Utilization of recombinant adenovirus and dominant negative mutants to characterize hepatocyte nuclear factor 4-regulated apolipoprotein AI and CIII expression.

Using recombinant adenoviral vectors and a dominant negative mutant of HNF-4, we have examined the contribution of hepatocyte nuclear factor 4 (HNF-4) to endogenous apolipoprotein AI and CIII mRNA expression. Overexpression of HNF-4 leads to a 7.4-fold increase in apolipoprotein CIII expression, while infection with the dominant negative mutant of HNF-4 reduces the level of apolipoprotein CIII mRNA by 80%, demonstrating that endogenous HNF-4 is necessary for apolipoprotein CIII expression. Experiments using the hepatoma cell lines, HepG2 and Hep3B, indicate that HNF-4 is also involved in the regulation of apolipoprotein AI expression in these lines. However, the effect of HNF-4 on apolipoprotein AI expression is much more dramatic in cell lines derived from intestinal epithelium. Infection of the intestinal-derived cell line IEC-6 with the HNF-4 adenovirus resulted in a greater than 20-fold increase in the level of apolipoprotein AI mRNA. These results indicate that HNF-4 does regulate apolipoprotein AI and CIII mRNA expression and suggest that HNF-4 is critical for intestinal apolipoprotein AI expression.

The apolipoproteins are lipid-binding polypeptides involved in the transport and metabolism of cholesterol, triglycerides, and phospholipids (1). These proteins regulate the structural characteristics of lipoprotein particles as well as their metabolism and uptake by cell surface receptors. While it has been demonstrated that genetic defects in apolipoprotein structure or biosynthesis can lead to severe disorders of plasma lipid transport and the development of atherosclerotic disease, recent clinical studies indicate that even relatively small imbalances in lipoprotein concentrations can increase the risk of atherosclerosis (2)(3)(4). Apolipoprotein gene transcription is a regulated process, and there have been significant advances in our understanding of the DNA elements that control apolipoprotein expression. However, for the most part, the identity of the factors that recognize these sequences and the signal transduction pathways that control the level of apolipoprotein expression remains unclear.
Hepatocyte nuclear factor 4 (HNF-4) 1 is a member of the steroid/thyroid superfamily of ligand-dependent transcription factors that was originally isolated from liver nuclear extracts (5). HNF-4 mRNA is present in the intestine, kidney, and pancreas as well as the liver (5,6). At present no ligand for HNF-4 has been identified, therefore HNF-4 is referred to as an orphan member of the intracellular receptor superfamily (7)(8)(9).
Whether an activity-modulating ligand for HNF-4 exists is not known, but HNF-4 is capable of activating transcription in the absence of exogenously added ligand (10 -12). The structure of HNF-4 is very highly conserved; there is 96% sequence identity at the amino acid level between the human and rat HNF-4s (13). Previous studies have identified several sequences capable of binding HNF-4 that are located within the promoters of apolipoproteins, including apolipoprotein AI (apoAI) and apolipoprotein CIII (apoCIII) (5,11,14,15). To examine the role of HNF-4 on apoAI and apoCIII expression we have utilized recombinant adenoviral vectors expressing wild-type and a dominant negative mutant of HNF-4. HNF-4 binds DNA as a homodimer and does not appear to form heterodimers with several other intracellular receptor family members, including retinoid X receptor ␣, ␤, ␥, retinoid acid receptor ␣, or thyroid hormone receptor ␣ (16). We have taken advantage of this and created a mutant of HNF-4 which lacks DNA binding activity and inhibits transcriptional activation by wild-type HNF-4 via the formation of heterodimers consisting of wild-type HNF-4 and the non-DNAbinding HNF-4 mutant that exhibit decreased DNA binding avidity. Since the apoAI and apoCIII genes are closely linked and appear to share regulatory elements (17), we have constructed recombinant adenoviral vectors for the dominant negative mutants and wild-type HNF-4 that has allowed us to examine the effects of modulating HNF-4 transcriptional activity on the endogenous expression levels of the apoAI and apo-CIII genes.

Expression Vectors and Reporter Constructs-The 1.4-kilobase pair
BglII-EcoRI apoAI (Ϫ1378 to ϩ11) promoter fragment was isolated from pGF1, which was a generous gift from Dr. Michael Saunders. This construct was sequenced and the apoAI promoter region was identical to that reported by Karathanasis and colleagues (18) This was ligated into BamHI and EcoRI digested Bluescript KSϪ (Stratagene) and digested with HindIII and SpeI. The apoAI promoter fragment was gelpurified and ligated into HindIII and NheI digested pGL2 (Promega). This construct (p1400 apoAI-luc) was digested with SmaI, the 5.9kilobase pair fragment containing the apoAI (Ϫ256 to ϩ11) and pGL2 sequences was isolated and religated to create p256A1-luc.
The construct p3XA was made by ligating together oligonucletides corresponding to Ϫ210 to Ϫ188 of the apoAI promoter that had BamHI overhangs (5Ј-GATCACTGAACCCTTGACCCCTGCCCT-3Ј) and then ligating the multimers into the BamHI site of pBL-tk-LUC (19).
The apoCIII reporter vector was created using 5Ј-ACGAGAGAAT-CAGTCCTGGT-3Ј and 5Ј-TGCCTCTAGGGATGAACT-3Ј as primers for polymerase chain reaction from human genomic DNA to create a frag-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: 9393 Towne Centre Dr., Ligand Pharmaceuticals Inc., San Diego, CA 92121. E-mail: jfraser@ligand.com. 1 The abbreviations used are: HNF4, hepatocyte nuclear factor 4; apoAI, apolipoprotein AI; apoCIII, apolipoprotein CIII; m.o.i., multiplicity of infection. ment containing from Ϫ810 to ϩ23 of the apoCIII promoter. The product was then ligated into Bluescript KSϪ via the BamHI and SpeI sites. The apoCIII promoter was sequenced and then the BamHI-SpeI fragment was ligated into pGL2 digested with BglII and SmaI to create p810CIII-luc.
The ⌬111HNF4 mutant was created using GATGGTACCGCCGC-CACCATGGACTTCCGGGCTGGCATGAAGAAAGAAGCC and T3 primer (Stratagene) for polymerase chain reaction from pK-HNF4␣1. The resulting truncated HNF-4 was also ligated into pCDNA3 to create pC-⌬111HNF4 ␣1. All polymerase chain reaction mutants were sequenced. Transient transfections of HepG2 and Caco2 cells were performed as described previously (17,19). Each transfection included Rous sarcoma virus-␤-galactosidase, and all luciferase values were normalized to ␤-galactosidase values. Each result is representative of at least three independent transfections.
Recombinant Adenovirus-Full-length HNF-4␣1 and ⌬111HNF-4␣1 were cloned into pACCMVpLpA (22) using the BamHI and HindIII sites to create pAC-HNF4␣1 and pAC-⌬111HNF4␣1, respectively. Recombinant adenovirus were prepared, purified, and titered as described previously (22). Multiplicity of infection (m.o.i.) that resulted in the highest percentage of cells being infected was determined by infection with Ad␤gal followed by staining for ␤-galactosidase activity.
HNF4 Antiserum-Peptides corresponding to amino acids 37-53 and 126 -148 of rat HNF-4 were synthesized along with a cysteine at the amino terminus for coupling. The peptides were then conjugated with keyhole limpet hemocyanin. New Zealand White rabbits were injected subcutaneously with 0.5 mg of linked peptide using standard procedures and subsequently boosted with 0.3-mg subcutaneous injections.
Western Blot Analysis-Nuclear extracts derived from adenovirusinfected HepG2 cells were separated on 10% SDS-polyacrylamide gels followed by electrophoretic transfer to polyvinylidene difluoride membrane (Bio-Rad). The membranes were blocked using 20 mM Tris (pH 8.0), 150 mM NaCl, 0.5% Tween 20 (TBST), and 5% non-fat dry milk for 1 h at room temperature and then incubated for 1 h with the rabbit anti-HNF4 126 -148 antisera. The blots were washed three times for 5 min with TBST and then incubated for 30 min with a horseradish peroxidase-conjugated goat anti-rabbit immunoglobulin antisera (Bio-Rad). After three washes the specific HNF-4 bands were detected using chemiluminescence (Pierce).
Northern Blot Analysis-RNA was extracted from the tissue culture cells using RNAzol (Tel-Test, Inc.). The RNA samples were electrophoresed (20 g/lane) and transferred to Hybond N (Amersham Corp.) as described in the protocol supplied by the manufacturer. The indicated mRNAs were detected using random primed probes generated from the 440-base pair PstI fragment of human apoCIII, the HindIII-PstI fragment of human apoAI, the 1,500-base pair BamHI-HindIII fragment from rat HNF-4, and the 350-base pair EcoRI-XhoI fragment fragment from human glyceraldehyde phosphate dehydrogenase. The hybridization conditions and washes were done using the Rapid-Hyb system (Amersham Life Sciences) as described by the manufacturer. To quantify differences in mRNA expression, results were analyzed by PhosphorImager (Molecular Dynamics) and normalized to the level of glyceraldehyde phosphate dehydrogenase signal.

Dominant Negative Mutants of HNF-4 -
To examine the role of HNF-4 on apolipoprotein gene transcription, we constructed expression vectors coding for full-length cDNAs of the HNF-4␣1 and HNF-4␣2 splice forms as well as a truncated, domi-nant negative form of HNF-4␣1 (Fig. 1A). ⌬111HNF-4␣1 contains a deletion of the amino-terminal 111 amino acids, which include the DNA-binding domain. ⌬111HNF4␣1 fails to display any detectable DNA binding activity (data not shown) and should act as a dominant negative of HNF-4-induced transcription due to dimerization with wild-type HNF-4. The subsequent reduction in DNA binding avidity has previously been demonstrated for DNA-binding domain mutants of several other members of the intracellular receptor superfamily (25)(26)(27)(28).
Cotransfection of HepG2 with a reporter construct containing three copies of the Ϫ210 to Ϫ188 portion of the apoAI promoter (p3XA-luc) along with expression vectors for either HNF-4␣1 or HNF-4␣2 resulted in 40 -50-fold increases in luciferase activity (Fig. 1B). Although there was some experimental variation, the two splice forms appeared to activate the p3XA-luc vector to aproximately the same degree. To determine whether the DNA binding domain mutant, ⌬111HNF-4␣1, inhibits HNF-4-induced transcription, submaximal levels of HNF-4␣1 or HNF-4␣2 were cotransfected into HepG2 cells along with p3XA-luc and increasing amounts of pC-⌬111HNF4 ␣1. The HNF-4␣1 deletion mutant blocked the transfectioninduced promoter activity of either HNF-4␣1 or HNF-4␣2 (Fig.  1C). These results indicate that this DNA-binding domain mutant of HNF-4␣1 behaves as a dominant negative and that HNF-4␣1 and HNF-4␣2 can dimerize with each other. This result appeared to be specific for HNF-4-induced activity as cotransfection of ⌬111HNF-4␣1 had no effect on peroxisome proliferator-activated receptor ␣ or retinoid X receptor ␣-induced promoter activity (data not shown).
ApoAI Expression-While these and previous studies have demonstrated that the Ϫ210 to Ϫ188 region of the apoAI promoter can bind HNF-4 and that artificial constructs containing multiple copies of this region express increased levels of reporter gene activity in response to cotransfected HNF-4, there is little evidence that increasing the level of HNF-4 leads to significant increases in apoAI promoter activity in hepaticderived cells. One possible explanation for this result is that the hepatoma cell lines used for these studies contain sufficient endogenous HNF-4 for maximal apoAI expression. To address this we cotransfected HepG2 cells with p256A1-luc, a construct that contains from Ϫ256 to ϩ11 of the apoAI promoter linked to firefly luciferase, and the individual HNF-4 expression vectors. Cotransfection of either pCHNF-4␣1 or pCHNF-4␣2 resulted in only small changes in apoAI promoter activity (Fig. 2, A  and B). However, cotransfection of the dominant negative pC⌬111HNF4␣1 reduced apoAI promoter expression by 51%, suggesting that HNF-4 does contribute substantially to apoAI gene expression in this cell line (Fig. 2C). In contrast to HepG2 cells, cotransfection of the intestinal cell line Caco2 with pCHNF4␣1 resulted in a 4-fold activation of apoAI-luc (Fig.  2D). Surprisingly, cotransfection with the dominant negative pC⌬111HNF4␣1 failed to significantly affect apoAI promoter activity in Caco2 cells (Fig. 2E).
ApoCIII Expression-Cotransfection of either pCHNF4␣1 or pCHNF4␣2 along with a promoter construct containing from Ϫ810 to ϩ23 of the apoCIII promoter linked to firefly luciferase (p810CIII-luc) resulted in 4 -5-fold increases in luciferase activity (Fig. 3, A and B). This indicates that, similar to the response observed using the construct consisting of multiple copies of the apoAI A site, both of these forms of HNF-4 can recognize the apoCIII HNF-4 response element(s) and that both have roughly the same activation potential for this promoter. To address whether or not endogenous HNF-4 is involved in the basal expression of the apoCIII promoter by HepG2, cells were cotransfected with p810CIII-luc and increasing amounts of pC⌬111HNF4␣1. Cotransfection of the dominant negative resulted in a 70 -80% decrease in p810CIII-luc promoter activity (Fig. 3C).

Generation of Recombinant Adenovirus Vectors for the Expression of HNF-4 -It
can be difficult to recapitulate the bona fide regulation of endogenous genes from recombinant promoter constructs that may lack critical regulatory sequences. To address this limitation, we generated recombinant adenovirus vectors to express HNF-4␣1 and the dominant negative deletion mutant ⌬111HNF-4␣1. In many cell lines, adenovirus vectors can achieve nearly 100% infection, thus allowing us to examine the effect of expressing either wild-type HNF-4␣1 or the dominant negative mutant on endogenous gene expression. The ability of the recombinant adenoviral vectors to express HNF-4␣1 was tested by infection of HepG2 cells at titers of adenovirus that resulted in 80 -95% of the cells being infected (data not shown) followed by extract preparation and immunoblotting using a polyclonal antiserum specific for HNF-4. Infection of HepG2 cells with AdHNF4␣1 resulted in greatly increased levels of a 55-kDa polypeptide recognized by this antiserum compared with cells infected with a recombinant adenovirus that expresses ␤-galactosidase (Fig. 4A, lanes 1 and  4). Coinfection of HepG2 with AdHNF4␣1 and Ad⌬111HNF4␣1 resulted in the expression of an additional immunoreactive polypeptide with an apparent molecular mass of 42 kDa (Fig.  4A, lanes 2 and 3). Coinfection with the dominant negative mutant of HNF-4 did not have any detectable effect on the level of HNF-4␣1 expression (Fig. 4A).
Nuclear extracts prepared from HepG2 cells 2 days after infection with AdHNF4␣1 show a large increase in binding to double-stranded oligonucleotides corresponding to the apoAI (Ϫ210 to Ϫ188) (Fig. 4B). The major complex formed in the nuclear extracts from AdHNF4␣1-infected cells contains HNF-4 as it is recognized by anti-HNF-4 antiserum (Fig. 4B,  lane 3). Coinfection of HepG2 cells with the dominant negative truncation mutant of HNF-4 and wild-type HNF4␣1 resulted in substantially decreased HNF-4 DNA binding compared with extracts obtained from cells infected with AdHNF4␣1 alone (Fig. 4C).
Infection of papoCIII-luc transfected HepG2 cells with the adenoviral vectors encoding HNF-4 or the dominant negative mutant resulted in similar effects on apoCIII promoter activity as those obtained in the cotransfection experiments. Infection with AdHNF4␣1 lead to a 3-4-fold increase in luciferase activity, while infection with Ad⌬111HNF4␣1 resulted in a 70% decrease in reporter expression (Fig. 5). Reporter activity in HepG2 cells transfected with the p256A1-luc vector was unaffected by infection with AdHNF4␣1 and decreased 27% after infection with the dominant negative Ad⌬111HNF4␣1 (data not shown).
To determine whether the observed effects of HNF-4 modulation on recombinant promoter constructs was reflective of endogenous gene expression as well as to determine whether the tight genetic linkage between apoAI and apoCIII resulted in common HNF-4 regulatory elements, the hepatoma lines HepG2 and Hep3B as well as the intestinal cell lines Caco2 and IEC-6 cells were infected with AdHNF4␣1, Ad⌬111HNF4␣1, or Ad␤gal. Adenoviral infection levels of 80 -95% were obtained for the HepG2, Hep3B, and IEC-6 cell lines with no detectable toxicity. However we were unable to achieve greater than 30% adenoviral infection of the Caco2 cells without significant cell death, therefore adenoviral studies with this line were discon- viruses, indicating that infection with the viral vectors does not have significant nonspecific affects on cellular gene expression (Fig. 6). As was previously observed in the transient transfection assays with the apoAI promoter construct in HepG2 cells, the level of apoAI mRNA was relatively unaffected by expres-  4). B, gel mobility and supershift assays were performed using an oligonucleotide corresponding to Ϫ210 to Ϫ188 of the apoAI promoter.

HNF-4-regulated Apolipoprotein Expression
sion of the recombinant HNF-4␣1 ( Fig. 6 and Table I). However, addition of AdHNF-4␣1 did result in a 3-fold increase in apoAI expression in Hep3B cells, and infections with AdHNF-4␣1 resulted in a dramatic induction of apoAI mRNA expression in the rat intestinal cell line IEC-6 ( Fig. 6 and Table I). Infection with the dominant negative Ad⌬111HNF-4 ␣1 caused a 33 and 54% decrease in HepG2 and Hep3B, respectively. The basal level of apoAI mRNA expression was quite low in the IEC-6 cells and addition of the dominant negative did not detectably alter apoAI mRNA levels in this cell line (Table I).
In contrast to the effect on apoAI expression in HepG2 cells, expression of exogenous HNF-4 increased the level of apoCIII mRNA, while infection with the dominant negative mutant Ad⌬111HNF4␣1 resulted in decreased message levels (Fig. 6). Quanitation of the apoCIII mRNA levels revealed a 7-fold increase in apoCIII mRNA after infection with AdHNF4␣1 and an 80% decrease after addition of the dominant negative mutant ( Table I); confirming that HNF-4 is a significant contributor to apoCIII gene expression. We were unable to detect basal apoCIII expression in either the Hep3B or IEC-6 cells and infection with the HNF-4 adenovirus failed to induce detectable apoCIII mRNA (data not shown). DISCUSSION HNF-4 is an orphan member of the steroid/thyroid superfamily that is expressed in the liver and small intestine, where the majority of apolipoproteins are synthesized. HNF-4 is capable of binding to sequences present in the apolipoprotein A1, apolipoprotein B, apolipoprotein CIII, and apolipoprotein AIV promoters and has been proposed to regulate the expression of these apolipoprotein genes (11,12,17,29,30). However, the evidence for HNF-4-regulated apolipoprotein expression has been primarily based on the existence of sequences within the apolipoprotein promoters to which HNF-4 binds in gel shift experiments and transient transfections utilizing overexpressed HNF-4 and recombinant promoter-reporter constructs.
To complement experiments utilizing overexpression of HNF-4, we have constructed a mutant of HNF-4␣1 that lacks the DNA-binding domain. Since HNF-4 binds to DNA as a homodimer and the ⌬111HNF-4␣1 mutant contains the sequences necessary for dimerization, ⌬111HNF-4␣1 should form dimers with full-length HNF-4 that have reduced DNA binding avidity similar to that previously described for mutants of c-erb-␣, thyroid hormone receptor, and the estrogen receptor (25)(26)(27)(28). HNF-4 fails to form heterodimers with several of the other intracellular receptors (16), suggesting that the dominant negative HNF-4 mutant should act as a specific inhibitor of HNF-4-induced transcription. The ⌬111HNF-4␣1 mutant demonstrates no detectable DNA binding activity and efficiently blocks both HNF-4␣1 and ␣2-induced transcription, indicating that the two splice forms can dimerize with each other despite the sequence changes within the carboxyl-terminal domain. While we could not detect any affect of ⌬111HNF-4␣1 cotransfection on promoter activity induced by other intracellular receptors, it is also possible that the carboxyl-terminal domain of HNF-4 sequesters out coactivators that are necessary for HNF-4-induced activity.
In agreement with previous studies (11,12,30), cotransfection of HepG2 cells with HNF-4␣1 or ␣2 had fairly modest affects on apoAI promoter expression. However, expression of the dominant negative in HepG2 cells does lead to a 40 -50% decrease in apoAI promoter-activated expression, suggesting that endogenous HNF-4 contributes to apoAI expression in this cell model. Utilizing the recombinant adenoviral vectors, we have extended these findings to examine endogenous apoAI gene expression. The effect of increasing HNF-4 expression on endogenous apoAI mRNA levels in HepG2 cells is simlar to the results of the recombinant promoter assays in that increasing HNF-4 levels did not change apoAI expression, whereas expression of the dominant negative mutant decreases expression of apoAI mRNA. Overexpression of HNF-4 leads to a 3-fold increase in apoAI mRNA levels in the Hep3B cell line, while HNF-4 overexpression appears to have a much more pronounced effect on apoAI expression in cell lines derived from intestinal epithelium. Transient transfection of HNF-4 up-regulated the promoter activity of apoAI-luciferase constructs by 3-5-fold in the enterocyte-like cell line Caco2 as has recently been described by Bisaha et al. (17). Infection of the IEC-6 cell line with the adenoviral vector encoding wild-type HNF-4 dramatically increased apoAI mRNA expression by greater than 20-fold. Both of these cell lines appear to have less HNF-4 than do the hepatic derived HepG2 or Hep3B, therefore magnifying the affect of adding exogenous HNF-4. But even so, the large induction using the adenovirus on the endogenous apoAI gene suggests a more critical role for HNF-4 on intestinal expression of apoAI. It is tempting to speculate that these differences are related to the previously described intestinal control region contained within the closely linked apoCIII gene, as this region does contain HNF-4 response elements that could contribute to apoAI gene expression by intestinal cells (17,30,31).
Unlike the apoAI promoter assays in the HepG2 cells, cotransfection of either the ␣1 or ␣2 splice forms of HNF-4 leads to increased reporter expression from an apoCIII reporter construct. The strong induction of apoCIII expression by HNF-4 is also observed at the level of apoCIII mRNA following infection with the adenovirus encoding wild-type HNF-4. More importantly, infection with the dominant negative mutant of HNF-4 results in an 80% decrease in apoCIII mRNA levels. This indicates that endogenous HNF-4 is required for apoCIII promoter activity and suggests that modulation of HNF-4 transcriptional activity would lead to significant changes in the level of apoCIII expression. High levels of apoCIII correlate with increased fasting triglycerides in both clinical hypertriglyceridemic patients and murine model systems (32)(33)(34). Recent clinical studies suggest that high serum triglycerides are directly atherogenic as well as being inversely correlated with high density lipoprotein cholesterol levels, a well documented risk factor for atherosclerotic disease (35,36). Therefore modulation of apo-CIII expression via HNF-4 might represent a potential therapeutic target. A primary question is whether HNF-4 activity is regulated or not. HNF-4 is capable of activating transcription in the absence of exogenously added ligand. Therefore, if HNF-4 requires a ligand for transcriptional activation, it must be present under normal tissue culture conditions as has been found to be the case for nuclear receptors such as peroxisome proliferator-activated receptor and farnesoid X-activated receptor (37,38). The question of whether or not an unidentified ligand actually regulates HNF-4-induced transcription awaits future studies as does elucidation of HNF-4s role in lipid metabolism and cardiovascular disease. P]cDNA hybridization was determined using a Phos-phorImager (Molecular Dynamics). Specific binding to each band was obtained by subtracting the average signal from three randomly selected control sections of equal size. The results are expressed as the specific binding divided by the specific binding of the same cell line infected with the control Ad␤gal virus. Infection with the Ad␤gal virus did not affect mRNA expression level for any of the tested cDNAs in any of these cell lines. These results are representative of three individual infections and RNA extractions.