Induction of hepatitis D virus large antigen translocation to the cytoplasm by hepatitis B virus surface antigens correlates with endoplasmic reticulum stress and NF-{kappa}B activation

Hepatitis D virus (HDV) is the smallest human RNA virus and has a genome of 1.7 kb in the form of a single-, negative-stranded circle (for a review see Lai, 1995). Although the genome structures of HDV and plant viroids share a common evolutionary origin, unlike viroids, which do not encode any proteins, HDV encodes two antigen isoforms, the small and large forms (SDAg and LDAg, respectively) during its replication cycle. The SDAg and LDAg share 195 aa at their N terminus, while the LDAg contains 19 extra aa at its C terminus (Weiner et al., 1988). SDAg is required for HDV RNA replication, whereas LDAg suppresses this process and interacts with hepatitis B virus (HBV) envelope proteins (HBsAg) to assemble a mature virion for secretion and infection (Kuo et al., 1989; Chao et al., 1990; Chang et al., 1991; Ryu et al., 1992).

Replication of the HDV genome and antigenome is dependent on host cell RNA polymerases and occurs via a rolling-circle method, which produces multiple copies of HDV RNA in a linear form (Modahl et al., 2000; Macnaughton et al., 2002). Ribozymes in both the genome and antigenome then self-cleave the linear RNA into single units, which are then ligated into a circular form by cellular ligases (Lai, 1995; Reid & Lazinski, 2000; Taylor, 2003). During the HDV replication cycle, host enzymes called ADARs (adenosine deaminases that act on double-stranded RNA) edit a portion of the HDV RNA to convert the amber stop codon (UAG) of SDAg to a tryptophan codon (UGG), which results in the production of LDAg (Casey & Gerin, 1995; Sato et al., 2001; Jayan & Casey, 2002). Thereafter, LDAg inhibits HDV RNA replication and then LDAg together with SDAg and the HDV genome assemble into a ribonucleoprotein (RNP) complex (Ryu et al., 1993), which is then transported to the cytoplasm to form a mature virion with the HBsAgs.

In addition to the host RNA polymerases and ADARs, many other cellular enzymes involved in the post-translational modifications of HDAgs are important for the execution of HDAg's function (Mu et al., 2004; Li et al., 2004; Lai, 2005). For example, farenyl-transferase is required for LDAg isoprenylation, which is crucial to the interaction with HBsAg for secretion (Glenn et al., 1992; Hwang & Lai, 1993; Sheu et al., 1996). Different kinases are also required for SDAg and LDAg phosphorylation because the SDAg is phosphorylated at both serine and threonine, while the LDAg is phosphorylated only at serine (Chang et al., 1988; Mu et al., 1999), which may account for their distinct functions. Furthermore, the serine residues at positions 2 and 177 of SDAg have been demonstrated to modulate HDV RNA replication but have no significant role in subviral particle formation (Yeh et al., 1996; Yeh & Lee, 1998; Mu et al., 1999, 2001). Recently, methylation at arginine-13 and acetylation at lysine-72 of HDAg have been reported to play an important role in virus replication (Mu et al., 2004; Li et al., 2004). These lines of evidence fully support the hypothesis that post-translational modifications of HDAgs can drive HDAgs to specific cellular compartments and functions (for a review see Lai, 2005). However, signals and mechanisms involved in HDAg post-translational modification have not yet been well studied.

Previously, we used a green fluorescent protein fused to LDAg (GFPLD) to demonstrate that translocation of GFPLD from the nucleolus to SC-35 speckles can be induced by treatment with the casein kinase II inhibitor, dichlororibofuranosyl benzimidazole (Shih & Lo, 2001), in which serine-123 of LDAg in the deposphorylated state is responsible for preferentially targeting to the SC-35 speckles (Tan et al., 2004). How cellular kinases and phosphotases are regulated in order to modify LDAg for translocation is still under question. We also demonstrated that the small form of HBsAg can facilitate the translocation of GFPLD from the nucleus to the cytoplasm (Tan et al., 2004), but the mechanism remains unclear. In this study, we explore what signal molecules are generated by HBsAgs, which reside in the endoplasmic reticulum (ER) and can induce GFPLD nuclear export.

Plasmids used in this study.
Based on the encoded proteins, plasmids used in this study can be grouped into four series: (i) pMTLD (Hu et al., 1996) and pN2LD encode authentic LDAg under the metallothionine (MT) and cytomegalovirus (CMV) promoter, respectively; (ii) pMTS, pMTMMS[S] and pMTLS encode HBV small, middle and large (S, M and L) surface antigens, respectively (Sheu & Lo, 1994), and pN2LS encodes L HBsAg; (iii) those expressing different versions of HDAg fused to a GFP, pGFPLD, pGFPLDM, pGFPLD(31214) and pSDGFP (Shih & Lo, 2001; Shih et al., 2004); and (iv) pIKKα, which encodes the wild-type IκB kinase α subunit; pKA, which encodes a mutant form of IKKα; and pNF-κB-Luc, which contains the NF-κB-response element (TGGGGACTTTCCGC)₅ fused to a luciferase gene (kindly provided by Dr Y. S. Chang, Chang Gung University, Republic of China). The characteristics of the GFP fusion proteins encoded by the series (iii) plasmids are summarized as follows: (a) pGFPLD encodes GFP fused to wild-type LDAg; (b) pGFPLDM produces GFP fused to a non-isoprenylated mutant of LDAg; (c) pGFPLD(31214) produces GFP fused to an N-terminal deletion (aa 130) mutant of LDAg; and (d) pSDGFP produces wild-type SDAg fused to GFP (Shih & Lo, 2001; Shih et al., 2004).

Cell culture and plasmid transfection.
Two human cell lines were used in this study; one is a well-differentiated hepatoma cell line, HuH-7, while the other is a cervical carcinoma cell line, HeLa. Most of the experiments were carried out using HuH-7 cells and a few were done with HeLa cells. Both cell lines were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10 % fetal bovine serum, penicillin (100 IU ml¹), streptomycin (100 µg ml¹), Fungizone (50 µg ml¹) and 2 mM L-glutamine, and grown at 37 °C under 5 % CO₂. Plasmids in a supercoiled form were obtained using the Qiagen Plasmid Maxi kit and then used for transfection. Cells at 6080 % confluence in a 10 cm Petri dish or six-well plate were transfected with 1020 µg of the indicated plasmids by the calcium phosphate-DNA precipitation method (Graham & van der Eb, 1973) or by adding lipofectamine (Invitrogen). At 1 or 2 days post-transfection, cells were treated with various drugs for different time intervals: (i) tunicamycin (TM; 25 µg ml¹) or brefeldin A (BFA; 25 µg ml¹) for 2 h, and (ii) with or without pre-treatment with cycloheximide (10 µg ml¹) for 30 min and then followed by treatment with tumour necrosis factor alpha (TNF-α; 10 ng ml¹) for 1 h.

Fluorescence microscopy.
To visualize the co-expression of HBsAgs and various forms of GFP fusion proteins, transfected cells were reseeded on 22x22 mm coverslips. After full attachment, cells were fixed and permeabilized using methanol for 30 min at 20 °C and then stained with anti-HBs antibody followed by the secondary antibody conjugated with rhodamine. In parallel, cells were stained with Hoechst 33258 for 10 min to show the nucleus. Finally, cells were mounted onto glass slides with mounting solution, and visualized using a fluorescence microscope (Olympus IX71) with a fluorescein isothiocyanate (FITC) or rhodamine filter. For quantification of the GFP fusion protein, the distribution of the nucleus only (N) versus the nucleuscytoplasm (N+C) pattern was observed, and between 100 and 500 GFP-positive cells were randomly selected and their patterns classified as described previously (Tan et al., 2004). For simplicity in this study, type I, II and III distribution patterns were redesignated N pattern, nucleus only, while the type IV was redesignated N+C, nucleus and cytoplasm.

Western blot analysis.
To detect the amount of GFP fusion proteins in the nucleus and cytoplasm, transfected cells were fractionated into nuclear and cytoplasmic fractions using a nuclear extraction kit (Active Motif), according to the manufacturer's instructions. Protein samples were separated by SDS-PAGE and then electro-transferred onto PVDF membranes. The membranes were incubated with 5 % non-fat milk for 1 h at room temperature for blocking and then reacted with anti-GFP, anti-HBsAg, anti-GRP78, anti-NF-κB p65 or anti-tubulin antibodies (depending on the purpose of the experiment) in the presence of 5 % non-fat milk overnight at 4 °C. This was followed by incubation with the secondary antibody conjugated with horseradish peroxidase and the blots were then developed by enhanced chemiluminescence using a commercial kit (Amersham).

Luciferase activity assay.
Measurement of NF-κB activity was carried out as described by Wu et al. (1998). Briefly, transfected cells were lysed using a buffer supplied in a commercial luciferase assay kit (Promega) according to the manufacturer's instructions. The luciferase activity of each cell lysate was determined by using an Autolumat LB953 luminometer (Berthold). All experiments were carried out in duplicate and repeated three times.

The small form of HBsAg can induce nuclear export of both authentic and GFP fusion LDAg
Previously, we have demonstrated that the small form of HBsAg facilitated GFPLD translocation from the nucleus to the cytoplasm (Tan et al., 2004). In this study, we first confirmed that HBsAg expression correlates with GFPLD in the cytoplasm by co-transfection of pMTS and pGFPLD into HuH-7 cells. A representative fluorescence microscopy image shows that three cells that have GFP appearing in an N+C pattern, while one cell has GFP located in the nucleus (N pattern) only [Fig. 1a(i) and a(ii)]. This particular N pattern cell had no HBsAg expression in contrast to the three N+C cells [Fig. 1a(iii)], indicating clearly that HBsAg can facilitate GFPLD translocation to the cytoplasm. Consistently, the presence of authentic LDAg in the cytoplasm was highly correlated with the expression of HBsAg in HeLa cells [Fig. 1b(i), (ii) and (iii)], indicating that HBsAg can affect similarly on both GFPLD and authentic LDAg. We then tested whether HBsAg can facilitate on other HDAg fusion proteins by performing single, double and triple plasmid transfection experiments, in which each of four plasmids [pGFPLD, pGFPLDM, pGFPLD(31214) and pSDGFP] expressing a different version of HDAg fused to GFP was co-transfected with or without pMTS and pMTLD. As shown in Table 1, the percentage of N+C pattern cells for expressing GFPLD (GFP fused to wild-type LDAg), GFPLDM (GFP fused to a non-isoprenylated mutant of LDAg), GFPLD(31214) [GFP fused to an N-terminal deletion (aa 130) mutant of LDAg] or SDGFP (wild-type SDAg fused to GFP) ranged from 1.1 to 5.1 % in single plasmid transfections. In the presence of HBsAg, the N+C pattern of GFPLD cells increased to 26.5 %, while that of GFPLDM, SDGFP and GFPLD(31214) expressing cells increased ranging from 2.5 to 5.1 %, suggesting that HBsAg only facilitates the full-length of LDAg translocation to the cytoplasm but not the single point mutation, which cannot be isoprenylated (GFPLDM), N-terminal deletion mutation [GFPLD(31214)] or SDAg. As compared with the result of double plasmid transfection, results of triple plasmid transfection showed 0.3 % or no increase of N+C pattern in cells expressing GFPLD(31214) and GFPLD, while 13.5 and 12.9 % increase in cells expressing GFPLDM and SDGFP, respectively (see the last column of Table 1). A higher increasing percentage of N+C pattern in GFPLDM and SDGFP expressing cells as compared with that of GFPLD(31214) cells in the presence of HBsAg and LDAg suggested that LDAg can form a complex with GFPLDM and SDGFP via the coiled-coil domain and this allows them export to the cytoplasm, in contrast, the coiled-coil domain of GFPLD(31214) is truncated (Chang et al., 1992; Shih & Lo, 2001). No significant increase of N+C pattern in GFPLD cells between double and triple plasmid transfections suggested that GFPLD behaves similarly to LDAg under the effect of HBsAg.

(65K): Fig. 1. The effect of HBsAg on GFPLD [a(i)(iii)] and authentic LDAg [b(i)(iii)] distribution. [a(i)(iii)] HuH-7 cells were co-transfected with pGFPLD and pMTS. After 24 h post-transfection, cells were permeabilized and reacted with anti-HBsAg and followed by rhodamine-conjugated antibody. The same cells were photographed using different fluorescent filters and show: (i) cells stained with DNA dye, Hochest 33258, to outline the nucleus, (ii) GFPLD pattern inside of cells using an FITC filter, and (iii) HBsAg distribution using a rhodamine filter. [b(i)(iii)] HeLa cells were co-transfected with pN2LD and pN2LS, which contain the CMV promoter to express L HBsAg and LDAg, respectively. After 24 h post-transfection, cells were fixed and stained with antibody conjugated with rodamine to indicate the expression of LDAg and with FITC-conjugated antibody to indicate the expression of HBsAg. The same cells were photographed using different fluorescent filters and show: (i) cells stained with DNA dye, Hochest 33258, to outline the nucleus, (ii) LDAg distribution using a rhodamine filter, and (iii) HBsAg distribution using an FITC filter. N indicates the nucleus pattern and N+C indicates the cytoplasm and nucleus pattern. Bars, 10 µm.

Table 1. Effect of S-HBsAg on distribution pattern of various GFP fusion proteins in HuH-7 cells These distribution patterns of various GFP fusion proteins were observed after 72 h post-transfection. Each number in the N and N+C column represents the mean value of three independent experiments. Numbers (1, 2 and 3) at the left represent for transfection conditions: 1, single plasmid transfection; 2, double plasmid transfection; and 3, triple plasmid transfection. N and N+C distribution pattern is described in the legend of Fig. 1.

The three forms of HBsAgs facilitate GFPLD translocation to the cytoplasm to different degrees
It is known that there are three different forms (S, M and L) of HBsAg in various lengths of amino acids and degrees of glycosylation modification (Sheu & Lo, 1992). We were thus interested to see whether or not the three different forms of HBsAg exert different capability on the translocation of GFPLD. Co-transfection of pGFPLD with pMTS, pMTMMS[S] or pMTLS, which express S, M and L HBsAg, respectively, into HuH-7 cells was performed. The N and N+C pattern of the GFPLD cells were analysed at 24, 48 and 72 h post-transfection. As shown in Table 2, GFPLD cells had an increased percentage of the N+C pattern when they were co-expressed with S, M and L HBsAg (10.4, 26.4 and 34.1 %, respectively) at 24 h post-transfection. A similar trend was also observed in the 48 h post-transfected cells (25.2, 36.3 and 47.0 %, respectively) as well as in the 72 h post-transfected cells (34.7, 47.8 and 56.5 %, respectively). Since the three forms of HBsAgs reside in the ER, it was speculated that the presence of HBsAg may induce ER stress and then in turn cause GFPLD translocation. Western blot analyses were performed to test this supposition. Results clearly show that the GRP78 protein/BiP, an ER stress-induced marker, which functions as a protein chaperone at the ER cisternae (Schröder & Kaufman, 2005; Lee, 2005), was increased in the total lysate of cells co-expressing GFPLD with S, M and L HBsAg (Fig. 2, lanes 35, bottom line) compared with cells that were mock transfected or expressed GFPLD alone (Fig. 2, lanes 1 and 2, bottom line). For an unknown reason, the increasing folds of GRP78/BiP induced by different HBsAgs varied and were not reproduced consistently. In contrast, the amount of p65 subunit of NF-κB consistently increased in the nuclear fraction of cells co-expressing S, M and L HBsAg (Fig. 2, lanes 35, line 2). A decreased amount of GFPLD in the nuclear fraction (Fig. 2, lanes 25, line 1) and, conversely, an increased amount of GFPLD in the cytoplasmic fraction (Fig. 2, lanes 25, line 4) were found in cells expressing GFPLD alone or co-expressing GFPLD with S, M, or L HBsAg. These Western blot results are highly correlated with the results of the fluorescence microscopic observation (Table 2), indicating that all three forms of HBsAg have the capability to induce GFPLD translocation, but to various degrees.

Table 2. Distribution pattern of GFPLD in HuH-7 cells Each number in the N and N+C column represents the mean value of three independent experiments. N and N+C distribution pattern is described in the legend of Fig. 1.

(20K): Fig. 2. Western blot analyses of GFPLD and NF-κB distribution in the presence of three different forms of HBsAgs (L, M and S). HuH-7 cells were co-transfected by pGFPLD with plasmids encoding L, M or S HBsAg. Cells were then fractionated into nuclear and cytoplasmic portions at 72 h post-transfection. Nuclear proteins (15 µg) were analysed with anti-GFP, anti-NF-κB p65 and anti-tubulin antibodies (the top three lines as indicated by nuclear fraction), while 15 µg cytoplasmic proteins were analysed with anti-GFP and anti-tubulin (as indicated by cytoplasmic fraction). Total cell lysate, in 20 µg protein, was analysed withGRP78/BiP and GFPLD (the last two lines). Tubulin served as a positive control for the cytoplasmic fraction and as a negative control for the nuclear fraction.

Treatment with TM and BFA increases GFPLD in the cytoplasm
Although an increase of GRP78/BiP in the presence of HBsAgs is shown in Fig. 2, to further correlate with ER stress and GFPLD distribution in the cytoplasm, we treated GFPLD expression cells with the ER stress inducers, TM and BFA, for 2 h and then determined the percentage of N and N+C pattern cells by fluorescence microscopy. Around 8.8 to 9.3 % N+C pattern increase was observed when they were treated with 5 µg TM and BFA ml¹ (see Supplementary Table S1 available in JGV Online). Consistently, Western blot results showed a decreased amount of GFPLD present in the nuclear fraction, while an increased amount of GFPLD present in the cytoplasm when cells were treated with TM (Fig. 3a, lanes 25, lines 1 and 3) and BFA (Fig. 3b, lanes 25, lines 1 and 3). In addition, a slightly increased amount of GRP78/BiP was also observed in a dose-dependent manner when cells were treated with lower doses of TM and BFA (24 µg ml¹) but not at higher doses (5 µg ml¹) (Fig. 3a and b, line 6). The highest dose was unable to induce the highest amount of GRP78/BiP and GFPLD in the cytoplasm (Fig. 3a and b lane 6, line 6), which could be due to a desensitization effect. Nevertheless, the data showing a good correlation between increasing GFPLD in the cytoplasm and expression of GRP78/BiP indicate that even in the absence of HBsAg, inducing ER stress by an exogenous drug also causes GFPLD translocation to the cytoplasm. Therefore, induction of ER stress is suggested to be an early event when HBsAg is expressed in cells to induce GFPLD translocation.

(32K): Fig. 3. Western blot analyses of GFPLD distribution after treatment with the ER stress inducers, TM (a) and BFA (b). HuH-7 cells were mock transfected (lane 1) or transfected with pGFPLD alone (lanes 26) and then treated with various amounts of drugs for 2 h at 72 h post-transfection. The amount of drug used is indicated above the gel (25 µg). Preparation of nuclear and cytoplasmic fractions was the same as described in Fig. 2. Both nuclear and cytoplasmic fractions were analysed by GFP and tubulin as indicated, while the total cell lysate in 20 µg was analysed by GFP and GRP78/BiP.

NF-κB plays a role in GFPLD transportation from the nucleus to the cytoplasm
Although an increasing amount of the p65 subunit of NF-κB in the nucleus has been demonstrated by Western blot in cells co-expressing GFPLD and S, M or L HBsAg (Fig. 2), whether the nuclear form of NF-κB can actively transcribe its target genes remains to be determined. Similar experiments described in Table 2 were therefore conducted, in which pNF-κB-Luc containing the luciferase gene followed by the NF-κB-response element was co-transfected with pGFPLD and plasmids encoding S, M or L HBsAg into HuH-7 cells. After 24, 48 and 72 h post-transfection, cells were lysed and luciferase activities were determined. Results clearly show that the luciferase activity (2.73.2x10⁵ luminescence intensity unit, liu) was higher in cells co-expressing GFPLD with S, M or L HBsAg as compared with in those expressing GFPLD without HBsAgs (1.3x10⁵ liu) or in pNF-κB-Luc transfection cells (8.2x10⁴ liu) at 24 h post-transfection (Fig. 4). A similar trend was also observed in 48 h post-transfection cells (3.14.1x10⁵ liu versus 1.41.7x10⁵ liu) and in 72 h post-transfection cells (4.37.3x10⁵ liu versus 2.1x10⁵ liu). Among the cells co-expressing various HBsAgs at 72 h post-transfection, the L HBsAg expression showed the highest luciferase activity (7.3x10⁵ liu), while the S HBsAg expression cells showed the lowest activity (4.3x10⁵ liu). These results are consistent with the Western blot results and indicate that the nuclear form of NF-κB is induced by three forms of HBsAgs (Fig. 2, line 2) and this varies to a range of different levels.

(28K): Fig. 4. NF-κB activity assay in transfected cells. HuH-7 cells were transfected with various combinations of plasmids as indicated in the key. After 24, 48 and 72 h post-transfection, cells were lysed and analysed for luciferase activity. Three independent experiments were performed and the mean±SD data are shown.

To obtain another line of evidence for the role of NF-κB in HBsAg-induced GFPLD translocation, we co-transfected plasmid pKA, which expresses a dominant negative form of IKKα that results in inactivation of NF-κB, to examine whether the functioning of HBsAg will be blocked or not. Western blot results show that the GFPLD remaining in the nucleus was increased proportionally to the amount of pKA transfected (Fig. 5a, line 1). Unlike the reciprocal change of GFPLD between the nucleus and cytoplasm detected and shown in Fig. 2, a constant and similar amount of GFPLD was present in the cytoplasm of cells with or without pKA transfection (Fig. 5a, line 3). However, the luciferase activity assay showed several fold decrease in cells expressing various amounts of mutant IKK (data not shown), suggesting that NF-κB was indeed inhibited in those cells and this resulted in a higher amount of GFPLD being retained in the nucleus. When cells co-expressed GFPLD and various amounts of active form of IκB kinase α subunit, an increasing amount of GFPLD, proportional to the amount of pIKKα that was transfected was observed in the cytoplasmic fraction of cells (Fig. 5b, line 3). Conversely, a decreasing amount of GFPLD was observed in the nuclear fraction of cells (Fig. 5b, line 1). Cells having the higher amount GFPLD in the nucleus were highly correlated with the higher luciferase activity (data not shown). Taken together the evidence shown in Figs 4, 5(a) and (b) conclude that NF-κB indeed plays a significant role in HBsAg-induced GFPLD translocation.

(23K): Fig. 5. Western blot analyses of GFPLD distribution under various degrees of NF-κB activation. (a) HuH-7 cells were co-transfected with pGFPLD and pMTS (lanes 14) with 4, 6 or 8 µg plasmid pKA (lane 2, 3 and 4, respectively), which produces a dominant negative form of IKK. Post-transfected cells (72 h) were fractioned into nuclear and cytoplsmic portions as indicated in the legend of Fig. 2 and GFPLD changes in the nuclear and cytoplasmic fractions were then analysed by anti-GFP. (b) HuH-7 cells were co-transfected with pGFPLD with various amounts (4, 6 or 8 µg) of pIKK, which produces an active form of IκB kinase α subunit. Post-transfected cells (72 h) were fractionated into nuclear and cytoplsmic portions and GFPLD changes in the nuclear and cytoplasmic fractions were analysed as indicated in (a).

TNF-α treatment also induces GFPLD translocation to the cytoplasm in the absence of protein synthesis
Since TNF-α induces cell responses largely through activation of NF-κB, to correlate further with activation of NF-κB and GFPLD distribution, GFPLD expressing cells were treated with TNF-α for 1 h and the percentage of N+C cells was determined by fluorescence microscopy. The results revealed that 1 h after TNF-α treatment induced more than 50 % of GFPLD expressing HuH-7 or HeLa cells appeared to have the N+C pattern, while treatment with TM and BFA induced only 8.8 and 9.3 % of the cells to have the N+C pattern (Supplementary Table S1 available in JGV Online). The robust effect of TNF-α allowed us to test whether the induction of GFPLD translocation into the cytoplasm requires newly synthesized proteins or not. This kind of experiment could not be performed in cells expressing HBsAg because the protein translation inhibitor will also block HBsAg production. After pre-treatment with cycloheximide for 30 min, cells were treated with TNF-α for 1 h and the N+C pattern of GFPLD was examined. Around 53.3 % of cells appeared to have the N+C pattern (Supplementary Table S1), indicating that GFPLD translocation into the cytoplasm does not require newly synthesized proteins. The inhibition effect by cycloheximide was also shown to occur by the luciferase activity assay, which showed that activity was greatly reduced in cells treated with both TNF-α and cycloheximide (data not shown). However, the Western blot result shows that the amount of NF-κB in the nucleus was similar after TNF-α treatment with or without cycloheximide but it was higher than that in TNF-α non-treated cells (Fig. 6, line 3). Western blot analysis also shows that the amount of GFPLD in the nucleus and cytoplasm (Fig. 6, lines 1 and 4) correlated well with the fluorescence microscopic observations.

(14K): Fig. 6. Western blot analyses to show the effect of TNF-α and cycloheximide on GFPLD distribution. HuH-7 cells were transfected with pGFPLD. After 72 h post-transfection, cells were untreated (lane 1), treated with 10 ng TNF-α ml1 for 1 h (lane 2) or pre-treated with cycloheximide (10 µg ml1) for 30 min followed by co-treatment of cycloheximide (10 µg ml1) and TNF-α (10 ng ml1) for 1 h (lane 3) and then fractionated into nuclear and cytoplasmic fractions as indicated in the legend ofFig. 2. GFPLD and NF-κB (p65) changes in the nuclear and cytoplasm fractions under various drug treatments were analysed as indicated in Fig. 2.

During the HDV replication cycle, the HDV RNP, which contains the HDV RNA genome and two isoforms of antigen (SDAg and LDAg), is known to be exported out of the nucleus in order to be enveloped by HBsAgs for production of mature viral particles. This process is very likely LDAg-mediated since there is a nuclear export signal (NES) localized at the C terminus of LDAg and a nuclear factor (NESI) that binds to the NES has been identified (Lee et al., 2001; Wang et al., 2005). However, signal molecules involved in HDV RNP export and the underlying mechanisms are largely unknown. In this study, we apply a previously established system, in which HBsAg can facilitate GFPLD translocation into the cytoplasm, to investigate signals participating in the GFPLD translocation. Several lines of evidence have demonstrated that ER stress and activation of NF-κB, which are induced by HBsAg, play a significant role (Figs 2, 4 and 5a). It is known that ER stress induced by many different viral proteins leads to various cellular effects (Su et al., 2002; Tardif et al., 2002; Dimcheff et al., 2003), including full-length and truncated forms of the HBV L and M surface antigens (Wang et al., 2003; Hsieh et al., 2004; Hung et al., 2004). The current study is the first report to show that induction of ER stress results in a nuclear protein translocation into the cytoplasm.

The current study also provides direct evidence that three different forms of HBsAgs induce various levels of NF-κB activities (Fig. 4) as well as 34.756.5 % cells having GFPLD in the cytoplasm (Table 2). The various capabilities exhibited by different forms of HBsAgs could be due to their intrinsic properties, i.e. capability of ER retention (Bruss & Ganem, 1991; Sheu & Lo, 1994; Chau et al., 2005). The effects of HBsAg on NF-κB activity and GFPLD translocation could be explained by either that they are two parallel events or that the GFPLD translocation is the downstream event following the NF-κB activation. We prefer the second explanation because in the absence of HBsAg, GFPLD cells treated with TNF-α, which induces a higher level of NF-κB, also appear to have a higher percentage of N+C pattern (Supplementary Table S1). At the present time, no direct evidence that HBsAg can trap the newly synthesized LDAg in the cytoplasm has been demonstrated. However, we favour the hypothesis that LDAg must enter into the nucleus and then translocate into the cytoplasm after post-translational modification, in which HBsAgs exert signals to facilitate the modification. This hypothesis is supported by results that cells co-expressing GFPLD(31214) and HBsAg with or without LDAg show a low N+C pattern from 1 to 3 days post-transfection (Table 1), while in 6 days post-transfected cells expressing GFPLD(31214) can be co-secreted with HBsAg (Shih & Lo, 2001). Therefore, the current study suggests that the sequence located between aa 1 and 30 may be important in helping LDAg translocation into the cytoplasm and post-translational modification of serine-2 and/or arginine-13 (Mu et al., 2004; Li et al., 2004) is likely to be involved in facilitating LDAg translocation into the cytoplasm.

In the absence of HBsAgs, GFPLD translocation can also be facilitated by adding ER stress inducing drugs, TM and BFA (Fig. 3), but the effect is fivefold lower than that of L HBsAg on 72 h post-transfected cells (Table 2), suggesting that many signal pathways and molecules may participate in the HBsAg-induced GFPLD translocation. This may explain why inactivation of NF-κB by co-transfection of dominant negative IKK does not significantly abolish the cytoplasmic distribution of GFPLD (Fig. 5a) and the expression of GRP78/BiP and activity of NF-κB is not perfectly matched in some cases. Interestingly, no new proteins are required for TNF-α to induce GFPLD translocation, indicating that a post-translational modification occurs to GFPLD to allow this translocation (Fig. 6 and Supplementary Table S1). Results of low translocation of GFPLD(31214) and GFPLDM in the presence of HBsAgs (Table 2) suggest that multimerization and farnesylation of GFPLD are two important modifications for facilitating GFPLD translocation. The present study cannot distinguish which step, farnesylation or multimerization, comes first when facilitating LDAg nuclear export but one report has shown that multimerization between SDAg and LDAg increases farnesylation of LDAg (O'Malley & Lazinski, 2005). Therefore, the farnesyl-transferase and other unidentified molecules that are directly or indirectly modified by the activated NF-κB are of interest for future exploration.

In conclusion, the current study presents the first report showing the presence of signals in the cross-talk between HDV and HBV using a system of GFP fusion proteins. However, to identify the target enzymes downstream to NF-κB and to understand how they are activated to modify LDAg are a great challenge for future studies. Moreover, whether the HDV RNP translocation into the cytoplasm can be facilitated by HBV and TNF-α remains to be tested.

We thank Dr Y. S. Chang (Chang Gung University) for providing us with the plasmids and Dr R. Kirby (National Yang Ming University) for editing the English in this article. Thanks are also due to lab members, C.-C. Lee and H.-J. Lai for their discussions and Y. L. Chen for figure preparation. This work was supported by grants from the National Science Council (NSC93-2320-B-010-096 and 92PS012) to S. J. L.