems.
LLC-PK1-FBPase+ cells are a glucogenetic substrain of LLC-PK1 cells that express fructose-bis-phosphatase (58, 59). Although they are different from the parental LLC cell line, for brevity we refer to them as LLC cells. The following cDNA probes were obtained: G6p, Pepck, FOXA2 (Hnf-3β), HNF-1α, and serum- and glucocorticoid-induced kinase 2 (SGK 2). Probes for Foxo1, Foxo3, and Foxo4 were described previously (45). Gapdh and β-actin were prepared by RT-PCR using Gene Amp RNA PCR kit (Applied Biosystems, Foster City, California, USA) according to the manufacturer’s protocol.
LLC cells were cultured at 37°C in DMEM with 5.5 mM D-glucose (Life Technologies Inc., Rockville, Maryland, USA), 2 mM glutamine, and 44 mM NaHCO3 supplemented with 10% FBS (Mediatech Inc., Herndon, Virginia, USA), 100 U/ml penicillin, and 100 μg/ml streptomycin. For induction of G6p and Pepck mRNA, cells were incubated with DMEM supplemented with 0.1% BSA, 0.5 mM 8-bromoadenosine 3′:5′-cyclic monophosphate (8-Br-cAMP) (Sigma Chemical Co., St. Louis, Missouri, USA), 0.5 mM 3-isobutyl-1-methylxanthine (IBMX) (Sigma Chemical Co.), and 1 μM dexamethasone (Sigma Chemical Co.) for 0–24 hours. This cocktail is referred to throughout as dex/cAMP. Anti–c-Myc mAb 9E10 and anti–hemagglutinin (anti-HA) mAB 12CA5 were obtained from Santa Cruz Biotechnology Inc. (Santa Cruz, California, USA) and Boehringer Mannheim GmbH (Mannheim, Germany),respectively. Rabbit polyclonal antisera against FOXO1-S256, FOXO4-T28, and FOXO1-S318 phosphopeptides were from Cell Signaling Technologies (Beverly, Massachusetts, USA). Ab’s were used for Western blot analysis at 1:1,000 dilution.
Hepatocytes were isolated according to the method of Honkakoski (60), with the following modifications: Complete protease inhibitor (Boehringer Mannheim GmbH) was added to digestion buffer according to the manufacturer’s specifications (61). Hormonal treatment was carried out in DMEM supplemented with dex/cAMP for 4 hours.
To facilitate identification of Foxo1 expressed with adenoviral vectors, a HA epitope tag was cloned at the 5′ of the murine cDNAs encoding wild-type (WT), T24A/S253D/S316A (ADA), and Δ256 mutant Foxo1. To this end, PCR reactions were performed using the respective Foxo1 cDNA as templates. The same upstream (sense) primer was used in all three constructs: 5′-ACT GGT ACC GCC ATG TAC CCA TAC GAT GTT CCG GAT TAC GCT GCC GAG GCG CCC CAG GTG GTG G-3′. Two different antisense primers were used: 5′-TTG CCC CAC GCG TTG CGG CGC GAC GAG C-3′ (to construct WT and ADA Foxo1) and 5′-AAT TCT AGA GTC CAT GGA CGC AGC TCT TCT CCG-3′ (to construct the Δ256 mutant) (47). The resulting PCR fragments were digested with KpnI and MluI and subcloned into KpnI/MluI–treated pCMV5/c-Myc WT and ADA Foxo1. For the construction of HA-tagged Δ256 Foxo1, after digestion with KpnI and XbaI, the PCR fragment was subcloned into KpnI- and XbaI-treated pCMV5/c-Myc. DNA encoding the HA-tagged WT and mutant Foxo1 was subcloned into pAxCAwt, and adenovirus vectors containing these cDNAs were generated by transfecting HEK 293 cells with the corresponding pAxCAwt plasmid, together with a DNA-terminal protein complex (62). Experiments were carried out 24 and 48 hours after infection in hepatocytes and LLC cells, respectively.
A full-length mouse cDNA clone encoding Foxo3 was assembled by subcloning two separate fragments into the expression vector pCMV5/c-Myc. The 5′ fragment was obtained by PCR using a cloned Foxo3 cDNA as template. The primers used are: upstream, 5′-GGG GAA TTC ATG GCA GAG GCA CCA GCC TCC-3′, nucleotide (nt) 326–343, and downstream, 5′-GTC GCC CTT ATC CTT GAA GTA-3′ (nt 922–902). The PCR product was digested with EcoRI and BglII and cloned into EcoRI/BglII–digested pCMV5/c-Myc. The 3′ fragment, encompassing nt 865–2886 of the Foxo3 sequence, was obtained by digestion of a full-length cDNA clone with BglII and HindIII and ligated into the same sites of the pCMV5/c-Myc. The expression vector was fully sequenced before transfection to confirm that no mutations had been introduced during the cloning procedure.
Cells were incubated in serum-free medium supplemented with dex/cAMP for 16 hours. Thereafter, insulin was added to the medium for up to 6 hours at a final concentration of 100 nM. Cells were harvested by trypsinization, and mRNA was isolated using Micro-Fast Track 2.0 kit (Invitrogen Corp., San Diego, California, USA). mRNA was size-fractionated on denaturing formaldehyde/agarose gels and transferred to a nylon membrane for Northern hybridization according to standard techniques.
Cells were harvested from a 10-cm culture dish and solubilized in Triton X-100. Detergent cell extracts were immunoprecipitated with anti-insulin receptor (anti-IR) Ab Ab-3 (Calbiochem-Novabiochem Corp., San Diego, California, USA) or anti–IGF-1R Ab C-20 (Transduction Laboratories, Lexington, Kentucky, USA) and blotted with anti–IR Ab C-19 or anti–IGF-1R Ab C-20 (Transduction Laboratories) as described previously (47). Antisera were used at a dilution of 1:1,000.
Cells were incubated in serum-free DMEM supplemented with 0.1% BSA or dex/cAMP for 16 hours and then stimulated with insulin (100 nM) for the indicated times. At the end of incubation, Foxo1-transfected cells were solubilized and immunoprecipitated with anti–HA mAb 12CA5 (Boehringer Mannheim GmbH), while Foxo3-transfected cells were immunoprecipitated with anti–c-Myc mAb 9E10 (Santa Cruz Biotechnology Inc.). Immune complexes were resolved on 8% SDS-PAGE and transferred to nitrocellulose filters. The membranes were sequentially reprobed with anti–phospho T24, S253, S316, and anti-Foxo1 or Foxo3 antisera (Upstate Biotechnology Inc., Lake Placid, New York, USA) to normalize the amount of phosphate incorporated into each band for the amount of protein applied to the gel. Western blot analysis was performed using detection of the immune complexes with horseradish peroxidase–conjugated anti–rabbit IgG (enhanced chemiluminescence [ECL]; Amersham Pharmacia Biotech AB, Piscataway, New Jersey, USA).
LLC cells were transiently transfected with pCMV5/c-Myc encoding Foxo1 or Foxo3 using Lipofectamine (Life Technologies Inc., Rockville, Maryland, USA) according to the manufacturer’s protocol (45, 47). After 24 hours, cells were seeded into a four-well slide culture chamber (Nalge Nunc Corp., Naperville, Illinois, USA). Cells were incubated in serum-free medium for 16 hours and then stimulated with insulin (100 nM) for the indicated times. Fixation of cells and incubation with anti–c-Myc mAb (9E10) and secondary Ab have been described previously (47).
Under basal culture conditions, Pepck and G6p mRNAs are expressed at very low levels in LLC cells (Figure (Figure1,1, a and b, lane 1). Addition of dex/cAMP to the culture medium results in approximately tenfold induction of both mRNAs. The increase in G6p occurs following 8 hours of the addition of dex/cAMP, peaks at 16 hours, and disappears by 24 hours. In contrast, the increase in Pepck is detectable by 2 hours and persists after 24 hours of dex/cAMP treatment (Figure (Figure1a).1a). To study the ability of insulin to affect this response, LLC cells were treated with dex/cAMP for either 4 or 16 hours (Figure1b, lanes 1–3 and 4–6, respectively), and then exposed to insulin for 6 hours. At both time points, the effect of dex/cAMP was not inhibited by insulin (Figure (Figure1,1, b, lane 3, and 6), indicating that Pepck and G6p are not regulated by insulin in LLC cells. Several experiments are summarized in Figure Figure11c.
The lack of insulin responsiveness cannot be ascribed to the absence of IRs, which are expressed in LLC cells in sizable amounts (Figure (Figure2,2, lane 2), as are IGF-1 receptors (lane 4).
Next, we compared expression patterns of Foxo isoforms in LLC cells and SV40 hepatocytes (5). As we have shown previously, Foxo1 is the main Foxo isoform in SV40 (tsA)–transformed hepatocytes (45). Its expression in LLC cells is reduced by approximately 70% compared with SV40 hepatocytes (Figure (Figure3a,3a, top panel, lanes 1 and 2). Foxo3 is the most abundant Foxo isoform in LLC cells. Its levels are approximately 60% of those of Foxo1 in SV40 hepatocytes (Figure (Figure3a,3a, second panel from top, lanes 1 and 2). Finally, Foxo4 is expressed at low levels in SV40 hepatocytes, but not in LLC cells (Figure (Figure3,3, a and b). The forkhead protein Foxa2 (HNF3β), which has been suggested to participate in regulation of Pepck and G6p expression (63, 64), is likewise absent from LLC cells. The transcription factor Hnf-1α has been implicated in the regulation of Pepck and G6p in the kidney (22, 65–70) and is expressed at similar levels in both cell types. Finally, the Foxo kinase Sgk-2 (41, 50) is not expressed in LLC cells. These data are consistent with the possibility that lack of insulin inhibition on gluconeogenic gene expression may result from reduced expression of Foxo1 and/or the Foxo kinase Sgk-2.
Insulin has been shown to alter Foxo subcellular distribution (47). Thus, we analyzed insulin-induced nuclear export of Foxo1 and Foxo3 in the presence and absence of dex/cAMP using immunofluorescence to detect c-Myc–tagged Foxo1 or Foxo3 following transient transfections. Individual cells were scored according to whether they showed exclusively cytoplasmic, exclusively nuclear, or diffuse immunostaining. In the absence of dex/cAMP, insulin treatment for 15 minutes decreased the percentage of cells showing exclusively nuclear Foxo1 staining from 45 to 12%, while the percentage of Foxo3-positive cells with nuclear staining decreased from 62 to 42% (Figure (Figure4a,4a, upper panels). In the presence of dex/cAMP, insulin-induced translocation was considerably delayed. Thus, results at 1 hour of treatment are shown. Nuclear staining decreased from 80 to 45% of Foxo1-expressing cells and from 80 to 60% of Foxo3-expressing cells (Figure (Figure4,4, lower panels). The difference in nuclear localization between Foxo3- and Foxo1-expressing cells was statistically significant (P < 0.01 by ANOVA) under all treatment conditions, except in the absence of insulin in dex/cAMP–treated cells (lower panels). These data show that insulin increases nuclear export of Foxo in kidney cells, while dex/cAMP antagonizes insulin’s effect. From these data we cannot conclude whether dex/cAMP affects insulin-induced Foxo1 translocation directly or indirectly. However, it is possible to suggest that Foxo1 is more efficiently translocated from the nucleus to the cytoplasm than Foxo3. The latter findings are consistent with the suggestion that these two proteins are differentially regulated by insulin (46).
Analysis of insulin-induced phosphorylation by way of anti-phosphopeptide Ab’s revealed that Foxo1 was phosphorylated on three Akt consensus sites — T24, S253, and S316 — in an insulin-dependent manner (Figure (Figure4b),4b), whereas Foxo3 was phosphorylated on two different sites, T24 and S253 (data not shown). We could not assess the phosphorylation of Foxo3 S316 for lack of a suitable phospho-specific Ab.
Next, we asked whether expression of Foxo1 would confer insulin regulation onto G6p and Pepck expression. In cells transduced with Foxo1, both Pepck and G6p were still detectable after 24 hours of dex/cAMP treatment, suggesting that Foxo1 extends the duration of this effect (Figure (Figure5a,5a, lanes 1–3). Foxo1 did not increase basal G6p and Pepck levels, but resulted in approximately 60% increase of dex/cAMP–induced G6p expression (P < 0.05) and approximately 25% increase of Pepck expression (P = NS) (Figure (Figure5b,5b, lane 2). Addition of insulin blunted the effect of dex/cAMP on G6p in a time-dependent manner, reaching the maximal effect after 6 hours of exposure to the hormone (Figure (Figure5b,5b, compare lanes 2–5 with 6–9). No significant effect was seen on Pepck (5b, middle panel). The effect of insulin was studied at two different time points of dex/cAMP treatment. After either 4 or 16 hours of dex/cAMP exposure, in Foxo1-expressing cells insulin inhibited the dex/cAMP–induced increment in G6p expression by 60% (P < 0.05 by ANOVA), while no inhibition was observed for Pepck. Data from several experiments are summarized in Figure Figure55c.
To determine whether Foxo1 induces gene expression through a dex/cAMP–dependent mechanism, we analyzed the ability of gain- and loss-of-function Foxo1 mutants to affect dex/cAMP–induced gene expression. It has been shown that a truncated Foxo1 mutant lacking the transactivation domain (Δ256) acts as a dominant negative inhibitor of Foxo1-mediated transcription of an Igfbp-1 reporter construct containing an insulin response element (IRE) (refs. 40, 71 and our unpublished observation). In contrast, a phosphorylation-defective mutant, in which all three potential Akt phosphorylation sites have been replaced by nonphosphorylatable amino acids (ADA), cannot be excluded from the nucleus in response to insulin and is constitutively active (39, 40, 47, 51, 71). Transduction of LLC cells with adenoviral vectors encoding Δ256-Foxo1 and/or ADA-Foxo1 resulted in readily detectable expression of the two mutant proteins (Figure (Figure6a,6a, lanes 2–4). In these loss-of-function experiments, we also studied the effect of Foxo1 in primary cultures of mouse hepatocytes. As shown in Figure Figure6b,6b, addition of dex/cAMP resulted in a brisk increase of both G6p and Pepck expression, which was readily inhibited by insulin treatment (lanes 2 and 3). Expression of the Δ256 mutant resulted in a dose-dependent decrease of dex/cAMP–induced G6p expression up to approximately 80% in LLC cells and 90% in primary hepatocytes (Figure (Figure6,6, c and d, lanes 1–3, and Figure Figure7,7, a and c). Likewise, Pepck expression decreased up to 70% in LLC cells and 80% in primary hepatocytes (Figures (Figures6,6, c and d, lanes 1–3, and Figure Figure7,7, b and d) (P < 0.001).
Similar results were obtained in cells cotransduced with the constitutively active mutant at a fixed moi and increasing concentrations of dominant negative mutant (Figure (Figure6,6, c and d, lanes 4–6, and Figure Figure7,7, a–d). These data indicate that a dominant negative Foxo1 protein can affect the dex/cAMP–induced rise of G6p and Pepck and effectively prevent the increase due to a constitutively active Foxo1 mutant.
In this study we demonstrate that cultured kidney epithelial cells lack insulin inhibition of Pepck and G6p expression. We seized upon this observation to investigate the role of the forkhead transcription factor Foxo1 in this process. We show that insulin’s failure to inhibit Pepck and G6p in LLC cells correlates with low Foxo1 mRNA levels and that transduction with a Foxo1 adenovirus confers insulin inhibition on dex/cAMP–induced G6p, but not Pepck expression. These data support the hypothesis that Foxo1 regulates expression of the endogenous G6p in an insulin-dependent manner. We also show that a dominant negative Foxo1 mutant effectively prevents the dex/cAMP–induced increases in G6p and Pepck expression in both LLC cells and primary hepatocytes, consistent with the possibility that Foxo1 functions as a bona fide insulin-regulated transcription factor on endogenous genes. It bears emphasizing that this point had, thus far, eluded demonstration.
The main Foxo isoform in LLC cells is Foxo3, which shares common structural features with Foxo1 (38). We show that Foxo3 is less responsive than Foxo1 to insulin-induced nuclear translocation. This difference is consistent with the observation that insulin inhibits Foxo3-dependent gene expression in hepatoma cells less markedly than Foxo1-dependent gene expression (46). A potential explanation for the lackluster insulin response of Foxo3 is that LLC cells lack one of the Foxo kinases, Sgk-2 (41, 50). We suggest that Foxo1 is the principal insulin-responsive Foxo isoform, consistent with its distribution in insulin-responsive tissues (unpublished observation). Moreover, loss- and gain-of-function Foxo1 mutations modulate insulin sensitivity in mice (72). We do not know whether differences in nuclear export rates between Foxo1 and Foxo3 can account for the different properties of these two proteins to mediate insulin inhibition of gene expression. From our limited analysis, we cannot conclude that nuclear exclusion is the sole or even the main mechanism by which insulin regulates Foxo1 (see below).
The ability of Foxo1 to suppress dex/cAMP–induced G6p, but not Pepck, expression in an insulin-dependent fashion suggests that different mechanisms regulate insulin inhibition of these two target genes. Since our studies analyzed expression of the endogenous genes, we are not in a position to comment on the cis-acting elements required for the Foxo1 effect. However, the magnitude of the Foxo1 effect on G6p inhibition by insulin is consistent with that observed in cultured liver cells or in animal models (30).
In the human and murine G6p promoters, two elements are required for transcriptional repression by insulin: A (nt –231 to –199) and B (nt –198 to –159). Region A binds the accessory factor hepatic nuclear factor-1α, which is expressed at high levels in LLC cells (refs. 70, 73 and this study). Region B contains three copies of a consensus IRS. This region has been shown to bind a GST-Foxo1 fusion protein (74). When used in a reporter gene assay in hepatoma cells, the same element bestowed insulin sensitivity and Foxo1 dependence, consistent with a role in Foxo1-mediated gene expression (57). Our data complement this body of observations and suggest that G6p is a physiologic target of Foxo1-mediated insulin inhibition.
The ability of Foxo1 to modulate G6p expression in a dex/cAMP–dependent manner is consistent with a model in which Foxo1 requires glucocorticoids and cAMP to regulate G6p transcription. This conclusion is supported by a variety of observations presented in this study, including the lack of effect of Foxo1 on basal G6p expression, the additive effect of Foxo1 and dex/cAMP to increase G6p mRNA, and the failure of the dominant negative Foxo1 to decrease G6p below the levels observed in the absence of dex/cAMP. This observation is similar to the reported requirement of HNF-1 in cAMP stimulation of G6p in LLC cells (68). Thus, Foxo1 appears to act as an accessory factor of the glucocorticoid/cAMP response that is inhibited by insulin, in part through nuclear exclusion.
The mechanism of Foxo1 action on Pepck appears to be quite different. In this case, Foxo1 is unable to confer insulin sensitivity on kidney Pepck expression, but the dominant negative Foxo1 can inhibit dex/cAMP–induced Pepck expression in both kidney and hepatocytes. These data are consistent with the possibility that Foxo1 regulates Pepck indirectly, as already suggested by Hall and coworkers based on reporter gene studies in hepatoma cells (56). The mechanism by which the dominant negative affects the glucocorticoid response is at present unclear. Based on the work of Nasrin and colleagues, one possible explanation is that the truncated Foxo1 prevents the recruitment of steroid receptor coactivator SRC to the glucocorticoid response unit (46). Interestingly, a different forkhead mutant, encoding a truncated FOXA2, has a similar inhibitory effect in hepatoma cells (63). The recent demonstration that Foxo1 can function both as a coactivator or a corepressor of transcription by interacting with nuclear hormone receptors provides an additional mechanism to explain the effect of the dominant negative Foxo1 on Pepck (75).
The kidney contributes a small, but significant, fraction to overall glucose release in humans and rodents. The exact magnitude of this contribution remains controversial, probably as a result of methodological differences (76). However, there is agreement that renal glucose production, like hepatic glucose production, is increased in diabetes (77–79). The ability of a dominant negative Foxo1 to partially inhibit G6p and Pepck expression in the kidney could be exploited in a gene therapy approach to reduce glucose production in diabetic patients without incurring the potentially lethal complication of hypoglycemia.
Supported by NIH grant DK-57539 and by a grant from the American Diabetes Association to D. Accili. We thank Norman P. Curthoys (Colorado State University) for his generous gift of LLC-PK1-FBPase+ cells and the following colleagues for providing cDNA probes: Janice Y. Chou (NIH), Richard W. Hanson (Case Western Reserve University), Markus Stoffel (Rockefeller University), Masato Kasuga and Wataru Ogawa (Kobe University), Karen Arden and William Biggs, III (University of California, San Diego). We thank Alan Tall (Columbia University) for making his laboratory available for the isolation of primary hepatocytes cultures.