Eukaryotic translation initiation machinery can operate in a prokaryotic-like mode without eIF2

Unlike prokaryotes, a specialized eukaryotic initiation factor 2 (eIF2), in the form of the ternary complex eIF2*GTP*Met-tRNA iMet is utilized to deliver the initiator tRNA to the ribosome within all eukaryotic cells 1 . Phosphorylation of eIF2 is known to be central to the global regulation of protein synthesis under stress conditions and infection 2 . Another distinctive feature of eukaryotic translation is scanning of mRNA 5’-leaders, whose origin in evolution may be relevant to the appearance of eIF2 in eukaryotes. Translation initiation on the hepatitis C virus (HCV) internal ribosome entry site (IRES) occurs without scanning 3,4 . Whether these unique features of the HCV IRES account for the formation of the final 80S initiation complex is unknown. Here we show that the HCV IRES-directed translation can occur without either eIF2 or its GTPase activating protein eIF5. In addition to the general eIF2- and eIF5-dependent pathway of 80S complex assembly, the HCV IRES makes use of a prokaryotic-like pathway which involves eIF5B, the analogue of bacterial IF2 5,6 , instead of eIF2. This switch from a eukaryotic-like mode of AUG selection to a “bacterial” one occurs when eIF2 is inactivated by phosphorylation, a way with which host cells counteract infection. The relative resistance of HCV IRES-directed translation to eIF2 phosphorylation may represent one more line of defense used by this virus against host antiviral responses and can contribute to the well known resistance of HCV to interferon based therapy.

Initiation factor eIF2 is a pivotal component of the translation initiation mechanism in all eukaryotic cells. It is absent from bacteria, thereby reflecting quite different modes of the start codon selection in pro-and eukaryotes. Strikingly, some viral mRNAs are efficiently translated under conditions when the activity of eIF2 is suppressed by phosphorylation as a result of host cells response to viral infection 2,7 . This is exactly the case of HCV RNA, reported to be refractory to reduced eIF2·GTP·Met-tRNA i Met ternary complex availability 8 . The assembly of fully functional 48S pre-initiation complexes on the HCV IRES only requires eIF2 and eIF3 3 . Addition of conventional ribosome joining factors eIF5 and eIF5B along with 60S subunits led predictably to 80S complex formation 9 (Fig. 1a). Surprisingly, omitting eIF5, a trigger of GTP hydrolysis on eIF2, does not abolish the 80S complex formation (Fig. 1a). The possibility of contamination with eIF5 within the system was ruled out by Western-blotting analysis of eIF2 and eIF5B preparations and by the inability of β-globin mRNA to form 80S complexes when eIF5, which is essential for this standard system 6 , was absent (Fig.1b). This finding could be explained by eIF5B functioning instead of eIF5 as was suggested for 80S complex formation on AUG-triplets 6,10,11 . However, control experiments in which the 48S to the 80S conversion was blocked by the presence of GMPPNP (a non-hydrolysable analogue of GTP) also resulted in 80S complex formation, although at a somewhat reduced level (Fig. 1c), indicating at least a partial independence of 80S assembly on GTP hydrolysis. The conversion of the 48S complex into the 80S is known to require GTP hydrolysis by eIF2 and thus should be completely blocked by GMPPNP. Therefore, the question arises of whether there is an obligatory requirement for any eIF2 if there was no strict requirement for GTP hydrolysis? Indeed, in the presence of only ribosomal subunits, eIF3, eIF5B, and Met-tRNA i Met we found that 80S translation initiation complexes were formed in the presence of GMPPNP, thereby eliminating the possibility of any residual eIF2 activity. However, eIF3 along with eIF5B were indispensable (Fig. 1d). The functional role of eIF3 in this very simplified process of 80S assembly is not yet clear.
Recently, a "factorless" translation initiation process has been reported for the HCV IRES at elevated Mg 2+ concentrations 12 . Although we performed the assays for 80S complex formation at physiological ionic concentrations, sucrose gradients employed to analyze translation initiation complexes routinely contained elevated concentrations of Mg 2+ (6 mM). However, decreasing the Mg 2+ concentration to 2 mM during centrifugation did not abolish eIF2-independent 80S assembly ( Fig. 1e). We failed to reproduce "factorless" 80S complex formation 12 either under our experimental conditions or at elevated Mg 2+ concentrations (data not shown).
The functionality of the 80S initiation complexes assembled from 40S and 60S ribosomal subunits, eIF3, eIF5B, and Met-tRNA i Met , i.e. their ability to be engaged into the subsequent step of translation, the elongation of the polypeptide chain, had to be proven. To this end, we changed the 7th triplet of the HCV ORF to a UAA stop codon and then reconstituted the translation elongation process from totally purified components 13 . The position of the stop codon was chosen in such a way that the ribosomes did not reach a putative frameshift site 14  Notably, in our in vitro experiments the presence of eIF2 in the reconstitution system strongly inhibited GTP hydrolysis-independent formation of the 80S complex (data not shown). This could be due to a higher affinity of the ternary complex for the 40S subunit, implying that eIF2 outcompetes eIF5B from the 40S. It also means that eIF2 inactivation should result in a switch of the translation initiation modes utilized by HCV from eIF2-dependent to eIF2-independent rather than in a severe inhibition of its translation.
To address the physiological relevance of these data we inactivated eIF2 via phosphorylation both in vitro and in vivo and studied the effect on HCV IRES-driven translation. Rabbit reticulocyte lysate (RRL) was treated with dsRNA to induce PKR 16 and supplemented with GMPPNP to block the eIF2-dependent pathway at the 48S stage. Control assays were not treated with dsRNA but also contained GMPPNP. It was found that no 80S complexes were assembled on the HCV IREScontaining mRNA in control lysates (Fig. 3a). In contrast, in dsRNA treated RRL 80S complexes were formed on the HCV IRES transcript. This was not the case for the ß-globin mRNA, a typical cellular mRNA which can only use the canonical eIF-2 dependent pathway of translation initiation ( Fig. 3a, b). It is clear that eIF2 had been inactivated, at least partially, since the amount of 48S complexes assembled on the ß-globin mRNA was reduced (Fig. 3b). The phosphorylation of eIF2 in the presence of dsRNA was confirmed by Western-blotting with anti phosphor-eIF2α(Ser51) antibodies (Fig. 3c). These data may explain why in the presence of GMPPNP in cell extracts no 80S complexes have been observed previously 3,17,18 . They indicate that under normal conditions, i.e. when eIF2 is fully active, the translation initiation on the HCV IRES proceeds primarily through the eIF2-dependent pathway. 5 The in vitro experiments were complemented with those performed using transfected cells.
HEK293T cells were first treated with various reagents which elicited eIF2 phosphorylation either by activation of various eIF2 kinases (PKR by dsRNA transfection, PERK by DTT 19 , or HRI by sodium arsenite 20 ) or by inhibition of PP1 phosphatase (by okadaic acid 21 ) and then transfected with a mixture of a standard cap-dependent mRNA and the HCV IRES containing monocistronic mRNA. As seen in Fig. 4, all these treatments affected much less the HCV IRES-driven translation than that directed by a standard 5'-UTR. Importantly, the translation directed by certain picornavirus IRESes, like HRV or EMCV, were inhibited by these treatments to the same extent as cap-dependent translation (Fig. 4, data not shown). Interestingly, treatment with IFNα, widely used in HCV therapy, also lead to relative resistance of the HCV RNA translation. Thus, consistent with other previously reported data 7,8,22 , eIF2-inactivation does not lead to severe inhibition of HCV IRES-driven translation. The data described above present for the first time the mechanistic explanation of this phenomenon, demonstrating the existence of the alternative, eIF2-independent, pathway of translation initiation for the HCV RNA. The switch from a eukaryotic-like mode of AUG selection to a "bacterial" one occurs when eIF2 is inactivated by phosphorylation, a way with which host cells counteract infection. Thus, presented data throw a bridge over evolutionary different modes of translation initiation.

Met
. Ribosomal subunits, initiation and elongation factors were purified from RRL or Krebs-2 ascites cells or were expressed in E.coli as described 13,23,24 .
Assembly and analysis of translation initiation complexes. Ribosomal complexes were assembled and analyzed by sucrose gradient centrifugation or toeprinting as described 3,23 . For in vitro translation elongation experiments factors eEF1H, eEF2, and total aminoacylated tRNA were added to preformed 80S complexes 13 . For in vitro eIF2 inhibition, 20 ng of synthesized dsRNA 13 were added to 14 μl of RRL (Promega) and incubated for 15 min at 30 o C, then lysate was supplemented with GMPPNP and ribosomal complexes were assembled as described previously 23 .

Met
. 40S, 60S, eIF4F, eIF5, eIF5B and eEF1H were isolated from RRL, eIF2 and eIF3 were purified from Krebs-2 ascites cells. eIF1, eIF1A, eIF4A, eIF4B, and MetRS were expressed in E.coli. Rabbit eEF2 was a kind gift from L.P.Ovchinnikov. Met-tRNA i Met was prepared as described 13 . The S100 supernatant from ascites cells, freed from nucleic acids by passing through DEAE-cellulose at 0.25 M KCl and ammonium sulfate precipitation, was used as a source of mammalian aminoacyl-tRNA synthetases. Total calf liver tRNA (Novagen) was aminoacylated using a protocol similar to that which was employed for the aminoacylation of tRNA i Met .