Analysis of the anti-apoptotic activity of four vaccinia virus proteins demonstrates that B13 is the most potent inhibitor in isolation and during viral infection

Vaccinia virus (VACV) is a large dsDNA virus encoding ~200 proteins, several of which inhibit apoptosis. Here, a comparative study of anti-apoptotic proteins N1, F1, B13 and Golgi anti-apoptotic protein (GAAP) in isolation and during viral infection is presented. VACVs strains engineered to lack each gene separately still blocked apoptosis to some degree because of functional redundancy provided by the other anti-apoptotic proteins. To overcome this redundancy, we inserted each gene separately into a VACV strain (vv811) that lacked all these anti-apoptotic proteins and that induced apoptosis efficiently during infection. Each protein was also expressed in cells using lentivirus vectors. In isolation, each VACV protein showed anti-apoptotic activity in response to specific stimuli, as measured by immunoblotting for cleaved poly(ADP ribose) polymerase-1 and caspase-3 activation. Of the proteins tested, B13 was the most potent inhibitor, blocking both intrinsic and extrinsic stimuli, whilst the activity of the other proteins was largely restricted to inhibition of intrinsic stimuli. In addition, B13 and F1 were effective blockers of apoptosis induced by vv811 infection. Finally, whilst differences in induction of apoptosis were barely detectable during infection with VACV strain Western Reserve compared with derivative viruses lacking individual anti-apoptotic genes, several of these proteins reduced activation of caspase-3 during infection by vv811 strains expressing these proteins. These results illustrated that vv811 was a useful tool to determine the role of VACV proteins during infection and that whilst all of these proteins have some anti-apoptotic activity, B13 was the most potent.


INTRODUCTION
Cell death is an essential biological process for development, cellular homeostasis and immune regulation, and can also restrict virus replication. Cell death occurs most commonly by apoptosis -an irreversible cascade of proteolytic events induced by the extrinsic or intrinsic activation of caspase proteases, particularly caspase-3 (Tait & Green, 2010). Extrinsic, or death receptor-mediated apoptosis, is initiated by Fas ligand or TNF, and induces the dimerization and activation of pro-caspase-8, which activates the effector caspases. Intrinsic, or mitochondrial apoptosis, is triggered by stimuli such as cell-cycle dysregulation, DNA damage or pathogen sensing, and causes mitochondrial outer membrane permeabilization (MOMP) and cytochrome c release, which, together with the apoptotic protease activating factor (APAF)-1 and caspase-9, form the apoptosome complex (Tait & Green, 2010). A tight regulation preceding MOMP occurs through a complex protein-protein interaction network involving the B-cell lymphoma (Bcl) family of proteins, which contains anti-apoptotic members, such as Bcl-2 and Bcl-x L , and proapoptotic members, such as Bax and Bak, and the BH3-only proteins Bid, Bad and Bim. The homo-oligomerization of the former is a pivotal step that triggers MOMP.
Apoptosis leads to morphological changes to cells and production of immunosuppressive cytokines (Griffith & Ferguson, 2011). In contrast, cell death deriving from caspase-1 activation and IL-1b production, termed pyroptosis, triggers inflammation. A third form of cell death called necroptosis occurs without caspase activity, but via the association of receptor interacting protein (RIP)-1 and -3 (Holler et al., 2000), and is the default route of cell death when the apoptotic signal cannot proceed canonically due to inhibition of caspase activity (Han et al., 2011). All forms of programmed cell death may be induced upon pathogen infection as part of innate immunity and consequently pathogens have evolved strategies to prevent their activation.
Vaccinia virus (VACV) is a member of the genus Orthopoxvirus. The centre of the VACV genome encodes proteins required for replication, whilst the terminal regions encode scores of proteins aiding immune evasion (Gubser et al., 2004;Smith et al., 2013), including apoptosis (for reviews, see Shisler & Moss, 2001;Taylor & Barry, 2006). Infection by VACV strain Western Reserve (WR) does not trigger apoptosis in many cultured cells and apoptotic pathways remained blocked (Lee & Esteban, 1994;Dobbelstein & Shenk, 1996;Kettle et al., 1997;Wasilenko et al., 2001). However, VACV strain Copenhagen (COP) infection induced apoptosis more readily (Heinkelein et al., 1996). In addition, mice infected with VACV activated necroptotic signalling to control infection in response to viral inhibition of caspase-3 (Cho et al., 2009). This indicated that VACV counteracts apoptosis efficiently during infection both in vitro and in animals.
Identification of additional VACV anti-apoptotic proteins by knocking out candidate genes from VACV strains such as WR is difficult because the effects are masked by the other anti-apoptotic proteins. An alternative approach is ectopic expression of proteins outside of viral infection and addition of pro-apoptotic stimuli to trigger cell death. However, whether these conditions mimic those under which viral proteins have evolved is unlikely. VACV strain vv811 lacks 55 genes from the parental strain COP, has a small plaque phenotype and replicates poorly in cell culture (Perkus et al., 1991). Genes missing include B13R, F1L and N1L, as well as GAAP (which is not present in COP), and consequently vv811 is unable to protect cells from apoptosis induced by pro-apoptotic stimulus (Wasilenko et al., 2003). Here, the potency of four different VACV anti-apoptotic proteins in isolation and in the context of viral infection was assessed using recombinant vv811 viruses expressing B13, F1, N1 and GAAP. Apoptosis was induced using extrinsic and intrinsic stimuli, and also by vv811 infection that alone caused caspase-3 activation. Results showed that although all proteins prevented apoptosis to some extent in different environments, B13 provided the greatest protection both in isolation and during viral infection.

Generation of U2-OS cells constitutively expressing VACV anti-apoptotic proteins
To study the anti-apoptotic capacity of individual VACV proteins, bi-cistronic lentivirus vectors were used to generate polyclonal U2-OS cell lines expressing GFP-and FLAG-tagged B13, F1, GAAP and N1 (Methods). An empty vector (EV) cell line expressing GFP only and a cell line expressing Bcl-x L were also generated as controls. Expression of B13, GAAP, F1 and Bcl-x L was verified by FLAG immunoblotting (Fig. 1a). Given its previously described degradation by the proteasome when expressed in isolation (Postigo et al., 2006), F1 was only detected after pharmacological inhibition of the proteasome with MG132 (Fig. 1b). Analysis of these cells by immunofluorescence using an anti-FLAG antibody demonstrated the expression and cellular localization of these proteins (Fig. 1c). The location of each protein was as described previously, except that N1 showed a greater nuclear localization than noted elsewhere (Bartlett et al., 2002;Maluquer de Motes et al., 2011), possibly due to higher expression.

Inhibition of extrinsic apoptosis in U2-OS cells expressing VACV anti-apoptotic proteins
To assess anti-apoptotic activity, extrinsic apoptosis was induced using cycloheximide (CHX) and TNF-a. This induced cell death in EV-transduced cells, as seen by morphological changes and fewer adherent cells remaining, and this was prevented in cells expressing B13 or Bcl-x L (Figs 2a and S1, available in the online Supplementary Material). Immunoblotting for cleaved poly(ADP ribose) polymerase (PARP)-1 confirmed induction of apoptosis in cells expressing F1, N1 and GAAP, and its inhibition by B13 or Bcl-x L (Fig. 2b, upper), and this was quantified in three independent experiments (Methods) (Fig. 2b, lower). A second analysis was obtained using the quantitative Caspase-Glo 3/7 assay. TNF-a treatment caused~8-fold increase in caspase-3 activity in EV-transduced cells, and this was reduced significantly only in cells expressing B13 and Bcl-x L (Fig. 2c). These results indicated that out of these four VACV proteins, only B13 blocked TNF-ainduced apoptosis in U2-OS cells, consistent with its inhibition of caspase-8 cleavage.

Inhibition of intrinsic apoptosis in U2-OS cells expressing VACV anti-apoptotic proteins
Next, the resistance of these cell lines to intrinsic apoptosis induced by staurosporine (STS) was investigated. This showed that B13 caused substantial reduction in cleavage of PARP-1 compared with EV-transduced cells, whereas cells expressing GAAP and N1 showed a milder effect (Fig.  3a). Quantitative analysis demonstrated significant reduction by all these VACV proteins except F1 (Fig. 3a). Using the Caspase-Glo 3/7 assay, STS treatment induced around fivefold increase in caspase-3/7 activity in EV-transduced cells and this was reduced by each VACV protein, particularly B13 (Fig. 3b). Likewise, around fivefold increase in active caspase-3 was observed in EV-transduced cell lines upon treatment with doxorubicin (DOX), a DNA-damaging agent causing mitochondrial apoptosis, and this was   3c). Taken together, each VACV anti-apoptotic protein was sufficient to reduce or inhibit drug-induced intrinsic apoptosis in U2-OS cells.

Induction of apoptosis using vv811
VACV strain vv811 lacks 55 genes, including all known VACV anti-apoptotic proteins except E3 (Fig. 4a), and so we hypothesized that vv811 infection might trigger apoptosis without additional stimuli. To examine this, the level of caspase-3/7 activation in U2-OS cells infected with vv811 or WR was measured. VACV strain vv811 infection induced caspase-3/7 activation from 2 h postinfection (p.i.) and this increased over 24 h (Fig. 4b), whereas apoptosis was not induced by VACV strains WR ( Fig. 4b) or COP (data not shown). The use of vv811 as an inducer of apoptosis provided a more physiological context to assess apoptosis without the need for exogenous drug treatments. (a) Cells were incubated overnight with CHX and TNF-a (10 mg ml "1 ) for 16 h and then photographed (see Fig. S1 for control). Bar, 200 mm. (b) Cells were incubated overnight with CHX and TNF-a (10 mg ml "1 ) for 16 h, and lysates were analysed by SDS-PAGE and immunoblotting against cleaved PARP-1 or a-tubulin (Tub). A representative blot of three repeats is shown. Quantitative data were obtained by integrating the intensity of the bands from the three repeats using a LI-COR system. Data were normalized to a-tubulin and are shown as the mean±SD.
(c) Cells were incubated overnight with either CHX (30 mg ml "1 ) or CHX and TNF-a (10 mg ml "1 ) for 16 h, and apoptosis was assessed quantitatively using Caspase-Glo. Data are shown as the mean±SD fold induction relative to EV and are representative of at least three experiments each performed in triplicate. Statistical differences between EV and anti-apoptotic protein-transduced cells were determined using an unpaired Student's t-test (*P,0.05, **P,0.01).
Next, the ability of the individual VACV anti-apoptotic proteins to inhibit apoptosis induced by vv811 and to enhance the poor growth of vv811 was examined. First, U2-OS cell lines expressing VACV proteins were infected for 24 h and the levels of apoptosis assessed by measuring caspase-3/7 activity using Caspase-Glo (Fig. 4c). F1 and B13 reduced vv811-induced apoptosis, whereas GAAP and N1 did not. Secondly, we produced multi-step growth curves of vv811 in the different U2-OS cell lines. Expression in trans of VACV anti-apoptotic proteins did not enhance viral growth (Fig. 4d). Taken together, these data indicated growth of vv811 was not enhanced in U2-OS cell lines expressing VACV proteins irrespective of their ability to prevent vv811-induced apoptosis.

Generation of recombinant vv811 expressing VACV anti-apoptotic proteins
To address whether expression of VACV anti-apoptotic proteins in cis could provide stronger effects, we generated recombinant vv811 expressing B13, F1, N1 or GAAP. To allow us to compare the efficiency of each protein independently of its expression level, each gene was cloned downstream of the VACV early/late promoter P7.5 (Mackett et al., 1984) and inserted by transient dominant selection into vv811 within non-essential genes encoding either the ribonucleotide reductase large subunit (I4L gene) (Tengelsen et al., 1988) or the thymidylate kinase (A48R gene) (Smith et al., 1989a) as described in Methods. The genomes of the derivative vv811 viruses were shown by PCR using primers annealed to the flanking regions to contain the inserted genes as expected (Fig. 5a). Expression of the inserted genes was confirmed by immunoblotting with specific antibody for B13 (Fig. 5b), F1 (Fig. 5c) and N1 (Fig. 5d), and by immunofluorescence for GAAP (Fig.  5e). For B13, F1 and N1, expression was detected as early as 2-4 h p.i., in agreement with the expression from the 7.5K promoter (Mackett et al., 1984), and was not detected in the vv811 parental strain. Equal infection of cells was confirmed by blotting for the VACV early protein A49 (Mansur et al., 2013). The localization of GAAP in the Golgi was shown by co-staining with GM130, a marker of the Golgi complex (Fig. 5e). The recombinant viruses were characterized further by measuring plaque sizes in U2-OS  Comparison of VACV anti-apoptotic proteins cells after 96 h of infection ( Fig. 5f) and by assessing viral replication in U2-OS cells infected for 24 h at 2.5 p.f.u. per cell (Fig. 5g). No statistically significant differences amongst the four recombinant viruses, and between these and the parental vv811, were observed in either of the assays. Taken together, these data indicated that the expression of these proteins in cis did not enhance viral replication or spread in the conditions tested.

Induction of apoptosis by the recombinant vv811 viruses
The ability of these anti-apoptotic proteins to inhibit apoptosis during VACV infection was investigated using either VACV strains that lack each gene individually or vv811 viruses engineered to express each gene individually. Viruses lacking the B13R (Kettle et al., 1995) and N1L (Bartlett et al., 2002) genes from the WR strain had been constructed previously, whereas a virus lacking the GAAP gene was engineered in the Evans strain, one of the few VACV strains to express GAAP (Gubser et al., 2007), and a F1L gene deletion virus was produced in the same WR strain background as we had used previously (see Methods). Deletion of the F1L gene from WR was confirmed by PCR using primers that annealed to the flanking regions of the F1L gene and by immunoblotting with anti-F1 antiserum (Fig. S2).
To determine the role of VACV anti-apoptotic proteins in both WT VACV strains and vv811, U2-OS cells were infected at 2.5 p.f.u. per cell and the cell viability was measured using the Cell Titre-Blue assay at 24 h p.i.. Infection with vv811 reduced viability by 50 %, and this was ameliorated by infection with vv811-B13, vv811-F1 and vv811-GAAP, but not vv811-N1 (Fig. 6a). Similar results were obtained in HeLa cells, where viability after vv811 infection reached only 25 % and could only be increased by B13, F1 or GAAP expression (Fig. 6b). Next, caspase activity was measured. U2-OS cells were infected for 24 h with parental WR and its individual deletion derivatives WRDB13, WRDF1 and WRDN1, with parental Evans and its derivative EvansDGAAP, and with the recombinant vv811 viruses (Fig. 6c). The level of infection was confirmed by immunoblotting for VACV protein D8. As expected, neither VACV strain WR nor Evans induced apoptosis in this system. With the exception of WRDB13, which induced significantly more caspase-3/7 activation than WR, deletion of each individual protein in each parental VACV strain did not change the level of caspase activity, presumably due to the presence of the other anti-apoptotic proteins. Conversely, after vv811 infection the anti-apoptotic effects of B13, F1 and GAAP were readily detectable, with each virus inducing significantly less apoptosis than parental vv811. As observed in assays where the proteins were expressed in trans, B13 was the most potent inhibitor. vv811-N1 did not, however, show any ability to block viral induction of caspase activation during infection. Furthermore, caspase activation after vv811 infection was also measured in HeLa cells (Fig. 6d), a cell line commonly used for apoptosis studies.
Results were similar to those in U2-OS cells with the exception of vv811-GAAP, which did not show any protection in this cell type. Collectively, these data showed that inserting anti-apoptotic proteins individually in vv811 provided a powerful tool to assess their potency and relative contribution to preventing apoptosis in the context of viral infection, and that this system could be used in multiple cell types.

DISCUSSION
The ability of cells to undergo apoptosis is an important host antiviral mechanism that limits the replication and spread of many viruses and, consequently, viruses have evolved countermeasures. VACV encodes several antiapoptotic proteins (see Introduction), but their antiapoptotic activity has not been compared in the same experimental system and not in a context where apoptosis is induced by viral infection alone. In this study, we show that infection by VACV strain vv811, which unlike WT VACV strains lacks B13, B22, F1, N1 and GAAP proteins, induces apoptosis in cells without additional stimuli and so can be used to study the inhibitory activity of individual anti-apoptotic proteins in the context of viral infection. Of the four anti-apoptotic proteins studied (B13, F1, N1 and GAAP), B13 was found to be the most potent inhibitor.
When expressed in isolation, all four proteins were able to block apoptosis mediated by STS or DOX to some degree, whereas only B13 blocked apoptosis mediated by TNF-a (Figs 2 and 3). This is consistent with F1 and N1 targeting cellular Bcl-2 proteins involved in mitochondrial apoptotic pathways (Wasilenko et al., 2003(Wasilenko et al., , 2005Postigo et al., 2006;Cooray et al., 2007;Kvansakul et al., 2008;Maluquer de Motes et al., 2011), B13 being a pan-inhibitor of caspases (Ray et al., 1992;Dobbelstein & Shenk, 1996;Kettle et al., 1997;Garcia-Calvo et al., 1998), and GAAP inhibiting intrinsic apoptosis (Gubser et al., 2007). Although GAAP was reported to have some activity against extrinsic apoptosis (Gubser et al., 2007), such activity was not found under the conditions tested here and this discrepancy might, in part, be explained by the efficiency of the method used to deliver GAAP into cells, which differed in these studies.
As vv811 infection is sufficient to induce apoptosis, the ability of the VACV proteins to block this was tested by infecting cell lines expressing these proteins and by constructing recombinant vv811 viruses expressing these proteins. After vv811 infection of the transduced cell lines, only B13 and F1 showed an anti-apoptotic effect (Fig. 4), indicating that B13 and F1 are more potent inhibitors than GAAP and N1 in this experimental system, and/or that vv811 infection triggered apoptotic responses in a manner that was not mimicked by the drugs employed in Fig. 3. Use of the recombinant vv811 viruses demonstrated that B13 and F1 had the greatest anti-apoptotic activity in both U2-OS and HeLa cells, highlighting their more potent inhibitory effects compared with GAAP and N1 (Fig. 6).
Interestingly, infection of U2-OS cells with vv811-GAAP increased cell viability similar to vv811-B13 or vv811-F1, despite being unable to reduce caspase-3/7 activation similar to those viruses. Notably, in HeLa cells, viability data were recapitulated despite no reduction of caspase-3/7 activity. In fact, vv811-GAAP conferred a viability phenotype similar to vv811-F1 in U2-OS and similar to vv811-B13 in HeLa cells despite intermediate or null inhibition of caspase activity. This was confirmed by a very weak cytopathogenic effect after infection compared with other vv811 viruses (data not shown). Taken together, these data suggest that GAAP may contribute to cell viability by multiple mechanisms in addition to inhibition of caspase activity, and this may be more prominent in specific cell types. Indeed, the very closely related hGAAP has also been linked to cell adhesion and migration (Saraiva et al., 2013a). In contrast, B13 proved to be the most potent anti-apoptotic protein under all the conditions tested. However, despite this drastic inhibition of apoptosis, vv811-B13 infection only increased cell viability mildly. Taken together, this indicates that caspase-3/7 inhibition is not sufficient to determine cell viability during infection and other factors are required.
Although the anti-apoptotic effect of N1 was observed in isolation upon drug treatment (Fig. 3), this was not mirrored in the context of viral infection (Fig. 6). N1 is a Bcl-2-like protein that in addition to modulating apoptosis, inhibits activation of the pro-inflammatory transcription factor NFkB (DiPerna et al., 2004;Cooray et al., 2007;Chen et al., 2008;Graham et al., 2008;Maluquer de Motes et al., 2011). Recently, a structurally informed mutagenesis study demonstrated that it is the anti-NFkB activity of N1 and not its anti-apoptotic ability that contributes to virulence in vivo (Maluquer de Motes et al., 2011). Therefore, although N1 has a relatively weak anti-apoptotic activity, its main function following infection is the inhibition of inflammatory signalling. Interestingly, a similar situation exists for F1 that contributes to virulence by targeting NLRP-1 and thereby blocking inflammasome activation (Gerlic et al., 2013). Although F1 has greater anti-apoptotic activity than N1, virulence in both cases is driven by inhibition of innate immune signalling rather than apoptosis.
VACV is a large virus that alters the cellular environment by multiple mechanisms. Here, we demonstrate that unlike its parental full-length strain COP, infection with vv811 causes caspase-3/7 activation in cells. Thus, VACV infection induces apoptosis and this is prevented by antiapoptotic proteins. Exactly how VACV infection is sensed to trigger apoptosis requires clarification. VACV strain COP lacks GAAP (Goebel et al., 1990;Gubser & Smith, 2002) and has a non-functional B13R gene (Kettle et al., 1995), indicating that these genes are non-essential for blocking apoptosis during COP infection. In contrast,  proteins F1 and N1 are highly conserved amongst VACV strains and other orthopoxviruses, and each interferes with mitochondrial apoptosis, suggesting that this apoptotic route may be prominent during viral infection. Data presented here show, however, that only B13 reduced caspase-3/7 activation during vv811 infection to levels observed with VACV strains WR or COP. Given that the anti-apoptotic activities of N1 and F1 do not contribute to virulence, it is likely that caspase activation occurs in vivo primarily via non-mitochondrial pathways including Fas ligand or TNF signalling, and this is blocked by B13. Although a few VACV strains express soluble TNF receptors (Alcamí et al., 1999), intracellular inhibitors of such extrinsic apoptosis other than B13 have not been described in VACV, but are present in other large DNA viruses, such as human cytomegalovirus (Skaletskaya et al., 2001).
In summary, this study shows that vv811 infection induces apoptosis, and so is a useful tool to identify the cellular factors that trigger recognition and apoptosis upon VACV infection. Such information will be useful in understanding VACV-host interactions, but also for developing VACV as a vaccine vector and oncolytic agent.

METHODS
Plasmids, cells and viruses. The F1L and B13R genes were amplified by PCR from VACV strain WR, and cloned into pcDNA4/ TO fused with an N-terminal tandem affinity purification (TAP) tag consisting of a streptavidin-binding sequence and FLAG epitope (Gloeckner et al., 2007). GAAP was amplified by PCR from VACV strain Evans and cloned as the C-terminal TAP fusion. N1-TAP was described previously (Maluquer de Motes et al., 2011). All plasmids were verified by DNA sequencing. Finally, tagged alleles, as well as Bcl-x L , were subcloned into the first cistron of a lentivirus vector also expressing GFP (Saraiva et al., 2013b). BSC-1 and U2-OS cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10 % FBS (Biosera), 100 U penicillin ml 21 (Invitrogen) and 100 mg streptomycin ml 21 (Invitrogen). HeLa cells were grown in minimum essential medium (Gibco) supplemented as above plus nonessential amino acids (Sigma). U2-OS cells expressing proteins B13, F1, GAAP, N1 or Bcl-x L , or EV control, were obtained after transduction with lentiviruses expressing GFP and the protein of interest, and sorting in a MoFlo MLS high-speed cell sorter (Beckman Coulter) to obtain .95 % GFP-positive cells. Lentivirus particles were generated after transient transfection of HEK 293T cells with the bicistronic genomic vector together with entry and packaging vectors using the calcium chloride method. Parental vv811 was provided by Michelle Barry (University of Alberta, Canada). VACV strain WR lacking either B13R or N1L and VACV strain Evans lacking GAAP have been described (Kettle et al., 1995;Bartlett et al., 2002;Gubser et al., 2007).
Construction of mutant viruses. For the construction of vv811-B13, vv811-F1 and vv811-N1, plasmids containing the B13R, F1L or N1L gene flanked by~250 bp regions of the A48R gene encoding thymidylate kinase (Smith et al., 1989a;Hughes et al., 1991) and containing the Escherichia coli guanylphosphoribosyltransferase gene (Ecogpt) fused in-frame with the EGFP gene were used (Ember et al., 2012). For the construction of vv811-GAAP, a similar plasmid was used containing the GAAP gene inserted between the flanking regions of the I4L gene encoding the ribonucleotide reductase large subunit (Tengelsen et al., 1988). VACV strain WR lacking the F1L gene (WRDF1) was constructed using a similar plasmid containing the 250 bp flanking region of the F1L gene. The mutant viruses were constructed using the transient dominant selection method (Falkner & Moss, 1990) by selection of EGFP-positive plaques in the presence of mycophenolic acid, hypoxanthine and xanthine, as described previously (Unterholzner et al., 2011). The genotype of the resolved viruses was analysed by PCR following proteinase K treatment of infected BSC-1 cells using primers that annealed within the flanking regions of each gene.
Immunofluorescence. Transduced U2-OS cells were seeded into sixwell plates containing sterile glass coverslips. After two washes with icecold PBS, cells were fixed in 4 % (v/v) paraformaldehyde, quenched in 150 mM ammonium chloride, permeabilized with 0.1 % (v/v) Triton X-100 in PBS and blocked for 30 min in 5 % (w/v) FBS in PBS. The cells were stained with rabbit anti-FLAG antibody (Sigma) for 1 h, followed by incubation with goat anti-rabbit Alexa Fluor 546 secondary antibody (Invitrogen). Coverslips were mounted in MOWIOL 4-88 (Calbiochem) containing DAPI. Images were taken on a Zeiss LSM780 confocal microscope using Zen2011 acquisition software. Following vv811-GAAP infection, U2-OS cells were stained using rabbit anti-GAAP polyclonal antiserum (Gubser et al., 2007) and mouse anti-GM130 antibody (BD Biosciences) to check GAAP expression and localization.
Plaque assays and growth curves. U2-OS cells were infected with vv811 and derivative viruses at 50 p.f.u. per well of a six-well plate for 5 days. The cells were washed once with PBS and stained for 1 h with crystal violet [5 % (v/v) crystal violet solution (Sigma), 25 % (v/v) ethanol]. The sizes of 20 plaques per well were measured using Axiovision acquisition software and a Zeiss AxioVert.A1 inverted microscope as described previously (Doceul et al., 2010;Ferguson et al., 2013). To measure virus growth, U2-OS cell lines were infected with vv811 at 0.1 p.f.u. per cell, and samples harvested at 0, 24 and 48 h. U2-OS cells were infected with recombinant viruses at 2.5 p.f.u. per cell for 24 h. Virus titres were determined as described by Chen et al. (2006), but using U2-OS cells.
Statistical analysis. Data were analysed using an unpaired Student's t-test, with Welch's correction where appropriate.