Searching for Interferon-Induced Genes That Inhibit Hepatitis B Virus Replication in Transgenic Mouse Hepatocytes

Abstract

We have previously shown that alpha/beta interferon (IFN-α/β) and IFN-γ inhibit hepatitis B virus (HBV) replication noncytopathically in the livers of HBV transgenic mice and in hepatocyte cell lines derived from these mice. The present study was designed to identify transcriptionally controlled hepatocellular genes that are tightly associated with the inhibition of HBV replication and that might, therefore, mediate the antiviral effect of these cytokines. Twenty-nine genes were identified, many of which have known or potential antiviral activity. Notably, multiple components of the immunoproteasome and ubiquitin-like proteins were strongly induced by both IFN-α/β and IFN-γ, as were a number of GTP-binding proteins, including GTPases with known antiviral activity, chemokines, signaling molecules, and miscellaneous genes associated with antigen processing, DNA-binding, or cochaperone activity and several expressed sequence tags. The results suggest that one or more members of this relatively small subset of genes may mediate the antiviral effect of IFN-α/β and IFN-γ against HBV. We have already exploited this information by demonstrating that the antiviral activity of IFN-α/β and IFN-γ is proteasome dependent.

Hepatitis B virus (HBV) is a hepatotropic, noncytopathic DNA virus that causes acute and chronic necroinflammatory liver disease and hepatocellular carcinoma (13). We and others have demonstrated that viral hepatitis during HBV infection is characterized by the production of inflammatory cytokines such as gamma interferon (IFN-γ) by HBV-specific T cells in the liver (4, 20, 21, 88). Furthermore, we have shown that IFN-γ is strongly expressed in the liver during viral clearance in acutely HBV-infected chimpanzees (88). Accordingly, we have suggested that these cytokines, especially IFN-γ, might play a role in viral clearance and disease pathogenesis during this infection (13, 32). To test this hypothesis, we have used HBV transgenic mice that replicate the virus in the liver (34) to explore the antiviral and pathogenetic potential of IFN-γ and various other cytokines (reviewed in reference 31). In these studies we have shown that HBV DNA replication is abolished noncytopathically by IFN-γ and tumor necrosis factor alpha (TNF-α) produced by adoptively transferred HBV-specific cytotoxic T lymphocytes (CTLs) (33). Experiments using antibodies against IFN-γ and TNF-α or utilizing either IFN-γ-deficient or TNF-α receptor-deficient mice indicated that IFN-γ mediates most of the antiviral effect of the CTLs (58). Similar IFN-γ-dependent antiviral mechanisms in these animals are observed following injection of interleukin-12 (IL-12) (12), IL-18 (47), anti-CD40 (an agonistic antibody activating antigen-presenting cells to produce IFN-γ) (47a) antibodies or α-galactosylceramide (a glycolipid antigen capable of specifically activating NKT cells) (44) or following infection of the mice with adenovirus (11), murine cytomegalovirus (11), or lymphocytic choriomeningitis virus (LCMV) (58).

In the LCMV system we also showed that the intrahepatic induction of IFN-α/β inhibits HBV DNA replication (30). The major contribution of IFN-α/β to this process was demonstrated by showing that the antiviral activity is completely blocked by antibodies to IFN-α/β (11, 30), and antiviral activity was not detectable in mice genetically deficient for the IFN-α/β receptor (58). Similarly, injection of HBV transgenic mice with polyinosinic-polycytidylic acid-poly(I-C) complex (58, 89) or infection with adenovirus inhibits HBV replication by an IFN-α/β-dependent mechanism (30). Furthermore, direct injection of IFN-α/β results in inhibition of HBV replication in the livers of HBV transgenic mice (58). In addition, we showed that IFN-α/β inhibits HBV replication in the transgenic mouse liver by inhibiting the formation and/or promoting the destabilization of immature HBV RNA-containing capsids (64).

Recently, we established immortalized and highly differentiated hepatocyte cell lines (HBV-Met.4) from these same HBV transgenic mice (64). These differentiated hepatocyte cultures support HBV gene expression and replication, and, importantly, they were shown to be sensitive to the antiviral activity of IFN-γ and IFN-α/β but not TNF-α (64).

IFN-γ and IFN-α/β are known to induce an intracellular antiviral state effective against a variety of viruses (for a review see reference 73). IFN-inducible genes such as the genes encoding RNA-dependent protein kinase (PKR), RNase L, and Mx GTPases have also been shown to inhibit the replication of many viruses (73). In addition, experiments with mice (91) and cell cultures (69) that are deficient in these factors suggest that additional pathways may also contribute to the antiviral activity of IFNs. In support of this notion, DNA microarray-based studies have already demonstrated transcriptional regulation of a broad range of novel host genes upon cytokine administration (17, 18) as well as during viral infections (8, 16, 24, 25, 43, 60, 67). Furthermore, we have recently shown that the antiviral activity of IFNs in HBV transgenic mice is not mediated by Mx, RNAse L, PKR, or IFN regulatory factor 1 (IRF-1), although mice deficient in PKR or IRF-1 replicate HBV in their livers at levels that are higher than those for the respective controls (35). Thus, although both PKR and IRF-1 may individually mediate signals that modulate HBV replication under basal conditions, it is possible that other pathways involving yet-undefined IFN-regulated genes mediate the antiviral activity of IFNs against HBV.

The present study was performed to identify hepatocellular genes whose transcriptional regulation is tightly linked with IFN-γ- and IFN-α/β-mediated inhibition of HBV replication. Using DNA microarrays and a high-throughput cDNA differential display method (total gene expression analysis [TOGA] [86]) we examined the gene expression profiles of HBV transgenic mouse livers before and after the intrahepatic induction of IFN-α/β and IFN-γ and the corresponding inhibition of HBV replication. Similarly, the gene expression profiles of highly differentiated hepatocyte cell lines that replicate HBV were monitored before and after administration of IFN-α/β and IFN-γ. Twenty-nine hepatocellular genes that are regulated by both IFN-α/β and IFN-γ in mouse liver and differentiated hepatocyte lines were identified. We suggest that one or more of these genes might mediate the antiviral effect of both IFN-α/β and IFN-γ on HBV replication.

MATERIALS AND METHODS

HBV transgenic mice.

The HBV transgenic mice used in this study, lineage 1.3.32, official designation (Tg[HBV 1.3 genome] Chi32) have been previously described (34). These mice replicate HBV in the hepatocytes from an integrated greater-than-genome-length HBV transcriptional template. The level of HBV replication in the livers of these mice is comparable to that seen in the infected livers of patients with chronic hepatitis, and there is no evidence of cytopathology (34). Experiments were performed with mice that were matched for age (8 to 10 weeks), sex (male), and level of HBV e antigen (HBeAg) in the serum (measured with commercially available reagents from Abbott Laboratories, Abbott Park, Ill.).

CTL injection and LCMV infection.

HBV transgenic mice were injected with an HBsAg-specific, H-2^d-restricted, CD8⁺ CTL clone (designated 6C2) that recognizes an epitope (IPQSLDSWWTSL) located between residues 28 and 39 of HBsAg (2, 40, 59). Clone 6C2 was maintained as previously described (59). Five days after the last stimulation, the cells were washed and counted and 10⁷ cells were injected intravenously (i.v.) into HBV transgenic mice. Mice were sacrificed at various time points after injection, and their livers were harvested and snap-frozen in liquid nitrogen and stored at −80°C for subsequent molecular analyses (see below). HBV transgenic mice were also infected by i.v. inoculation of 2 × 10⁶ PFU of LCMV isolate WE clone 2.2 as described previously (30) and sacrificed 12 h later.

Anticytokine antibodies.

Hamster monoclonal antibody H22, specific for murine IFN-γ (74), and hamster monoclonal antibody TN3 19.12, specific for murine TNF-α (77) (both were generously provided by Robert Schreiber, Washington University, St. Louis, Mo.) were used as described previously (33). Mice were injected intraperitoneally with a cocktail of these antibodies (250 μg of each/mouse) 16 h before the i.v. injection of the CTLs. A cocktail of TN3 19.12 (250 μg/mouse) and neutralizing sheep immunoglobulin (Ig) against murine IFN-α/β (200 μl/mouse; generously provided by Ion Gresser [27]) was also used as described previously (30). One dose of each antibody was administered i.v., first 1 h before LCMV infection and then simultaneously with LCMV infection as described previously (30). Irrelevant hamster IgG (Jackson Immune Research, West Grove, Pa.) and normal sheep Ig were used as control antibodies.

HBV-Met cells.

Immortalized HBV transgenic hepatocyte cell lines were established from HBV-Met doubly transgenic mice as described elsewhere (64). HBV-Met.4 cells were grown to confluence and subsequently kept in complete medium supplemented with 2% dimethyl sulfoxide for 10 days to allow for hepatocyte differentiation and HBV gene expression and replication (64). Recombinant murine IFN-β (mIFN-β; provided by M. Moriyama, Toray Industries, Tokyo, Japan), recombinant mIFN-γ, and recombinant murine TNF-α (mTNF-α) (provided by S. Kramer, Genentech, South San Francisco, Calif.) were then added at 100 U/ml to the culture medium.

HBV DNA analysis.

Frozen liver tissue from HBV transgenic mice was processed as described previously (34), and 30 μg of total DNA was analyzed by Southern blotting after HinDIII digestion (34). HBV-Met.4 cells were lysed in the culture dish by adding 500 μl of DNA lysis buffer (50 mM Tris-HCl [pH 8.0], 20 mM EDTA, 1% sodium dodecyl sulfate). Samples were then digested overnight at 37°C with proteinase K (1 mg/ml), and total DNA was extracted as described previously (34). Twenty micrograms of total DNA was analyzed by Southern blotting with a ³²P-labeled full-length HBV DNA probe after HinDIII digestion (34).

RNA isolation.

RNA from frozen liver tissue from HBV transgenic mice and from HBV-Met.4 cells was isolated by the guanidine thiocyanate method (14) as described previously (34).

DNA microarray analysis.

Liver RNA and HBV-Met.4 cell RNA were analyzed with Mu11K-A and -B and Mu74A oligonucleotide arrays, respectively (Affymetrix, Inc., Santa Clara, Calif.). The biotinylated cRNA was synthesized according to the Affymetrix protocol. The sample hybridization procedure had the following modifications. One hundred ten microliters of a 3× hybridization mixture was prepared and hybridized for 16 h on a TEST2 chip. The hybridization mixture was collected the next day and kept at 4°C until the quality of the sample was determined. A good hybridization mixture was brought up to 310 μl with 1× NTMES buffer (1 M NaCl, 0.1% Triton, 1 × MES [morpholineethanesulfonic acid; 12× MES according to Affymetrix]) and hybridized on a murine chip. All chips were completely filled with sample and laid flat in an oven at 45°C for Mu11kA and -B and 50°C for Mu74A. Posthybridization the chips were double stained by the following procedure. The hybridization mixture was removed from the array and stored at −20°C. The chip was incubated with 0.1× NTMES buffer (0.1 M NaCl, 0.1% Triton, 1× MES) for 30 s at 45°C to remove nonspecific hybridization. The chip was stained with SAPE (R-phycoerythrin-streptavidin at 10 μg/ml, 0.5 mg of bovine serum albumin [BSA]/ml, 1× MES) at 37°C for 15 min. The amplification of the fluorescence signal was obtained with the antibody stain (goat biotinylated antistreptavidin antibody at 1.25 μg/ml, 0.5 mg of BSA/ml, 1 × MES) at 37°C for 30 min, followed by another SAPE stain. The chip was filled with 1× MES every time a solution was removed, and a wash was done with 6× SSPE (1× SSPE is 0.18 M NaCl, 10 mM NaH₂PO₄, and 1 mM EDTA [pH 7.7]) after each stain.

Image files of the hybridization signal were thoroughly checked for scratches and specks. The arrays were analyzed with Genechip, version 3.1 (Affymetrix, Inc.). Quality reports were created for all absolute and comparison files. Duplicate experiments were checked for consistency. This analysis yields values for fold change in gene expression between experimental and control samples. If the absolute gene expression level in the control or experiment samples is not different from the hybridization background, the background signal is used to calculate an approximate fold change. The Genechip data for Mu74A was first filtered with Nfueggo, version 4.5 (Lockhart, GNF, La Jolla, Calif.). The filtering criterion was to select genes called PRESENT in at least one file with an absolute average difference change (AvgDiffChange) of at least 75.

For further data filtering, the Genechip data files were imported into Panorama (ProVUE Development, Inc., Huntington Beach, Calif.). For the Mu11K data set, genes were considered induced when the absolute call in the experimental sample (ExpCall) was “not absent” (A), the difference call (DiffCall) was “increased” (I) or “medium increase” (MI), and the fold change was ≥2 in duplicate experiments. Genes were considered to be repressed when the absolute call of the saline sample (AbsCall) was not A, the DiffCall was “decrease” (D) or “medium decrease” (MD), and fold change was ≥2 in duplicate experiments.

For induced genes in the Mu74A data set the ExpCall had to be different from A. In addition, the AvgDiffChange had to be ≥100 with a fold change of ≥2, or ≥500 if the fold change was between 1 and 2. Genes were considered to be repressed when their AbsCall in the saline sample was different from A and the average difference (AvgDiff) was ≥50. In addition, the fold change had to be ≥2, or the AvgDiffChange had to be ≤−200 if the fold change was between 1 and 2.

TOGA.

TOGA was performed as described previously (86) on mRNA samples isolated from the livers of HBV transgenic mice as described above. cDNA was prepared from each RNA sample following reverse transcription with a degenerate pool of biotinylated oligo(dT) anchor primers and digested with MspI. The 3′ end fragments were isolated by streptavidin bead capture and subsequently modified at their 5′ ends to harbor a T3 polymerase start site by ligation of an oligonucleotide adapter to the MspI overhang. A cRNA pool was produced from each sample following in vitro transcription. For the initial PCR step, first-strand cDNA was prepared from the cRNA pool by reverse transcription and used in four separate PCRs, in which a 5′ primer that extends by one of four possible nucleotides beyond the MspI site (N1 position) was paired with a universal 3′ primer to generate an N1-specific double-stranded DNA template. In the final step, 256 primers corresponding to all possible permutations of the four nucleotides immediately adjacent to the MspI site (N1N2N3N4) were matched with the appropriate N1 template and paired with a fluorescent primer to produce 256 separate reactions in individual robotically performed PCRs. PCR products were resolved by electrophoresis on ABI Prism 377 DNA sequencers. The peak amplitudes of the fluorescent PCR products correspond to the initial concentrations of their parent mRNAs; following normalization, the data were stored in a database, indexed, and queried to identify mRNAs whose concentrations differed among the experimental samples.

Based on regulation patterns, some PCR products were selected for further characterization. The PCR product was isolated and either directly sequenced or cloned into a TOPO vector (Invitrogen, Carlsbad, Calif.) and sequenced on both strands. By comparing the sequence of the PCR product with GenBank and expressed sequence tag (EST) databases, it was possible in most cases to assign a gene or EST identity.

Quantitative real-time PCR.

Transcript levels in total liver and in HBV-Met cells were determined by quantitative real-time reverse transcription-PCR (RT-PCR) using an iCycler system (Bio-Rad Laboratories, Inc., Hercules, CA). PCR primers were designed with Oligo software (Molecular Biology Insights, Inc., Cascade, Colo.). Briefly, 1 μg of total RNA was reverse transcribed in a 50-μl reaction mixture by using hexamer primers and the Taqman reverse transcription reagents (Applied Biosystems, Foster City, Calif.) according to the manufacturer's instructions. Subsequently, 5 μl of cDNA was subjected to quantitative real-time PCR in a 50-μl reaction mixture containing 200 nM concentrations of both upper and lower primers and SYBR Green PCR master mixture (Applied Biosystems). Samples were run in triplicate, and the results were normalized to the GAPDH (glyceraldehyde 3-phosphate dehydrogenase) transcript levels that were determined in separate quantitative PCRs.

RESULTS

Cytokine-induced inhibition of HBV replication in vivo and in vitro.

Southern blot analysis of HBV DNA replicative intermediates in total liver DNA confirms our previous reports that HBV replication in the livers of HBV transgenic mice is inhibited within 24 h after the injection of virus-specific CTLs (33, 45) (Fig. 1A). Furthermore, the antiviral effect of the CTLs is blocked by coinjection of a cocktail of anti-IFN-γ- and anti-TNF-α-specific antibodies (33) (Fig. 1B). Figure 1C illustrates that infection of HBV transgenic mice with LCMV also inhibits HBV replication and that this antiviral mechanism is blocked by the administration of anti-IFN-α/β and anti-TNF-α antibodies (30). Together with subsequent studies utilizing HBV transgenic mice deficient in the cellular receptors for TNF or IFN-α/β or the IFN-γ gene (data not shown) (58), these results suggest that IFN-γ or IFN-α/β induced in these systems may regulate hepatocellular genes that, in turn, inhibit HBV replication. These findings were further validated when we added each of the individual cytokines to a highly differentiated immortalized HBV transgenic mouse hepatocyte cell clone (HBV-Met.4) that replicates HBV (64). Figure 1D shows that HBV DNA replication is suppressed in these cells within 24 h of mIFN-β or mIFN-γ treatment but not after mTNF-α treatment. Taken together, these results suggest that IFN-α/β and IFN-γ modulate common or converging intracellular pathways that ultimately suppress HBV DNA replication in the hepatocyte (58, 64).

FIG. 1.

Inflammatory cytokines inhibit HBV replication in HBV transgenic mice and in immortalized hepatocyte cultures. Total liver DNA isolated from HBV transgenic mouse liver or HBV-Met.4 cells was subjected to Southern blot analysis. (A) Virus-specific CTLs inhibit HBV replication. HBV transgenic mice were i.v. injected with saline (NaCl) or CTLs, and total liver DNA was harvested at the indicated time points after CTL administration. (B) Virus-specific CTLs inhibit HBV replication through an IFN-γ- and TNF-α-dependent pathway. HBV transgenic mice were i.v. injected with CTLs 24 h after intraperitoneal injection of irrelevant hamster IgG or 250 μg of anti-IFN-γ and anti-TNF-α monoclonal antibodies, and the mice were sacrificed 48 h after the CTL administration. (C) LCMV infection induces IFN-α/β and TNF-α, which mediate the inhibition of HBV replication. HBV transgenic mice were i.v. injected with a combination of anti-IFN-α/β and anti-TNF-α or IgG antibodies 1 h before LCMV injection and simultaneously with LCMV injection and were sacrificed 12 h later. (D) Cytokine-mediated inhibition of HBV replication in immortalized HBV-Met.4 hepatocyte cell cultures. The HBV-Met.4 cells were grown to confluence and then maintained in 2% dimethyl sulfoxide for 10 days. At this point parallel dishes were treated with mIFN-β (100 U/ml), mIFN-γ (100 U/ml), or mTNF-α (1,000 U/ml), and total cellular DNA was harvested 6 or 24 h later. The bands for the HBV transgene (Tg) and double-stranded (DS) and single-stranded (SS) HBV DNA replicative intermediates are indicated.

Experimental strategy to identify hepatocellular genes whose expression is tightly associated with inhibition of HBV replication.

Based on these results we attempted to identify genes that are induced or repressed by both IFN-γ and IFN-α/β and that are temporally linked to the disappearance of HBV DNA. Accordingly, we analyzed the gene expression profile in the livers of HBV transgenic mice before and after the adoptive transfer of HBV-specific CTLs or infection with LCMV, which inhibit HBV replication by IFN-γ- and IFN-α/β-dependent mechanisms, respectively. The genes identified by this strategy can be expressed in any or all of the cells in the liver, not only the hepatocyte. Since HBV replication is confined to hepatocytes in vivo (34), candidate antiviral genes must be regulated in the hepatocyte. This subset of genes was identified by determining the impact of IFN-α/β and IFN-γ on the gene expression profiles of highly differentiated immortalized hepatocyte cell lines that replicate HBV. Genes that are transcriptionally regulated under all of these conditions are strongly associated with cytokine-dependent antiviral activity. To confirm this association, additional experiments were performed to determine if the regulation of these genes in the liver was blocked by the administration of anticytokine antibodies that block the antiviral effects of CTL injection and the LCMV infection.

Identification of cytokine-regulated liver genes associated with inhibition of HBV DNA replication in vivo.

Livers from CTL-injected and from LCMV-infected HBV transgenic mice were analyzed with the Affymetrix Mu11K oligonucleotide microarray set comprising approximately 11,000 unique murine genes and ESTs. As shown in Fig. 1A and B, between 4 to 24 h must elapse for HBV-specific CTLs to inhibit HBV DNA replication in the livers of HBV transgenic mice and inhibition is even more profound at 48 h after CTL injection. Since antiviral genes may be induced before and during this period, we performed DNA microarray analysis on a pool of liver RNA that was isolated from mice shown in Fig. 1A that were sacrificed 4, 24, or 48 h after CTL injection. These results were compared with the gene expression profiles derived from liver RNA isolated from a control group of mice that were injected with saline (Fig. 1A).

Since HBV DNA replication is maximally inhibited 12 h after LCMV infection (Fig. 1C, lane LCMV + IgG), we performed DNA microarray analysis on total liver RNA from the mice shown in Fig. 1C that were sacrificed 12 h after LCMV infection, and these results were compared with those obtained from the RNA isolated from the saline-injected control mice. Induced or repressed liver genes were selected from individual hybridization experiments using the selection criteria described in the Materials and Methods.

As shown in Table 1, 176 genes were induced in the liver between 4 and 48 h after injection of virus-specific CTLs whereas 111 genes were induced 12 h after LCMV infection of HBV transgenic mice, and 65 of these genes were induced under both conditions. DNA microarray analysis also revealed 19 and 27 genes that were transcriptionally repressed 4 to 48 h after CTL injection and 12 h after LCMV infection, respectively, and 7 of these genes were suppressed under both conditions (Table 1). The identity and the complete profiling data for these and all the other regulated genes in the livers of CTL-injected or LCMV-infected HBV transgenic mice are available online at http://www.scripps.edu/mem/expath/wieland/supplement.html.

TABLE 1.

Selection of IFN-regulated hepatocellular genes associated with inhibition of HBV replication

Condition	No. of genes or ESTsa
	Induced	Repressed
In vivob
CTL	176	19
LCMV	111	27
Both	65	7
In vitroc
mIFN-γ	80	48
mIFN-β	149	18
Both	47	2
In vitro and in vivo	22	0

^aInduced or repressed compared to saline-injected control mice (in vivo) or untreated hepatocyte cultures (in vitro).

^bPool of liver RNA harvested 4, 24, and 48 h after CTL injection (CTL) or 12 h after LCMV infection (LCMV).

^cHepatocyte RNA harvested 6 h after cytokine addition.

IFN-α/β- and IFN-γ-regulated hepatocellular genes associated with inhibition of HBV replication in vitro.

Since potential antiviral genes must be expressed in the hepatocyte in order to inhibit HBV replication, we identified genes that are transcriptionally regulated by IFN-γ and IFN-α/β in the highly differentiated hepatocyte line HBVMet.4, which replicates HBV. Based on the kinetics of clearance of HBV replicative intermediates from the HBVMet.4 cells (Fig. 1D) (64), labeled cDNA produced from total cellular RNA harvested from HBVMet.4 cells 6 h into cytokine treatment or from untreated control cultures was used to hybridize Affymetrix oligonucleotide microarrays (murine genome U74A) containing essentially the same 11,000 genes and ESTs present on the former Mu11K set.

By the selection criteria described in Material and Methods, 80 genes were induced by mIFN-γ and 149 genes were induced by mIFN-β and 47 were induced under both conditions (Table 1). Furthermore, 48 and 18 genes were repressed by mIFN-γ and mIFN-β, respectively, and 2 genes were repressed under both conditions (Table 1). The identity and the complete profiling data for these and all the other cytokine-regulated genes in the hepatocytes are shown in online supplementary Table A. Based on our previous report (64) that HBV replication in the HBVMet.4 hepatocyte line is probably inhibited by the same mechanism(s) that is operative in the livers of HBV transgenic mice, we further selected the cytokine-regulated genes that were shared between the in vivo and in vitro gene expression profiling analyses. As shown in Table 1 this resulted in a minimal set of 22 hepatocellular genes that are induced after cytokine induction in vivo or cytokine administration in vitro under conditions in which HBV replication is inhibited. These genes are shown in Table 2, and the complete profiling data for these genes are shown in online Table B.

TABLE 2.

IFN-γ- and IFN-α/β-regulated hepatocellular genes tightly associated with inhibition of HBV replication

Gene or EST	Description	Unigene	Fold change in transcriptiona in:
			Hepatocytesb		Liver
			DNA array		DNA array		TOGA
			mIFN-γ	mIFN-β	CTLc	LCMVd	CTLe	LCMVd
Protein degradation
Psmb10	Proteasome (prosome, macropain) subunit, beta type 10 (MECL-1)	Mm.787	4	2	∼47	∼8	—	—
Psmb9	Proteasome (prosome, macropain) subunit, beta type 9 (large multifunctional protease 2)	Mm.16251	∼184	∼53	19	∼12	5	5
Psmb8	Proteasome (prosome, macropain) subunit, beta type 8 (large multifunctional protease 7)	Mm.13913	11	5	6	4	6	5
Psme2	Proteasome (prosome, macropain) 28 subunit, beta	Mm.626	2	2	2	—	3	3
Usp18	Ubiquitin specific protease 18	Mm.27498	∼39	∼247	2	15	—	—
Isg15	IFN-stimulated protein (15 kDa)	Mm.4950	∼77	∼458	10	60	—	—
GTP-binding proteins
Gtpi	IFN-γ-induced GTPase	Mm.33902	21	7	ND	ND	5	14
Tgtp	T-cell specific GTPase	Mm.15793	∼609	∼388	41	∼47	—	—
Ifi1	IFN-inducible protein 1 (LRG47)	Mm.29938	11	10	5	7	—	—
Iigp	IFN-inducible GTPase (IIGP)	Mm.29008	∼943	∼554	4	5	4	2
Ifi47	IFN-γ-inducible protein, 47 kDa	Mm.24769	∼547	∼248	9	13	—	—
Gbp2	Guanylate nucleotide binding protein 2	Mm.24038	∼873	∼288	53	∼8	—	—
Gbp3	Guanylate nucleotide binding protein 3	Mm.1909	22	17	13	∼11	—	—
Signaling
Scyb10	Small inducible cytokine B subfamily (Cys-X-Cys), member 10 (IP-10)	Mm.877	39	32	33	∼58	—	—
Pbef	Pre-B-cell colony-enhancing factor	Mm.28830	3	3	2	3	2	4
Stat1	Signal transducer and activator of transcription 1	Mm.8249	∼176	∼149	∼27	∼28	26	44
Miscellaneous and ESTs
Abcb2	ATP-binding cassette, subfamily B (MDR/TAP), member 2	Mm.16122	20	9	7	∼10	—	—
Mpeg1	Macrophage expressed gene 1	Mm.3999	3	3	8	7	—	—
Xdh	Xanthine dehydrogenase	Mm.11223	8	8	4	4	—	—
DLM-1	Tumor stroma and activated macrophage protein	Mm.116687	426f	577f	ND	ND	2	13
Tbrg1	Transforming growth factor beta-regulated gene 1	Mm.28689	2	3	ND	ND	17	10
Ifit 1	IFN-induced protein with TPR repeats 1	Mm.6718	∼28	∼554	3	∼37	—	—
Ifit 2	IFN-induced protein with TPR repeats 2	Mm.2036	∼65	∼319	∼16	∼14	—	—
Ifit 3	IFN-induced protein with TPR repeats 3	Mm.951	∼84	∼570	3	∼138	2	20
EST	ESTs, weakly similar to B chain human acyl protein thioesterase 1	Mm.532	3	8	7	∼7	—	—
EST	Expressed sequence AW112010	Mm.6381	6	5	4	3	—	—
EST	RIKEN cDNA 2010008K16 gene	Mm.45558	∼33	∼74	—	—	6	11
EST	RIKEN cDNA 4933435C21 gene (similar to equi- librative nucleoside transporter 3 [Ent3])	Mm.82642	1,000f	53f	ND	ND	11	18
EST	Weakly similar to contraspin (Mus musculus)	Mm.90033	−2f	−2f	ND	ND	−4	−3

^aCompared to saline-injected control mice (liver) or untreated hepatocyte cultures (hepatocytes).—, no change in gene expression; ND, sequence not present on Mu11K DNA array; ∼, approximate value.

^b6 h after cytokine addition.

^cPool of liver RNA harvested 4, 24, and 48 h after CTL injection.

^d12 h after LCMV infection.

^e48 h after CTL injection.

^fResults from quantitative real-time RT-PCR.

TOGA of HBV transgenic mouse liver RNA during LCMV- or CTL-induced hepatic cytokine induction.

DNA microarray analysis will miss genes that are not represented on the arrays used. To identify cytokine-regulated liver genes that are not represented on the DNA microarrays, we compared the gene expression profiles of pooled liver RNA harvested from groups of three mice that were sacrificed either 48 h after CTL injection (Fig. 1B, lane CTL + IgG) or 12 h after LCMV infection (Fig. 1C, lane LCMV + IgG) with pooled liver RNA from saline-injected control mice (Fig. 1B and C) by TOGA (86). TOGA is an automated, high-throughput cDNA display technique that produces 3′-cDNA fragments which are resolved by electrophoresis into discrete bands of measurable length, allowing their association with mRNAs of known genes and ESTs, with the peak height reflecting the relative expression level of the gene queried. Using the single enzyme MspI to produce the cDNA fragments results in about 40 to 60% coverage of all the mRNAs in a given sample. cDNAs associated with candidate peaks are subsequently cloned and verified. Computer-assisted peak recognition revealed 76 and 243 cDNA fragment peaks that were induced 48 h after CTL injection and 12 h after LCMV infection, respectively, 25 of which were common to both conditions (Table 3). Similarly, 45 and 115 cDNA peaks were repressed 48 h after CTL injection and 12 h after LCMV infection, respectively, 11 of which were common to both conditions (Table 3). The cloning of the cDNA associated with the candidate peaks yielded the 21 induced and 2 repressed unique genes and ESTs shown in Table 4. The complete profiling data for these genes are shown in online Table C and are also incorporated in online Table A.

TABLE 3.

TOGA of HBV transgenic mouse liver RNA

Treatment	No. of cDNA peaks		No. of cloned and verified genes or ESTs
	Induceda	Represseda	Induced	Repressed
CTLsb	76	45
LCMVc	243	115
Both	25	11	21	2

^aInduced or repressed compared to saline-injected control mice.

^bLiver RNA harvested 48 h after CTL injection.

^cLiver RNA harvested 12 h after LCMV infection.

TABLE 4.

Comparison of TOGA and DNA microarray analysis of HBV transgenic mouse liver RNA

Description	Unigene	Fold change in transcriptiona
		CTL		LCMV
		TOGA	DNA arrayb	TOGA	DNA arrayb
Proteasome (prosome, macropain) subunit, beta type 9 (Lmp2)	Mm.16251c	5	19	5	∼12
Proteasome (prosome, macropain) subunit, beta type 8 (Lmp7)	Mm.13913c	6	6	5	4
IFN-inducible GTPase (IIGP)	Mm.29008c	4	4	2	5
Signal transducer and activator of transcription 1	Mm.8249c	26	∼27	44	∼28
IFN-induced protein with TPR repeats 3	Mm.951c	2	3	20	∼138
Serum amyloid A3	Mm.14277	4	111	5	∼11
Small inducible cytokine B subfamily (Cys-X-Cys), member 9 (MIG)	Mm.766	19	∼416	8	∼71
Metallothionein 1	Mm.192991	3	∼333	12	∼92
IFN regulatory factor 7	Mm.3233	2	3	6	12
Pre-B-cell colony-enhancing factor	Mm.28830c	2	2	4	3
Proteasome (prosome, macropain) 28 subunit, beta	Mm.626c	3	2	3	—
Heptoglobin	Mm.26730	3	—	4	2
RIKEN cDNA 2010008K16 gene	Mm.45558c	6	—	11	—
Fibulin 2	Mm.6120	5	—	10	—
EST	Mm.29815	2	—	2	—
IFN-γ-induced GTPase	Mm.33902c	5	ND	14	ND
Tumor stroma and activated macrophage protein	Mm.116687c	2	ND	13	ND
Transforming growth factor beta-regulated gene 1	Mm.28689c	17	ND	10	ND
RIKEN cDNA 4933435C21 gene (similar to Ent3)	Mm.82642c	11	ND	18	ND
EST	Mm.35837	3	ND	3	ND
EST	Mm.56926	3	ND	3	ND
Weakly similar to contraspin (Mus musculus)	Mm.90033c	−4	ND	−3	ND
EST	BI148879	−3	ND	−4	ND

^aCompared to saline-injected control mice. ∼, approximate fold change.

^b—, no change in gene expression detected; ND, sequence not present on Mu11K DNA microarray set.

^cGene expression also regulated in hepatocytes.

For the first 10 genes in Table 4, TOGA confirmed the transcriptional regulation that was also detected by microarray analysis of liver RNA. Interestingly, TOGA revealed five cytokine-regulated liver genes that were represented on the DNA microarrays but that were not recognized as cytokine regulated (Table 4). Furthermore, the last eight liver genes in Table 4 that were shown to be cytokine regulated by TOGA of HBV transgenic mouse liver RNA were not represented on the oligonucleotide arrays used in this study. The hepatocyte-specific expression of the genes detected by TOGA was assessed by comparison with the results from the oligonucleotide array analysis of the HBVMet.4 cells if possible. For novel genes, quantitative real-time RT-PCR was performed on the same HBVMet.4 RNA samples used for the DNA microarray analysis. Overall, TOGA revealed 12 induced hepatocellular genes and 1 repressed hepatocellular gene (Table 4, identified by footnote c). Seven of these genes/ESTs were not detected by microarray analysis, so they were included in Table 2, thereby increasing the minimal set of cytokine-controlled hepatocellular genes that are associated with cytokine-dependent antiviral activity to 28 induced genes and ESTs and 1 repressed EST (Table 2).

Tight association of candidate antiviral active genes with inhibition of HBV replication in vivo.

We have previously shown that the antiviral activity induced by injection of HBV-specific CTLs (Fig. 1B, lane CTL + IgG) or infection with LCMV (Fig. 1C, lane LCMV + IgG) can be blocked by simultaneous administration of an antibody cocktail of anti-IFN-γ and anti-TNF-α (Fig. 1B) or anti-IFN-α/β and anti-TNF-α (Fig. 1C), respectively. These results suggest that the regulation of potential antiviral genes should also be blocked under these conditions. To verify this, we determined the relative induction or repression of the 29 genes in Table 2 in the livers of HBV transgenic mice in which the inhibition of HBV replication was blocked by anticytokine antibodies. Accordingly, we prepared cDNA for DNA microarray analysis from total liver RNA isolated from groups of three mice sacrificed 48 h after the injection of CTLs plus either an irrelevant antibody (Fig. 1B, lane CTL + IgG) or a cocktail of anti-IFN-γ and anti-TNF-α (Fig. 1B, lane CTL + αIFN-γ αTNF-α), and the results were compared with those for a control group of mice injected with saline (NaCl). In addition we prepared cDNA from total mouse liver RNA isolated from anti-IFN-α/β- and anti-TNF-α-treated HBV transgenic mice 12 h after they were infected with LCMV (Fig. 1C), and these results were compared to those from the microarray analysis of liver RNA from the saline-injected control group. Induced or repressed liver genes were selected from individual hybridization experiments by using the selection criteria described in Materials and Methods. In addition, the same samples were also subjected to TOGA as described in Materials and Methods.

Online Table B shows the fold induction or repression of the 29 genes in these samples (columns LCMV+ [LCMV plus anti-IFN-α/β and anti-TNF-α], CTL− [CTL plus IgG], and CTL+ (CTL plus anti-IFN-γ and anti-TNF-α]). Most importantly, the induction or repression of all of these genes and ESTs was blocked twofold or more by the anticytokine antibodies (online Table B; compare column LCMV with LCMV + and column CTL or CTL − with CTL +). Since the antiviral effect of the CTL injection and the LCMV infection was also blocked by the anticytokine antibodies (Fig. 1B and C), these data demonstrate a very tight association between the cytokine-controlled expression of the genes listed in Table 2 and inhibition of HBV replication in vivo.

Thus, we conclude that the 29 genes and ESTs listed in Table 2 represent hepatocellular genes whose transcriptional control by both IFN-α/β and IFN-γ is tightly associated with inhibition of HBV replication in vivo and in vitro. These genes can be broadly assigned to four categories as shown in Table 2. Importantly, all but 2 (the Gbp2 and Gbp3 genes) of the 13 genes in the first two categories have been shown to be involved in host-pathogen interactions and/or in host defense mechanisms against viral and/or microbial infections.

DISCUSSION

In this study we demonstrate the use of global gene expression profiling to identify a specific and limited set of candidate hepatocellular antiviral genes that might interfere with HBV replication in HBV transgenic mice. Previous studies have demonstrated that IFN-α/β and IFN-γ inhibit HBV DNA replication in the livers of HBV transgenic mice and in highly differentiated HBV transgenic hepatocyte lines (58, 64). Based on the hypothesis that IFN-α/β and IFN-γ induce common or at least converging antiviral pathways in hepatocytes, we searched for genes that were transcriptionally regulated by both of these cytokines in the livers of HBV transgenic mice and in hepatocyte cell lines derived from these animals. To do so, two generations of DNA microarrays (Mu11K set and Mu74A) as well as automated cDNA display (TOGA) were used to analyze HBV transgenic mouse liver RNA and immortalized hepatocyte RNA harvested after induction (in vivo) or administration (in vitro) of IFN-α/β or IFN-γ. Hence, selected genes were detected in four independent samples using different DNA microarrays as well as cDNA display analysis. Thus, we are confident that our results are not confounded by false-positive signals.

We cannot, however, rule out the possibility that some candidate antiviral genes were missed in this study. Gene expression profiling does not allow for the detection of posttranscriptionally regulated genes that might be involved in cytokine-mediated inhibition of HBV replication. Furthermore, the detection of transcriptionally regulated genes might be limited by the time points of sample collection, incomplete gene representation on the arrays, or detection sensitivity. Using the pooled liver RNA harvested at different time points after CTL injection we minimized the risk of missing genes regulated early after cytokine induction. The incomplete representation of the murine transcriptome on the DNA microarrays used was at least partially compensated for by including cDNA display-based TOGA (86). Indeed, we found five genes and ESTs expressed in hepatocytes in HBV transgenic mice which were not yet present on the DNA microarrays used (Table 4 and online Table C, which is available at http://www.scripps.edu/mem/expath/wieland/supplement.html). Furthermore, there were two genes and ESTs that were selected based on the TOGA result but that failed to pass the selection criteria used for the DNA microarray analysis. Also, we cannot rule out the possibility that some of the cytokine-regulated genes that were detected in the hepatocyte cell lines but not in vivo might inhibit HBV replication. This group involves 24 genes and ESTs, and they are shown in online Table D. There was very little change in transcription for some of these genes upon cytokine treatment in vitro, and this might explain why they were missed in the liver, where the hepatocyte contributes only a fraction of the total RNA in each sample. While we focus our discussion below and the initial screening for antiviral proteins on the minimal gene list presented in Table 2, we recognize that these additional genes and ESTs (online Table D) might also be important.

A complete list of all the 453 genes and ESTs that were regulated in at least one of the samples analyzed is available in online Table A. In the following we discuss the function and possible antiviral activity of the 29 genes that were regulated in all the samples and that are listed in Table 2. The first group contains genes involved in protein degradation. Among these are genes encoding the regulatory proteasome subunit PA28β as well as the active proteasome subunits β1i (LMP2), β2i (MECL-1), and β5i (LMP7) (28). These subunits are known to be transcriptionally induced by IFN-γ and TNF-α and to replace their constitutive homologues β1 (δ, Y), β2 (MC14, LMP9, Z), and β5 (MB1, X), respectively, during proteasome neosynthesis during an antiviral and antibacterial immune response in the liver (46). The change in active subunits changes the proteolytic specificity of the proteasome and results in the production of different fragments from polypeptides (reviewed in reference 29). Thus, it has been shown that the immunoproteasome enhances the processing and presentation of adenovirus (79), LCMV (76), and HBV CTL epitopes (78). Changing the half-life of one or more viral or host proteins by the immunoproteasome might affect HBV replication, since we have shown that the cytokines inhibit HBV replication either at the level of HBcAg-dependent capsid assembly or by accelerating the degradation of early capsids (89). Based on the present results, we recently demonstrated that cytokine-mediated inhibition of HBV replication requires proteasome activity (71). At present we do not know, however, if proteasome activity is important at the level of HBV protein turnover or whether it influences other IFN-induced intracellular antiviral mechanisms.

The remaining two proteins in this group are Isg15 (19) and Usp18 (UBP43 [51]). Isg15 has been found to have homology to ubiquitin (36) and to be conjugated to cellular proteins (53). Usp18 shows homology to ubiquitin-specific proteases and has been shown to specifically remove Isg15 from conjugated proteins (56). Interestingly, influenza B virus apparently inhibits conjugation of Isg15 to target proteins, suggesting that Isg15 conjugation might be important in the host defense against viral infections (90). However, the cellular and/or viral proteins that are targeted for Isg15 conjugation have not been identified (42, 52). It is possible that transient IFN-induced Isg15 conjugation to viral or cellular proteins may interfere with HBV replication.

The second group of genes in Table 2 comprises genes encoding GTP-binding proteins (GBP) belonging to the 47- (Gtpi, Tgtp, Ifi1, Iigp, and Ifi47) and 65-kDa (Gbp2 and Gbp3) GBP families (6). Transcription of these genes in macrophages and fibroblasts is strongly induced by IFN-γ and in some cases (Tgtp and Gbp2 genes) by IFN-α/β (6, 10). We now show that transcriptional induction of the proteins also occurs in hepatocytes, where it is mediated by both IFN-γ and IFN-α/β. Structural and biochemical analysis of these proteins suggests that they belong to the group of GTP-binding proteins that includes Mx and dynamin (68). The human MxA protein has been shown to have antiviral activity against several RNA viruses including influenza virus and vesicular stomatitis virus (22, 37, 66) and to block the nuclear import of Thogoto virus nucleocapsids (48). The role of MxA in host defense against HBV, however, is controversial (26, 70). Note also that Mx proteins are exclusively induced by IFN-α/β (3, 9, 17, 80), and therefore their genes are not included in the list of 29 genes shown in Table 2. We find the mouse Mx1 and Mx2 genes transcriptionally induced by IFN-α/β in our system (online Table A), but the associated mRNAs do not encode functional Mx proteins in most inbred mouse strains including the ones used in this study, so they cannot contribute to the antiviral effect (83, 84). Taken together, these findings strongly suggest that inhibition of HBV replication can occur by an Mx-independent mechanism(s). This mechanism(s) might involve GBP proteins of both the 47- and 65-kDa families, as it has been recently demonstrated that these proteins display not only antimicrobial activity (15, 87) but also, at least for Tgtp and Gbp1, antiviral activity against RNA viruses such as vesicular stomatitis virus and encephalomyocarditis virus (1, 10). Whether one or more of the GTP-binding proteins in Table 2 mediate the antiviral effect of IFN-α/β and IFN-γ in our system remains to be determined, and experiments designed to test that hypothesis are under way in our laboratory.

The next group of genes in Table 2 are involved in extracellular and intracellular signaling. Scyb10 (IP-10) is a chemokine shown to have antiviral activity against vaccinia virus in vivo through activation of natural killer cells (NK cells) and induction of both IFN-α/β and IFN-γ (55). For HBV, however, we have recently shown that, although IP-10 is highly induced by CTL injection in the livers of HBV transgenic mice and although neutralization of IP-10 in these mice reduces the inflammatory infiltrate that is induced by the CTLs, IP-10 itself does not inhibit HBV replication (45). To our knowledge, hepatocellular expression of the pre-B-cell colony-enhancing factor (Pbef) has not been previously observed. Pbef promotes early B-cell proliferation (65, 72), but whether it plays a role in the inhibition of HBV replication in our system remains to be determined. The intracellular signaling molecule signal transducer and activator of transcription 1 (Stat1) has been shown to be transcriptionally induced by IFN-γ (50, 57) and IFN-α/β (50) and to be essential in establishing an IFN-induced antiviral state (38).

Genes and ESTs that did not fall into one of the above groups are shown in the “miscellaneous” category in Table 2. The first gene in this group, the gene encoding the transporter associated with antigen processing (TAP), is known to be transcriptionally induced by IFN-γ (54) and IFN-α/β (41). TAP translocates peptides produced by the proteasome across the endoplasmic reticulum (ER) membrane into the ER lumen, where the peptides are loaded onto major histocompatibility complex class I molecules (reviewed in reference 62). It remains to be determined whether increased TAP activity might contribute to the antiviral effect of IFN-γ and IFN-α/β.

Although macrophage proliferation-specific gene 1 (encoding Mpeg1/MPS-1) was isolated specifically from mature macrophages (82), we show here that it can be transcriptionally induced by IFN-γ and IFN-α/β in hepatocytes. Nothing is known about the function of the gene's product except that it contains a domain with sequence homology to perforin, a granule protein in CTLs and natural killer cells that polymerizes in target cell membranes and causes cell destruction (85). Thus, it is possible that Mpeg1 plays a hitherto-unsuspected role in hepatocyte physiology or in the HBV life cycle.

The transcriptional induction of xanthine dehydrogenase increases the intracellular levels of not only xanthine dehydrogenase but also, through protein processing, xanthine oxidase (61). Thus it influences purine metabolism in the hepatocyte, which could affect the priming of HBV reverse transcription and thus could prevent the formation of replication-competent capsids. Additional experiments are needed to verify this interesting hypothesis.

DLM-1 was identified as a tumor-associated protein but was also found to be upregulated in IFN-stimulated macrophages (23). Recently, structural analysis has revealed that, like the vaccinia virus (VV)-encoded IFN resistance protein E3L, DLM-1 has the ability to bind left-handed DNA (Z-DNA) (75). Interestingly, the Z-DNA-binding domain of E3L is necessary for full pathogenesis in VV infection in vivo (7). Further analysis of the role of DLM-1 in HBV-host interactions may be warranted.

The Ifit-1, -2, and -3 proteins all contain tetratricopeptide (TPR) domains and are IFN inducible (81). The human homologue TPR domain-containing protein (ISG-54K) is the product of one of the first IFN-inducible genes described (49). TPR domains promote protein-protein interactions and are present in a number of proteins that are functionally unrelated (reviewed in reference 5). It is known that TPR-containing proteins act as cochaperones for heat shock proteins 70 and 90 (Hsp70 and Hsp90) (5). Thus, it is possible that the TPR-containing proteins listed in Table 3 interfere with HBV replication through interaction with heat shock proteins that are thought to be involved in HBV capsid assembly (39, 63).

The transforming growth factor beta-regulated gene (encoding Tbrg1) and the ESTs are poorly characterized, and their antiviral capacity remains to be determined.

In summary, the gene expression profiling experiments described herein have identified a limited set of genes, many of which could potentially mediate the antiviral effects of IFN-α/β and IFN-γ in HBV-infected hepatocytes. Thus, experiments in which each of these genes is systematically overexpressed or suppressed in HBV-Met.4 cells or other suitable model systems may yield important new insights into the host-virus interactions that control HBV replication. Indeed, the identification of the genes involved in protein degradation shown in Table 2 prompted experiments demonstrating that proteasome activity is required for cytokine induced inhibition of HBV DNA replication (71).