ERK MAP kinase-activated Arf6 trafficking directs coxsackievirus type B3 into an unproductive compartment during virus host-cell entry

Coxsackievirus type B3 (CVB3) infects polarized intestinal epithelium via caveolae in order to traverse the tight junction (Coyne & Bergelson, 2006) and subsequently gain access to the host circulation from the gut. Once in the circulation, CVB3 can infect a myriad of cell types, including those in the gut, liver, pancreas and heart. Non-polarized cells can be infected, and virus entry into these cells may occur via clathrin-coated pits and the early endosome network (Chung et al., 2005).

Decay-accelerating factor (DAF), the CVB3 co-receptor, has been shown to traffic into high-lipid raft compartments called glycosylphosphatidylinositol (GPI)-anchored protein-enriched early endosomal compartments (Sabharanjak et al., 2002), and coxsackie–adenovirus receptor (CAR) has been shown to reside with low-density lipoprotein receptor in a separate lipid-raft compartment (Ashbourne Excoffon et al., 2003). This discordance of trafficking and compartmentalization is consistent with the GPI-linked nature of DAF and the fact that CAR is a transmembrane protein. Recently, ADP-ribosylation factor 6 (Arf6)-dependent endocytosis has been shown to traffic GPI-linked proteins (Karacsonyi et al., 2007), such as the CVB3 co-receptor DAF. Arf6-mediated endocytosis is a lipid-raft and non-raft endocytosis pathway that cycles at the plasma membrane and is involved in the recycling of major histocompatibility complex class I to and from the cell surface. Activation of Arf6 can also lead to membrane ruffling, cell spreading, cell migration and turnover of adherens junctions (reviewed by Donaldson, 2003). We investigated Arf6, as DAF has been shown to traffic via this endocytosis pathway (Karacsonyi et al., 2007; Naslavsky et al., 2004) and we wanted to know what effect(s) alterations in Arf6 activity had on CVB3 entry and infectivity.

Caveolae and clathrin-coated pits are common pathways for the entry of viruses (Sieczkarski & Whittaker, 2002). However, other endocytic pathways exist and, indeed, may internalize some virus simultaneously into dead-end or non-productive virus entry pathways, as has been described previously (Marchant et al., 2005). We investigated Arf6-mediated endocytosis as a candidate for the diversion of CVB3, during entry, into a non-productive pathway. Recently, Mercer & Helenius (2008) reported that Arf6 can inhibit the entry of vaccinia virus into host cells.

The ability of endocytosis to direct viruses into non-productive pathways has been described for human immunodeficiency virus (HIV) types 1 and 2 (Marchant et al., 2005). This pathway was described as a pH-independent lipid-raft endocytic process, but no particular pathway was identified. Here, we used CVB3 infection of HeLa cells as a model to demonstrate that extracellular signal-regulated kinase (ERK)-mediated signalling can activate Arf6-dependent endocytosis and cause restriction of virus entry.

Viruses and cells.
CVB3 Kandolf strain (CVB3-Kan) and a molecular clone of CVB3 Woodruff (pH3; GenBank accession no. U57056[GenBank] ) containing an enhanced green fluorescent protein (eGFP) expression cassette (eGFP–CVB3-Woodruff) described previously (Feuer et al., 2002; Slifka et al., 2001) were used in this study. The eGFP–CVB3-Woodruff virus was provided by Ralph Feuer and J. Lindsay Whitton (The Scripps Institute, La Jolla, CA, USA). eGFP expression from virus molecular clones allows rapid detection of virus replication, and detection of eGFP with a flow cytometer is more objective than manual counting of plaques. Results for infectivity experiments were performed on a series of virus dilutions from an m.o.i. of 0.001 to 1. Imaging experiments were conducted with both the pH3 molecular clone and Kandolf virus to confirm that these observations represented CVB3 replication in general.

HeLa and 293T cells were grown, and all treatments were performed, in Dulbecco's modified Eagle's medium with 10 % fetal bovine calf serum.

Viral vector and pseudotype production.
The HIV vector pCSGW and packaging construct p8.91 have been described previously (Demaison et al., 2002; Zufferey et al., 1997). The vesicular stomatitis virus envelope G protein (VSV-G) construct pMDG has been described previously (Yee et al., 1994). VSV-G-pseudotyped HIV-1 vectors [HIV(VSV)] were produced by three-plasmid transfection of 293T cells (Towers et al., 2000).

Flow cytometry.
Cells infected with 10-fold dilutions of eGFP–CVB3-Woodruff and lentivirus vector pseudotypes (m.o.i. 0.001–1) were harvested with 5 mg trypsin ml^–1/5 mM EDTA and fixed with 3.6 % formaldehyde at the peak of GFP expression, prior to cell lysis. This optimal time frame proved to be between 8 and 10 h post-infection (p.i.) for eGFP–CVB3-Woodruff and at 48 h for HIV(VSV) vector pseudotypes, and 2.0x10⁴ cells were enumerated with a Beckman Coulter EPICS XL-MCS flow cytometer.

Antibodies and inhibitors.
Rabbit anti-haemagglutinin (HA) polyclonal antibody was purchased from Sigma, anti-enterovirus VP1 monoclonal antibody (mAb) from Dako and a rabbit anti-β-catenin antibody from Cell Signaling Technology. Total ERK and phospho-ERK [phospho-p44/42 mitogen-activated protein (MAP) kinase (Thr202/Tyr204) rabbit mAb] antibodies were purchased from Cell Signaling Technology. The ERK inhibitor U0126 (10 µM working concentration) was from Cell Signaling Technology. Anti-DAF and anti-CAR mAbs used in ERK signalling neutralization studies were from Santa Cruz Biotechnology and Upstate Biotechnology, respectively. HA expression by the Arf6 constructs was detected with a rabbit polyclonal antibody and an Alexa Fluor 405-conjugated goat anti-rabbit secondary antibody (Invitrogen). Polyclonal anti-DAF (Santa Cruz) and polyclonal anti-early endosome antigen 1 (EEA1; Calbiochem) were detected with either Alexa Fluor 488 or 594 conjugated to donkey anti-goat (for DAF) or goat anti-rabbit (for EEA1) secondary antibodies.

Receptor neutralization of ERK signalling.
Experiments were performed at 37 °C only. HeLa cells were treated with an anti-DAF or anti-CAR mAb at a 1 : 1000 dilution (200 ng ml^–1) for 30 min prior to and during CVB3 infection at an m.o.i. of 10. A higher concentration (1 : 50; 4 µg ml^–1) of antibody was added to cells, alone, to activate signalling via antibody–receptor binding. Cells were harvested with protein lysis buffer [10 mM HEPES (pH 7.4), 50 mM Na₄P₂O₇, 50 mM NaF, 50 mM NaCl, 5 mM EDTA, 5 mM EGTA, 1 mM Na₃VO₄, 0.5 % Triton X-100, 10 µg leupeptin ml^–1 and 1 mM PMSF] on ice at 15 min p.i. and lysates were probed by Western blotting for phospho-(pThr202/pTyr204) ERK MAP kinase and total ERK.

Plasmid and RNAi transfection.
Plasmids were transfected into 80 % confluent cells using Fugene (Roche) following the manufacturer's instructions. The plasmid transfection efficiency was determined as 91.3±6.2 % by flow cytometry of a GFP-expressing plasmid, from three independent experiments run in parallel with the wild-type (WT) and mutant Arf6 experiments.

Dominant-negative (DN) T27N Arf6, constitutively active (CA) Q67L Arf6 and DN dynamin K44A mutants were provided by Mark Marsh (MRC Laboratory for Molecular Cell Biology and Cell Biology, London, UK). Wild-type Arf6 was constructed from the DN mutant by reversion mutation of the DN asparagine mutation at aa 27 back to a WT threonine using the GeneTailor Site-Directed Mutagenesis System (Invitrogen). The DN-Arf6 and CA-Arf6 mutants tagged with HA have been described previously (Peters et al., 1995), cloned into pSRα (Dyer et al., 2007).

Oligofectamine (Invitrogen) was used for transient RNAi transfection, following the manufacturer's instructions. The Arf6 small interfering (si)RNA oligonucleotides used were as follows: Arf6 oligonucleotide 1, 5'-GGGAGAUGAGGGACGCCAUAAUCCT-3'; Arf6 oligonucleotide 2, 5'-GGAGGCAGCGCACCGAGUUCCCGCG-3'; and a scramble negative oligonucleotide, 5'-CUUCCUCUCUUUCUCUCCCUUGUGA-3'.

Immunofluorescence microscopy.
Immunofluorescence of coxsackievirus has been described previously (Coyne & Bergelson, 2006). Briefly, cells were fixed by addition of 3 : 1 methanol : acetone at room temperature for 2 min. Cells were blocked with 1 % BSA in PBS or blocked and permeabilized with 0.1 % Triton X-100/1 % BSA in PBS for 10 min at room temperature. Primary antibodies were added at a 1 : 100 dilution at room temperature for 2 h. Cellular markers were stained with goat anti-rabbit or donkey anti-goat secondary antibodies conjugated to Alexa Fluor 488 or 594, at a dilution of 1 : 400. The HA tag of the Arf6 constructs was detected with a rabbit polyclonal antibody against HA (Sigma) and stained with a goat anti-rabbit Alexa Fluor 405-conjugated secondary antibody (Invitrogen). For the experiments that required goat (DAF) and rabbit (CA-, DN- and WT-Arf6–HA) antibody detection, the goat anti-rabbit secondary antibody was added after the donkey anti-goat antibody, with an additional washing step, to prevent secondary antibody cross-reactivity. Virus was stained with a goat anti-mouse biotinylated secondary antibody (Vector Technologies) at a dilution of 1 : 400. Streptavidin-conjugated 655 quantum dots (Invitrogen) were added at a dilution of 1 : 1000 for 1 h at room temperature. Each step was followed by three washes in PBS. Alternatively, to improve clarity, virus was stained using Alexa Fluor 488-conjugated secondary antibodies and cellular markers were stained using Alexa Fluor 594- and 405-conjugated secondary antibodies as described above. Samples were mounted under a coverslip in Vectashield containing DAPI (Vector laboratories).

Image acquisition, three-dimensional modelling and rendering.
Immunofluorescence confocal microscopy was conducted using an AOBS Leica Confocal TCS SP2 microscope (63x, NA 1.2). Lambda scans were performed throughout the experiments to confirm the specificity of the Alexa Fluor and quantum dot signal profiles. Image stacks (90–140x512x512 pixels; three-frame mean) were collated using Improvision Volocity software, with minor brightness and contrast adjustment (not greater than 15 % of original values).

Pearson's correlations were obtained using Volocity software. Channels 2 [green (VP1)] and 3 [red (EEA1)] from 90–140x512x512 image stacks were subject to correlation.

Statistical analysis.
SD was used to represent data distribution. Student's t-test was used to determine statistical significance.

Arf6 uptake of CVB3 during entry is detrimental to infection
CVB3 enters HeLa cells via a dynamin-dependent pathway (Chung et al., 2005). To confirm the endocytic uptake of CVB3 during entry, we monitored the entry of a dynamin-dependent VSV-G-pseudotyped retroviral vector, (VSV)HIV, in parallel with CVB3 during WT dynamin and DN dynamin K44A expression. Fig. 1 shows that dynamin K44A significantly decreased CVB3 infectivity from 5.0x10⁵ to 1.8x10⁵ (P=0.05) infectious units (IU) ml^–1. This constituted a 2.8-fold reduction in infectivity. The infectivity of the (VSV)HIV vector, as described previously (Pelkmans et al., 2005; Sun et al., 2005), was correspondingly decreased 3.6-fold (P=0.0004). These results suggested that our results are consistent with what has been shown previously for CVB3 regarding dynamin dependence (Chung et al., 2005).

(11K): Fig. 1. Dynamin-dependent entry of CVB3 into cells. Entry of a dynamin- and pH-dependent VSV-G-protein pseudotyped retroviral vector, (VSV)HIV, was monitored in parallel with eGFP–CVB3-Woodruff during WT and K44A dynamin expression. CVB3 and (VSV)HIV were added to dynamin K44A- and WT-transfected cells and fixed at 8 and 48 h p.i., respectively, and the number of IU ml–1 was enumerated by flow cytometry. Note the statistically significant decrease in eGFP–CVB3-Woodruff in dynamin K44A-transfected cells (*P<0.05; ‡P<0.001). Results are representative of two independent experiments.

Endocytosis can result in the productive entry of viruses into host cells (Sieczkarski & Whittaker, 2002). However, lipid raft-mediated endocytosis pathways have also been implicated in virus entry restriction (Marchant et al., 2005). We focused on Arf6-mediated endocytosis as it has been shown to traffic at the plasma membrane and Arf6-mediated endocytosis is thought to mediate the trafficking of lipid rafts at the cell surface (Balasubramanian et al., 2007). Thus, we expressed DN-Arf6 (T27N), CA-Arf6 (Q67L) and WT-Arf6 in HeLa cells and monitored CVB3 infection (Fig. 2). The control for Arf6 activity confirmed that CA- and DN-Arf6 inhibited the internalization of β-catenin upon exposure to 5 mM EDTA (Fig. 2a) (Palacios et al., 2001).

(52K): Fig. 2. Arf6 activity mediates restriction of CVB3 during host-cell entry. (a) DN-Arf6 and CA-Arf6 inhibit the internalization of β-catenin after treatment with 5 mM EDTA. HeLa cells were treated with 5 mM EDTA for 30 min prior to fixation and staining. Images were captured using a Leica SP2 confocal microscope. Bar, 10 µm. (b) DN-Arf6-expressing cells recover WT-Arf6-mediated restriction of CVB3 relative to control cells. HeLa cells were transfected with DN-, CA- or WT-Arf6, washed the following day and infected at 48 h post-transfection. Cells were fixed and the number of IU ml–1 was enumerated by flow cytometry 8 h later. Alternatively, cells were harvested with protein lysis buffer and subjected to an anti-HA and anti-actin Western blot (shown as a positive image from a ChemiGenius blot and gel imaging system). (c) Silencing of endogenous Arf6 expression in HeLa cells recovers eGFP–CVB3-Woodruff infection in HeLa cells. HeLa cells were silenced for Arf6 expression using siRNA oligonucleotides directed against Arf6. Cells were transfected with Arf6 RNA oligonucleotides 1 or 2 or the scramble oligonucleotide (CON), washed the following day and infected with CVB3 at 48 h post-transfection. Alternatively, cells were harvested with protein lysis buffer and subject to Western blotting to detect Arf6 and actin. †P<0.01; ‡P<0.001. Results are representative of three independent experiments.

HeLa cells transfected with WT-, DN- and CA-Arf6 were infected with eGFP–CVB3-Woodruff. We found that CA-Arf6 decreased infection 3.0-fold, from 1.0x10⁷ to 3.3x10⁶ IU ml^–1 (P=0.009; Fig. 2b). We recovered CVB3 infection using the reduced Arf6 trafficking activity of the GDP-bound DN-Arf6 mutant to 8.1x10⁶ IU ml^–1, which constituted a 2.5-fold recovery of infection over the CA-Arf6 mutant. Overexpression of Arf6 via a WT-Arf6 mutant decreased infection to 4.4x10⁶ from 1.0x10⁷ IU ml^–1 in control HeLa cells (P=0.009), which constituted a 2.3-fold reduction in infectivity compared with control infection.

We silenced endogenous Arf6 expression in HeLa cells to determine the effect of endogenous Arf6 levels on CVB3 infection (Fig. 2c, inset). Arf6 was silenced using siRNA and cells were infected 2 days later (Fig. 2c). We observed a 3.0- and 2.6-fold recovery of CVB3 infection in Arf6-silenced HeLa cells using oligonucleotides 1 and 2, respectively, compared with the scramble oligonucleotide-treated control cells.

A fraction of co-localized CVB3 and DAF is sequestered to discrete regions of HeLa cells overexpressing WT-Arf6
Arf6-mediated endocytosis is responsible for some GPI-linked protein trafficking activity (Naslavsky et al., 2004). Therefore, we imaged the interaction of CVB3 with its GPI-linked co-receptor, DAF, at early time points p.i. in HeLa cells overexpressing WT-Arf6. We used a VP1 mAb, a biotin-conjugated secondary antibody and quantum dots to enhance the signal-to-noise ratio and narrow the emission wavelengths of CVB3 imaging during virus entry. At 10 min p.i., we observed CVB3 and DAF sequestered into discrete regions, distant from the nucleus (Fig. 3a, white arrows), in WT-Arf6-expressing cells. We noted that the untransfected cells demonstrated proximal virus/DAF nucleus localization (arrow head) at 10 min p.i. As WT-Arf6 expression resulted in a loss of infectivity, we considered the sequestering of CVB3 into pockets away from perinuclear regions as a mechanism of virus entry restriction.

(78K): Fig. 3. Arf6 expression results in sequestering of CVB3 away from perinuclear regions during entry. HeLa cells expressing WT-Arf6 were infected with CVB3 on ice for 1 h, warmed to 37 °C and fixed, and the virus was stained with 655 nm quantum dots at 0 and 10 min p.i. (a) or with Alexa Fluor 488 at 0, 30 and 60 min p.i. (b) to confirm the imaging method in (a). Images were acquired with a Leica SP2 confocal microscope. Arrows indicate virus sequestered into tubular processes, distant from the nucleus, in WT-Arf6-transfected cells. Arrowheads demonstrate virus/DAF proximal to the nucleus of untransfected cells. Results are representative of two independent experiments for each quantum dot and Alexa Fluor 488 virus-staining method.

We also imaged CVB3 entry using secondary antibodies conjugated with Alexa Fluor 488. This dye has a lower signal-to-noise ratio than quantum dots, but we noted that the clarity of the images was greater. Fig. 3(b) is a 60 min time course of CVB3 entry. At 0 min p.i., we saw a similar pattern of VP1 staining on WT-Arf6-transfected and -untransfected cells. However, at 30 and 60 min p.i., there was a brighter pattern of CVB3 VP1 staining on WT-Arf6-transfected cells (arrows). The untransfected cells showed a fainter but proximal nuclear localization of CVB3 (arrow heads).

Virus ligation of the CAR receptor, but not the DAF co-receptor, activates ERK MAP kinase
Trafficking of the Arf6-dependent endocytosis pathway can be enhanced by ERK MAP kinase activation (Robertson et al., 2006) and ERK MAP kinase activity can be triggered by Arf6 (Tushir & D'Souza-Schorey, 2007). The importance of ERK signalling during CVB3 infection has been demonstrated in vitro (Luo et al., 2002) and in vivo (Opavsky et al., 2002), but the precise mechanism of activation has not been elucidated.

Fig. 4 shows a Western blot of phospho-(pThr202/pTyr204) ERK MAP kinase and total ERK at 15 min p.i. The results showed that virus activated ERK during entry but that ERK activation was inhibited with an anti-CAR mAb at a 1 : 1000 dilution (200 ng ml^–1). In contrast, the anti-DAF mAb had little effect on activation of ERK. By raising the concentrations of the DAF and CAR antibodies to 1 : 50 (4 µg ml^–1), we were able to activate ERK with anti-CAR antibody alone but not with anti-DAF antibody. These results show for the first time that CVB3 activates ERK via its primary CAR but not its DAF co-receptor on HeLa cells during virus entry.

(35K): Fig. 4. ERK activation can be inhibited by blocking the virus–CAR interaction or activated by adding high concentrations of anti-CAR, but not anti-DAF, antibody. HeLa cells were treated with anti-DAF or anti-CAR mAb for 30 min prior to and during CVB3 infection. Cells were harvested with protein lysis buffer on ice at 15 min p.i. and lysates were probed for phospho-(pThr202/pTyr204) ERK MAP kinase (upper panels) and total ERK (lower panels). The image shown is the product of one Western blot and was rearranged for ease of presentation. Results are representative of three independent experiments.

Additionally, we noted in every experiment conducted that the anti-DAF mAb treatment decreased the basal ERK activity to undetectable levels (Fig. 4).

Inhibition of ERK activation during CVB3 entry reduces VP1 to EEA1 co-localization during virus entry
A previous study has shown how ERK activation can trigger uptake by clathrin-mediated endocytosis (Khundmiri et al., 2004). As CVB3 has been reported to enter HeLa cells via clathrin-mediated endocytosis and EEA1-positive endosomes (Chung et al., 2005), we analysed a time course of CVB3 infection by staining, including a DMSO vehicle control and U0126 (10 µM)-treated cells, for EEA1 and VP1, at 0, 30 and 60 min p.i. (Fig. 5). Transport of CVB3 into EEA1 during entry has been demonstrated previously (Chung et al., 2005). Consistent with these findings, we noted that EEA1 coalesced in the control cells at 30 and 60 min p.i. (Fig. 5b and c). The U0126-treated infection time courses in Fig. 5(b and c) showed that there was less coalescence of EEA1 at 30 and 60 min p.i., respectively. In addition, there also appeared to be less VP1 to EEA1 co-localization at these same time points.

(82K): Fig. 5. Inhibition of virus-induced ERK MAP kinase activation by U0126 inhibits coalescence of, and virus entry into, EEA1 endosomes at 60 min p.i. (a–c) CVB3 was bound to cells on ice for 60 min in the presence or absence of U0126 and warmed to 37 °C for the indicated period of time. Control cells were treated with DMSO vehicle only. Cells were fixed at 0 min (a), 30 min (b) and 60 min (c) p.i. and stained for virus (Alexa Fluor 488), EEA1 (Alexa Fluor 594) and nuclei (DAPI). Cells were imaged using a Leica SP2 AOBS confocal microscope. (d) Graph of Pearson's correlation of EEA1 and VP1 co-localization (y-axis) at 0 and 60 min p.i. from two independent experiments. *P=0.05.

We used Pearson's correlation to quantify the co-localization of pixels in the CVB3 VP1 channel to pixels in the EEA1 channel during CVB3 entry. A Pearson's correlation of 1 indicates that every pixel in a given channel is correlated/co-localized with the pixels in the second channel. A correlation of 0 indicates no co-localization of pixels between the different channels. Fig. 5(d) shows the Pearson's correlation of the VP1 channel (Alexa Fluor 488) to the EEA1 channel (Alexa Fluor 594) at 0 and 60 min p.i. As expected, the Pearson's correlation of VP1 to EEA1 was similar between the control and U0126-treated cells at 0 min p.i. However, at 60 min p.i., the Pearson's correlation of the control cells increased to 0.39 compared with 0.22 in the U0126-treated cells (P=0.05). These results suggested that ERK inhibition by U0126 treatment inhibits the uptake of CVB3 into the early endosome.

We treated HeLa cells, expressing WT-, DN- and CA-Arf6, with U0126 for 30 min prior to infection and for the first 1 h of CVB3 infection. This experiment was designed so that ERK activation would be targeted during the entry phase of virus infection only. Cells were harvested 8–10 h later for enumeration of infection by flow cytometry. As discussed earlier, and shown in Fig. 6(a), the DN-Arf6 mutant recovered infection (2.3x10⁵ IU ml^–1) restricted by the WT-Arf6 mutant (1.2x10⁵ IU ml^–1; P=0.02). Fig. 6(a) also shows how infection was decreased when ERK activation was inhibited during entry of CVB3, from 4.1x10⁵ to 1.5x10⁵ IU ml^–1, a 2.7-fold loss of infectivity, when ERK was inhibited (Fig. 6b). However, the WT-Arf6- and DN-Arf6-expressing cells exhibited similar titres of 1.1x10⁵ and 1.3x10⁵ IU ml^–1 upon treatment with U0126 (P=0.5; Fig. 6a). These titres were within 1.3- and 1.2-fold of the control HeLa cells, respectively. This suggests that CVB3 is not further restricted by Arf6 in the absence of ERK activation. However, CA-Arf6 should remain active even in the absence of ERK stimulation and should therefore restrict CVB3 further with U0126 treatment. Indeed, U0126-treated CA-Arf6-expressing cells were infected to a titre of 5.3x10⁴ IU ml^–1 whereas the U0126-treated control cells were infected to a titre of 1.5x10⁵ IU ml^–1 (P=0.02; Fig. 6a). This constituted a further 2.8-fold reduction in infectivity in the U0126-treated CA-Arf6-expressing cells compared with the U0126-treated control cells (Fig. 6b).

(18K): Fig. 6. Inhibition of ERK activation with U0126 eliminates Arf6-mediated restriction of CVB3. (a) HeLa cells were transfected with WT-, DN- or CA-Arf6 and infected with CVB3 2 days later. U0126 ERK inhibitor (filled bars) or DMSO vehicle (empty bars) was added for the first hour of infection to target the virus-entry window. Infection was enumerated by flow cytometry at 8 h p.i. (b) Fold changes in infection relative to the corresponding U0126-treated or untreated transfection control in (a). Infection (IU ml–1) in the control transfection cells was divided by infection in Arf6-transfected cells (empty bars) or U0126-treated cells (filled bars). *P<0.05. Results are representative of four independent experiments.

CVB3 can be sequestered by Arf6 trafficking activity into a compartment that does not support replication. By imaging early virus entry events, it appears that virus is caught within Arf6-dependent compartments that sequester virus and prevent trafficking to perinuclear regions that better support virus replication. However, virus also triggered, via CAR ligation, ERK MAP kinase signalling activity, which enhanced the productive internalization of CVB3 into the cell. Paradoxically, this same ERK signalling activity is responsible for Arf6-mediated restriction of virus infection.

Our results support the notion that CVB3 enters EEA1-positive endosomes during entry (Chung et al., 2005), and that this activity is dependent on the activation of ERK MAP kinase. However, this is not the first time that endosomal trafficking activity has been attributed to MAP kinase of ERK (MEK)/ERK activation. Phosphorylation and subsequent endocytosis of the Na⁺,K⁺-ATPase α₁-subunit into clathrin-coated vesicles is mediated by ERK (Khundmiri et al., 2004), and the negative feedback of this process may even be mediated by localization of MEK2 to endosomes (Galperin & Sorkin, 2008).

We noted that cells expressing a WT-Arf6 construct demonstrated the greatest degree of lamellipodia-like protrusions as early as 10 min p.i., which were not evident with U0126 treatment. Results from our laboratory have shown that 10–20 min after CVB3 infection is the time when the majority of entry-phase ERK activation takes place (Luo et al., 2002; Si et al., 2005) and when ERK has been shown to enhance Arf6 activity (Robertson et al., 2006), resulting in cortical actin rearrangement and cell spreading (Klemke et al., 1997; Song et al., 1998). Imaging experiments showed VP1 protein sequestered away from perinuclear localized VP1 at 10 min p.i. in WT-Arf6-expressing cells, which we did not see in untransfected cells, so we postulated that this virus-sequestering activity by Arf6 results in CVB3 entry restriction.

The CVB3 restriction activity of the Arf6 mutants was discordant with the inhibition of β-catenin internalization during EDTA treatment. Our results showed that the DN-Arf6 mutant recovered infection restricted by the CA- and WT-Arf6 mutants; however, the DN and CA mutants both perturbed the internalization of β-catenin. These data indicate different activities of Arf6-mediated internalization. The CA-Arf6 mutant has been reported to produce an accumulation of Arf6-positive vesicles, not normally seen in cells expressing WT-Arf6, blocking recycling and vesicle convergence (Donaldson, 2003). Any mutant that blocks the cycle may block Arf6 function (Donaldson, 2003) and our results of both CA and DN mutants blocking β-catenin internalization are certainly supported by these observations. Thus, merely perturbing the action of Arf6 with the DN and CA mutants is sufficient to inhibit internalization of β-catenin, probably because the pathway into which β-catenin is normally routed is interrupted by both mutants. Therefore, we propose that the activity level of Arf6, rather than any particular compartment, can draw virus away from a clathrin-mediated endocytosis pathway and productive regions of virus replication, resulting in restriction.

The CAR-dependent activation of ERK MAP kinase appears to activate two conflicting pathways: there was elimination of virus restriction by Arf6 when we treated target cells with the ERK inhibitor U0126, but productive clathrin-mediated pathways were inhibited. Arf6 and ERK activation states may well be co-dependent as Arf6 activity can result in ERK activation (Tushir & D'Souza-Schorey, 2007), but the results obtained here support the observation that Arf6 activity is dependent on ERK activation (Robertson et al., 2006).

Virus restriction by internalization into a detrimental endocytosis pathway has been described for HIV-1 and -2 (Marchant et al., 2005; Marsh & Helenius, 2006) and for vaccinia virus via Arf6 (Mercer & Helenius, 2008), which are very different viruses in morphology and life cycle compared with CVB3. Interestingly, vaccinia virus (de Magalhaes et al., 2001) and HIV (via CD4 ligation) (Popik et al., 1998) also activate ERK MAP kinase during virus entry, which may also activate Arf6-mediated restriction during internalization. The congruity of these studies suggests that equilibrium between productive and non-productive pathways may be a common trait of virus entry. Finally, non-productive endocytosis pathways of entry may be determinants of virus tropism through restriction of virus infection.

This work was supported by grants from the Heart and Stroke Foundation of British Columbia and Yukon (B. M. M.) and the Canadian Institutes of Health Research (CIHR) (B. M. M. and R. G. H.). X. S. is supported by a CIHR IMPACT Post-Doctoral Fellowship and CIHR Michael Smith Post-Doctoral Fellowship. D. M. is supported by CIHR, IMPACT and a British Columbia Lung Association Post-Doctoral Fellowship.