Determinants for membrane association and permeabilization of the coxsackievirus 2B protein and the identification of the Golgi complex as the target organelle.

The 2B protein of enterovirus is responsible for the alterations in the permeability of secretory membranes and the plasma membrane in infected cells. The structural requirements for the membrane association and the subcellular localization of this essential virus protein, however, have not been defined. Here, we provide evidence that the 2B protein is an integral membrane protein in vivo that is predominantly localized at the Golgi complex upon individual expression. Addition of organelle-specific targeting signals to the 2B protein revealed that the Golgi localization is an absolute prerequisite for the ability of the protein to modify plasma membrane permeability. Expression of deletion mutants and heterologous proteins containing specific domains of the 2B protein demonstrated that each of the two hydrophobic regions could mediate membrane binding individually. However, the presence of both hydrophobic regions was required for the correct membrane association, efficient Golgi targeting, and the membrane-permeabilizing activity of the 2B protein, suggesting that the two hydrophobic regions are cooperatively involved in the formation of a membrane-integral complex. The formation of membrane-integral pores by the 2B protein in the Golgi complex and the possible mechanism by which a Golgi-localized virus protein modifies plasma membrane permeability are discussed.

mediates, the exact function of most of the viral proteins in the replication cycle is still largely unknown.
Host cell membranes are subject to a number of profound alterations upon enterovirus infection. Enteroviruses gradually modify host cell membrane permeability and rearrange intracellular membranes during infection. The modification of the plasma membrane permeability is most likely important for the lysis of the cell and the release of virus progeny. Modifications of secretory pathway membranes are connected to viral functions such as genome replication. The modification of host cell membrane permeability is such that initially calcium is released from intracellular stores (1). Ionic gradients maintained by the plasma membrane are also disrupted (2). Also, later in infection, small compounds such as hygromycin B, a small non-permeative translation inhibitor, can efficiently enter the cell. The 2B protein has been identified as the viral protein that is responsible for the alterations in host cell membrane permeability that take place in enterovirus-infected cells. Individual expression of the 2B protein was shown to be sufficient for the release of calcium from intracellular stores and the increase in plasma membrane permeability to both calcium and hygromycin B (1,(3)(4)(5)(6).
Another important membrane modification that can be observed in enterovirus-infected cells is the massive proliferation and accumulation of membrane vesicles in the cytoplasm. These vesicles are derived from the secretory pathway (7) and were shown to be the site at which viral RNA replication takes place (8,9). Individual expression of the viral proteins has shown that the 2BC protein is responsible for the induction of the vesicle formation (9 -12), possibly together with the 3A protein (13). The virus-induced inhibition of protein secretion is another important modification, affecting trafficking of host cell membranes. This inhibition may be advantageous for the virus by preventing anti-viral host cell responses such as major histocompatibility complex expression and secretion of interferon and interleukines (14,15). Individual expression of the enterovirus non-structural proteins has shown that both the 2B protein and the 3A protein are endowed with the ability to inhibit protein secretion (1,3). The 3A protein was shown to inhibit ER-to-Golgi transport (16). The step blocked by the 2B protein is still unknown.
The mechanism by which the enterovirus 2B protein exerts its effects on membranes is as yet unknown. Moreover, it remains to be established whether the different activities of the 2B protein (i.e. membrane permeabilization, membrane rearrangement, and secretion inhibition) represent different functions or whether these activities are the result of a general membrane-disturbing activity of the 2B protein. Our aim is to gain more insight into the structure and function of the coxsackievirus 2B protein. The coxsackievirus 2B protein is a small protein of 99 amino acids that, in the infected cell, is present at the virus-induced, secretory pathway-derived membrane vesicles at which the viral replication takes place (17). The protein contains two hydrophobic regions (Fig. 1), of which one is predicted to form a cationic amphipathic ␣-helix with characteristics typical of the group of the so-called membranelytic peptides (18). Mutations in the amphipathic ␣-helix or the second hydrophobic region were shown to have deleterious effects on viral RNA replication and virus growth, indicating that the integrity of these regions is essential for an early function in the viral life cycle (18,19). Mutations in these domains were also found to interfere with the ability of the 2B protein to increase membrane permeability and to inhibit protein secretion (6). This strongly suggests that the membranemodifying activities of the 2B protein are in some way required for the accumulation of the membrane vesicles at which viral RNA replication takes place.
At present, the determinants for the membrane interaction and the subcellular localization of the 2B protein have not been defined. In this study, we have investigated the mode of membrane association in vivo and the subcellular localization of the 2B protein. Furthermore, by testing 2B deletion mutants, we have analyzed the importance of specific domains for the mode of membrane interaction, the subcellular localization, and the ability to increase membrane permeability. Moreover, the correlation between the subcellular localization and the membrane-active character of the 2B protein was determined.

EXPERIMENTAL PROCEDURES
Cells and Media-Buffalo green monkey (BGM) 1 cells were grown in minimal essential medium (Invitrogen) supplemented with 10% fetal bovine serum, 100 units of penicillin/ml, and 100 mg of streptomycin/ ml. Cells were grown at 37°C in a 5% CO 2 incubator.
Antibodies and Conjugates-Rabbit polyclonal anti-EGFP antiserum was a kind gift from Dr. J. Fransen, Department of Cell Biology and Histology, University Medical Center Nijmegen, Nijmegen, The Netherlands. Rabbit polyclonal anti-2B antiserum was obtained by immunizing specific pathogen-free rabbits with a recombinant MBP-2B fusion protein produced in Escherichia coli. Western blot analysis showed that the immune sera specifically recognized the MBP-2B and 2B-EGFP fusion proteins, whereas no reactivity of the preimmune sera was detected (data not shown). Mouse monoclonal anti-C-Myc (clone 9E10) and mouse monoclonal anti-Golgi 58K protein (clone 58K-9) antisera were obtained from Sigma. Rabbit polyclonal anti-calreticulin was obtained from Affinity Bioreagents, Inc. Fluorescein isothiocyanate-conjugated goat anti-rabbit polyclonal antibody, Texas Red-conjugated goat anti-mouse polyclonal antibody, and Texas Red-conjugated goat anti-rabbit polyclonal antibody were obtained from Jackson Immu-noResearch Laboratories. Peroxidase-conjugated goat anti-rabbit immunoglobulin was obtained form Dako Diagnostika.
2B-EGFP and EGFP-2B Fusion Proteins-For the construction of the p2B-EGFP plasmid, the 2B coding sequence was amplified using a forward primer that introduced a SaII restriction site (restriction sites are in italics) and a start codon (underlined) preceded by a Kozak sequence (p115-16, 5Ј-ctcctgtggctggctagcgtcgacgccaccatgggagtgaaggactatgtggaa-3Ј) and a reverse primer that introduced a BamHI restriction site (p115-20, 5Ј-ccagctggatccttggcgttcagccatagg-3Ј). The PCR product was cloned into pEGFP-N3 digested with SalI and BamHI to yield p2B-EGFP. For the construction of the pEGFP-2B plasmid, the 2B coding sequence was amplified by PCR using primer p115-16 and a reverse primer that introduced a stop codon (underlined) and a SmaI restriction site (p115-7, 5Ј-aagccacccgggctattggcgttcagccatagg-3Ј). The PCR product was cloned into pEGFP-C1 digested with SmaI and SalI to yield pEGFP-2B.
For the construction of the p2B(⌬20N)-EGFP plasmid, the coding sequence of aa 21-99 of the 2B protein was amplified using a forward primer that introduced a SalI restriction site and a start codon (underlined) preceded by a Kozak sequence (p115-30, 5Ј-ggggggggggtcgacgccatcatggtatgtgagcaagtcaacctc-3Ј) and a reverse primer p115-20. The PCR product was cloned into pEGFP-N3 digested with SalI and BamHI to yield p2B(⌬20N)-EGFP.
For the construction of the plasmids pHR1-EGFP and pHR2-EGFP, the coding sequences of the regions encompassing these regions (aa 30 -62 and aa 56 -85, respectively) were amplified using primers that introduced a SalI restriction site and a start codon (underlined) preceded by a Kozak sequence at the 5Ј-end and a BamHI restriction site at the 3Ј-end. The HR1 coding sequence was amplified using p115-47 (5Ј-aacctcgtcgacgccaccatgtcactagtgggtcaagactcc-3Ј) and p115-49 (5Ј-agtcacggattcgtcatcgtggttcctcaccac-3Ј), and the HR2 coding region was am- A, hydropathy plot of the 2B protein of CVB3, according to Kyte and Doolittle (45), using a window size of 9 residues. B, amino acid sequence of the two hydrophobic regions shown in the hydropathy plot. C, helical wheel diagram of the putative amphipathic ␣-helix formed by HR1. The hydrophobic residues (boxed) are positioned at one face of the ␣-helix, and the charged and polar residues are positioned at the other face of the ␣-helix. plified using p115-48 (5Ј-tcagccgtcgacgccaccatggtggtgaggaaccacgat-3Ј) and p115-50 (5Ј-atattgggatccctgtttgagccaccgccacgg-3Ј). The region encompassing both HR1 and HR2 (aa 30 -85) was amplified using p115-47 and p115-50. The pHR1-EGFP, pHR2-EGFP, and p(HR1ϩHR2)-EGFP plasmids were made by cloning the PCR products digested with SalI and BamHI into the pEGFP-N3 vector, digested with the same enzymes. All sequences that were amplified by PCR were checked by sequence analysis.
Golgi-targeted 2B Construct-For the construction of Golgi-targeted 2B-EGFP, a construct was generated that contained an additional SmaI restriction site between the 2B and the EGFP coding regions. The coding sequence of the 2B protein was amplified using p115-16 and a reverse primer that introduced a SmaI restriction site to the 5Ј-end of the BamHI restriction site (p318-25, 5Ј-gggggggggggatcccccgggttggcgttcagccatagg-3Ј). The PCR product was cloned into p2B-EGFP digested with SalI and BamHI to yield p2B-EGFP(ϩSmaI). The coding sequence of the N-terminal 103 amino acids of ␤-1,2-N-acetylglucosaminyl-transferase I (NGAT-I) was amplified from pNGAT-I-GFP (a kind gift from Dr. Graham Warren, Cell Biology Laboratory, Imperial Cancer Research Fund, London, UK) using a forward primer that introduced an EcoRV restriction site (p318-14, 5Ј-gggggggatatcctgaagaagcagtctgcagg-3Ј) and a reverse primer that introduced a BamHI restriction site (p381-13, 5Ј-ggggggggatcccgccggcgcgggggtcacagg-3Ј). The PCR product was cloned into p2B-EGFP(ϩSmaI) that was digested with SmaI and BamHI to yield p2B-Golgi-EGFP.
EGFP-Golgi-The plasmid pEGFP-Golgi was constructed by replacing the ECFP coding region of pECFP-Golgi (Clontech) with the EGFP coding region of pEGFP-N1 using the BamHI and NotI restriction sites. In this construct, EGFP is fused to the N-terminal 81 amino acids of human ␤-1,4-galactosyltransferase, which targets the EGFP protein to the trans-medial region of the Golgi apparatus.
Transfections-BGM cell monolayers were grown to 70% confluency in 6-well plates or on coverslips in 24-well plates and transfected with 5 g of plasmid DNA/well of a 6-well plate or with 1 g of plasmid DNA/well of a 24-well plate. Transfections were carried out using Fu-GENE 6 reagent (Roche Molecular Biochemicals) according to the instructions of the manufacturer. For each transfection reaction, the Fugene reagent was mixed with serum-free medium and incubated for 5 min at room temperature. This mixture was added dropwise to the plasmid DNA preparation and incubated for 15 min at room temperature. The DNA/FuGENE mixture was added dropwise to the cells. Cells were grown at 37°C until further analysis.
Membrane Association-BGM cells grown in 6-well plates were transfected with plasmids encoding the proteins EGFP, wild-type 2B-EGFP, or the 2B-EGFP deletion mutants. At 40 h after transfection, cells were washed twice with ice-cold TES (20 mM Tris (pH 7.4), 1 mM EDTA, 100 mM NaCl). Samples were kept on ice during the entire procedure. Cells were scraped in 0.5 ml of ice-cold 1:10 TES, collected by centrifugation for 10 min at 4,500 ϫ g, and resuspended in 250 l of 1:10 TES. Cells were incubated for 15 min and broken by 30 strokes in a Dounce homogenizer. Nuclei and cell debris were removed by centrifugation for 10 min at 4,500 ϫ g. Membrane and cytoplasmic fractions were separated by centrifugation for 1 h at 150,000 ϫ g at 4°C. Supernatants were removed and stored at Ϫ80°C until further analysis. Pellet fractions were either resuspended in one supernatant volume of buffer (50 mM Tris (pH 7.4), 150 mM NaCl, 1 mM EDTA, 0.1 M phenylmethylsulfonyl fluoride, 1% Nonidet P-40, 0.05% SDS) and stored at Ϫ80°C until further analysis or resuspended in 200 l of PBS, 0.1 M Na 2 CO 3 (pH 11.5), 0.5 M EDTA, 1 M NaCl, 4 M urea, or 1% Triton X-100. Resuspended pellet fractions were incubated on ice for 1 h and centrifuged for 1 h at 150,000 ϫ g and 4°C. The supernatant fractions and the pellet fractions, which were resuspended in 200 l of lysis buffer, were analyzed by Western blot.
Western Blot Analysis-Supernatant and pellet fractions of each sample were run on a 12.5% SDS-polyacrylamide gel and transferred to a nitro-cellulose membrane (Bio-Rad). EGFP fusion proteins were stained with the anti-EGFP polyclonal antiserum (dilution 1:10,000) and peroxidase-conjugated goat anti-rabbit immunoglobulins (dilution 1:1,000) and visualized using Lumi-Light plus Western blotting substrate (Roche Molecular Biochemicals) according to the instructions of the manufacturer.
Hygromycin B Assay-BGM cells grown in 6-well plates were transfected in duplicate with plasmids encoding the proteins EGFP, wildtype 2B-EGFP, or the 2B-EGFP deletion mutants. At 40 h after transfection, one of the duplicate wells was starved of methionine for 15 min in the presence of 500 g/ml hygromycin B, whereas the other well was starved of methionine for 15 min in the absence of hygromycin B. The cells were then pulse-labeled for 1 h with [ 35 S]methionine (50 Ci/well) in the presence or absence of hygromycin B, washed with ice-cold PBS, and lysed in 1 ml of lysis buffer (50 mM Tris (pH 7.4), 150 mM NaCl, 1 mM EDTA, 0.1 M phenylmethylsulfonyl fluoride, 1% Nonidet P-40, 0.05% SDS). Anti-EGFP rabbit polyclonal antiserum (1:1,000) was added to the cell lysate, and the lysate/antiserum solution was incubated at 4°C for 18 h. To collect the antibody-protein complexes, the samples were incubated for 2 h with protein A-Sepharose (Amersham Biosciences), washed two times with dilution buffer (0.01 M Tris (pH 8.0), 0.14 M NaCl, 0.1% bovine serum albumin, 0.1% Triton X-100), washed once with TSA (0.01 M Tris (pH 8.0), 0.14 M NaCl), washed once with 0.05 M Tris (pH 6.8), and precipitated. Samples were resuspended in 35 l of SDS-sample buffer, boiled for 5 min, and analyzed by SDS-PAGE. The amount of radiolabeled anti-EGFP-precipitated protein was quantified using a phosphorimaging device (Bio-Rad Multi-Analyst version 1.0.1), and the ratios of the amount of protein synthesized in the presence and absence of hygromycin B were determined.

Construction and Characterization of N-and C-terminal Fu-
sions of EGFP to the 2B Protein-Tagged 2B proteins were used because (i) an anti-2B antiserum might not recognize all 2B deletion mutants due to possible loss of epitopes and (ii) the deletion mutants may become too small (Ͻ10 kDa) to be efficiently detected. Fusion proteins of wild-type and mutant 2B to EGFP were used because they may be used in subcellular localization studies in living cells, and the resulting 36-kDa fusion protein is efficiently recognized and detectable by an anti-EGFP serum.
The fusion of a fluorescent protein at the N terminus or C terminus of a protein may affect its biological properties. Therefore, the possible effects of the EGFP fusion on the membrane-active character and the subcellular localization of the 2B protein were studied. The ability of 2B-EGFP and EGFP-2B to increase plasma membrane permeability of transfected cells was studied by analyzing the entry of hygromycin B, a small inhibitor of translation that under physiological conditions poorly passes the plasma membrane. Cells were transfected in duplicate wells with the indicated constructs and pulse-labeled in the absence or presence of hygromycin B. The radiolabeled EGFP fusion proteins were immunoprecipitated and analyzed by SDS-PAGE to measure the entry of hygromycin B ( Fig. 2A). In EGFP-expressing cells, the amount of protein synthesis was similar in the absence or presence of hygromycin B, indicating that the EGFP protein does not increase plasma membrane permeability to hygromycin B. Expression of the 2B-EGFP protein resulted in the increased entry of hygromycin B, as was reflected by the almost complete inhibition of protein synthesis in the presence of hygromycin B. The increase in plasma membrane permeability induced by the 2B-EGFP protein was similar to that induced by the untagged 2B protein. The EGFP-2B protein showed a reduced ability to increase plasma membrane permeability as compared with the untagged 2B protein.
The subcellular localization of the 2B-EGFP and EGFP-2B fusion proteins was compared with that of the untagged 2B protein, which was stained with the anti-2B antiserum (Fig.  2B). The 2B-EGFP fusion protein was predominantly present in the juxtanuclear Golgi region, similar to the untagged 2B protein. In contrast, a small portion of the 2B-EGFP was observed in a reticular, ER-like pattern. The subcellular localization of the EGFP-2B protein differed substantially from that of the untagged 2B protein and showed a resemblance to the ER. Based on these results, it was decided to use the 2B-EGFP protein in the subsequent experiments.
In Vivo Membrane Association of 2B-EGFP-The membrane association properties of the 2B protein were studied using an in vivo system in which cytosolic and membrane fractions of EGFP-or 2B-EGFP-expressing cells were collected and analyzed on Western blot. As expected, the cytosolic EGFP protein was found exclusively in the supernatant fraction (Fig. 3A). The 2B-EGFP fusion protein was found exclusively in the pellet fraction, indicating that the 2B protein is associated with membranes. The mode of membrane association was further investigated by extracting the membrane fractions with buffers that discriminate between peripheral membrane proteins (which are attached to membranes by ionic interactions with membrane-integral proteins or phospholipid head groups) and integral membrane proteins (which are embedded in the phospholipid bilayer). Fig. 3B shows the supernatant and pellet fractions of the membrane fractions following extraction with the indicated buffers. The 2B-EGFP protein remained attached to membranes upon extraction with PBS (control) and buffers that extract peripheral membrane proteins by increasing the ionic strength of the buffer (1 M NaCl), creating mild chaotropic salt conditions (4 M urea), or chelating divalent cations (0.5 M EDTA). Moreover, the protein was still detected in the pellet fraction when membranes were extracted using alkaline conditions (0.1 M Na 2 CO 3 , pH 11.5), one of the most potent methods to discriminate between peripheral and membrane-integral proteins (22). In addition, alkaline conditions convert membranes to sheets (but do not solubilize membranes) and release soluble proteins that are trapped inside membranous vesicles. The resistance of 2B-EGFP to alkaline extraction therefore argues against a peripheral association of the 2B-EGFP protein at the lumenal side of endo-membranes. The 2B-EGFP protein was found in the supernatant fraction only upon solubilization of membranes using 1% Triton X-100, a non-ionic detergent that releases integral membrane proteins. Taken together, these findings provide evidence that the 2B protein is embedded in the phospholipid bilayer, rather than evidence that it is peripherally associated with membranes.
In Vivo Membrane Association of 2B-EGFP Deletion Mutants-Deletion mutants of the 2B protein were designed to identify the regions that are important for its mode of membrane association. Wild-type 2B (99 aa) and deletion mutants of 2B are depicted in Fig. 3C. Protein 2B(⌬59C) contains only the N-terminal 40 amino acids of 2B, protein 2B(⌬14C) lacks the C-terminal 14 amino acid residues, and protein 2B(⌬20N) has a deletion of the N-terminal 20 amino acids. In addition, deletion mutants were generated that lacked either the amphipathic ␣-helix (aa 34 -56), which will further be referred to as the first hydrophobic region HR1 or the second hydrophobic region (aa 64 -80), which will further be referred to as HR2.
Cytosolic and membrane fractions of cells transfected with 2B-EGFP deletion mutants were prepared and analyzed as described above for the wild-type 2B-EGFP protein. Fig. 3C shows that protein 2B(⌬59C)-EGFP is predominantly present in the cytosolic fraction, indicating that the N-terminal 40 amino acids of the 2B protein are not sufficient to mediate membrane association. The 2B(⌬14C)-EGFP and 2B(⌬20N)-EGFP proteins were found predominantly in the pellet fraction, demonstrating that the N and the C terminus of the protein do not contribute to the membrane association of the 2B protein. Also, remarkably, the 2B(⌬HR1)-EGFP and 2B(⌬HR2)-EGFP proteins were found predominantly in the membrane fraction, indicating that the presence of either one of the hydrophobic regions of the 2B protein was sufficient to mediate membrane binding.
The membrane-associated 2B-EGFP deletion mutants were further investigated to determine their mode of membrane association. Membrane fractions were extracted using PBS, 0.1 M Na 2 CO 3 , pH 11.5, and 1% Triton X-100 (Fig. 3D). The 2B(⌬14C)-EGFP and 2B(⌬20N)-EGFP proteins behaved like the wild-type 2B protein and were found predominantly in the membrane fraction after the extraction with 0.1 M Na 2 CO 3 , suggesting that the N and C terminus of the 2B protein have little influence on its mode of membrane association. Both protein 2B(⌬HR1)-EGFP and protein 2B(⌬HR2)-EGFP behaved differently upon Na 2 CO 3 extraction. Substantial amounts of these proteins were found in the supernatant fraction upon Na 2 CO 3 extraction, whereas both proteins were almost exclusively detected in the pellet fraction upon PBS extraction. These results indicate that both hydrophobic regions are required for the correct membrane association of the 2B protein. When the membranes were solubilized using 1% Triton X-100, the wild-type 2B-EGFP protein as well as 2B(⌬14C)-EGFP, 2B(⌬20N)-EGFP, and 2B(⌬HR2)-EGFP were predominantly present in the supernatant fraction. In contrast, the majority of the 2B(⌬HR1)-EGFP protein remained present in the pellet fraction upon Triton X-100 extraction, indicating that 2B(⌬HR1)-EGFP was pelleted in a membrane-independent manner upon solubilization of membranes. This result suggests that the 2B(⌬HR1)-EGFP protein is prone to form aggregates in the absence of membranes (due to the Triton X-100 extraction). It cannot be excluded that the 2B(⌬HR1)-EGFP protein also (partially) formed aggregates in the Na 2 CO 3 extraction experiment. It is therefore difficult to draw conclusions regarding the membrane association properties of 2B(⌬HR1), other than that it behaves differently from the wild-type 2B protein, suggesting that this region is an important determinant for the membrane association properties of the 2B protein. In the subcellular localization experiments documented below, however, evidence will be provided that the 2B(⌬HR1)-EGFP protein is present as a membrane-associated protein in vivo. The results of the in vivo membrane association experiments are summarized in Table I.
Subcellular Localization of Protein 2B Deletion Mutants-The importance of specific domains of the 2B protein for its subcellular localization was investigated in living cells expressing 2B-EGFP deletion mutants (Fig. 4). The wild-type 2B-EGFP protein was predominantly present in the juxtanuclear Golgi region and was additionally found in an ER-like reticular pattern, similar to what was observed in fixed cells (Fig. 2B). The 2B(⌬59C)-EGFP was present throughout the cell, staining both the cytoplasm and the nucleus, consistent with the observation that the protein is not membrane-associated (Fig. 3C). Localization of the 2B(⌬14C)-EGFP and 2B(⌬20N)-EGFP was essentially the same as that of the wild-type 2B-EGFP protein (data not shown). The 2B(⌬HR1)-EGFP and 2B(⌬HR2)-EGFP proteins showed an altered localization as compared with the wild-type 2B-EGFP protein. The 2B(⌬HR1)-EGFP protein was present in an ER-like network, and staining of the juxtanuclear Golgi region was virtually absent. The 2B(⌬HR2)-EGFP protein was observed in the juxtanuclear Golgi region; however, staining of the ER-like network was more apparent than for wild-type 2B-EGFP.
A more detailed examination of the subcellular localization was undertaken in fixed cells by investigating colocalization of the 2B protein with markers for the Golgi and the ER. In these experiments, the 2B deletion mutants were analyzed as Myctagged proteins. This procedure was followed because the subcellular localization of the Myc-tagged 2B protein better resembled that of the untagged 2B protein (compare Figs. 2B and 5A). Colocalization experiments with the Golgi complex were performed using the EGFP-Golgi fusion protein in which EGFP is targeted to the trans-medial region of the Golgi complex by fusion to the N-terminal 81 amino acids of human ␤-1,4-galactosyltransferase. Colocalization experiments with the ER were performed using the anti-calreticulin antiserum. The 2B-Myc, 2B(⌬14C)-Myc, and 2B(⌬20N)-Myc proteins colocalized with the EGFP-Golgi marker and were virtually absent outside of the Golgi region (Fig. 5, A-C, respectively), indicating that the N and the C terminus of the 2B protein have little or no influence on its subcellular localization. Interestingly, 2B(⌬HR1)-Myc staining was completely different from the wild-type 2B-Myc protein. The 2B(⌬HR1)-Myc protein did not colocalize with EGFP-Golgi (Fig. 5D) but was exclusively present in a network-like pattern that was identical to the anticalreticulin-stained ER (Fig. 5F). Also, the 2B(⌬HR2)-Myc

FIG. 3. Membrane association of the wild-type and mutant 2B-EGFP proteins.
Cells were transfected with the indicated constructs. At 40 h after transfection, cells were harvested and broken in a Dounce homogenizer. Cytosolic (supernatant (s)) and membrane (pellet (p)) fractions were prepared and analyzed by Western blot using the anti-EGFP polyclonal antiserum. A, analysis of cytosolic and membrane fractions of cells expressing either EGFP or 2B-EGFP. The EGFP protein is present in the cytosolic fraction, and the 2B-EGFP protein is present in the membrane fraction. B, mode of membrane association of the 2B-EGFP protein. Membrane fractions were extracted with the indicated buffers. The 2B-EGFP protein is only released using Triton X-100 (TX-100). C, schematic representation and membrane association characteristics of the 2B-EGFP protein and 2B-EGFP deletion mutants. Lines indicate the deleted regions of the protein. All proteins, except 2B(⌬59C)-EGFP, are present in the membrane fractions. wt, wild type. D, mode of membrane association of 2B-EGFP deletion mutants. The 2B(⌬HR1)-EGFP and 2B(⌬HR2)-EGFP proteins behave differently from the wildtype 2B-EGFP protein and are partially present in the supernatant fractions upon Na 2 CO 3 extractions. All proteins, except the 2B(⌬HR1)-EGFP protein, are predominantly present in the supernatant fraction upon TX-100 extraction. For each panel, one of three representative experiments is shown. staining differed from that of the wild-type 2B-Myc protein.
The 2B(⌬HR2)-Myc protein was present in the Golgi complex (Fig. 5E), but also a substantial part was observed in a network-like pattern that was identified as the ER (Fig. 5G).
To further investigate the role of HR1 and HR2 in the Golgi localization, the ability of these domains to provide Golgi localization in the absence of other determinants of the 2B protein was analyzed. Initially, Myc-tagged HR1 and HR2 constructs were generated. However, it was found that HR1-Myc and HR2-Myc proteins were too small to be preserved during fixation (data not shown). Therefore, heterologous proteins were constructed in which regions of the 2B protein encompassing either HR1 or HR2, or both HR1 and HR2, were fused to the N terminus of the EGFP protein. The Golgi complex was stained using the anti-Golgi 58K marker. Colocalization studies of these proteins with this endogenously expressed Golgi marker showed that the presence of HR1 could indeed partially target the EGFP protein to the juxtanuclear Golgi region (Fig. 6A), indicating that HR1 acts as a Golgi targeting signal in the absence of other domains of the 2B protein. The HR2-EGFP protein was present in a reticular network that showed a close resemblance to the ER (Fig. 6B). Its absence from the Golgi complex suggests that HR2 does not contain any Golgi targeting signals. In contrast to the HR1-EGFP protein, which was partially localized in the ER, the (HR1ϩHR2)-EGFP fusion protein was predominantly present in the Golgi complex (Fig.  6C), giving support for an additional role of HR2 in Golgi localization. Taken together, these findings indicate that, although the amphipathic ␣-helix is the Golgi targeting signal, the second hydrophobic region is an important additional determinant for the Golgi localization of both the 2B protein and the heterologous EGFP protein. The results of the subcellular localization experiments are summarized in Table I.
Membrane-permeabilizing Activity of 2B-EGFP Deletion Mutants-In the experiments described above, we have shown that the 2B protein is an integral membrane protein that is predominantly localized at the Golgi complex and that these properties are largely dependent on the presence of both hydrophobic regions of the protein. To investigate the importance of different regions of the 2B protein for its membrane-active character, we also tested the 2B-EGFP deletion mutants for their ability to increase membrane permeability to hygromycin B.
First, a time course experiment was performed to define the time point at which plasma membrane permeability induced by the 2B-EGFP protein was maximal (Fig. 7A). From 24 h after transfection to 40 h after transfection, an increased influx of hygromycin B was observed, resulting in almost complete inhibition of protein synthesis in its presence at 40 h after transfection. When cells were analyzed at 48 h after transfection, no further increase in plasma membrane permeability to hygromycin B was observed. Therefore, all subsequent experiments were performed at 40 h after transfection. Fig. 7B shows the results of the hygromycin B assay for the 2B-EGFP deletion mutants. All mutants were tested in three independent assays, the amount of protein synthesized was determined by phosphorimaging, and the ratio of protein synthesis in the presence and absence of hygromycin B was calculated. Cells expressing wild-type protein 2B-EGFP showed an increased entry of hygromycin B, reflected by the reduction in protein synthesis to 10% of protein synthesis in cells expressing the wild-type 2B-EGFP protein in the absence of hygromycin B. Cells expressing protein 2B(⌬59C)-EGFP showed no influx of hygromycin B, indicating that the N-terminal 40 amino acids of the 2B protein are not sufficient to increase plasma membrane permeability. Hygromycin B entry in cells expressing the 2B(⌬14C)-EGFP protein was approximately equal to that of cells expressing the wild-type 2B protein. In cells expressing 2B(⌬20N)-EGFP, protein synthesis in the presence of hygromycin B was ϳ45% of that in the absence of hygromycin B, reflecting that plasma membrane permeability was reduced as compared with that of cells expressing the wild-type 2B-EGFP protein. Cells expressing either the 2B(⌬HR1)-EGFP or 2B(⌬HR2)-EGFP fusion protein showed no entry of hygromycin B, indicating that the ability of these proteins to increase plasma membrane permeability was completely disrupted. Thus, the combined presence of both hydrophobic regions is not only re- The resistance to Na 2 CO 3 extraction is scored ϩϩϩϩ when 80 -100% of the protein was present in the pellet fraction upon extraction, ϩϩϩ when 60 -80% was present in the pellet fraction, ϩϩ when 40 -60% was present in the pellet fraction, ϩ when 20 -40% was present in the pellet fraction, and Ϫ when less than 20% was present in the pellet fraction.
b The membrane-active character is scored ϩϩϩϩ when protein synthesis in the presence of hygromycin B was reduced by 80 -100% as compared with protein synthesis in the absence of hygromycin B, ϩϩϩ when protein synthesis was reduced by 60 -80%, ϩϩ when protein synthesis was reduced by 40 -60%, ϩ when protein synthesis was reduced by 20 -40%, and Ϫ when protein synthesis was reduced by less than 20%. c wt, wild type. d NT ϭ not tested. quired for the membrane-integral character of the 2B protein but also for its membrane-active function, suggesting that the membrane-integral topology is essential for its ability to modify membrane permeability. The results of the membrane permeabilization experiments are summarized in Table I.
Functional Characterization of Organelle-targeted 2B Proteins-The 2B protein is predominantly present in the Golgi complex of transfected cells. We investigated the importance of the Golgi localization for the ability to increase plasma membrane permeability. For this purpose, the 2B-EGFP protein was targeted exclusively to the ER or to the Golgi complex by fusion of specific targeting signals. It was shown previously that membrane proteins acquired ER localization by the addition of a tag sequence containing the KKAA dilysine motif (AAAAAAKKAA), which acts as an ER retention signal in mammalian cells (23). The fusion of the KKAA motif to the 2B-EGFP protein indeed abolished the Golgi localization and resulted in a typical ER localization pattern (Fig. 8A). The presence of a control tag of 10 alanine residues, containing no additional localization signals, did not affect the subcellular localization of the 2B-EGFP protein (Fig. 8A). To confer complete Golgi localization to the 2B-EGFP protein, a construct was generated in which the 2B protein was fused at the N terminus of the NGAT-I-GFP construct (which contains the targeting information of the ␤-1,2-N-acetylglucosaminyl-transferase I, a type I med-trans Golgi membrane protein) (24). The resulting 2B-Golgi-EGFP protein was exclusively present in the Golgi complex (Fig. 8A).
To analyze the importance of the subcellular localization of the 2B protein for its membrane-active function, the entry of hygromycin B was analyzed in cells expressing the targeted 2B proteins. Fig. 8B shows that the ER-targeted 2B-EGFP-KKAA protein was severely impaired in its ability to increase plasma membrane permeability to hygromycin B. The 2B-EGFP-AAAA protein behaved like the untagged protein, indicating that the decrease in hygromycin B entry was not the result of the fusion of a tag to the C terminus of 2B-EGFP. Rather, the altered subcellular localization of the protein was responsible for this result, suggesting that the Golgi localization of the 2B protein is important for its effect on plasma membrane permeability. This finding was further substantiated by the observation that the 2B-Golgi-EGFP protein increases plasma membrane permeability to hygromycin B to the same extent as the untagged FIG. 8. Functional characterization of organelle-targeted 2B-EGFP proteins. The 2B-EGFP-KKAA construct encodes the wild-type 2B-EGFP protein carrying the AAAAAAKKAA sequence (containing the KKAA ER retention signal) at its C terminus. The control 2B-EGFP-AAAA construct contains the AAAAAAAAAA sequence (lacking targeting information) at its C terminus. The 2B-Golgi-EGFP construct contains the N-terminal 103 aa of NGAT-I between the 2B and EGFP coding sequences. A, subcellular localization of the targeted 2B-EGFP proteins. Cells were transfected with the indicated constructs, fixed at 24 h after transfection, and analyzed by CLSM. The 2B-EGFP protein is predominantly present in the juxtanuclear Golgi region and additionally in a reticular pattern. The 2B-EGFP-KKAA protein is exclusively present in the ER, whereas the subcellular localization of 2B-EGFP and 2B-EGFP-AAAA is identical. The 2B-Golgi-EGFP is exclusively present in the Golgi complex. B, analysis of plasma membrane permeability to hygromycin B (HB) of cells expressing the targeted 2B-EGFP proteins. The experimental setup is described in the legend for Fig. 1. Values represent means Ϯ standard errors of measurements of three independent experiments. The 2B-EGFP-AAAA and 2B-Golgi-EGFP proteins increased plasma membrane permeability to a level that is similar to that of the untagged 2B-EGFP protein. The 2B-EGFP-KKAA protein is severely impaired in its ability to increase plasma membrane permeability. 2B-EGFP protein. These data indicate that the Golgi localization is a critical determinant for the ability of the 2B protein to modify permeability of the plasma membrane. DISCUSSION The enterovirus 2B protein is responsible for the alterations in the permeability of secretory membranes and the plasma membrane that are observed in infected cells (1,3,5,6). The molecular mechanism by which the 2B protein disturbs host cell membranes is largely unknown. In this study, we have investigated the structural and functional requirements for the in vivo membrane association and the membrane-active function of this important virus protein.
The membrane association of the 2B protein was resistant against extraction with buffers that release peripherally associated membrane proteins, indicating that one or more domains of the protein are embedded in the phospholipid bilayer rather than that the protein associates with membranes through electrostatic interactions with membrane proteins or polar lipid head groups. Expression of deletion mutants and heterologous proteins showed that each of the hydrophobic regions HR1 and HR2 alone could mediate membrane binding. However, in contrast to the wild-type 2B protein, a substantial amount of the 2B proteins lacking either HR1 or HR2 was released upon Na 2 CO 3 extraction, indicating that both hydrophobic regions are required for the correct membrane association. Moreover, the presence of both HR1 and HR2 was absolutely required for the membrane-active function of the 2B protein, consistent with previous reports (6,11). Together, these findings suggest that the two hydrophobic regions are cooperatively involved in the formation of a membrane-integral complex and that the formation of this complex is required for the membrane association and membrane-active function of the 2B protein.
These findings provide important new insights into the molecular architecture of the 2B protein in the membrane. The first hydrophobic region of the 2B protein has been predicted to form a cationic amphipathic ␣-helix that shows similarities with the group of lytic polypeptides (18). Amphipathic ␣-helical lytic polypeptides may form multimers that build membraneintegral pores by exposing their hydrophobic sides to the lipid bilayer, whereas their hydrophilic faces form an aqueous interior (25)(26)(27). The 2B protein has been shown to form homomultimers by yeast two-hybrid analysis (28), mammalian twohybrid analysis (21), and in living mammalian cells using fluorescence resonance energy transfer microscopy (29). The behavior of the 2B protein as an integral membrane protein and the observation that each of the hydrophobic regions HR1 and HR2 can independently mediate membrane binding lend support to the idea that multimers of protein 2B build membrane-integral pores and that each of these hydrophobic regions is involved in the formation of these pore complexes. Fig.  9 shows a model of a multimeric pore complex formed by 2B proteins. According to this model, HR1 and HR2 interact with each other to form a "helix-loop-helix" hairpin motif that traverses the lipid bilayer. The hydrophobic regions are separated by a hydrophilic loop sequence (RNHDD in the CVB3 2B protein). Presumably both the N and C terminus of the 2B protein reside at the cytosolic side of the membrane. This suggestion is supported by the observation that a cytosolic ER retention signal (KKAA) (23,30) at the C terminus of the 2B protein is functional in conferring ER retention to the 2B protein. Furthermore, the lack of glycosylation at NX(S/T) glycosylation signals introduced at different positions in the region upstream HR1 argues that the N terminus is localized at the cytosolic side of the membrane. 2 Finally, localization of the N and C terminus at the cytosolic side of the membrane is consistent with the need of proteolytic liberation of the 2B protein from the viral polyprotein by the viral protein 3C pro , a cytosolic protease. Both the membrane association and the membraneactive function were independent of the C-terminal 15 amino acids. Although dispensable for the membrane association, the N-terminal 20 amino acids were found to be of importance for the membrane-active function of the 2B protein, indicating that the molecular structure and function of the 2B protein depend on more than just the presence of HR1 and HR2 and therefore require further investigation.
Upon expression in the absence of other viral proteins, the CVB3 2B protein was localizing in the Golgi complex, consistent with the Golgi localization described for the 2B protein of the closely related poliovirus (31). Retention of the 2B protein in the ER by fusion of an ER retention signal severely inhibited the ability to increase plasma membrane permeability. This activity, however, was unaffected when the 2B protein was exclusively targeted to the Golgi complex by fusion of a Golgi targeting signal. This strongly suggests that the Golgi complex is the target organelle of the 2B protein to exert its effect on the plasma membrane. It is remarkable that a virus protein that is localized at the Golgi complex can modify plasma membrane permeability. The mechanism underlying this activity is as yet unknown. We predict that this activity is somehow linked to the ability of the 2B protein to increase Golgi membrane permeability. Previously, we have shown that expression of the 2B protein resulted in a decrease in the amount of calcium that could be released from thapsigargin-sensitive stores (1). Because at that time only the ER was recognized as a thapsigargin-sensitive store (32), we concluded that the 2B protein modified ER membrane permeability. Recently, however, the Golgi complex was also identified as a thapsigargin-sensitive intracellular calcium store (33). This finding, together with the identification of the Golgi complex as the main target organelle of the 2B protein, suggests that the 2B protein causes an increased Golgi membrane permeability and a general disruption of the ionic gradients maintained by the Golgi membrane. Alterations in ionic gradients maintained by membranes of the Golgi complex have been demonstrated to result in functional disturbances of the Golgi complex. Disruption of cationic gradients by monensin, an H ϩ /Na ϩ ionophore that mediates cation diffusion over Golgi membranes, results in the inhibition of complex glycosylation in the Golgi complex (reflected by the inability of secretory proteins to acquire resistance to endo-␤-N-acetylglucosaminidase H treatment) and inhibition of protein secretion (34). Analogous to the effects of monensin, expression of the 2B protein also results in the inhibition of complex glycosylation (as demonstrated by the endo-␤-N-acetylglucosaminidase H sensitivity of glycoproteins in 2B-expressing cells) and inhibition of protein secretion (3). We propose that the disruption of the ionic content of the Golgi complex by 2B is responsible for this effect. How the disturbance of the Golgi milieu leads to an increase in plasma membrane permeability remains to be established. Possibly, the 2B-induced defect in transport of membrane proteins or lipids beyond the Golgi complex results in destabilization and permeabilization of the plasma membrane. However, it cannot be excluded that the effect on the plasma membrane is caused by the disturbance of another Golgi function(s).
Having identified the Golgi complex as the main target organelle of the 2B protein, the question of what determines the Golgi localization of the 2B protein arises. In this study, the amphipathic ␣-helix was identified as a Golgi targeting signal that could confer partial Golgi localization to the heterologous EGFP protein in the absence of other domains of the 2B protein. However, the presence of both the amphipathic ␣-helix and the second hydrophobic domain was found to be required for efficient Golgi targeting. The Golgi localization of the 2B protein might depend on the length of the transmembrane domain(s), as suggested by the "bilayer-thickness model" (35)(36)(37)(38)(39)(40). HR1 and HR2 are predicted to consist of 20 and 18 aa respectively, which is similar to the transmembrane domain of some Golgi resident proteins but shorter than the transmembrane domain of plasma membrane proteins (35,36,41). Alternatively, the "kin-recognition" model suggests that Golgi residents form large, oligomeric complexes that become too large to be transported as they reach the correct Golgi compartment (42,43). It is unlikely that homo-multimerization reactions of the 2B protein (21,28,29) are involved in determining its Golgi localization as 2B deletion mutants that lacked either HR1 or HR2 were not transported beyond the Golgi complex, although they were unable to form homo-multimers (28). Another possible explanation is that the 2B protein is not transported further along the secretory pathway because of the lack of specific signals, as was suggested for some trans-Golgi network markers (35). The possibility that trafficking of the 2B protein out of the Golgi complex is blocked as a consequence of the inhibition of protein secretion that is induced by the 2B protein itself (3,6) is unlikely as 2B deletion mutants lacking either HR1 or HR2 are not transported beyond the Golgi complex, although their function was severely impaired.
How do our findings relate to the situation in virus-infected cells, where the 2B protein is present at the virus-induced membrane vesicles that build the viral RNA replication complex? Initially, these membrane vesicles are derived from the ER (9). Later in infection, however, the Golgi complex has disappeared (12,31,44), and Golgi membranes contribute to the vesicle population (7,44), suggesting that the Golgi complex is gradually used up to produce the membrane vesicles that build the viral replication complex. Therefore, our findings that identified Golgi membranes as the main target of the 2B protein are in agreement with the localization of the 2B and 2BC proteins at the ERand Golgi-derived membrane vesicles in infected cells.
The exact function of the 2B protein in the early steps in viral RNA replication remains to be established. The viability of viruses that carry mutations in the 2B protein closely correlates with the ability of the 2B protein to increase membrane permeability and inhibit protein secretion (1, 6), suggesting that the destabilization and/or permeabilization of secretory pathway membranes by the 2B protein is somehow required for the accumulation of membrane vesicles or the formation of the membranous replication complex. Additional studies are needed to gain better insight into the membrane-active functions of the 2B and 2BC proteins and how these functions might aid in the processes that are required to facilitate the formation of membrane-associated replication complexes.