Viral and cellular membranes each consists of two leaflets and in principal membrane fusion could occur by two different pathways as discussed by Chernomordik 
. The direct fusion model posits that pores form in each of the apposing membranes and the pore rims join forming a fusion pore that allows lipid and content mixing in a single step. In contrast, the 2-step model posits fusion of the outer leaflets of the apposing membranes to form a hemifusion intermediate followed by merging of the inner leaflets to form the fusion pore. In the latter model, lipid mixing and content mixing occur sequentially. Evidence to support the second model involving a hemifusion intermediate has been obtained for several different viruses by demonstrating membrane lipid mixing without content mixing by mutation of viral fusion proteins, slowing or interrupting fusion with inhibitors and decreasing the surface density of viral fusion proteins 
. In the present study of VACV, we showed that membrane lipid mixing could occur without core entry under three circumstances: depletion of certain EFC proteins (A28, L1 or L5), neutralization of VACV with a MAb to the L1 EFC-associated protein, and partial cholesterol depletion of the cell membrane. These findings are consistent with a 2-step entry model with a hemifusion intermediate for VACV.
In the first part of the Results, we described the effects of inhibitors of cell processes on virion attachment, membrane fusion and core entry. Most of the inhibitors had previously been shown to reduce entry as determined by reporter gene expression or detection of cytoplasmic cores 
. We found that none of these inhibitors prevented binding of virions to cells, many reduced membrane fusion, while others only acted at the core entry step (). The membrane fusion inhibitors were either directly involved with actin polymerization or remodeling (CK-636, CK-548, latrunculin A, cytochalasin D) or blocked tyrosine kinases that can modulate actin cytoskeletal changes (genestein, Iressa, 324674). The action of dynasore, a specific inhibitor of dynamin GTPase, could be due to its known effect on actin since there is evidence against a role for caveolae-mediated endocytosis in VACV entry 
. Further evidence for dynamin2 in VACV core entry has been obtained with siRNA 
. Extensive actin remodeling and mobilization has been observed during MV binding to cell surfaces 
suggesting that actin-enriched membrane protrusions increase the intimacy of membrane contact and promote virus-cell membrane fusion. Actin remodeling has been suggested to facilitate fusion by forcing membranes together and enlarging pores in a variety of systems 
including virus entry and viral protein-induced cell-cell fusion 
. With human immunodeficiency virus, actin remodeling appears to have a more important role in pore expansion and content mixing than in hemifusion 
. We found that cytochalasin D and latrunculin A had a greater inhibitory effect on core entry (determined by LUC expression) than membrane fusion as determined by lipid mixing, suggesting that actin dynamics may be required for multiple steps in VACV entry.
Summary of effects of inhibitors on VACV entry.
In contrast to the role of actin rearrangement, inhibitors that prevented membrane blebbing involved in virus surfing and macropinocytosis or that interfered with the reduction in pH of endosomes, had a much greater effect on core entry than membrane lipid mixing (). It will be important to determine whether lipid mixing is occurring at the plasma membrane or in endosomes at neutral pH. Similarly, a 74% reduction of cellular cholesterol with mßCD had little effect on membrane fusion but had a major effect on core entry as measured by LUC expression. A previous study had shown that MVs associate with cholesterol-rich regions of the plasma membrane and that cholesterol depletion reduced VACV entry as measured by visualizing cores in the cytoplasm 
. In studies with influenza virus and Semliki Forest virus in insect cells, which can be more stringently depleted of cholesterol than mammalian cells, both hemifusion and pore widening were affected 
. The cell surface receptors for certain viruses reside in cholesterol-rich lipid rafts, but receptors for VACV have not been identified.
The VACV EFC proteins were previously shown to be required for virus core entry and cell-cell fusion but evidence for a role in the fusion of viral and cell membranes had been indirect. Of the ten EFC or EFC-associated mutants tested in the present study, all were blocked in core entry as determined by infectivity or transmission electron microscopy and seven of these were unable to mediate membrane fusion. The three proteins apparently not required for membrane fusion were A28, L1, and L5. It is possible that these proteins have a specific role at a later step in entry such as pore formation. However, in other systems it has been shown that the density of activated fusion proteins has to be higher for the formation and expansion of a fusion pore than for hemifusion 
. Although these three mutants each display stringent repression of EFC protein expression as shown by Western blotting, undetectable differences could affect the sensitive lipid-mixing assay. Therefore, our main conclusion is that the EFC is required for membrane fusion and that additional studies are required to conclude that A28, L1 and L5 have a specific role at a later step of entry such as pore formation.
The L1 protein is a target of potent neutralizing and protective antibodies 
. The structure of L1 alone and in association with a conformation-specific MAb has been solved to high resolution 
. The Fab fragment binds to a discontinuous epitope containing two loops that are held together by a disulfide bond. Here we showed that the MAb prevents VACV entry at a step beyond lipid mixing, consistent with the effect on entry of virions deficient in the L1 protein.
Since our inhibitor studies had shown that actin dynamics are required for membrane fusion and core entry, we considered the possibility that the EFC indirectly promotes entry by inducing cell signaling. Indeed, such a role could contribute to the need for multiple EFC proteins. Since Mercer and Helenius 
had shown that cell signaling requires few virus particles, we tried to rescue EFC protein-deficient virions in trans by coinfecting with wild type VACV. Although wild type virus enhanced core entry by four-fold as measured by LUC expression, this value was still only 3% of that achieved by the control virus, suggesting that the EFC proteins have a direct role in membrane fusion and entry. Nevertheless, whether EFC protein interactions also cause signaling is an interesting question for future studies.
Why so many different proteins are needed for poxvirus entry remains an enigma. None of the proteins resemble type I or type II viral fusion proteins by sequence so that determination of the 3-dimensional structure of the VACV EFC may be needed to define putative fusion loops, if the mechanism of entry involves such structures. At this time, only the structure of the L1 EFC-associated protein has been solved