Expression of Semliki Forest Virus E1 Protein in Escherichia coli LOW pH-INDUCED PORE FORMATION*

Exposure of Semliki Forest virus 1 to mildly acidic conditions results in conformational changes of the viral spike proteins, which in turn leads to a pore formation across its membrane. The ability to form a pore has been ascribed to the ectodomain of the Semliki Forest virus (SFV) E1 spike protein. To elucidate whether the E1 protein per se is sufficient for low pH-dependent pore formation, we expressed E1 in Escherichia coli in an inducible manner using the pET11c expression system. The data obtained clearly showed that the E1 protein was expressed in the bacterial cell membrane and that exposure of E. coli expressing the SFV E1 protein to low pH ( < 6.2) resulted in a permeability change of the mem- brane. Thus, we conclude that the E1 protein of SFV per se is sufficient to promote pore formation under mildly acidic conditions.

Exposure of Semliki Forest virus 1 to mildly acidic conditions results in conformational changes of the viral spike proteins, which in turn leads to a pore formation across its membrane. The ability to form a pore has been ascribed to the ectodomain of the Semliki Forest virus (SFV) E1 spike protein. To elucidate whether the E1 protein per se is sufficient for low pH-dependent pore formation, we expressed E1 in Escherichia coli in an inducible manner using the pET11c expression system. The data obtained clearly showed that the E1 protein was expressed in the bacterial cell membrane and that exposure of E. coli expressing the SFV E1 protein to low pH (<6.2) resulted in a permeability change of the membrane. Thus, we conclude that the E1 protein of SFV per se is sufficient to promote pore formation under mildly acidic conditions.
The entry of a virus into a host cell is an essential step in the chain of events leading to infection. A multitude of viruses use the endocytotic pathway to access host cells. As a model for the entry of enveloped animal viruses into cells, the ␣ virus Semliki Forest virus has been extensively studied (1). Once attached, the virus is internalized via coated vesicles and transferred to the endosome. Due to the acidic conditions within this organelle, the lipid envelope of SFV fuses with the endosomal membrane of the target cell (2). This low pH-induced fusion is mediated by the so called virus spikes (3)(4)(5). Each spike is a heterotrimer, being composed of the type I integral membrane glycoproteins E1 (50.786 kDa) and E2 (51.855 kDa), plus the peripheral glycoprotein E3 (11.369 kDa), which is associated with E2 (6). Several functions have been ascribed to the spike proteins; e.g. the E2 and E3 precursor protein p62 forms a heterodimer with E1 in the endoplasmic reticulum and is responsible for the transport of the complex to the plasma membrane (6). The E1 protein is involved in the acid-induced fusion of the viral and endosomal membranes (7)(8)(9).
Under mildly acidic conditions (pH 5.8) the spike proteins undergo an irreversible conformational change that results in the dissociation of the E1/E2/E3 complex, the formation of an E1 homotrimer and the exposure of a fusion peptide on the E1 protein (7). This conformational change also leads to the formation of a pore, which causes an alteration in the permeabil-ity of the virion membrane or of a cell membrane expressing the spike proteins (10 -14). It has been suggested that this acid-induced pore formation plays a crucial role in the penetration and uncoating process of SFV (15).
Several experiments have shown that pore formation is dependent on the ectodomain of the E1 spike protein (14,16,17).
It has been speculated that the E1 protein per se would be sufficient for triggering acid-induced pore formation. So far, all attempts to express isolated E1 protein on the surface of eukaryotic cells have failed. The E1 protein was synthesized but retained in the endoplasmic reticulum, since efficient transport of the glycoproteins to the plasma membrane requires heterodimerization of E1 with E2/E3 (18). Therefore, to further investigate the role of E1 during pore formation, we decided to express E1 in the prokaryotic host Escherichia coli.
Our data clearly demonstrate that the E1 protein, although lacking a signal sequence, is transported to the plasma membrane of E. coli. Furthermore, we have shown that E1 is capable of modifying the membrane permeability of E. coli in a pH-dependent manner, leading to pore formation.

Construction of the SFV E1 Expression
Plasmid-DNA manipulations were performed by the use of standard cloning procedures (19). Polymerase chain reaction was used to amplify the SFV 1 E1 gene from pSKm-E1Ј. This plasmid contains the entire E1 gene and was derived by subcloning a 1951-bp SpeI-EcoRV fragment from pSP6-SFV4, which contains the full-length cDNA sequence of SFV (20,21).
The primers were designed to introduce a start and an additional stop codon for translation. To facilitate cloning of the E1 gene into the expression vector pET11c (Stratagene AG, Amsterdam, Netherlands), the oligonucleotides encoded a unique NdeI and BamHI restriction site at their 5Ј and 3Ј end, respectively. Since the E1 gene contains an intragenic NdeI restriction site, the 3Ј primer was designed to introduce a silent point mutation (AϾG) that eliminated this NdeI cutting site.
The 1340-bp E1 polymerase chain reaction fragment was purified using a polymerase chain reaction purification kit (Qiagen AG, Basel, Switzerland), digested with the restriction enzymes BamHI and NdeI, and ligated into the pET11c vector. The ligation mixture was used to transform competent XL-1 blue E. coli cells and the resulting colonies screened by digestion of the isolated plasmid DNA with the appropriate restriction enzymes. The plasmid containing the E1 gene (pET11c-E1) was then used to transform BL21(DE3) E. coli cells.
Construction of the pET11c_E1/23_E2/432 Plasmid-Most of the E2 protein sequence was cut out from the plasmid pSp6-SFV4 (21) using the restriction enzymes BglI and SgrAI. The ends of the resulting fragment of 1262 bp were polished with Klenow enzyme and mung bean nuclease. The plasmid pMSEH1_E1/23_phoA (Nyfeler et al.) 2 was digested with PstI and SalI, resulting in a 5371-bp fragment whose ends were blunted, too. This fragment was ligated with the 1262-bp-long * This work was supported in part by Swiss National Science Foundation Grant 31-49217.96 (to C. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Present address: Dept. of Pathology, University Hospital, Schmelzbergstrasse 12, 8091 Zü rich, Switzerland.
ʈ To whom correspondence should be addressed: Dept. of Chemistry and Biochemistry, University of Bern, Freiestrasse 3, 3012 Bern, Switzerland. Tel.: 41-31-6314339; Fax: 41-31-6314887; E-mail: kempf@ibc.unibe.ch. fragment encoding the "E2" protein (432 amino acids, no signal sequence), and the resulting plasmids were transformed into XL-1 blue. Plasmids derived from growing colonies were analyzed for correct orientation of the insert. The correctly oriented plasmid was designated as pMSEH1_E1/23_E2/432. The region encoding for the "E1/"E2 fusion protein (1385 bp; 467 amino acids) was cut out with XbaI and HindIII and subcloned into a pET11c vector. This newly constructed plasmid pET11c_E1/23_E2/432 encoded for a fusion protein composed of the 23 N-terminal amino acids of E1, 12 amino acids resulting from the cloning strategy, and the C-terminal 432 amino acids of the E2 protein.
In another set of experiments bacteria harboring the pET11c-E1 plasmid were grown to an A 600 of ϳ0.1 before the pH was adjusted to either 7.4, 6.4, or 5.0, respectively, and the cells induced by addition of 50 M IPTG. Bacterial growth was recorded by measuring the A 600 .
Expression of E1 was tested in bacteria that had been induced with IPTG for 3.5 h.
Purification of E. coli Membrane-Cells were pelleted and resuspended in 0.10 of the original volume of buffer A (50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 1 mM EDTA pH 8, 1 mM Dithiothreitol), disrupted with a French press, and centrifuged at 12,000 ϫ g for 5 min. The resulting pellet was washed twice with buffer A and resuspended in 400 l of buffer B (10 mM Tris-HCl, pH 9.3, 1 mM ␤-mercaptoethanol). The supernatant of the 12,000 ϫ g centrifugation was centrifuged at 300,000 ϫ g for 50 min and the pellet resuspended in 1 ml of buffer B, homogenized, and centrifuged at 300Ј000 g for 60 min. Finally, the pellet containing the membranes was resuspended in 400 l of buffer B.
Localization of the E1 Protein-To demonstrate that the E1 protein was inserted into the E. coli plasma membrane, different control experiments were performed. E. coli expressing the E1 protein were mixed with E. coli harboring a plasmid pTSGH11 encoding the enzyme IICB of the glucose transporter (23). Then the cell membranes of this E. coli mixture were isolated and the proteins analyzed by SDS-PAGE and subsequent immunoblotting using antibodies against E1 and enzyme IICB, respectively. Alternatively the enzyme IICB was co-expressed in pET11cE1 harboring cells using the plasmid pAGB421 (24). Cells containing both plasmids were selected based on the kanamycin (pAGB421) and ampicillin resistance (pET11cE1). The presence of both plasmids in the growing colonies was verified by plasmid isolation and subsequent restriction analysis. Membranes of E. coli containing both plasmids were isolated and analyzed for the presence of the two proteins as described above.
Treatment of Purified Membranes with 8 M Urea-Urea treatment has been used to distinguish between peripheral and integral membrane proteins (25,26) and to solubilize inclusion bodies. Hence, to determine the localization of the E1 protein, urea was added to the purified membrane fraction and to the cell debris pellet to a final concentration of 8 M. The samples were incubated at room temperature for 45 min and then centrifuged for 18 min at 300,000 ϫ g. The pellets were washed with 8 M urea, centrifuged as above, and resuspended in 200 l of buffer A. The resulting fractions were dialyzed against 1% SDS, electrophoretically separated, and immunoblotted as described above.
Electrophoretic Analysis of Proteins-The proteins of the purified membranes and of the pellet consisting of cell debris, respectively, were separated by SDS-PAGE (10 or 20%) under reducing conditions. Expression of E1 and enzyme IICB was detected by Western blot analysis using a polyclonal rabbit anti-SFV and rat anti-enzyme IICB antibody, respectively, followed by horseradish peroxidase-conjugated anti-rabbit IgG (Bio-Rad AG, Glostrup, Switzerland) and anti-rat IgG (Dako, Glattbrugg, Denmark). Visualization occurred by either using a BM chemiluminescence kit (Roche Diagnostics Ltd., Rotkrenz, Switzerland) or diaminobenzidine as substrate for the horseradish peroxidase.
Preparation of Spheroplasts-E. coli BL21DE3 containing the pET11cE1 plasmid were grown in M9 medium. Expression was induced by addition of 1 mM IPTG. After 2-3 h of induction cells were harvested, and spheroplasts were prepared as described by Osborn et al. (27). Conversion of bacteria to spheroplasts was monitored by phase contrast microscopy and was generally between 95 and 99%.
Labeling of Spheroplasts with Antibodies and FACS Analysis-Spheroplasts were fixed with 1% paraformaldehyde immediately after isolation, washed, incubated with rabbit anti-SFV antibodies, washed, subsequently incubated with fluorescein isothiocyanate-labeled anti-rabbit IgG, and finally subjected to FACS analysis. FACS analysis was performed on a Becton Dickinson flow cytometer. To test the integrity of the spheroplasts FACS analysis was performed in the presence of propidium iodide. For controls anti-SFV antibodies were replaced with a preimmune serum.
Proteinase K Treatment-E. coli cells expressing the pET11cE1 plasmid were converted into spheroplasts as described above. The spheroplasts were washed and resuspended in phosphate-buffered saline supplemented with 20 mM glucose and subsequently incubated with 0.1 mg/ml proteinase K for 20 min on ice either in the presence or absence 0.4% Triton X-100. Control samples consisted of spheroplasts kept under the same conditions but in the absence of proteinase K.
Choline Flux Experiment and pH Optimum-To test changes in membrane permeability, cells were loaded with [ 14 C]choline, and the released radioactivity was measured. Cells were grown and induced as described above, loaded with [ 14 C]choline chloride (Amersham Pharmacia Biotech, Dü bendorf, Switzerland) at a concentration of 2 Ci/ml and incubated for 45 min. Cells were then pelleted, washed three times, and resuspended in M9 medium (pH 7.5 or 5.85). Aliquots of 1 ml were taken at different time points and cells separated from the supernatant either by centrifugation through a silicon oil cushion or by filtration through a 0.2-m Dyna guard filter (Spectrum Laboratories Inc.). Aliquots form supernatants were mixed with a phenylxylylethane-based scintillation fluid (Beckman AG, Zü rich, Switzerland), and radioactivity was measured in a Kontron MR 300 ␤-counter.
Cells were grown, preloaded, and washed as described above. Cells were then resuspended in M9 medium at different pH values and choline efflux measured after 10-min incubation as described previously.

RESULTS
Expression of SFV E1 Protein in E. coli Cells-Comparison of the growth of noninduced bacteria harboring either the pET11c or pET11c-E1 at various pH (5.4 -7.2) showed no significant difference. However, induction with 50 M IPTG strongly hampered the growth of bacteria containing the pET11c plasmid, but not bacteria containing the pET11c-E1. Under mildly acidic condition (pH 5) proliferation of induced cells containing the pET11c-E1 was strongly impeded, whereas at pH 6.4 and 7.4, respectively, growth appeared to be normal (data not shown). These results would be in agreement with a functional expression of the E1 protein in the E. coli membrane.
Expression of E1 was detected by analyzing whole cell lysates by SDS-PAGE and subsequent immunoblotting using a polyclonal rabbit anti-E1 antibody (data not shown). Separation of the bacterial plasma membrane from other cell components followed by Western blot analysis revealed that E1 was not solely localized in the membrane but also associated with cell wall fragments or other protein aggregates (Fig. 1A). To further determine whether the E1 protein found in the cell membrane fraction is indeed a membrane-bound protein and does not represent a contamination by inclusion bodies, we used 8 M urea (25,26) to treat the fraction containing the bacterial membrane as well as the "cell debris" pellet. SDS-PAGE and subsequent Western blot analysis of the different fractions showed that the E1 protein remained associated with urea-treated membranes (Fig. 1B, lane MiUp), whereas the corresponding supernatant fraction revealed no signal for E1 (lane MiUu). Treatment of the cell debris pellet with 8 M urea demonstrated that a significant amount of E1 protein was most probably localized intracellularly in inclusion bodies or was associated with cell wall components (Fig. 1B, lane PiUu). These results were further strengthened by demonstrating the co-localization of the E1 protein and the enzyme IICB in isolated membranes obtained from either a mixture of E. coli expressing E1 and enzyme IICB, respectively, or membranes isolated from E. coli that expressed both proteins simultaneously (Fig. 1, C and D).
Incorporation of the E1 protein into the plasma membrane of E. coli was further assessed by FACS analysis of spheroplasts generated from E. coli containing the pET11c-E1 plasmid that were labeled with rabbit anti-SFV and fluorescein isothiocyanate-anti-rabbit antibodies. FACS analysis in the presence of propidium iodide showed that the membranes of the spheroplasts were intact and non leaky (Fig. 2, A2 and B2). Hence, the positive signal depicted in Fig. 2, B1 proves that the E1 protein faces the extracellular space of the spheroplasts. To further analyze the orientation of the protein in the membrane, spheroplasts were exposed either in the presence or absence of Triton X-100 to proteinase K and the remaining membrane-associated protein fragments analyzed by SDS-PAGE and Western blot using anti-SFV antibodies. The results (Fig. 3) clearly show that a residual fragment of 3 kDa (lanes 2 and 3) was unaffected by the digestion. The size of the fragment and its reactivity with antibodies strongly suggest that it is the anchoring region (25 amino acids) of the E1 protein. Hence, the E1 protein was correctly oriented in the membrane.
Modification of the E. coli Membrane Permeability at Mildly Acidic pH-It has been shown previously that under acidic conditions, virus spike proteins can alter the host cell mem-brane permeability by pore formation (11). Several lines of evidence lead to the assumption that the SFV E1 protein is responsible for this process (14,20). To test this hypothesis we have performed efflux experiments, as described under "Materials and Methods," by using E. coli containing either the pET11c-E1, or as controls the pET11c_E1/23_E2/432 or pET11c Lane 1, membranes isolated from E. coli harboring the pET11cE1 plasmid. Lane 2, membranes isolated from a mixture of E. coli expressing the E1 protein and enzyme IICB, respectively. Lane 3, membranes isolated from E. coli co-expressing E1 and enzyme IICB. Lane 4, membranes isolated from E. coli expressing enzyme IICB. D, immunoblot using anti-enzyme IICB as first antibody. Lanes 1-4 correspond to the lanes in Fig. 3C). Gray scale pictures were produced from the original immunoblots using a Hewlett-ackard Scan Jet 4c and the corresponding software Desk Scan II.

FIG. 2. FACS analysis of rabbit anti-SFV-labeled spheroplasts
containing the pET11c-E1 plasmid. Spheroplasts that had been fixed with paraformaldehyde as described under "Materials and Methods" were incubated with anti-SFV (B) or as a control with preimmune rabbit serum (A) followed by fluorescein isothiocyanate goat anti-rabbit antibody. Prior to measurement on the flow cytometer, the DNA stain propidium iodide was added to assess viability of the spheroplasts. As shown in the histogram, ϳ20% of the spheroplasts gated in a forward scatter/90°side scatter plot (not shown) were stained with the anti-SFV antibody (B versus control A). This staining was surface-located: no DNA staining was obtained with propidium iodide as can be seen from the dot plots 2B and 2A which display simultaneously propidium iodide staining (FL-2) and anti-SFV staining (FL-1). As a positive control for propidium iodide staining, spheroplasts permeabilized by mild detergent treatment were used. These fixed and leaky spheroplasts were readily stainable with propidium iodide (not shown). plasmids, respectively. At neutral pH, the [ 14 C]choline efflux of E. coli cells expressing the E1 protein remained unchanged compared with controls. As depicted in Fig. 4, lowering the pH (pH ϭ 5.85) resulted in an increase in choline release within the first 10 min in E. coli expressing E1. In contrast E. coli exposed to mildly acidic pH and harboring either the fusion protein E1/E2 or the pET11c plasmid only showed no or only a marginal increase in choline release compared with pH 7.5.
To test the pH influence on choline efflux, preloaded cells expressing the E1 protein were resuspended in medium ranging from pH 4 to 7. After 10 min of incubation, choline release was measured. As shown in Fig. 5, choline efflux in E1-expressing cells was strongly dependent on the pH of the extracellular medium, starting at a pH Ͻ 6.2 and reaching a maximum efflux rate at a pH of ϳ5.2.

DISCUSSION
The entry of many enveloped animal viruses into cells is mediated by conformational changes of the viral envelope proteins. These changes are triggered by binding of the virion to the receptor and/or by low pH, e.g. within the endosome, leading to fusion of the viral with the endosomal membrane. This membrane fusion is essential for a successful infection. In the case of the Semliki Forest virus this membrane fusion in the acidic milieu of the endosome is catalyzed by the envelope spike proteins (15). Previous findings indicate that these spike proteins may be responsible not only for membrane fusion but also for pore formation across the viral envelope (13,14) and the membrane of infected insect cells (10,11) under slightly acidic conditions.
It was postulated that this pore formation plays a crucial role in the penetration process of SFV (15), but the question which of the viral structural proteins plays the key role in this process remained open.
One possible candidate is the small structural membrane protein 6K, which is present only in small amounts in the viral membrane. Upon expression in E. coli 6K was capable of increasing the membrane permeability leading to cell lysis (22). Similar results have been demonstrated for poliovirus protein 3A (28). However, a deletion mutant of SFV lacking the 6K protein showed unaltered behavior with respect to low pH-induced pore formation in infected eukaryotic cells (20), although there was a reduction in virus release (21). This may indicate that the 6K protein plays a role in the budding process rather than in pore formation, as suggested by Loewy et al. (29).
It has been demonstrated previously that low pH-induced pore formation is dependent on the ectodomain of the viral spike (13). Furthermore, other findings using various mutant viruses strongly suggest that the E1 protein plays a crucial role in this process (20). Experiments showing that pore formation also takes place in the so-called E1 particles, where the E2 ectodomain has been removed by proteolysis (14) supported this notion. However, since E1 particles still contain the transmembrane part of the E2 protein, an involvement of E2 could not be entirely excluded.
Independent expression of E1 protein on the cell surface of vertebrate cells has not yet been achieved (the protein is produced, but not transported to the plasma membrane) (18). Thus, it has not been possible to prove that the E1 protein per se is sufficient for the formation of pores across the membrane.
In this study we have therefore expressed the E1 protein in E. coli in an inducible manner using the pET11c expression system (30) and showed that these E1 proteins are indeed integrated into the plasma membrane of E. coli cells in an identical orientation as in SFV or SFV-infected cells. Fig. 3 clearly shows that in spheroplasts exposed to proteinase K the E1 protein is digested. An E1-derived peptide of ϳ3 kDa is protected from the proteinase digestion, i.e. is localized inside the periplasmic membrane. The size of this peptide is in agreement with the expected size of the anchoring region (2.9 kDa) and strongly supports our findings that the E1, in the E. coli cytoplasmic membrane, is correctly oriented.
To investigate whether pH-dependent pore formation occurs, [ 14 C]choline release assays were performed. An enhanced choline efflux at pH 6 -6.2 and below was found. It was maximal at a pH around 5.2, which is a pH similar to the one prevailing in endosomes. Lanzrein et al. (10) have reported corresponding results using SFV-infected insect cells: upon lowering the extracellular pH, efflux of a radiolabeled tracer molecule started at a pH of ϳ6.2 and reached a maximal level at a pH of ϳ5.5. Furthermore, these data are in accordance with what is known about the pH dependence of the SFV membrane fusion reaction (5) that in turn is dependent on the conformational change of the spike proteins.
The fact that exposure of E. coli expressing the E1 protein to mildly acidic pH results in a change of the membrane permeability, which shows the characteristics of previously described acid-induced pore formation by the virus proteins, further strengthens the notion that the E1 protein is indeed incorporated into the cell membrane. The formation of pores at a pH below 6.2 also explains the observations that the growth of bacteria containing the pET11c-E1 plasmid was strongly hampered at pH 5, but not at pH 6.4 or 7.4, respectively.
In conclusion, the data presented demonstrate that the E1 per se is sufficient to form acid-induced pores. Thus, these results confirm the previously proposed hypothesis (15).
However, these data also raise the interesting question how the E1 protein gets inserted into the E. coli cell membrane. In a regular infection of eukaryotic cells the viral spike proteins are synthesized as a polyprotein, which is cleaved co-translationally into the single proteins, and the signal for the insertion of the E1 protein into the membrane is contained within the 6K protein that precedes E1 on the polyprotein (6). Within the membrane the E1 protein has a type I orientation (C terminus inside, N terminus outside). The membrane anchor sequence is located within the C-terminal 25 amino acids with just two arginines on the inside. With the selected strategy of cloning the E1 protein coding sequence into the pET11c vector, all of the 6K sequence was omitted and replaced by a start codon.
Analysis of the E1 sequence (439 amino acids) for both prokaryotic and eukaryotic signal sequences using the PSORT program (31,32) predicted the protein to reside in the cytoplasm. Hence, the cloned DNA containing the sequence encoding the E1 protein lacks a known signal that would govern the insertion of the protein into the cell membrane. To identify the sequences within the E1 protein responsible for protein insertion into the membrane, further experiments are needed.