Mutational Analysis of the EMCV 2A Protein Identifies a Nuclear Localization Signal and an eIF4E Binding Site

Abstract

Cardioviruses have a unique 2A protein (143 aa). During genome translation, the encephalomyocarditis virus (EMCV) 2A is released through a ribosome skipping event mitigated through C-terminal 2A sequences and by subsequent N-terminal reaction with viral 3C^pro. Although viral replication is cytoplasmic, mature 2A accumulates in nucleoli shortly after infection. Some protein also transiently associates with cytoplasmic 40S ribosomal subunits, an activity contributing to inhibition of cellular cap-dependent translation. Cardiovirus sequences predict an eIF4E binding site (aa 126–134) and a nuclear localization signal (NLS, aa 91–102), within 2A, both of which are functional during EMCV infection. Point mutations preventing eIF4E:2A interactions gave small-plaque phenotype viruses, but still inhibited cellular cap-dependent translation. Deletions within the NLS motif relocalized 2A to the cytoplasm and abrogated the inhibition of cap-dependent translation. A fusion protein linking the 2A NLS to eGFP was sufficient to redirect the reporter to the nucleus but not into nucleoli.

Introduction

Encephalomyocarditis virus (EMCV) and Theiler’s murine encephalitis virus (TMEV) are species in the Cardiovirus genus of the Picornaviridae family. Like all picornaviruses, the 7.8 kb positive-sense RNA genomes encode single, large open reading frames (~2200 aa). Viral translation is mediated through a cap-independent type II internal ribosome entry site (IRES) located immediately 5′ of the open reading frame. The encoded polyprotein is cleaved by viral proteases in co-translational and post-translational reactions to produce a spectrum of mature viral proteins and partially processed precursors (Hahn and Palmenberg, 1996; Palmenberg, 1990; Parks, Baker, and Palmenberg, 1989). The mature proteins and their precursors are named according to their sequential locations in the polyprotein. The P1 region includes 3–4 viral capsid proteins (e.g. 1A, 1B, 1C and 1D). The P2 and P3 regions include multiple nonstructural proteins (2B, 2C, 3A, 3B, 3C^pro, 3D^pol), conserved among the viruses and responsible for RNA replication (Rueckert and Wimmer, 1984). Unique to the genome organization of each picornavirus genus are variable length Leader proteins (L) encoded 5′ of P1 region, and 2A proteins at the N-terminus of the P2 region. The L and 2A, separately or in combination, provide key virus anti-host activities and/or primary polyprotein cleavage activities, producing distinct and characteristic patterns of polyprotein processing, particularly with regard to initial co-translational cleavages. For cardioviruses, the primary scission reaction between 2A and 2B, is carried out by a monomolecular ribosome skipping mechanism dependent on the C-terminal 18 amino acids of 2A (Hahn and Palmenberg, 2001). Subsequent cleavage by viral 3C^pro (from P3 region) produces the majority of secondary polyprotein cleavages (Palmenberg, 1990; Parks and Palmenberg, 1987), including scissions between L/P1 and P1/2A (Jackson, 1986).

Since picornaviral translation is cap-independent by virtue of the 5′ IRES, many of these viruses have evolved potent mechanisms to inhibit cellular cap-dependent translation during infection, thereby thwarting detrimental antiviral responses. The enteroviruses and aphthoviruses, for example, encode secondary proteases at their 2A and L positions respectively, which target eIF4G (Guarné et al., 1998; Lloyd, Grubman, and Ehrenfeld, 1988), an essential scaffolding protein for the assembly of cap-dependent eIF4F complexes. Normally, within eIF4F, the eIF4G bridges interactions between eIF4E (cap-binding protein), eIF4A (an RNA helicase) and the incoming 40S ribosomal subunit (Gingras, Raugnt, and Sonenberg, 1999). Cleavage of this factor precludes productive association of capped mRNAs with preinitiation complexes, preventing their translation.

Cardioviruses do not have secondary proteases. Their L and 2A proteins have essential host shut-off roles, but use non-proteolytic mechanisms to achieve them. The EMCV L (67 aa) contributes to the inhibition of cap-dependent translation by triggering dramatic disruption of nucleocytoplasmic trafficking during infection. The Leader binds Ran-GTPase, an essential trafficking control regulator. This binding correlates with hyper-phosphorylation of multiple nucleoporins (Nups) in the central nuclear pore (NPC) transport channel as well as additional phosphorylation events on regulatory proteins throughout the cell (Lidsky et al., 2006; Porter and Palmenberg, 2009; Ricour et al., 2009). As a consequence, all active nuclear import and export, including that of nascent cellular mRNAs is stopped, and only very small molecules and proteins (>50 KDa) can still exchange by passive diffusion through the NPC. In the absence of L, or before it exerts effects on Nup phosphorylation, protein/RNA transport back and forth across the NPC is signal-dependent (NLS) and requires interaction with importin/exportin receptors (karyopherins) to chaperone traffic (Fahrenkrog and Aebi, 2003; Gorlich and Kutay, 1999; Stewart and Semler, 1997). Classical NLS sequences, epitomized by the SV40 T antigen, contain one or more contiguous motifs, highly enriched in positively charged residues. Other NLS sequences can be bipartite (e.g. nucleophosmin), or more hydrophobic, such as the monpartite c-myc NLS or the NLS of the hnRNP K protein (Mattaj and Englmeier, 1998). Regardless of the specific sequence, NLS-mediated binding with cognate karyopherins, defines a cargo’s destination (in or out) and can even define its subcellular localization (e.g. nucleolus). Our current model of EMCV L activity proposes that Nup phosphorylation prevents shuttling of karyopherins. The consequence of this inhibition is that nuclear proteins are not replenished in the nucleus and only large proteins which can’t diffuse, or which are anchored to chromatin, remain nuclear and do not efflux into equilibrium with the cytoplasm. The consequence of unregulated relocalization affects availability of multiple translation and transcription factors, helping to abrogate many critical host antiviral functions.

The cardiovirus 2A protein is not required for L-dependent NPC shutoff. Yet it plays an ancillary, albeit poorly understood role in support of cellular translational shutoff. The protein is small (17 KDa), and in contrast to L (pI ~3.2) is highly basic (pI ~10.3). There is no overt sequence similarity to other database proteins and only limited motif-dependent sequence conservation among various strains of cardioviruses (Fig 2). From the earliest times after infection, EMCV 2A accumulates in nucleoli, where it associates with nascent ribosomal RNA (Aminev, Amineva, and Palmenberg, 2003b). A smaller cytoplasmic pool of 2A transiently associates with 40S ribosomal subunits, in reactions that correlate with the appearance of excess 80S monosomes in cells and a concomitant decrease in polysome abundance (Groppo and Palmenberg, 2007). These unusual 80S are virtually devoid of cellular (but not viral) mRNA. When 2A is expressed in cells in the absence of infection, it subtly modulates the cellular translational environment in an unknown manner, to increase the ratio of IRES-dependent translation to cap-dependent translation of reporter constructs. Deletions within EMCV 2A, in a viral context, prevent the initial polyprotein cleavage (requires the C-terminal 18 amino acids) (Hahn and Palmenberg, 1996; Hahn and Palmenberg, 2001) and are reported to affect the phosphorylation status of eIF4E binding protein 1 (4E-BP1) during Mengo infection of BHK cells (Svitkin et al., 1998). This is an important finding because hypo-phosphorylated 4E-BP1 is a direct competitor of eIF4G binding to eIF4E, suggesting that cardioviruses, like aphtho- and enteroviruses may similarly identify eIF4G and/or its interactions with eIF4E as a vulnerable targets in their respective host translation shutoff mechanisms. In the absence of infection, the competitive reactions between eIF4G (translation turned on) and 4E-BP1 (translation turned off) are mediated by common surface motifs of the format, YXXXXLΦ, where Φ is a hydrophobic amino acid (Rhoads, 2009). Since these proteins cannot interact simultaneously with eIF4E, the competition regulates cap-dependent translation activity according to the needs of the cell. Given this functional context, it was of great interest to recognize the same motif near the C-terminus of every cardiovirus 2A protein. We now report for EMCV, this motif indeed binds eIF4E, and moreover, that a nearby 2A NLS sequence is required for virus shutoff of cap-dependent host protein synthesis.

Figure 2.

2A proteins. Cardiovirus 2A protein sequences (Table 2) encoding the putative NLS and eIF4E binding sites are aligned relative to known eIF4E binding sequences and yeast ribosomal protein NLS (YRS-NLS) sequences (Gritz, Abovich, and al., 1985; Moreland, Nam, and al., 1985; Schaap et al., 1991; Underwood and Fried, 1990). The eIF4E binding motifs are reviewed in (Rhoads, 2009).

Results

2A Localization During Infection

HeLa cells infected with vEC9, a recombinant EMCV-R with a shortened poly C tract (Hahn and Palmenberg, 1995), were examined for the localization of 2A relative to cell stains for membrane lectins (WGA) and DNA (DAPI). Consistent with previous experiments in L cells (Aminev, Amineva, and Palmenberg, 2003b) the earliest 2A signals (0–3 hrs) became visible as small punctuate nuclear speckles, proceeding to definite nucleolar accumulation after about 3 hr PI (Fig 1A). During the log phase of viral replication (3–6 hrs PI), nucleoli consistently displayed bright, steady 2A signals, while the more diffuse cytoplasmic signals also continued to build. Cell lysis initiates at 6–8 hrs PI under these infection conditions. The relative nuclear/cytoplasmic 2A distributions described by multiple microscopy trials were confirmed by subcellular fractionation and quantitative Western analyses (data not shown).

Figure 1.

2A localization. A. Infected (vEC9) HeLa cells were harvested at the indicated times, fixed and stained for 2A (red), membranes (WGA, green) and DNA (DAPI, blue). Images of all stains are merged. B. HeLa cells were infected with the indicated viruses (MOI of 30), fixed (4 hrs PI) and stained for 2A (green), membranes (WGA, red), and DNA (DAPI, blue). Right image column shows all stains (merged). Left column show 2A stain only.

2A Sequence Motifs

Aligned cardiovirus 2A protein sequences share less than 58% identity for the genus as a whole and only 88% and 82% within the EMCV and TMEV species, respectively. The 2A structure has not been resolved, but sequence-based prediction algorithms suggest the COOH half of these proteins (Fig 2) have strong alpha-helical propensities, particularly within the terminal 18 amino acids required to achieve co-translational polyprotein scission (Hahn and Palmenberg, 1996; Hahn and Palmenberg, 2001). Terminal NPG/P-type ribosome skipping elements are common to many genera of picornaviruses (Atkins et al., 2007), but unique to the cardioviruses, a short segment within the required element (vEC9, aa 126–134), fits the profile of YXXXXLΦ binding motifs used by eIF4G and 4E-BP proteins to interact with eIF4E (Rhoads, 2009). About 30 amino acids N-terminal to this segment, cardiovirus 2A proteins also display short, highly basic segments, consistent with general consensus required for (yeast) ribosomal protein nuclear localization (YRP-NLS). In yeast this sequence directs cargo entry into the nucleus but typically is not sufficient for nucleolar localization (NoLS), a targeting destination where a universal consensus sequence is much harder to define by bioinformatics (Gritz, Abovich, and al., 1985; Moreland, Nam, and al., 1985; Schaap et al., 1991; Underwood and Fried, 1990).

eIF4E Pull-down Experiments

The ability of vEC9 2A to bind eIF4E in vivo was tested in pull-down assays with anti-eIF4E antibodies linked to agarose beads. After appropriate lysates were reacted with the beads, the bound proteins were fractionated by SDS-PAGE then identified by Western analyses (Fig 3A). As expected, eIF4E itself reacted with the beads regardless of whether the lysates came from infected or uninfected cells. Likewise, some eIF4G was captured as predicted for a primary binding partner of eIF4E. A portion of the 2A protein was also brought down with beads exposed to infected cell lysates. We have reported a minor fraction of 2A typically associates with 40S ribosome subunits during infection (Groppo and Palmenberg, 2007). To rule out the possibility the 2A pull-down signal was due to eIF4E-linked ribosomes, an antibody to the small ribosome subunit protein S6, was included among the probes. No S6 was associated with the beads, indicating the 2A detected by this the assay was not from ribosome contaminants and that our bead-washing procedures were sufficiently stringent.

Figure 3.

Affinity assays. Clarified lysates were prepared from mock-infected or infected HeLa cells (vEC9, MOI of 30), harvested 5 hours PI. A. Agarose-conjugated eIF4E antibody was added to a portion of each lysate and the bound proteins identified by Western analyses with the indicated reagents. “Lysate” is a diluted (10%) starting sample without resin fractionation. B. Similar to A, except resin was sepharose conjugated to 7^me-GTP (cap). The bound and unbound proteins from the same lysate samples were analyzed in parallel. C. Similar to A except the sepharose resin was conjugated to GMP.

In the preferred format of pull-down experiments, reciprocal antibodies are used to cross-test and confirm interactions. But when mAb-2A was coupled to beads, reacted with lysates and boiled to release the proteins, the IgG light chain (27 kD) masked potential signals from eIF4E (27 kD), rendering the results inconclusive (data not shown). Instead, to verify the putative interaction between eIF4E and 2A the tests were repeated using 7^meGTP beads, a 5′ cap mimic, as an alternate method for capturing eIF4E. As with antibody-mediated pull-downs, proteins associated with eIF4E were also captured (Fig 3B), but they were not captured equivalently by control beads bound to GMP (Fig 3C), since eIF4E does not react with the unmodified nucleotide (Marcotrigiano et al., 1999). In agreement with Fig 3A, eIF4E and 2A (when present) were extracted from lysates only with the 7^meGTP beads. The absence of S6 and 4E-BP1 among the resin-bound signals indicates the captured 2A was not mediated by eIF4E-bound ribosomes or by indirect 2A:4E-BP1 interactions.

eIF4E Binding Motif Mutations

If the conserved segment near the COOH terminus of 2A is responsible, mutagenic interference should disrupt eIF4E binding. Single (L₁₃₄A) and double (LI₁₃₄AA) substitution mutations were engineered into this motif within a vEC9 context (Fig 4A). The resultant viruses grew more slowly than vEC9, requiring ~35 and 48 hrs respectively, to achieve plaques similar to wild-type at 29 hrs PI (Fig 6A). The mutations were stable throughout virus growth and did not revert (sequencing not shown). Infected cell extracts harboring the wild-type or mutant 2A proteins were retested in assays with the α-eIF4E beads (Fig 4B) and with the 7^meGTP beads (Fig 4C). Predictable, eIF4E and eIF4G were again extracted from all samples. The 2A from vEC9-infected lysates also bound both sets of resins. The L₁₃₄A and LI₁₃₄AA proteins, although synthesized in their respective lysates and detectable by Westerns (Fig 4B) were not extracted by the eIF4E resin (Fig 4B). The 7^meGTP beads showed some (lower) reactivity with L₁₃₄A but none with LI₁₃₄AA (Fig 4C). Again, the results are consistent with reduced eIF4E binding by the mutant 2A proteins relative to wild-type. Therefore, the 2A region impacted by these mutations must contribute to the ability of eIF4E to pull down 2A.

Figure 4.

Mutations in eIF4E binding site. A. Recombinant substitution mutations in the putative eIF4E-binding site are mapped. Wild-type virus (vEC9) or virus encoding the single (L₁₃₄A) or double (LI₁₃₄AA) mutations were infected into HeLa cells. Lysates were harvested and reacted with α-eIF4E resin (B) as in Fig 3A, or with 7^me-GTP resin (C) as in Fig 3B. “M” samples are from mock-infected cells carried in parallel.

Figure 6.

Point mutation phenotypes. A. Wild-type virus (vEC9) and those encoding the single (L₁₃₄A) or double (LI₁₃₄AA) 2A mutations were infected into HeLa cells. Plaques were visualized at 29 or 48 hr PI. B. Matched samples of infected or mock-infected cell lysates were harvested at 4 hrs or 7 hrs PI, fractionated by SDS-PAGE, then subjected to Western analyses using the indicated reagents.

While such experiments provide effective assessments of mutual affinities, they do not provide overtly quantitative measures of binding stoichiometries. Over the course of multiple such experiments, including those shown here, we found eIF4E and eIF4G were typically extracted with approximately equivalent Western signals, regardless of the source of starting lysate (mock or infected). But variability between experiments and in bead saturation levels, made it difficult to conclude whether 2A (vEC9) was augmenting or displacing any eIF4E:eIF4G interactions.

The Western analyses with mutant infected lysates did suggest lower levels of 2A were present in these cells, especially for the LI₁₃₄AA mutation (Fig 4BC, “Lysate” lanes). If true, 2A-dependent mechanisms invoking an overall reduction in viral protein synthesis, enhanced 2A turnover rates or a consequent defect in host-protein shutoff schemes might be consistent with the observed slower plaque development for these mutant viruses (Fig 6A). Surprisingly, cells pulsed with ³⁵S-Met at 7 hr PI labeled the full panel of non-2A viral proteins at roughly equivalent levels (Fig 5A), and moreover indicated that effective host translational shutoff was readily achieved by these mutants as well as the wild-type virus. Infected lysates probed with virus-specific antibodies after 4 and 7 hr PI, confirmed these observations (Fig 6B). The viral polymerase 3D, and capsid proteins 1AB and 1D, gave similar signals for all 3 viruses. The diminished 2A mutant signals (Fig 4) were protein-specific, because 2A but not its precursor P12A, was of lesser intensity only in the double mutation sample.

Figure 5.

Protein synthesis during infection. A. Infected (vEC9) and uninfected (M) L-929 cells were pulse-labeled with ³⁵S-methionine then harvested at 5 hr PI. B. Similar to A, vEC9 and viruses encoding the single (L₁₃₄A) or double (LI₁₃₄AA) 2A point mutations were infected into HeLa cells (MOI of 100). At 6 hrs PI the cells were treated with ³⁵S-methionine and then harvested 1 hr later. C. Similar to B, HeLa cells were infected with vEC9 or the Δ94–100 and Δ40–97 NLS deletion mutant viruses, then pulsed with ³⁵S-methionine 30 min before harvest, at 4 hr or 7 hr PI. For all samples, the proteins were fractionated SDS-PAGE and visualized by phosphoimaging.

The NLS Motif

The mechanism by which 2A protein enters nuclei and accumulates within nucleoli is unknown. A conserved region centering around residue R₉₇ matches the consensus pattern for yeast ribosomal protein nucleolar localization signals, specialized derivatives of monopartitate NLS (Fig 2). The activities of this region were probed with single (R₉₅A, R₉₇E, Q₁₀₅A), double (YY₉₂AA) and triple (KRR>AAA) point substitutions with the intention of targeting positive-charged or highly conserved residues (Fig 7A). Additional deletion viruses excising the full NLS motif (Δ94–100) or the central portion of the protein (Δ40–97) completed the panel. All these sequences proved viable as transfected genome RNAs or derivative virus, although their growth characteristics fell into different categories (Fig 7A). The Q₁₀₅A mutant behaved like wild-type vEC9 (not shown). The R₉₅A, R₉₇E and YY₉₂AA mutants developed plaques within 29 hrs, but their growth titer was about 5–10 fold lower and their plaque sizes were a bit smaller than wild-type. The deletion viruses and KRR>AAA viruses proved difficult to grow. They required at least 48 hrs for visible plaques, with the larger deletion having the most severe phenotype. Virus titers for all of these averaged about 100 fold lower than vEC9.

Figure 7.

Mutations in NLS. A. Recombinant vEC9 with single (R₉₅A, R₉₇E, Q₁₀₅A), double (YY₉₂AA), triple (KRR>AAA) point mutations, and 7 (Δ94–100) or 58 (Δ40–97) amino acid deletions in the 2A protein, spanning the putative NLS site were constructed. Italics denote residues changes relative to vEC9. B. The viruses were infected into HeLa cells at the indicated dilutions. Plaques were visualized at 29 or 48 hr PI. C. Matched samples of infected (MOI of 10) or mock-infected HeLa lysates were harvested at 4 hrs or 7 hrs PI, fractionated by SDS-PAGE, then subjected to Western analyses using the indicated reagents.

Characteristics of the large Δ40–97 deletion during infection of BHK cells have been described (Svitkin et al., 1998). The mutant polyprotein is poorly expressed and has inefficient primary processing. Because of this, or in addition to these phenotypes, Δ40–97 is defective in host translational shutoff. The same parameters held true in HeLa cells (Fig 5C and 7C), and moreover were displayed in part by the much smaller Δ94–100 deletion excising only the putative NLS region. When pulsed with ³⁵S-Met 30 minutes prior to harvest (Fig. 5B), both deletion mutants showed a decreased ability to inhibit cap-dependent protein synthesis relative to vEC9, as evidenced by the continued strong presence of non-viral background bands, even as late as 7 hrs PI. Both deletions also had delayed P1/2A processing, showing stronger precursor bands (e.g. P12A, 1CD2A, 1D2A) and diminished mature capsid bands (1AB and 1D) at 4 hr PI (Fig 7C). Primary and secondary processing was occurring however, because by 7 hr PI, despite the continued presence of P1 precursors, 1AB, 1D and free 2A were also detected. The very small Δ40–97 2A (<8 kd) is not retained on these gels, but it does contain epitopes recognized by the mAb used in these assays. As a rule, the tested point mutations including KRR>AAA did not show the same processing delays. Mature capsid proteins and 2A appeared in a timeframe similar to vEC9 (all mutants not shown) but the lower overall fecundity of the more severe mutations (e.g. YY₉₂AA, KRR>AAA) reduced the total yield of 2A.

NLS contribution to 2A Localization

When infected into cells each of NLS region point mutants showed 2A nucleolar staining, although not necessarily as bright or persistent as that induced by vEC9 (Fig 1B). In contrast both deletion mutants had almost exclusively cytoplasmic 2A signals, even at late time points (8 hrs PI), when their delayed processing allowed for mature proteins (Fig 7C). Only when multiple cell fields were examined, were occasional examples with slight nuclear staining observed for the deletion mutants (e.g. Fig 1B). The weaker patterns were always less distinct than the strong punctuate nucleolar staining by vEC9, or even the point mutants. Sub-cellular fractionation studies and Western analysis confirmed the results obtained by microscopy (not shown).

Certainly, active nuclear transport or passive diffusion across the NPC have equivalent potential to give the same apparent nuclear phenotypes for a protein as small as 2A. Discrimination between these mechanisms used a test sensitive to active transport, using a larger protein reporter. The 2A sequence was subdivided (codons 1–50, 51–100, 101–143) then linked in-frame to an eGFP gene within an expression cDNA (Fig 8A). The variants were sequenced to confirm their identity, especially at the fusion junction. Western analyses after transfection of HeLa cells gave strong signals for eGFP and for the fusions with 2A_1–50 and 2A_51–100 (Fig 8C). Reporters linked to full-length 2A or the COOH-terminal 2A fragment (2A_101–143) expressed below the detection limit of this antibody (Fig 8C). Localization of the fusion proteins within transfected cells therefore relied on live cell imaging of eGFP fluorescence. The parent protein (eGFP) was found to be distributed in both the nucleus and cytoplasm (Fig 8B). Linkage to 2A_1–50 did not significantly change this distribution, with the fluorescent signal showing evenly throughout the cytoplasm and nucleus. In the cytoplasm eGFP-2A_1–50 in some cells localized to discrete foci that were not present during viral infection and are likely an accumulation phenomenon due to the inability of 2A_51–100 to efficiently localize to nuclei. In a few cells, eGFP-2A_1–50 demonstrated marginal nucleolar localization. However, these cells were uncommon and diffuse nuclear cytoplasmic distribution was the more representative phenotype as evidenced by subcellular fractionation (Fig 8C). Linkage of eGFP to the 2A_51–100 fragment converted the pattern into almost exclusively nuclear but the protein was still excluded from the nucleolus. For eGFP, eGFP-2A_1–50 and eGFP-2A_51–100, there was sufficient GFP antibody signal to confirm the distributions by subcellular fractionation and Western analyses (Fig 8C). As suggested by the distributions within live cells, the relative Western signals (N/C ratio) measured about a 5-fold higher in the nucleus with the 2A_51–100 linkage, than for eGFP alone. Fluorescent microscopy also detected the eGFP-2A and eGFP-2A_101–143 fusion proteins, although adequate recording of these signals required increased exposure (2×) and gain (4×) relative to the other tested (less toxic) samples. Clearly, the full 2A fusion redirected eGFP from the cytoplasm into the nucleus, and looked very much like the 2A_51–100 signals with the exception of accumulation of full length 2A in the nucleolus. Reporter linkage to 2A_101–143 had the opposite effect, rendering the protein almost exclusively cytoplasmic. The different phenotypes for all these proteins are consistent with the idea that full length 2A, or 2A_51–100 have NLS-like sequences influencing protein accumulation in the nucleus, independent of diffusion.

Figure 8.

Expression of eGFP-2A fusion proteins. A. Map of fusion protein variants with expected molecular weights. B. HeLa cells were transfected with the indicated plasmid cDNA. At 20 hrs post transfection, live cells were imaged with bright field and fluorescence filters. C. The eGFP plasmid series was transfected into HeLa cells. At 24 hrs post-transfection, cells were disrupted, then either lysed directly into SDS buffer, or fractionated into nuclear and cytoplasmic components before SDS-PAGE and Western assays. The N/C Ratio evaluates comparative pixel densities for the nuclear/cytoplasmic eGFP bands (nd, not done)

Cell Protein Response to Infection

Nuclear sequestration of eIF4E is one possible model by which the 2A NLS and the 2A factor-binding motif could potentially exert linked activities. When tested in cells however, this simple hypothesis could not be substantiated. Nuclear and cytoplasmic extracts from infected or uninfected cells (4 hrs PI) gave no indication of eIF4E redistribution (Fig 9A). The infected extracts did have stronger nuclear 2A than cytoplasmic 2A signals, as expected. But the eIF4e, eIF2α and tubulin signals were predominantly cytoplasmic and remained so after infection. Likewise, there was no infection-induced relocalization of B23 (nucleolin) or eIF3n, a cytosolic ribosome dissociation factor.

Figure 9.

Subcellular Protein Distribution. A. Infected (+, vEC9, MOI of 30) and uninfected (−) HeLa cells were harvested at 4 hr PI. Equivalent cell counts for whole cell extracts (Whole), cytoplasmic (Cyto) or nuclear (Nuc) extracts were gel fractionated, then subjected to Western analyses using the indicated reagents. B. Infected (lane 2, vEC9, MOI of 100) or uninfected (lane 1) HeLa cells or L-929 cells were harvested at 5 hr PI, fractionated by SDS-PAGE then probed by Western analyses for 4E-BP1 phosphorylation status. P+ indicates hyper-phosphorylation, P− indicates hypo-phosphorylation. Control samples (lane 3) are from uninfected cells after serum starvation, or (lane 4) from similarly starved cells, resupplemented with serum 1 hr before harvest.

Although experiments of this type cannot characterize every cellular protein, we also monitored reagents peripheral to the 4E-BP1 pathways, since 2A competition for eIF4E binding is another potential mechanism for host-translational shutoff activities. Reports that 4E-BP1 becomes hypophosphorylated in a 2A-dependent manner after infection of BHK cells with EMCV-R (Svitkin et al., 1998), were not supported by our results in HeLa and L cells (Fig 9B). With an appropriate antibody, the four mTOR-induced 4E-BP1 phosphorylation states are readily distinguished in Western assays (Mothe-Satney et al., 2000). Whole cell extracts from our experiments showed no significant shift in the distribution of these states following infection with vEC9 for either cell type (lanes 1 vs 3). Control cells deprived of serum for 48 hours (starved) did indeed dephosphorylate 4E-BP1 in HeLa but not L cells (lanes 3), a status that could be reversed when serum was re-added (lanes 4). Parallel infections with ³⁵S-Met pulse-labeled HeLa and L cells, affirmed that host-protein synthesis was in fact shutoff in both cell types (Fig 5A and 5B), but without triggering apparent changes in the 4E-BP1 profile. It is of interest that mTOR kinase, the reactive agent controlling the phosphorylation status of 4E-BP1 did partially relocalize to the nucleus during vEC9 infection (Fig 9A). Moreover, there were additional changes in the phosphorylation status of eIF4E, S6 and S6 kinase (S6K).

Discussion

Each of the 12 genera of picornaviruses encode one or more mid-polyprotein genes designated “2A”. Seven of the genera also encode N-terminal Leader proteins, which work in synergy with 2A, to confer a variety of anti-host defensive properties to genomes that contain them. The 2A and L of cardioviruses are neither homologs nor analogs to the other genera, and even among cardioviruses, these proteins have variable sequences. The structure of EMCV L was recently resolved (Cornilescu et al., 2008). Parallel function studies assigned its potent cellular toxicity to nucleocytoplasmic trafficking disruption, mediated by aberrant L-induced MAPK cellular phosphorylation cascades directed at NPC nucleoporin proteins and related cellular targets (Porter et al., 2006; Porter and Palmenberg, 2009).

Studies on the cardiovirus 2A protein are not nearly as advanced. When introduced independently into cells via cDNA, 2A is highly inhibitory to self-expression from cap-containing mRNAs (Fig 8C). Unfortunately, the consequent corollary is that low protein yields make it hard to follow what’s actually happening in compromised cells. Expression-competent mutant 2A sequences from equivalent cDNAs, are by definition defective in 2A toxicity, and therefore less than helpful for many experiments. Switching research tactics to virus mutation studies creates additional technical problems. The C-terminus of 2A is required for primary polyprotein processing. If this sequence or its function are interfered with, the resulting L-P1-2A region is not recognized or cleaved by 3C^pro (Hahn and Palmenberg, 1996; Hahn and Palmenberg, 2001). In turn, this means the P1 capsid proteins are cleavage-defective and assembly incompetent, the L protein is not released for its anti-cellular activities, 2B lacks a proper N-terminus to be functional in replication complexes, and the compromised 2A may or may not be released for nucleolar trafficking and its own unspecified tasks. Virus failure at any of these points can be lethal, or at a minimum give small plaque phenotypes, perhaps erroneously assigned to “2A function” instead of the mechanistically simpler, malfunction of co-translational ribosome skipping. As an example, a key study linking the 4E-BP1 phosphorylation status to 2A function, was based on data suggesting rapamycin and wortmannin, mTOR inhibitors, could rescue a 2A-mutated virus, defective in cap-dependent shutoff (Svitkin et al., 1998). We now know the virus used in those studies, 2A_Δ58 (i.e. 2A_Δ40–97) has significant processing delays (Fig 7C), tying up the L protein in a precursor form and preventing its required activities. In other words, for this particular deletion, rapamycin could have rescued L instead of 2A pathways, and we would not have distinguished a functional difference.

That 4E-BP1 is involved in 2A mechanisms remains an intriguing hypothesis. We did not detect changes in this factor’s phosphorylation status during vEC9 infections in HeLa or L cells. On the other hand, it can’t be fortuitous that 2A sequences mimic the eIF4E binding motifs of 4E-BP1 and eIF4G (Fig 2), another required factor for cap-dependent translation. Antibody beads and 7^meGTP-resins specific for eIF4E, each pulled 2A out from infected cell extracts, suggesting direct motif-protein contacts rather than third-party connections. Single or double point mutations within this motif did not compromise polyprotein processing patterns (Fig 6B), but the resultant 2A was only poorly reactive with eIF4E. Therefore, this site must be active and somehow relevant to the 2A protein. Viruses with these mutations grew slower than vEC9 (Fig 6B), although surprisingly, they were still effective at the shut off of cap-dependent cellular translation (Fig 5B).

For eIF4E, 4E-BP1, and perhaps also 2A, cellular localization may provide a key to functionality. The role of eIF4E can vary depending on its location (Culjkovic, Topisirovic, and K.L., 2007). When sited in the cytoplasm, eIF4E functions in translation initiation. Its availability as an intermediate between 5′ cap recognition and eIF4G is regulated by the phosphorylation status of 4E-BP1. In the nucleus, a much smaller pool of eIF4E is responsible for export of mRNAs containing 4E-sensitive elements (SE) (Culjkovic et al., 2008; Culjkovic et al., 2005). Normally, the special RNAs with SE elements (e.g. cyclinD1), are under tight regulation and inefficiently transported through the NPC without concomitant eIF4E interactions (Culjkovic et al., 2005). Over-expression of eIF4E deregulates these mRNA pools, creating potent cell cycle abnormalities, cell transformation, and typically correlates with a poor outcome in certain cancers (Culjkovic, Topisirovic, and K.L., 2007).

Conceivably, if the relevant form of 2A:eIF4E interactions were nuclear and not cytoplasmic, it might explain some of our observed phenomena. 2A accumulates in the nucleus, preferentially in nucleoli and in additional small perinuclear pockets (Fig 1A). Mutagenic disruption of 2A:eIF4E interactions had little effect on the shutoff of cap-dependent translation initiation, nor did it affect virus IRES-dependent translation, consistent with idea that 2A exerts its influence in the nucleus or nucleolus, rather than cytoplasm. The bulk of eIF4E (cytoplasmic), did not relocalize to nuclei after infection (Fig 9A), and in our cell types, the 4E-BP1 phosphorylation status did not change sufficiently to account for cytoplasmic sequestering (Fig 9B). These results all point to an undefined role for 2A:eIF4E interactions outside of canonical cytoplasmic translation initiation.

Within the virus motifs we explored by mutation, only those which prevented nucleolar accumulation of 2A were defective in host translational shutoff. Nucleolar localization requires a protein to gain nuclear access through an NLS or by passive diffusion, then to be targeted more specifically to nucleoli with auxiliary sequence patterns (NoLS) providing retention interactions with nucleolar component such as DNA, RNA or resident proteins. Cardiovirus 2A amino acids 90 -100, form a conserved motif closely resembling the NLS and NoLS used by yeast ribosomal proteins (Fig 2). Single, double or even triple charge change mutations in this motif gave smaller plaque viruses (Fig 7B) without causing significant primary processing defects. But unexpectedly, these 2A proteins still accumulated in nucleoli (Fig 1B). Only excision of the full motif (i.e. 2A_Δ94–97) prevented this localization pattern, and by extension, prevented shutoff of cap-dependent translation (Fig 5B). As with the larger 2A_Δ40–97 deletion, even the precise motif excision deletion was severely defective in virus growth and had delayed L-P1-2A processing (Fig 7C). Therefore, assigning unambiguous 2A motif culpability requires the same analytical caveats as the previously reported 4E-BP1 experiments (Svitkin et al., 1998). Namely, we can’t be absolutely sure if the cellular cap-shutoff phenotypes of these particular deletions were imposed because of, or in spite of their 2A NLS locations.

As a start towards this answer, we subdivided the 2A sequence, linked each fragment to eGFP then asked whether the resulting fusion could drive the reporter into nuclei after cDNA transfection. Full-length 2A and the 51–100 fragment encoding the putative NLS, both had this property, but only the full-length protein (2A-linked) was strongly nucleolar. Independently, the NH₂- and COOH 2A fragments were ineffective for sustained relocalization. This identifies the central 2A NLS motif (94–100) as active under conditions where nucleocytoplasmic trafficking is functional (i.e. early during viral infection or after transfection of cDNA). Nucleolar retention however, must require additional 2A sequences. Since the protein needs to gain nuclear access (NLS or diffusion) then stick there, we suspect there must be additional (currently) unmapped rRNA or protein binding tethers yet to be identified (N-terminus?), a 2A functionality suggested by some of the very original work on this protein (Medvedkina et al., 1974). We are currently initiating further protein mapping studies to identify whether such a motif exists and if so, to describe its specificity.

Interestingly, the 101–143 fragment even when linked to eGFP proved highly toxic to cDNA cap-dependent self-expression, as was the full-length 2A-linked fusion. It cannot be a coincidence that this C-terminal fragment, unique in sequence among picornaviruses, contained the intact eIF4E binding motif. Perhaps the overall purpose of cardiovirus 2A nucleolar targeting may be to deliver this toxic segment to its required locale for disruption of nucleolar eIF4E activities. If so, the recombinant eGFP-2A_101–143 might have inadvertently depleted nuclei (cytosol?) of eIF4E, synthetically achieving the same effect without nucleolar retention. To test this idea, we are attempting to devise better assays for all these proteins so we can begin to follow the fates and phosphorylation status of nuclear eIF4E during infection or transfection experiments. Sensitive methods to measure rRNA synthesis and SE-containing mRNA fates when 2A is (or is not) allowed nucleolar accumulation are also required. Moreover, the potential role of L-induced MAPK phosphorylation cascades and/or cellular NPC trafficking inhibition cannot be discounted for possible synergy with 2A activities during infection. The functional status of mTOR and 4E-BP1 are most certainly linked to L-dependent regulation. The data we have presented here on the 2A NLS and eIF4E binding motifs are just the most recent steps towards unraveling these complicated mechanisms.

Materials and Methods

Viruses and Cells

HeLa cells (ATCC CCL-2) were grown in suspension culture with modified Eagle’s medium, 10% calf serum (NBCS) and 1% fetal bovine serum (FBS). Virus vEC9 is a recombinant EMCV-R containing a shortened poly-C tract (Hahn and Palmenberg, 1995). Cell or mouse infections with this sequence or its mutant derivatives are well described (Hoffman and Palmenberg, 1995; Martin et al., 1996). Briefly, subconfluent (90–95%) plated HeLa cells or L-929 cells were washed with phosphate buffered saline (PBS) then overlaid with virus diluted in PBS. After an attachment period (30 min, 25°C) Eagle’s medium containing 10% NBCS or 1XP5 (37°C) was added and cells were incubated (37°C, 5% CO₂). Plaque analysis was as described (Dvorak et al., 2001). During experiments to measure 4E-BP1 phosphorylation status, some control (uninfected) cells were starved of serum for 48 hours (induced hypo-phosphorylation), or starved (48 hr) then resupplemented with serum (1 hr) before harvest (induced hyper-phosphorylation) (Svitkin et al., 1998).

Subcellular Fractionation

Infected (vEC9, MOI of 30) and uninfected HeLa cells were counted (on plates) at 4 hr PI, washed with PBS, harvested, resuspended in buffer (10mM hepes, pH 7.7, 1.5mM MgCl₂, 10mM KCl, 1mM DTT, 1mM PMSF) then incubated on ice for 15 minutes. Cells were pelleted (400 × g), resuspended in buffer then lysed by passing through a 26G syringe (10×). A fraction of the whole cell lysate (Whole) was retained. Nuclei were pelleted (10,000 × g, 10 min., 4°C). The supernatant was saved as the cytoplasmic lysate (Cyto). The nuclei were washed (3×, 10mM Tris, pH 7.4, 2mM MgCl₂, 0.3M sucrose), repelleted, suspended in wash buffer then lysed by sonication (Nuc). Samples normalized to contain equivalent cell counts were re-suspended in loading buffer, boiled, fractionated by SDS-PAGE (Laemmli) and analyzed in Western assays.

Recombinant Constructions

Full-length recombinant EMCV genomes encoding single (L₁₃₄A, R₉₅A, R₉₇E, Q₁₀₅A) double (LI₁₃₄AA, YY₉₂AA) or triple (KRR>AAA) 2A point mutations or a 7 aa deletion (2A_Δ94–100) were engineered using the primers in Table 1 and standard cloning procedures. Positive-sense primers for each mutation were added to PCR reactions along with primer 526 and pEC9 cDNA. Similar reactions were performed with corollary negative-sense mutagenic primers and primer 256. The amplicons were combined then used as templates in secondary PCR reactions with primers 526 and 256. The products were digested with Sac I and Sac II (New England Biolab, NEB), as was the pEC9 cDNA. Appropriate fragments were gel purified (agarose) then ligated (T7 ligase, NEB). Recombinant plasmid pE-2A_Δ40–97 (previously, pEC9-2A_Δ58) has a deletion of 58 residues within 2A removing the middle third of the protein (Svitkin et al., 1998).

Table 1.

Primers used for 2A variants.

Primer		Sequence
	526 +	TGGTAGTCGAAACCATCTGACAG
	256 −	GTTCCTCATTGCCTACAC
L134A	987 +	GCGGACCTAGCGATTCATGACATTGAGACAAAT
L134A	988 −	GTCATGAATCGCTAGGTCCGCAAAGTAACCAGCG
LI134AA	1080 +	GACCTAGCGGCTCATGACATTGAGACAAAT
LI134AA	1081 −	TGTCATGAGCCGCTAGGTCCGCAAGTAACC
R95A	711 +	TATATATTATAAGGCAGTCAGGCCTTTTAGACTGC
R95A	712 −	AAGGCCTGACTGCCTTATAATATATATCCATGG
R97E	560 +	GAGAGTCGAGCCTTTTAGACTGCCCCTGG
R97E	561 −	CAGTCTAAAAGGCTCGACTCTCTTATAATATATATCC
Q105A	562 +	CTGGTTGCGAAGGAATGGCCCGTGCGAG
Q105A	563 −	CATTCCTTCGCAACCAGGGGCAGTCTAAA
YY92AA	747 +	TATAGCTGCTAAGAGAGTCAGGCCTTTTAG
YY92AA	748 −	TCTCTTAGCAGCTATATCCATGGTCACTCTAC
KRR>AAA	743 +	TATATTATGCGGCAGTCGCGCCTTTTAGACTGCCCCTGG
KRR>AAA	744 −	AAAAGGCGCGACTGCCGCATAATATATATCCATGGTCAC
Δ94–100	745 +	CATGGATATATATTATCTGCCACTGGTTCAGAAGGA
Δ94–100	746 −	TTCTGAACAGTGGCAGATAATATATATCCATGGTCAC
peGFP	1206 +	TGTAAATCCATGGGCGGCTCCCACCATCATCATCA-TCATATGGTGAGCAAGGGCGAGGAGCTG
peGFP	1208 −	GAGATCGGATCCCGGACTTGTACA
peGFP-2A	1211 +	TGATGGTGGTGCTCGAGCTACCCT
peGFP-2A	1212 −	CCGAAAAAGCTTTCCAAATGCCCTAGA
peGFP-2A1–50	1217 +	GTTACCGAAAGCTTAGTCCAAAT
peGFP-2A1–50	1218 −	CAGGAACTCGAGTTATTTGGTCTTTGATCT
peGFP-2A51–100	1219 +	TCAAAGAAGCTTCAGGTCTCTTTCCTGAGC
peGFP-2A51–100	1220 −	CTGAACCTCGAGTTATCTAAAGGCCTGACTCTCTT
peGFP-2A101–143	1221 +	AAGCCTAAGCTTCTGCCCTGGTTCAGAAGGAA

Plasmid peGFP was prepared by amplifying the eGFP gene from pt-GFP (Mir and Panganiban, 2008) using primers 1206 and 1207. The product amplicon was digested with Nco I and Bam HI, gel purified, then ligated into pTriex-1.1 (Novagen) using the same restriction sites. Plasmid peGFP-2A was from an EMCV 2A amplicon (primers 1211 and 1212), digested with Hind III and Xho I, then ligated into the same sites of peGFP. Protein expression after plasmid transfection is driven by a β-globulin promoter, and produces an abutted in-frame fusion between eGFP and 2A. Plasmids peGFP-2A_1–50, peGFP-2A_51–100 and peGFP-2A_101–143 are deletion derivatives created by the same strategy, albeit with different primers (Table 1). Each encodes a full eGFP but only the listed 2A amino acids. The insert configurations of all cDNAs were confirmed by sequence analysis.

Western Assays

Denatured protein samples were fractionated by SDS-PAGE then transferred onto PVDF membranes (Immobilon P, Millipore) using transfer buffer MT (25 mM Tris, pH 8.0, 0.19 M glycine, 20% methanol). After treatment with blocking buffer (5% w/v nonfat dry milk, 0.05% w/v Tween-20 in TBS, 1 hr, 20°C), the membranes were rinsed with TBS (20mM Tris-HCl, pH 7.6, 140mM NaCl) and reacted overnight with primary antibody (in TBS with 2.5% w/v nonfat dry milk, 0.05% w/v Tween-20 in TBS) at 4°C with agitation. The membranes were rinsed (3×, TBS) and treated with horseradish peroxidase-conjugated secondary antibody (in MT), then rinsed (3×, TBS) again before the bands were visualized by chemiluminescence (ECL kit, Amersham Bioscience). Murine monoclonals (5A12; 1:2000 dilution) against Mengo 2A (Aminev, Amineva, and Palmenberg, 2003a) and Mengo 3D^po (8D10) as well as murine polyclonal against the Mengo capsid (Frolov, Duque, and Palmenberg, 1999) have been described. Polyclonal antibodies against tubulin, GFP, eIF3n, B23 (1:1000, Santa Cruz Biotech), S6, eIF4E, 4E-BP1, eIF2α and phosphospecific antibodies against S6, S6K, eIF4E and Mnk1 (1:1000, Cell Signaling Technology) were commercial, as were appropriate secondary antibodies (1:8000) and stains for WGA (rhodamine or fluorescein) and DAPI (Sigma). On all figures “α” designates the target of the primary antibody.

Confocal Microscopy

HeLa cells (7×10⁵) grown on coverslips in 30mm dishes or in 24 well plates were infected with virus (MOI of 10 PFU/cell). For transfection experiments, the cells (2.2×10⁵ cells/well), were reacted with Lipofectamine 2000 (Invitrogen) and appropriate cDNAs (1:1, μg cDNA/Lipofectamin), purified using Maxi-Prep techniques (Qiagen). At the time of harvest (24 hrs for transfections), cells were washed (2×, PBS), fixed for 10 minutes at room temperature with 4% formaldehyde, washed again (3×, PBS) and permeabilized (0.2% Triton, 5 min, 25°C). After further washing (3×, PBS), blocking solution (10% FBS in TBS) was added (1 hour at 25°C), followed by another wash (1×, TBS). The cells were reacted with primary antibody (diluted with 3% FBS in TBS) overnight at 4°C, then rewashed (3×, TBS) before the secondary antibody, WGA (rhodamine-conjugated or fluorescein -conjugated) and/or DAPI (diluted in 3% FBS in TBS) were added (1 hr, room temperature). After a final wash (3×, TBS) the samples and mounted on slides using Vectorshield mounting media (Vector Labs). Visualization used a Nikon Eclipse TE2000-V laser scanning confocal microscope with excitation lines of 408 (blue), 488 (green) and 543 (red) nM. EZ-CL1 and Photoshop software was used for image analysis.

eIF4E Affinity Assays

Infected HeLa cells (vEC9, MOI of 30) were harvested 5 hr PI, washed with PBS then resuspended in IP buffer (20mM Tris-HCl pH 7.4, 150mM NaCl, 1mM EDTA, 1mM EGTA, 1% Triton X-100, 1mM beta-glycerol phosphate, 1mM Na₃VO₄, 1mM PMSF). After 5 min on ice, samples were lysed by sonication (3×, 5 sec.) and clarified by centrifugation (4°C, 10 min., 14,000× g). Protein concentrations were determined (Bio-Rad assay) and equivalent total protein samples were reacted with anti-eIF4E antibody conjugated to agarose beads (~20 μl, Santa Cruz Biotech). After incubation (16 hrs, 4°C) the beads were washed extensively (1× TBS), then boiled in SDS, before the collected proteins were fractionated by SDS-PAGE and visualized by Western analyses.

Cap Affinity Assays

Infected cells were collected, washed with PBS then resuspended in buffer C (20mM Tris, pH 7.4, 100mM KCl, 20mM beta-glycerol phosphate, 1mM DTT, 0.25mM Na₃VO₄, 10nM NaF, 1mM EDTA, 1mM EGTA, 10nM okadaic acid, 1mM PMSF) before being subjected to 3× freeze/thaw cycles. Samples were clarified by centrifugation (10,000× g, 4°C, 10 min). Protein concentrations were determined (Bio-Rad assay) and equivalent total protein samples were reacted with 7^meGTP-sepharose 4B resin (~30 μl, GE Healthcare) or GMP-agarose resin (Sigma) for 3 hrs, 4°C before the resin samples were extensively washed with buffer C then boiled in SDS, before the collected proteins were fractionated by SDS-PAGE.

Metabolic Labeling

Confluent HeLa monolayers (30 mm plates) were infected with virus (MOI of 100 PFU/cell). After an attachment period (30 min, 20°C) the plates were washed (2×, PBS) then overlaid with P5 medium (Rueckert and Pallansch, 1981). Ninety minutes prior to harvest the monolayers were washed again (2×, PBS) then overlaid with MEM lacking methionine (Sigma). One hour later, [³⁵S]-Met was added (100 μCi/plate, Amersham), and after an additional 30 min, the cells were washed (2×, PBS) and lysed (30mM Tris-HCl, pH 7.4, 140mM NaCl, 0.5% NP-40). The incorporation of label into acid-insoluble material was quantified by scintillation counting. The protein content was visualized by autoradiography after fractionation by SDS-PAGE.

Bioinformatic Analysis

Cardiovirus genome sequences are available from NCBI (Table 2). The 2A protein alignment used MEGA software (Tamura et al., 2007) with Blosum matrix settings, a gap opening penalty of 3 and gap extension penalty 1.8. The output file was examined and adjusted for high and low road considerations consistent with the biology of these proteins.

Table 2.

Cardiovirus Sequences

Sequence	GeneBank accession number
TMEV (GDVII)	P08545
Theiler’s-like virus of rats	BAC58035
Mengo virus (M)	P12296
EMCV (B)	P17593
Human TMEV-like cardiovirus	YP_001950226
Saffold virus	YP_001816886
EMCV (vEC9)	P03304