Fusion Induced by a Class II Viral Fusion Protein, Semliki Forest Virus E1, Is Dependent on the Voltage of the Target Cell

Abstract

Cells expressing the low pH-triggered class II viral fusion protein E1 of Semliki Forest virus (SFV) were fused to target cells. Fusion was monitored by electrical capacitance and aqueous dye measurements. Electrical voltage-clamp measurements showed that SFV E1-induced cell-cell fusion occurred quickly after acidification for a trans-negative potential across the target membrane (i.e., negative potential inside the target cell) but that a trans-positive potential eliminated all fusion. Use of an ionophore to control potentials for a large population of cells confirmed the dependence of fusion on voltage polarity. In contrast, fusion induced by the class I fusion proteins of human immunodeficiency virus, avian sarcoma leukosis virus, and influenza virus was independent of the voltage polarity across the target cell. Initial pore size and pore growth were also independent of voltage polarity for the class I proteins. An intermediate of SFV E1-induced fusion was created by transient acidification at low temperature. Membranes were hemifused at this intermediate state, and raising the temperature at neutral pH allowed full fusion to occur. Capacitance measurements showed that maintaining a trans-positive potential definitely blocked fusion at steps following the creation of the hemifusion intermediate and may have inhibited fusion at prior steps. It is proposed that the trans-negative voltage across the endosomal membrane facilitates fusion after low-pH-induced conformational changes of SFV E1 have occurred.

Two major classes of viral fusion proteins have been identified to date (9). A class I viral protein is characterized, in its final, postfusion state, by the presence of a trimer of hairpins based on a central triple-stranded α-helical coiled coil in which each of the three grooves is packed by C-terminal hairpin segments (43). This structure is often referred to as a six-helix bundle. Once the final hairpin structures have formed, the fusion peptides are near the transmembrane domains. The structures of two class I proteins have also been crystallographically determined with respect to their state prior to activation of fusion. For one protein (influenza virus hemagglutinin [HA]), this initial state does not exhibit the bundle (47), whereas for the other (simian virus 5 fusion protein) a six-helix bundle is already present (48), but this structure is not the final bundle. In both cases, in transiting from their initial to their final state the proteins undergo changes in secondary structure that cause parts of the protein, notably fusion peptides, to move long distances (1, 5). The two virus genera that are known to express class II proteins are the alphaviruses and flaviviruses. Their fusion proteins all exhibit common three-dimensional structures rich in β-strands in their initial states (27, 36, 39) and final states (3, 18, 37). Bundles of α-helices do not form. The conformational changes necessary to transit from the crystallographically determined initial state to the final state do not involve substantial changes in secondary structure. Instead, the domains of class II proteins rotate at “pivot points” so that large-scale movements bring fusion loops and transmembrane domains into proximity, as is the case for class I protein.

Semliki Forest virus (SFV) is a prototypic alphavirus. The SFV membrane contains heterodimers of two transmembrane glycoproteins, the receptor-binding E2 protein and the fusion protein E1. The structure of the E1 ectodomain has been determined by crystallography in its initial conformation (27, 40) and its post-low-pH conformation (18). SFV E1 has been extensively characterized: its assembly, interactions with membranes during fusion, and requirement for low pH have all been well studied (20, 23, 42). The receptor necessary for the uptake of SFV into an endosome has not been identified, but the fusion event itself can occur with any target (Tg) membrane that contains large amounts of cholesterol and small amounts of sphingolipids. These lipids are also necessary for fusion to proceed, but it is not known whether their presence is simply required or whether they actually trigger some conformational changes that lead to fusion.

For all class II proteins, low pH induces fusion, and low pH has been assumed to be the only trigger—but this has never been explicitly shown. Over 20 years ago it was suggested that voltage across the Tg membrane could be involved in SFV E1-induced fusion: it was found that when an extracellular high-Na⁺ solution, as is found in the normal state, was replaced by a low-Na⁺, high-K⁺ solution, infection by SFV was inhibited (21). Because the membrane potential should be small whenever extracellular Na⁺ is replaced by K⁺, it appeared that this depolarizing effect inhibited viral fusion within the endosome. But SFV showed efficient fusion with liposomes in the absence of a membrane potential or ion gradient (21). The idea that voltage could play a direct role in fusion of SFV was thus, for all intents and purposes, abandoned. In the course of studies investigating the dependence of fusion upon lipids, however, we noticed that during the fusion of cells expressing SFV E1 to voltage-clamped planar bilayer membranes, the bilayer quickly ruptured after lowering pH at trans-positive potentials, a polarity that corresponds to positive potentials inside target cells. In contrast, for trans-negative potentials, as would exist across the endosomal membrane, fusion proceeded efficiently (41). This difference suggested that the interactions of the fusion protein with the Tg planar bilayer were dependent on the polarity of voltage across the bilayer. These observations led to our present, definitive investigation.

In this study, we used rigorous electrophysiological measurements to directly determine whether fusion between cells expressing SFV E1/E2 and Tg cells is under the control of the voltage of the Tg membrane. We found that fusion readily occurs for trans-negative voltages but is completely blocked by a trans-positive potential across the Tg membrane. Further, we were able to apply electrophysiological measurements in the cell-cell fusion system to identify steps in the fusion process that utilize voltage. The findings of this study open up the question of voltage dependence as a more generalized mechanism for class II viral fusion.

MATERIALS AND METHODS

Reagents.

The fluorescent dyes calcein acetoxymethyl ester (calcein-AM) and CMAC (7-amino-4-chloromethylcoumarin) were obtained from Molecular Probes, Inc. (Eugene, OR). Chlorpromazine (CPZ), bovine serum albumin (BSA), and poly-l-lysine (∼95 kDa) were purchased from Sigma Chemical Co. (St. Louis, MO). Cosmic calf serum was purchased from HyClone (Carlsbad, CA).

Cells and transfection.

HEK 293T cells (2), as well as cells stably expressing influenza virus HA (HAb2) (29) or human immunodeficiency virus (HIV) Env (TF228.1.16) (35) or CD4 and CXCR4 (HelaT4⁺) (35), were maintained as previously described. Cells (Tva950) stably expressing avian sarcoma leukosis virus (ASLV) Env (33) or the receptor (Tva) of ASLV Env (33) were also maintained as previously described. To express the SFV fusion protein, HEK293T cells were transfected with the pCB3-wt vector by use of a standard calcium phosphate transfection method (41). This vector encodes the SFV structural proteins E1, p62 (precursor of E2), 6K, and capsid; transfection produces transient cell surface expression of the E1 fusion protein and the mature E2 protein.

Fluorescent labeling and following fusion by dye transfer.

Tf228.1.16 cells were grown in suspension. All other cell lines were adhered to culture dishes, and in all cases except one, cells were removed from the dishes with an EDTA/EGTA divalent-free phosphate-buffered saline (PBS) solution. The exception occurred when HAb2 cells were used as effector (Ef) cells. Here, the cells adhered to culture dishes were bathed in divalent-containing PBS; 10 μg/ml trypsin was added at room temperature in order to cleave HA0 into HA1-HA2 subunits (29). After 10 min, most of the cells had detached from the dish; these cells were diluted into Dulbecco's modified Eagle's medium (DMEM) containing 10% Cosmic calf serum to terminate the activity of trypsin. After two washes in divalent-containing PBS, the HAb2 cells were incubated with 0.2 mg/ml neuraminidase for 10 min at room temperature and the enzyme was then washed out.

To characterize SFV E1-induced fusion, HEK 293T cells transfected with SFV E1/E2 were used as Ef cells. They were removed from the culture dish 48 h after transfection, and their cytosols were labeled with (1.3 μM) calcein-AM (green emission at 515 nm). HAb2 cells were used as Tg cells in this case, and they were loaded with 30 μM aqueous dye CMAC (blue emission at 466 nm). For influenza virus HA-mediated fusion, HAb2 cells were the Ef cells and HEK 293T cells were the Tg cells. ASLV Env-expressing Ef cells and TVA-expressing Tg cells were similarly fluorescently labeled and used as cell pairs. For HIV Env-mediated fusion, Tf228.1.16 cells were the Ef cells and HelaT4⁺ cells were the Tg cells. For all cell pairs, the Ef cells were loaded with calcein-AM and the Tg cells were labeled with CMAC as described previously (35).

For fluorescent dye mixing experiments, equal numbers (10⁵ cells/ml) of the Ef and Tg cells were mixed in microcentrifuge tubes containing DMEM supplemented with 0.1% BSA and then transferred to eight-well slides that were precoated with 1 mg/ml poly-l-lysine and allowed to settle for 10 min. For SFV E1/E2-, ASLV Env-, or influenza virus HA-induced fusion, the cells were maintained for 45 min at 37°C and neutral pH before fusion was induced by changing the solution within the wells to a low pH solution: pH 5.7 for SFV E1/E2, pH 5.4 for ASLV Env, and pH 4.8 for influenza virus HA. After 5 min, the solution was again changed to neutral pH. After 10 min at 37°C and pH 7.2, dye spread was assessed by fluorescence microscopy (35). For HIV Env-mediated fusion, bound cells were maintained for 2 h at 37°C before dye spread was measured.

Electrical measurements.

For electrophysiological measurements, Ef and Tg cells mixed in microcentrifuge tubes were added at room temperature to a 35-mm-diameter culture dish containing a DMEM-BSA solution. A no. 1.5 coverslip coated with poly-l-lysine had been placed at the bottom of the culture dish, and the cells were allowed 30 min to attach to the coverslip. The coverslip was then placed on the bottom of an experimental chamber that was maintained at 4 to 6°C. Bound Ef-Tg cell pairs were microscopically identified by the calcein fluorescence of the Ef cell and the CMAC florescence of the Tg cell. An Ef-Tg cell pair was patch clamped, and the whole-cell voltage-clamped mode was established. A 200 Hz sine wave voltage was superimposed upon voltages of either +40 mV or −40 mV and applied across the Tg cell membrane. Measurements of time resolved admittance (i.e., “capacitance”) were used to determine the conductance of a fusion pore starting from its formation through its growth (35). Capacitance measurements allow the conductance of the fusion pore to be calculated because this pore connects the area of the Ef cell membrane to that of the voltage-clamped Tg cell. The components of the admittance (i.e., current) that are in phase (Y₀) and 90° out of phase (Y₉₀) with respect to the applied sine wave voltage as well as the DC conductance (Y_DC) that results from the application of ±40 mV are measured. The solution inside the patch pipette consisted of 135 mM cesium glutamate-5 mM MgCl₂-5 mM BAPTA [1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetate]-10 mM HEPES (pH 7.2). The external solution was 135 mM N-methyl-glucamine aspartate-5 mM MgCl₂-2 mM HEPES (pH 7.2). Average conductances over time were determined by aligning pores at their initial opening and calculating mean pore conductance every 20 ms. The initial pore conductance was set equal to the first conductance jump observed for that pore.

For SFV E1/E2-induced fusion, after a whole-cell mode was achieved, temperature was increased to 32°C and cell-cell fusion was triggered by local application for 30 s of a pH 5.7 solution from another glass pipette placed near the cell pair. The same procedure was used for ASLV Env-induced fusion except that the temperature was raised to 37°C and a pH 5.4 solution was then used for acidification. For influenza virus HA, the temperature was raised to 37°C followed by acidification to pH 4.8. For HIV Env-induced fusion, Ef and Tg cells were maintained at room temperature for 2.5 h before the whole-cell mode was established. Fusion was triggered by raising the temperature to 37°C. In all cases of raising the temperature, a temperature-jump procedure employing an infrared diode laser was used, as previously described (28).

Use of ionophores.

The sodium ionophore SQI-Pr (Teflabs, Austin, TX) was used to generate trans-positive potentials across cells. This ionophore fluoresces somewhat in the blue color range, so we labeled the Ef 293T cells with the green aqueous dye calcein-AM and the Tg HAb2 cells with the red aqueous dye CMTMR (Molecular Probes). For cell fusion experiments, cell pairs were allowed to bind for 30 min at room temperature onto the bottom of eight-well slides in a solution consisting of 135 mM NaMeSO₃, 5 mM MgSO₄, and 2 mM HEPES (pH 7.4). Fusion was induced by replacing the solution with 135 mM NaMeSO₃-5 mM MgSO₄-20 mM MES (pH 5.7) for 5 min at 4°C. A 10 μM concentration of SQI-Pr was always present for experiments in which a trans-positive potential was desired. For each experiment, conditions were duplicated in two wells of the slide and the presence or absence of spreading of aqueous dye was assessed for ∼100 cell pairs/well.

Creating CAS.

A cold-arrested stage (CAS) of fusion for SFV E1/E2 was created by incubating Ef and Tg cells for 45 min at room temperature and acidifying to pH 5.7 for 10 min at 4°C followed by reneutralization at 4°C for 10 min. As determined by aqueous dye spread assessments, fusion does not occur at this CAS intermediate. To determine whether “restricted hemifusion” (6) had occurred by this point, 0.5 mM chlorpromazine was added for 2 min at 4°C to rupture any hemifusion diaphragms (34). The extent of rupture was quantified by measuring the percentage of Ef-Tg cell pairs that exhibited aqueous dye spread.

RESULTS

SFV E1-induced fusion requires a trans-negative potential.

A high percentage of the individual Ef HEK 293T cells expressing SFV E1 and the Tg cells in contact must fuse in order for electrophysiological techniques to be practical, because in voltage-clamp experiments a cell pair must be chosen before triggering fusion, and one does not know beforehand which pairs will fuse. We tested HEK 293T and HAb2 cells as Tg cells by means that allowed us to monitor a large population of cells to determine which revealed a high degree of fusion. Syncytia formed for each Tg cell after a transient 3 min, low-pH treatment and overnight incubation. Monitoring a large number of cell pairs for fusion by fluorescence microscopy, we found that aqueous dye had mixed for a high percentage of the cells within a short time after the pH was reduced only when HAb2 cells were used as the target. For these Tg cells, the dye-mixing experiments showed that 45 to 65% of the cell pairs fused with a time course of a half-life of approximately 30 to 60 s after the pH was reduced. The higher fusion efficiency when HAb2 cells were used as the target compared to HEK 293T cell results was presumably due to its more intimate contact with the Ef cells (49): HAb2 cells constitutively express the uncleaved form of influenza virus HA (HA0). HA0 binds to terminal sialates on proteins and lipids on Ef cell membranes, and this promotes adhesion between an Ef cell and an HAb2 cell.

We determined the voltage requirement for fusion of the Tg cell by placing it in whole-cell voltage-clamp mode at 4°C and then raising the temperature to 32°C. We used capacitance measurements to monitor the formation and growth of fusion pores and alternated, in successive experiments, the holding potential of the Tg cell between −40 mV and +40 mV. For electrophysiological experiments, cells were exposed to pH 5.7 for 30 s (see Materials and Methods). At −40 mV across the Tg cell membrane, a pore formed and enlarged in 20 out of 30 experiments after the pH was reduced (Fig. 1). This is consistent with the percentage of cell pairs that exhibited aqueous dye spread. The monotonic increases in Y₉₀ and biphasic Y₀ are precisely the characteristics of fusion pore formation and growth. The total conductance (G_P) between the cells rapidly increased (upper trace). G_P often increased in steps, indicating that several pores formed in succession between active cell pairs. Often, Y_DC did not increase despite the rapid increase in total pore conductance and probable formation of multiple pores. In fact, in some experiments Y_DC did not increase even after the G_P value became quite large. Thus, leaks during fusion for the Ef cell or the Tg cell were not obligatory (12). Fusion was fast after the pH was reduced to 5.7. When a pore did form, it opened within 30 s of acidification (Fig. 2); more than half the pores formed within 10 s of lowering the pH to 5.7.

FIG. 1.

Raw electrical traces of fusion pore formation and growth mediated by SFV E1. The solution surrounding a selected cell pair was lowered to pH 5.7 via local application of an acidified solution. (A) The pore conductance, G_P, calculated from Y₀ and Y₉₀ is shown. (B) Fusion occurred within 5 s of acidification, as evidenced by the increase in Y₉₀ and Y₀. The mean value of Y_DC did not increase, showing that membrane leaks did not accompany pore formation or growth. The orientations of trans-positive and trans-negative potentials are illustrated in the upper-left-hand portion of the figure.

FIG. 2.

Kinetics of fusion mediated by SFV E1. The times after acidification (time = 0) were rank ordered for all experiments for which electrically measured fusion occurred after acidification for a constantly maintained −40 mV.

In contrast to the high percentage of fusion seen at −40 mV, at +40 mV no pore was ever observed (n = 29) after the pH was reduced. Membranes could become leaky while +40 mV was maintained, but any leaks were usually small and did not interfere with our ability to detect fusion when that had occurred. For these trans-positive voltage experiments, after 3 min of low pH, voltage was shifted to −40 mV to test whether a pore would now form; it did so in only 1 of the 29 experiments. We assume that SFV E1 became inactive over the 3-min low-pH period in which trans-positive potential prevented fusion. The HAb2 target cells express the uncleaved influenza virus HA0, but not the HA cleaved into subunits (10), and only the cleaved form induces fusion. We verified through our functional assays that HA on the surface of the HAb2 cells was not the cause of fusion: in control experiments using mock-transfected 293T cells bound to HAb2 cells, fusion did not occur for Tg cells voltage clamped at either −40 or +40 mV; similarly, when aqueous dye spread measurements were performed, we determined that fusion did not occur for the mock-transfected cells. Thus, HA0 was not the cause of the low-pH-induced fusion observed for the SFV E1-transfected Ef cells. Our experiments thus unambiguously showed that SFV E1 fusion occurs when there is a trans-negative, but not a trans-positive, potential across the Tg membrane.

The only potential drawbacks of the voltage-clamp method are (i) that in practice, only a relatively small number of cell pairs can be tested and (ii) that as a result of rupturing the cell membrane under the pipette to create the whole cell mode, cytosolic components transfer from the cell to the pipette, and this, however unlikely, could cause the voltage dependence. We therefore used a nonelectrophysiological method to test for voltage dependence. In principle, a cell bathed in a high-Na⁺ solution that contains a sodium ionophore should have a trans-positive potential across its plasma membrane. We initially used electrophysiological measurements of isolated cells to explicitly check this and found that under our conditions, the average potential was +34.6 ± 0.8 mV (mean ± standard error of the mean; n = 7) in the presence of the ionophore SQI-Pr. In the absence of ionophore the average potential was −28.9 ± 0.9 mV (n = 7). Then, instead of electrophysiology, we used the sodium ionophore to create trans-positive potentials for fusion experiments. Fusion was triggered by acidifying at 4°C for 5 min, raising the temperature to 37°C for 10 min, and determining the percentage of cell pairs that exhibited aqueous dye spread. In the absence of ionophore (i.e., for a trans-negative potential), fusion was extensive (Fig. 3, first column). For a trans-positive potential (i.e., in the presence of ionophore), fusion was greatly depressed (second column). In these experiments, over 1,000 cell pairs were assayed for each potential. Thus, this additional system also demonstrates that a trans-positive potential inhibits fusion.

FIG. 3.

Fusion mediated by SFV E1 at positive and negative voltages created by using a sodium ionophore. Fusion proceeded well at the trans-negative potentials (∼−30 mV) of the control high-Na⁺ extracellular solution (first column, n = 8). Fusion was substantially inhibited by trans-positive potentials (∼+35 mV) produced by the ionophore SQI-Pr (second column, n = 8). Error bars represent standard deviations.

Class I protein-induced fusion is independent of voltage.

We performed similar electrophysiological experiments to evaluate fusion of Ef cells expressing one of three different class I viral fusion proteins: influenza virus HA cleaved into HA1-HA2 subunits, ASLV Env, and HIV Env (Fig. 4). Specific receptors are required for ASLV Env- and HIV Env-induced fusion, but it is not practical to express reproducible levels of a membrane protein in HAb2 cells by transfection. Thus, we used stable cell lines that express appropriate receptors as target cells for the cognate viral fusion protein. For ASLV Env and influenza virus HA, fusion was induced by low pH by the same protocol as used for SFV E1 except that the temperature was raised to 37°C after the whole-cell electrical configuration was achieved at 4°C. The low-pH value was chosen to be in accord with the known pH thresholds of the specific fusion protein; fusion was triggered with a pH 5.4 solution for ASLV Env and with a pH 4.8 solution for influenza virus HA. For cleaved HA, fusion occurred in 6 out of 10 experiments conducted with HEK 293T target cells voltage clamped at −40 mV; fusion occurred in 10 out of 17 experiments when the target cells were clamped at +40 mV. For ASLV Env, fusion to Tg cells expressing Tva, the receptor for Env, proceeded rapidly after the pH was reduced in 6 of 9 experiments at −40 mV and in 9 of 13 experiments at +40 mV. These experiments demonstrate an absence of voltage dependence in fusion and show that application of low pH per se was not somehow the cause of the voltage dependence of SFV E1/E2-induced fusion. Ef cells expressing HIV Env fuse, at neutral pH, to Tg cells expressing CD4, the Env receptor, and CXCR4, the coreceptor. Capacitance measurements showed that fusion proceeded in four of six experiments at −40 mV and in five of eight experiments at +40 mV. Thus, it is clear that for all three class I proteins, fusion proceeds independently of the polarity of voltage.

FIG. 4.

Dependence of probability of fusion on the polarity of the membrane potential of the target cell for the fusion proteins of SFV, HIV, ASLV, and influenza virus. Fusion only occurred for a trans-negative voltage for SFV E1 but was independent of voltage polarity for ASLV Env, HIV Env, and influenza virus HA. The voltages were ±40 mV.

The voltage dependence for fusion was not due to unusual properties of the SFV E1-induced pore. The initial pore size was comparable to that induced by ASLV Env and was between those of the larger HIV Env and smaller influenza virus HA pores (Fig. 5). Pore growth for SFV E1 (Fig. 5 inset) was slower than that seen with HIV Env but faster than that seen with influenza virus HA (data not shown). For the class I proteins, the conductances of the initial pores were not dependent on whether a trans-positive or trans-negative potential was maintained across the Tg cell. The unique voltage dependence of SFV E1/E2 compared to the class I proteins is therefore not accounted for by the kinetics of formation, size, or enlargement of the fusion pore itself. In other words, the voltage dependence is due to the response of the fusion protein to the voltage across the target membrane.

FIG. 5.

Initial conductance of pores. Initial pore conductances are shown for SFV E1 for −40 mV across the target membrane. For influenza virus HA, ASLV Env, and HIV Env, initial conductances are shown for −40 mV and +40 mV. These initial conductances were independent of voltage. The numbers (n) of pores for each condition were as follows: for SFV, n = 20; for HA, n = 6 for −40 mV and n = 10 for +40 mV; for ASLV, n = 6 for −40 mV and n = 9 for +40 mV; and for HIV, n = 4 for −40 mV and n = 5 for +40 mV. Error bars represent standard errors of the means. The inset shows the profile of average pore conductance induced by SFV E1 over time.

Voltage-dependent steps occur after hemifusion is achieved.

Fusion proceeds through sequential states in a multistep process, and voltage could influence fusion at any step. The fusion process is dynamic, and each state is fleeting, but there are means to artificially arrest the process at individual states. The consequences of voltage dependence can then be assessed downstream of that arrested state.

We took advantage of the temperature dependence of fusion to determine the appearance of voltage dependence in the cascade of fusion intermediates. Specifically, we lowered pH for 10 min at 4°C followed by a 10 min reneutralization, also at 4°C, generating CAS of fusion (see also reference 49). After achieving CAS, aqueous dye transferred between only a small percentage of the paired cells (Fig. 6A, column 2). In contrast, about half the cell pairs fused when pH was lowered at 4°C for 10 min and reneutralized at 4°C, but instead of waiting 10 min to create CAS, we immediately raised the temperature to 37°C (control, column 1). Raising the temperature to 37°C 10 min after creating CAS led to a high extent of fusion (column 3), but the extent was lower than that seen with the control (column 1). We used CPZ to test whether CAS is a state of hemifusion. CPZ is a membrane-permeable weak base that destabilizes membranes. It partitions more strongly into inner than into outer monolayer leaflets of cell membranes. Because hemifusion diaphragms are composed solely of inner leaflets, CPZ ruptures hemifusion diaphragms at concentrations that do not make cell membranes permeable (34). We found that the addition of CPZ (at 4°C) to cells after achieving CAS led to aqueous dye transfer (column 4), indicating that CAS is a state of hemifusion. The extent of fusion after adding CPZ at CAS was almost as great as that achieved by raising the temperature to 37°C (column 3). The demonstration that a smaller percentage of cells fused after addition off CPZ (column 4) than was the case with the control (column 1) indicates that hemifusion is reversible and that cells can revert to separate membranes over time; this was previously found for influenza virus HA-mediated fusion (26).

FIG. 6.

Use of CAS to test for voltage dependence after SFV E1-mediated hemifusion. (A) Establishment and characterization of CAS. Fusion was quantified by aqueous dye transfer using HAb2 target cells, which have a trans-negative membrane potential. For the control (first column), extracellular pH was maintained at pH 5.7 for 10 min at 4°C and then changed to a pH 7.2 solution at 37°C. To generate CAS, cells were incubated at pH 5.7 for 10 min followed by pH 7 for 10 min, all at 4°C. Under these conditions the amount of fusion was negligible (second column) compared to the control results. Raising the temperature to 37°C in pH 7 buffer for 10 min after creating CAS resulted in fusion (third column), but it was somewhat less than that seen with the control. The addition of CPZ at CAS (4°C) resulted in an extent of fusion (fourth column) that was comparable to fusion when temperature was raised to 37°C. n = 7 for each condition. Error bars represent standard deviations. (B) Test of voltage dependence of fusion before and after CAS. When a trans-negative potential (−40 mV) was maintained across the target membrane both before and after creating CAS, fusion was observed in approximately one-third of the experiments (column 1). Fusion was reduced but was still significant when a trans-positive potential (+40 mV) was maintained while creating CAS, and the polarity of the potential was then switched to negative (column 2). A chi-square test showed that in light of the control results, the null hypothesis that a trans-positive potential prior to CAS does not affect fusion cannot be rejected at a significance level of 10%. In more colloquial terms, there is about a 10% chance that some voltage-dependent steps occur prior to CAS. When CAS was created while a trans-negative potential was maintained and the potential was then switched to positive, fusion never occurred (column 3). Fusion did not occur if the potential was trans-positive before and after creating CAS (column 4). The abscissa shows the voltage polarities before/after CAS.

We then created CAS and used capacitance measurements to determine whether voltage-dependent steps occur prior to hemifusion, subsequent to hemifusion, or both. The Tg cell of a selected pair of Ef and Tg cells (each loaded with an aqueous fluorescent dye) was put into the whole-cell voltage-clamp mode. We generated CAS by ejecting an acidified solution (at 4°C) from a second pipette positioned near the selected cell pair. Several other cell pairs within the cone of the ejected solution were also transiently exposed to low pH and thus also reached CAS. We raised the temperature of the selected cell pair and the adjacent ones to 37°C. When a trans-negative potential (−40 mV) was maintained both before and after creating CAS (i.e., the control), fusion was observed to result in 6 out of 17 experiments (Fig. 6B, column 1). In contrast, when a trans-positive potential (+40 mV) was maintained while CAS was being created and the polarity of the potential was then switched to negative before the temperature was raised, fusion occurred in 4 of 28 experiments (column 2). Thus, fusion, albeit reduced, occurred when a trans-positive potential was maintained across the target membrane during the establishment of CAS.

A chi-square test showed that there is only about a 10% chance that some voltage-dependent steps occur prior to CAS. Not all cell pairs are in precisely the same state at CAS. Cells are heterogeneous, and the density of SFV E1 proteins on the cell surfaces varies. Because of heterogeneities inherent in a cell-cell fusion system, CAS represents a spectrum of states. It is thus possible that the cell pairs that were in the most advanced downstream states of CAS were the ones affected by the pre-CAS trans-positive potential and that those cells that were less advanced were not affected by the potential across the Tg cell. If this were the case, the creation of hemifusion would be independent of voltage. When CAS was created while a trans-negative potential was maintained and the potential was then switched to positive before the temperature was raised, fusion never occurred in any of 20 experiments (column 3). As expected, constant maintenance of a trans-positive potential before and after creation of CAS abolished fusion for all 12 experiments (column 4). Our experiments thus show that voltage-dependent steps definitely occur subsequent to hemifusion, although the possibility that some voltage-dependent steps might occur prior to it cannot be ruled out. Based on studies of E1-mediated red blood cell binding at acid pH (49), the fusion loop of SFV E1 has inserted into the target membrane prior to the point of hemifusion. We can therefore conclude that voltage-dependent steps reside downstream of fusion loop insertion.

In the above-described experiments, we could monitor fusion of cell pairs surrounding the pair that was voltage clamped because aqueous dye had been loaded into all of the cells. This provided internal controls showing that fusion proceeded normally when membrane potential was not manipulated. After CAS was established, a trans-positive potential completely eliminated fusion, but 34% of the surrounding cell pairs exhibited dye spread. Although trans-positive potential before CAS reduced the probability of fusion of the voltage-clamped pair, 36% of the surrounding cell pairs exhibited aqueous dye spread. The same percentage (36%) of adjacent cells exhibited dye spread when a trans-negative potential was maintained before and after creation of CAS. These observations of the surrounding cells demonstrate that conditions were reproducible from experiment to experiment and that fusion was inhibited only for reasons of voltage polarity.

DISCUSSION

We have unambiguously shown, by electrophysiological means, that the class II fusion protein SFV E1 induces cell-cell fusion in a voltage-dependent manner. It remains to be determined whether any other class II proteins—or class II proteins in general—utilize the trans-negative potential across the target membrane, but this is definitely the case for SFV E1. In contrast to SFV E1 findings, the results determined for three different class I proteins were shown to be independent of the voltage polarity of the target cell. Influenza virus HA utilizes only low pH for fusion (31), ASLV Env uses receptor binding at neutral pH followed by low pH (38), and HIV Env employs receptor binding followed by binding to a coreceptor at neutral pH (11). Because the natures of the triggers for these three proteins are quite different, it is likely that class I proteins generally do not utilize the transmembrane potential of the target membrane for fusion. No evidence has previously supported the possibility that voltage is a major influence on viral fusion, although the notion has been occasionally suggested (21). However, there is precedent for tapping trans-negative endosome membrane potentials in infection: many multisubunit bacterial toxins, such as diphtheria toxin (19) and tetanus toxin (14, 22), deliver their lethal subunits by utilizing both the low pH of endosomes and the trans-negative potential across these membranes. It does make a certain logical sense that voltages could be utilized as a mechanism in viral fusion, since they naturally occur across membranes in a manner regulated by small changes in membrane permeabilities. Having rigorously established the voltage dependence of SFV E1 fusion in a cell-cell system, we propose that in addition to the low pH within the interior of an endosome, the trans-negative potential that is naturally present across its membrane is utilized in fusion of the virus. Of course, there are differences between cell-cell and virus-cell fusion systems, including lower densities of SFV E1 in cells than in virus and possibly arrangements of E1 in cells that do not match the icosahedral arrangements on the surface of SFV. Experiments that control voltage and monitor fusion of virus within endosomes will thus be needed to rigorously test our proposal that voltage is utilized in SFV infection.

Fusion has been studied much more widely for class I than for class II proteins. Although class I and class II structures are very different, the two classes share certain mechanistic principles. Significantly, fusion intermediates with lipid configurations similar to those for class I proteins have been extensively demonstrated for the class II protein of the alphavirus Sindbis, a protein very similar to that of SFV E1, and for SFV E1 (49). We have now extended these findings to show that the intermediate of CAS for SFV E1 is similar to that found for influenza virus HA (Fig. 6A). Commonalities should tend to be imposed by physical properties of lipid bilayer membranes that govern the lipid rearrangements of two membranes merging into one.

We have shown that fusion induced by class I viral proteins is independent of voltage and we expect that this is also the case for SNAREs (45). The use of voltage in the fusion process may not, however, be limited to a class II viral fusion protein: it has been reported that for fusion of mitochondria, the naturally occurring potential across the inner membrane is necessary (32), although this is in dispute (8). Since class II viral structures are similar to each other, it is certainly possible that SFV E1 may serve as a paradigm for all class II proteins.

It has been repeatedly shown by a variety of methods that SFV and other viruses with class II fusion proteins efficiently fuse with liposomes (4, 7, 24, 46). Presumably the potentials across the liposomal membranes in these studies were zero. It is questionable whether liposomes as targets could be reliably used to test for voltage dependence of viral fusion: while one could design experiments using ionophores to vary voltage across liposomal membranes, the miniscule leaks that would occur during the fusion process would probably collapse all membrane potentials. Also, zero potential across a liposomal membrane is not equivalent to zero potential across a cell membrane: acidic lipids preferentially reside on the inner monolayer of plasma membranes, so a negative potential on the order of −20 mV occurs across a cell membrane even when the voltage-clamped potential is 0 mV. Consequently, the precise voltages that permit fusion will be somewhat dependent on the experimental system; results showing that SFV fuses to liposomes are not, in any way, counter to our findings. In addition, since our data show that significant hemifusion can proceed independently of voltage (Fig. 6B), assays of voltage dependence should be based on rapid aqueous content, rather than lipid dye, mixing assays.

Any region of SFV E1 that could sense membrane potential would have to reside within the target membrane prior to the action of voltage, and its insertion would presumably be dependent on the presence of low pH. Because it is known that low pH causes the apolar fusion loop (residues 83 to 100, the cd loop) to insert into the target membrane (16), we envision that voltage across the target membrane controls movement of segments of SFV E1 soon after the fusion loop inserts into the target membrane.

Low pH initiates the conformational changes of SFV E1. One thus expects that the driving force of a membrane potential would be utilized for some of the later conformational changes necessary for fusion that would occur within the membrane. Our experiments demonstrating voltage dependence downstream of hemifusion, and possibly before hemifusion (but not before application of low pH), verify this expectation. The requirement for a transmembrane potential across the target membrane suggests that the refolding of SFV E1 to a hairpin as identified by crystallography is not sufficient to induce fusion. Membrane potential affects conformational changes of E1 necessary for fusion subsequent to the low-pH-induced changes; these voltage-dependent changes could in turn affect fusion protein cooperative interactions, contacts between fusion loops, stem packing, or other aspects of the fusion protein machinery. It should be appreciated, however, that even a modest difference in the configuration of a peptide within a membrane can result in significantly different lipid arrangements (25). The presence of particular voltages may thus be responsible for inducing protein movements that, although small, are nevertheless enough to complete the fusion reaction.

It is possible that among the fundamental differences between class I and class II viral proteins are the ways that these very different structures derive energy. There is considerable evidence to support the commonly held theory that class I proteins store enough energy within their own TM subunits to carry out fusion—an external source of energy is not needed. For class I proteins, the SU subunits clamp the TM subunits in place after cleavage. The clamp allows for a “spring-coil” mechanism in which the triggers “loosen” the clamp, allowing the TM subunits to spontaneously reconfigure to their lowest energy state. In addition to the spring-coil release of energy, the spontaneous conformational changes of individual fusion proteins allow several copies to form complexes with each other, another potentially important energy-releasing event (30).

Structural determinations (3, 18, 27, 36, 37, 39) show that class II proteins do not use a clamp or spring-coil mechanism. There are, however, several energy sources available to class II fusion proteins that are not available to class I fusion proteins. Class I fusion proteins are composed of three identical monomers. In contrast, class II proteins are not trimeric in their prefusion, neutral pH form, but in response to low pH their subunits resort to create a homotrimer of fusion subunits (E1 in the case of Semliki Forest virus). This resorting should release energy, and experiments show that the homotrimer is much more stable than the initial fusion protein (15, 44). Energy may also be released as class II proteins reorient from parallel to perpendicular with respect to the viral envelope (23); this source of energy would not be available to class I proteins, because they remain more or less perpendicular to their respective membranes after conformational changes are triggered. Multiple copies of SFV E1 can, in a manner similar to that seen with class I proteins, associate with each other (17), providing yet another possible source of energy for fusion. Our discovery that a trans-negative voltage is needed across the target membrane for SFV E1-induced fusion suggests that the trans-negative potential across the endosomal membrane (13) is an external energy source. If this is generally the case for class II viral proteins, then they derive a portion of the energy necessary for fusion from the cells they infect. If true, this would mean that the cell membrane plays a fundamentally passive role in class I fusion but would actively be involved in supplying energy in class II fusion.