S. de Breyne & T. Ohlmann, INSERM U758, Ecole Normale Supérieure de Lyon, 46 Allée d’Italie, 69364 Lyon Cedex 07, France Fax: +334 72 72 81 37 Tel: +334 72 72 88 93; +334 72 72 89 53 E-mail: firstname.lastname@example.org; email@example.com Website: http://hvd.ens-lyon.fr/teams/TEVA
Expression of the two isoforms p55 and p40 of HIV-1 Gag proteins relies on distinct translation initiation mechanisms, a cap-dependent initiation and two internal ribosome entry sites (IRESs). The regulation of these processes is complex and remains poorly understood. This study was focused on the influence of the 5′-UTR and on the requirement for the eukaryotic initiation factor (eIF)4F complex components. By using an in vitro system, we showed substantial involvement of the 5′-UTR in the control of p55 expression. This highly structured 5′-UTR requires the eIF4F complex, especially RNA helicase eIF4A, which mediates initiation at the authentic AUG codon. In addition, the 5′-UTR regulates expression in an RNA concentration-dependent manner, with a high concentration of RNA triggering specific reduction of full-length Gag p55 production. HIV-1 genomic RNA also has the ability to use a strong IRES element located in the gag coding region. We show that this mechanism is particularly efficient, and that activity of this IRES is only poorly dependent on RNA helicase eIF4A when the viral 5′-UTR is removed. HIV-1 genomic mRNA exhibits in vitro translational features that allow the expression of Gag p55 protein by different mechanisms that involve different requirements for eIF4E, eIF4G, and eIF4A. This suggests that HIV-1 could adapt to its mode of translation according to the availability of the initiation factors in the infected cell.
Translation initiation promotes the attachment of the small ribosomal subunit (40S) to mRNA and then the selection of the start site codon (usually AUG). Most mRNAs use the scanning mechanism, in which the 43S preinitiation complex, composed of a 40S ribosomal subunit associated with a set of eukaryotic initiation factors (eIFs) and Met-tRNAiMet, interacts with mRNA via the eIF4F complex. This complex consists of three subunits: eIF4E (cap-binding protein), eIF4A (a DEAD-box RNA helicase), and eIF4G (a scaffold protein that binds eIF4E, eIF4A, and also the 43S preinitiation complex via eIF3). eIF4F allows 5′-end cap structure recognition, and unwinds local RNA structures to favor ribosome landing on the mRNA. Both eIF4G and eIF4A participate in the scanning of the 5′-UTR to reach the initiation codon. Recognition of the initiation codon leads to the release of the eIFs from the preinitiation complex, and association of the 40S and 60S subunits to form an elongation-competent 80S ribosome .
Some viral infections can induce the cleavage of eIF4G by viral proteases or the sequestration of eIF4E, resulting in the shut-off of cellular mRNA expression initiated by a cap-dependent mechanism . Some positive-strand RNA viruses have developed a sophisticated mechanism to initiate translation. Their mRNAs contain an internal ribosome entry site (IRES), allowing direct ribosome recruitment in a cap-independent manner that requires fewer eIFs than the scanning mechanism . Viral IRESs can be divided into different classes according to their need for both canonical components of the translational machinery and noncanonical components called IRES trans-acting factors. The first IRESs to be described belong to the Picornaviridae family [epitomized by poliovirus and encephalomyocarditis virus (EMCV)]. Initiation on these IRESs requires the C-terminal part of eIF4G and eIF4A but not eIF4E . eIF4G and eIF4A interact specifically with the IRES, and induce conformational changes of the RNA structure to promote attachment of the 43S ribosome [5,6]. IRESs from group III, epitomized by hepatitis C virus (HCV) and classical swine fever virus, have the ability to bind directly to the 40S ribosome. Addition of the GTP–eIF2–Met-tRNAiMet ternary complex and eIF3 is sufficient to enable formation of the 48S initiation complex . The last group is represented by the cricket paralysis virus-like IRESs, on which ribosome recruitment occurs directly without any eIFs or the Met-tRNA initiator . Thus, according to the structure of the IRES, the mechanism of ribosomal recruitment differs greatly and exhibits various levels of requirement for the components of the eIF4F complex.
IRESs have also been described in the Retroviridae family . HIV-1 belongs to the lentivirus subfamily, and is the major agent of AIDS. The HIV-1 particles contain two identical full-length positive-stranded RNA molecules. The full-length RNA molecule, once produced by the host cell, has a dual function, serving both as a genome to be encapsidated into newly formed virions and as an mRNA encoding the Gag and Gag-Pol polyproteins. Translation of the HIV-1 genomic mRNA can be initiated via the classical cap-dependent mechanism, via an IRES present in the 5′-UTR, or via an IRES located in the Gag ORF. The presence of these three different mechanisms to ensure ribosomal recruitment indicates that Gag translation is complex and under tight regulation [10–16].
HIV-1 cap-dependent initiation has been shown to occur in vitro and ex vivo [10,12,15]. However, the highly structured 5′-UTR is expected to inhibit ribosomal scanning, and may require RNA helicases and/or RNA chaperone proteins to assist ribosome movement through the 5′-UTR [17,18]. In addition, during infection, the HIV-1 Vpr protein can induce cell cycle arrest in G2/M, which results in decreased cap-dependent translation [19,20]. The HIV-1 5′-UTR, spanning nucleotides 104–336 (position of the AUG1 codon), was shown to contain an IRES element, the activity of which is cell cycle-regulated and enhanced when the cells are arrested in G2/M [12,21]. The Gag coding region, which spans nucleotides 336–761 (from the AUG1 to the AUG2 codon), also contains a functional IRES  that promotes the expression of an additional N-terminally truncated isoform of Gag polyprotein of unknown function but whose deletion severely affects growth kinetics in cell culture . This IRES, which lies exclusively in the Gag coding region, can also be found in HIV-2 genomic mRNA, and in the simian and feline homologs [22–24]. The HIV-2 Gag IRES was recently studied by generating synthetic leaderless mRNAs that start at the AUG1 codon [22,25,26]. The leaderless mRNA is used as a tool with which to study how the IRES in the Gag coding region can recruit ribosomes in the physical absence of the 5′-UTR, as the latter could interact and influence ribosome entry from the coding sequence. Initiation from synthetic leaderless mRNA is usually poorly efficient. Indeed, efficient translation requires a distance of at least four to eight nucleotides from the RNA 5′-end , and can be enhanced by a short motif in the 5′-UTR . However, because of the presence of a very particular type of IRES, embedded within the Gag coding region, the synthetic HIV-2 leaderless mRNA is able to promote robust translation from the first initiation codon in vitro and in cells . These characteristics have been extended to synthetic HIV-1 and simian immunodeficiency virus leaderless mRNAs . Enhancement of translation upon removal of the 5′-UTR suggests the use of a distinct translation initiation mechanism that may require a different set of eIFs.
By using a spectrum of in vitro transcripts, we showed that the presence, or absence, of the 5′-UTR greatly modulates protein production at the two initiation sites (AUG1 and AUG2). More detailed investigation revealed that this reflects a different requirement for eIFs of the eIF4F group, and notably for eIF4A, one of the RNA helicases required for ribosomal scanning. Taken together, our results reveal great adaptability of HIV-1 genomic RNA to the utilization of the translational machinery.
The HIV-1 5′-UTR regulates initiation at the AUG1 codon
To characterize the two initiation events on leader or leaderless RNA, we investigated the effects of RNA concentration on the translation of various HIV-1-derived mRNAs. For this, the region spanning from the transcription start site (nucleotide 1) to the second in-frame initiating AUG codon (AUG2; nucleotide 761) was added in front of the Renilla luciferase coding region to yield the UTR–AUG2 construct (Fig. 1A). Nuclease-treated rabbit reticulocyte lysate (RRL) was programmed with different amounts of capped UTR–AUG2 mRNAs (from 12.5 to 400 fmol), and this yielded two distinct proteins (Fig. 1B, lanes 2–7) that result from initiation events at the AUG1 and AUG2 codons (respectively, p55 and p40 Gag isoforms [11,12]). Expression from the AUG2 codon increased progressively to reach a plateau phase (Fig. 1B, lanes 2–7, gray curve). Similar expression profiles were observed when the reporter gene was placed under the control of the globin 5′-UTR (cap-dependent mechanism) or the EMCV IRES (Fig. S1), and most probably corresponds to saturating concentration of mRNAs in the translational assay. In contrast, the expression profile at the AUG1 codon was completely different, and translation was quite efficient at low RNA concentrations, but became weaker at higher concentrations (Fig. 1B, lanes 2–7, black curve). Surprisingly, from ∼ 200 fmol of RNA, the shorter Gag isoform became the predominant product. We next investigated the specific role of the viral 5′-UTR. For this, the HIV-1 5′-UTR (nucletides 1–336) was appended upstream of the Renilla luciferase gene to produce the UTR–AUG1 construct (Fig. 1A). The translation of different concentrations of capped UTR–AUG1 RNAs led to a similar expression profile (Fig. 1B, lanes 8–13), confirming that the presence of the 5′-UTR renders expression of the construct sensitive to RNA concentration. Translation at the AUG1 codon became more efficient in the absence of the Gag coding region, which is consistent with the fact that the Gag coding region is known to inhibit translation initiated at the AUG1 codon [11,12]. To confirm that the 5′-UTR was involved in translational control, it was removed to create a leaderless mRNA starting with the AUG codon at the 5′-extremity of the mRNA (AUG1–AUG2 or leaderless construct; Fig. 1A). In vitro translation of capped leaderless mRNA was very efficient for both isoforms (Fig. 1B, lanes 14–19), and this was particularly striking for initiation at the AUG1 codon, with a two-fold increase at low RNA concentration, and up to an eight-fold increase at high RNA concentration [Fig. 1B, compare lanes 14–19 (AUG1–AUG2) with lanes 2–7 (UTR–AUG2) and the black curves]. Enhancement of activities upon removal of the viral 5′-UTR has been previously reported for HIV-1, and also for the HIV-2 and simian immunodeficiency virus sequences [22,25], and appears to be specific to lentiviral sequences. However, expression profiles generated by the leaderless mRNA showed a progressive increase in protein synthesis until a plateau phase was reached indicating that the Gag coding region is not affected by the RNA concentration.
As it seems that the presence of the 5′-UTR influences the translational behavior of the constructs, we added, in trans, different amounts of competitor HIV-1 5′-UTR to the nuclease-treated RRL programmed with nonsaturating amounts (50 fmol) of the UTR–AUG2, AUG1–AUG2 or globin RNAs (Fig. 1C). Increasing amounts of competitor led to the inhibition of translation from the AUG1 codon but stimulated protein synthesis from the AUG2 codon (Fig. 1C, lanes 6–9). However, this effect was specific to constructs harboring the HIV-1 5′-UTR, because no significant variation in expression occurred with nuclease-treated RRL programmed with the control globin RNA (Fig. 1C, lanes 2–5) or the leaderless AUG1–AUG2 RNA (Fig. 1C, lanes 10–13). These results demonstrate that the presence of the HIV-1 5′-UTR modulates, in cis and in trans, expression at both the AUG1 codon and the AUG2 codon. The different effects of the 5′-UTR on initiation at the AUG1 codon and the AUG2 codon suggest a different mode of regulation at these sites.
Differential requirements for eIF4E and eIF4G
It has been previously suggested that the HIV-1 5′-UTR can recruit the translation machinery by both a cap-dependent mechanism and an IRES-dependent mechanism (for review, see [9,16]). Nevertheless, the proportion of ribosomes that are recruited by each of these two mechanisms and the initiation factors required remain to be determined. In this context, and in order to determine the relative contribution of the cap-dependent and Gag IRES modes of initiation, we first used different amounts of capped and uncapped mRNAs with and without the 5′-UTR to program the nuclease-treated RRL (Fig. 2A) and the untreated RRL (Fig. 2B). In contrast to the nuclease-treated RRL, the untreated RRL contains endogenous mRNAs (mainly globin and lipoxygenase mRNAs) and better reproduces the poly(A) and cap synergy . First, translation of our RNA constructs in the untreated RRL was weaker than in the nuclease-treated RRL. This is because of the endogenous mRNAs, which compete with the mRNAs of interest. However, translation in both systems gave similar patterns of expression (compare Figs 2A,B and S2A,B). Indeed, the effect of the RNA concentration previously shown (Fig. 1B, lanes 2–7) could still be observed in the untreated RRL (Fig. S2C). As can be observed, expression at the AUG2 codon was not affected by the presence and/or absence of the cap structure, which is consistent with the use of an IRES mechanism (Fig. S2A,B ). This contrasts with initiation at the AUG1 codon, which showed cap dependence whether or not the 5′-UTR was present (Fig. 2A,B; compare lanes 1–6 with lanes 7–12, and lanes 13–18 with lanes 19–24; Fig. S2A,B). Next, we investigated the initiation factors that are required for HIV-1 translation. For this, we used a spectrum of inhibitors targeting the components of the eIF4F complex: eIF4G, eIF4E, and eIF4A.
Cap-dependent and picornaviral IRES-dependent initiation [5,6,30], but not the hepacivirus, pestivirus and dicistronic virus IRESs [7,8], require at least the p100 fragment of the eIF4G scaffold protein to recruit the ribosome on the mRNA. Thus, we studied the dependence on eIF4G by using the L protease from foot-and-mouth-disease virus . In the untreated RRL, addition of in vitro-translated L protease resulted in the cleavage of eIF4G (Fig. 2C), and created conditions in which cap-dependent translation (Glo-Renilla and lipoxygenase mRNAs) was inhibited (Fig. 2D, lanes 7–9; Fig. 2E), whereas the EMCV IRES-mediated protein synthesis was not affected or stimulated (Fig. 2D, lanes 10–12; Fig. 2F) [4,32]. Addition of increasing concentrations of L protease to the untreated RRL resulted in the inhibition of initiation from the AUG1 codon (Fig. 2D, lanes 1–6; Fig. 2G,H, black columns). Interestingly, the presence or absence of the leader did not cause any change to the translational pattern. In contrast, expression initiated at the AUG2 codon was not inhibited by addition of L protease, and was even strongly stimulated in the presence of the 5′-UTR (Fig. 2D, lanes 1–3; Fig. 2G, gray columns). In its absence, initiation at the AUG2 codon became relatively sensitive to L protease (Fig. 2D, lanes 4–6; Fig. 2H, gray columns). These data suggest that initiation from the AUG2 codon occurs exclusively in an IRES-dependent manner, whereas initiation from the AUG1 codon may also depend on another mechanism requiring eIF4G integrity.
Requirement of eIF4A for HIV-1 genomic RNA initiation
Given the fact that translation of the constructs used in this study showed a similar pattern in the untreated and nuclease-treated RRLs, we chose to use the latter for the remainder of this study, as it does not synthesize lipoxygenase and globin mRNAs, which may lead to confusion. Next, we evaluated the requirement for DEAD box RNA helicase eIF4A, which is involved in RNA unwinding and scanning . Chemical and enzymatic probing analysis of the HIV-1 5′-UTR [21,34–36] and the leaderless mRNA  has shown that these regions are highly structured, and this could interfere with ribosome recruitment or scanning, and may require eIF4A . To test this, we used the trans-dominant negative eIF4AR362Q mutant , which strongly inhibits ribosomal scanning. Recombinant eIF4AWT and eIF4AR362Q expressed in Escherichia coli were tested for their activity on the expression mediated by the 5′-UTR of globin and by the HCV IRES. As expected, the eIF4AR362Q mutant suppressed cap-dependent translation, whereas the activity of the HCV IRES remained unaffected (Fig. 3A, compare lanes 5 and 6 with lane 2, and lanes 10 and 11 with lane 7; Fig. S3A), as previously described . Pretreated RRL was then programmed with UTR–AUG2 and AUG1–AUG2 RNA (200 fmol), and quantification of expression products showed that initiation at the AUG2 codon was always repressed in the presence of the dominant negative eIF4AR362Q mutant, whether or not the 5′-UTR was present (Fig. 3A, lanes 12–21; Fig. S3A, gray columns). The extent of this repression was comparable to that obtained for globin mRNAs. However, to our surprise, expression from the AUG1 codon was less affected by the trans-dominant effect of eIF4AR362Q (between 20% and 40%) in the absence of the 5′-UTR (Fig. 3A, compare lanes 20 and 21 with lane 17; Fig. S3A), suggesting that initiation on this leaderless mRNA would be less dependent on eIF4A. Persistence of expression upon eIF4AR362Q treatment was also conserved when lower amounts of RNA (50 and 100 fmol) were used (data not shown). The relative inhibition of translation by eIF4AR362Q was partially restored by the addition of recombinant eIF4AWT (Figs 3B and S3B).
To confirm this dependence on eIF4A, we used a natural chemical inhibitor of the eIF4A RNA-binding activity called hippuristanol . Interestingly, we observed a similar pattern of expression (Fig. 3C) as when we used the eIF4AR362Q mutant (Fig. 3A), showing that initiation events at the AUG1 codon became more resistant to the inhibition induced by hippuristanol treatment when the HIV-1 5′-UTR was removed (Figs 3C and S3C).
We then monitored ribosome complex formation on these RNAs by sucrose density gradient centrifugation (Fig. 3D). As a control, globin mRNA was incubated in the presence of cycloheximide, which induced stalling of the 80S complexes and their accumulation at the initiation codons. After separation of the initiation complexes, the RNA sedimented in two populations: the heterogeneous nuclear riboproteins (fractions 0–4 mL), and as a peak corresponding to the 80S (fractions 5–8 mL). As expected, addition of hippuristanol abrogated stable 80S ribosome formation on globin mRNAs. Sucrose density gradient analysis revealed that two high molecular mass complexes are paused on HIV-1 leaderless RNA (Fig. 3D). This observation has been attributed to the possible presence of two 80S complexes, one at the AUG1 codon and the other at the AUG2 codon. Thus, the first peak (fractions 5–6 mL) would correspond to one 80S complex paused on one of the two initiation triplets, whereas the heavier complex corresponds to two 80S complexes on each of the two AUG codons [25,26]. Upon hippuristanol treatment, the amount of the two complexes is reduced, but there is still a significant amount of the single 80S complex. This confirms that at least some of the initiation events on the leaderless RNA are less dependent on eIF4A than the standard cap-dependent initiation. Similar results have been reported with the Gag coding region IRES of HIV-2. Taken together, our results show that ribosome recruitment at each initiation codon is influenced by the presence of the viral 5′-UTR, and can occur even at a limiting concentration of eIF4A. This implies that the presence or absence of the HIV-1 5′-UTR greatly modifies the mechanism of ribosome recruitment.
Translation from HIV-1 clinical isolates
All of the above data were obtained with sequences derived from the NL4.3 laboratory strain of HIV-1 . To extend our results, we assayed these features on wild-type circulating HIV-1. The region from the transcription start site up to the AUG2 codon was extracted from the genomic DNA obtained from blood cells of naïve HIV-1-positive patients, and cloned in frame with a Renilla luciferase reporter gene. We deliberately chose three sequences from patients that exhibit very different patterns of translation when expressed in the nuclease-treated RRL. These sequences (patients 4, 8, and 18) are presented here, and share more than 80% similarity with NL4.3 (Figs 4A and S4).
Translation of these mRNA sequences in the nuclease-treated RRL gave rise to the two Gag isoforms observed previously with pNL4.3. However, the relative ratio of expression (AUG1 versus AUG2) varied among the different patients, with some showing a similar pattern as pNL4.3 (patient 18; Fig. 4B, lane 8 versus lane 6), and others showing a significant increase in initiation events at the first AUG codon (patients 4 and 8; Fig. 4B, lanes 2 and 4 versus lane 6). Differences in the size of the full-length Gag protein were also observed, and these corresponded to the insertion of additional sequences at positions 213, 217, 691, and 717 (Figs 4A and S4). It is noteworthy that translation from the derived sequence from patient 8 gave rise to an additional Gag isoform (Fig. 4B, lanes 2 and 3; Fig. 4C, lanes 11–20), resulting from the insertion of an AUG codon at position 426 (denoted AUG*) relative to the NL4.3 sequence.
Removal of the 5′-UTR induced stimulation of expression from both AUG codons to a level similar to that observed with NL4.3 in some patients (Fig. 4B, lanes 3, 5 and 9 versus lane 7). These data indicate that enhancement of translation upon removal of the leader sequence is not restricted to the NL4.3 sequence, and can be found within clinical HIV-1 isolates. This is particularly striking for patient 8, whose translation from AUG* was stimulated approximately fivefold, without affecting expression from the AUG1 codon, suggesting that initiation from these two start site codons is completely independent, and that they use distinct ribosome recruitment mechanisms.
Finally, we tested the effect of adding the trans-dominant negative eIF4AR362Q mutant on natural HIV-1 isolates. In all cases, initiation from the AUG2 codon was repressed in the presence of the eIF4AR362Q mutant (Fig. 4C), confirming the requirement for eIF4A to initiate translation at this site. Translation from the AUG1 codon also required eIF4A when the 5′-UTR was present (patients 4, 8, and 18; Fig. 4C, lanes 4 and 5 versus lane 1, lanes 14 and 15 versus lane 11, and lanes 24 and 25 versus lane 21, respectively). Interestingly, when an extra AUG codon was present (AUG*, patient 8), translation was also resistant to eIF4AR362Q treatment to the same extent as that observed with the AUG1 codon when the 5′-UTR was deleted (Fig. 4C, lanes 16–20). These data indicate that initiation on the HIV-1 leaderless mRNAs is less dependent on eIF4A, confirming that it uses an alternative mechanism to drive translation.
HIV-1 genomic mRNA contains two initiation start sites that allow expression of both the p55 Gag protein (AUG1 codon, position 336) and the p40 isoform Gag protein (AUG2 codon, position 761), whose function is unknown . Interestingly, initiation at both AUG codons is mediated by several and distinct mechanisms. Expression at the AUG1 codon is highly regulated by the 5′-UTR, and occurs both by a cap-dependent mechanism [10,13–15] and by a cell cycle-regulated IRES element located in the 5′-UTR [12,21]. In addition, the removal of the 5′-UTR creates a leaderless mRNA that allows efficient and specific translation at the AUG1 codon by a mechanism that has not yet been characterized (Fig. 1B ). In contrast, translation initiated at the second AUG codon (AUG2) is strictly IRES-dependent (Fig. 2 [11,13]). These facts provided the rationale for this work, and we tried to answer at least two questions: (a) is translation initiation at both AUG codons independent; and (b) what are the cis-acting and trans-acting regulatory elements that control initiation at each site?
The expression mediated from the AUG2 codon is conserved from the human to the simian homologs [11,22,24], and also from sequences derived from clinical isolates (Fig. 4B). Our results strongly indicate that initiation at the AUG2 codon takes place independently of the translation initiation at the AUG1 codon in the in vitro system. Indeed, strong variations in expression from the AUG1 codon did not affect translation initiated at the AUG2 codon (Figs 1B and 2A,B). Supplementation of the RRL with RNA coding only for the HIV-1 5′-UTR stimulated expression from the AUG2 codon, but inhibited that from the AUG1 codon (Fig. 1C). This increase in expression may be attributable to a partial redistribution of the ribosomes between both AUG codons. We further showed that ribosome recruitment at the AUG2 codon was strictly dependent on the IRES (Fig. 2), and involved the RNA helicase eIF4A (Figs 3A–C and 4C). Such a need for eIF4A and eIF4G (p100 fragment) is typical of ribosomal recruitment by type I and II picornaviral IRESs [5,6,40]. However, the HIV-1 Gag IRES can directly recruit both the 40S subunit and eIF3 , which is a property of type III IRESs . In this context, we suggest that the role of eIF4A and eIF4G is to remodel the IRES to allow proper positioning of the 43S preinitiation complex in the vicinity of the AUG codon. Interestingly, the effect of L protease treatment on expression from the AUG2 codon differs according to whether or not the 5′-UTR is present (Fig. 2D,G,H). A possible explanation for this could be that the removal of the 5′-UTR results in placement of the AUG2 codon closer to the 5′-extremity of the mRNA, which may favor the use of a cap-dependent mechanism. Alternatively, it could also be argued that removal of the 5′-UTR modifies the IRES structure, which renders ribosome recruitment more sensitive to the integrity of eIF4G.
Expression from the AUG1 codon (p55) is more complex, and appears to be controlled incis and in trans by the 5′-UTR of HIV-1 genomic mRNA. Indeed, we showed that, at a high RNA concentration, expression from the AUG1 codon was specifically inhibited (Fig. 1B,C), suggesting that ribosome recruitment or scanning is blocked. One explanation could be that high levels of HIV-1 RNA favor the RNA–RNA kissing complex via the dimerization signal [41,42], which blocks ribosome scanning and subsequent initiation at the AUG1 codon. However, mutation in the dimerization site did not modify the expression profile (data not shown), ruling out such a hypothesis, although we did not further study the influence of other elements, such as the TAR structure . By using an mRNA that contains two AUG initiation sites, Dasso et al.  have shown that high concentrations of RNA (> 40 μg·mL−1) could inhibit expression from the first AUG codon (∼ 20%). Under our experimental conditions, we obtained a decrease in translation of ∼ 40%, and this was reached with a maximum RNA concentration of 27 μg·mL−1, which remained under the threshold observed by Dasso et al. This suggests that HIV-1 mRNA exhibits particular features. Another hypothesis could be that structural elements in the 5′-UTR may interact with host cell factors such as eIF2, PKR, TRBP, and La, which also regulate HIV-1 translation [45,46]. An increase in RNA concentration could titrate one of these proteins for translation; however, the number and nature of these proteins remain undetermined.
The 5′-UTR, in an in vitro system, exhibits strong cap dependence and requires the use of eIF4G, as was previously shown (Fig. 2 ). We also confirmed the involvement of the RNA helicase eIF4A (Fig. 3A–C). This requirement could result from the highly structured 5′-UTR of HIV-1. However, eIF4A may not be sufficient, and other RNA helicases may be needed, such as the DExH-box protein DHX29, which is involved in translation initiation on cellular and viral mRNAs with structured 5′-UTRs [47,48], RNA helicase A , or even DDX3, which is required for the HIV-1 RNA export and translation of selected cellular mRNAs [49–51].
Interestingly, the total removal of the 5′-UTR completely changes both the pattern of expression and the need for eIFs. As such, this leaderless mRNA allows efficient expression from the AUG1 codon located at the cap structure (Fig. 1B). This characteristic is conserved between human and simian homologs [22,25] and in sequences derived from clinical isolates (Fig. 4B). This is consistent with the fact that RNA structures in the 5′-UTR can repress initiation from the IRES in the Gag coding region . Expression from leaderless mRNAs or those with very short 5′-UTRs is rare in the eukaryotic kingdom, and generally requires an unstructured RNA region  or at least a distance of four nucleotides between the cap structure and the AUG codon [27,53], which is not the case for the HIV-1 leaderless mRNA. Translation of this mRNA was robust even at limiting concentrations of both eIF4E and eIF4A (Figs 3 and 4). This suggests that the ribosome may be directly recruited via the IRES, with no need for RNA unwinding to access the AUG1 codon. This is similar to the expression observed with the very short 5′-UTR mRNAs that contain the translation initiator of the short 5′-UTR [TISU: SAASATGGCGC (in which S is G or C)] motif, and recruit ribosomes by a cap-dependent but eIF4A-independent mechanism [28,54]. As previously described, the HIV-1 leaderless mRNA does not have a sufficient distance between the cap structure and the AUG codon, and requires the IRES elements downstream of the AUG codon. Andreev et al. have shown that mammalian 40S and 60S ribosomal subunits possess the ability to initiate translation on a prokaryotic leaderless mRNAs only in the presence of the Met-tRNA in a reconstituted system of translation. This initiation does not require eIFs, and is independent of the cap structure . Surprisingly, initiation from the synthetic HIV-1 leaderless mRNA is enhanced by the cap structure. The dual influence of the cap structure and the IRES suggests that initiation on this mRNA occurs via an unconventional and efficient translation initiation mechanism that requires eIF4G and, to a lesser extent, eIF4A.
In conclusion, the p55 Gag protein can be expressed by different translation initiation mechanisms that require specific and different sets of canonical and noncanonical eIFs. This suggests that HIV-1 genomic mRNA has the intrinsic ability to adapt and regulate its own expression according to eIF availability. Interestingly, the integrity of some eIFs is modified during viral infection [2,56–58]. In addition, signals in the 5′-UTR induce full-length genomic RNA dimerization and its packaging in the viral particle via an interaction with the Gag protein (for review, see ). These events should inhibit the ribosomal entry or scanning along the 5′-UTR. The expression of both Gag isoforms may require the IRES located in the ORF. Recently, novel HIV-2 genomic-spliced mRNA isoforms have been identified , suggesting that some novel RNA species could be identified. Thus, it is tempting to speculate that a leaderless HIV-1 transcript could be generated by alternative splicing or the use of a cryptic promoter. Indeed, this pool of mRNA, even at low concentrations, would be able to express the Gag protein, but would be not encapsidated, because packaging signals are missing.
Sequences of the globin UTR, EMCV IRES (nucleotides 265–836; M81861.1) and HIV-1 (strain NL4.3) from positions 1 to 761 (UTR–AUG2), from positions 336 to 761 (AUG1–AUG2) and from positions 1 to 336 (UTR–AUG1) were amplified by PCR, digested by EcoRV and BamHI, and inserted into the pRenilla vector  previously digested by PvuII and BamHI.
Sequences of the circulating wild-type HIV-1 were obtained from blood samples of naïve HIV-1-positive patients (Hospices Civils de Lyon). The peripheral blood mononuclear cells were isolated by Ficol density gradient centrifugation and resuspended in NaCl/Pi. Nucleic acids were then extracted with NucliSens EasyMag technology (bioMérieux France 5, Craponne, France). Finally, HIV-1 genomic DNA was amplified by PCR, analyzed by sequencing, and then cloned in the pRenilla vector.
Expression and purification of eIF4AWT and eIF4AR362Q
Vectors for the expression of eIF4AWT and eIF4AR362Q [7,40] were used to transform E. coli BL21 Star (DE3). Expression of recombinant proteins was induced for 4 h with 0.5 mm isopropyl thio-β-d-galactoside when the culture had reached 0.4 < A600 nm < 0.6. After centrifugation (4500 g for 20 min at 4 °C) of bacterial cultures, pellets were resuspended in 20 mm Tris/HCl (pH 7.5), 300 mm KCl, and 10% glycerol, and then sonicated on ice. Supernatants derived from the bacterial lysate by centrifugation were added to an Ni2+–nitrilotriacetic acid column that has been prepared by adding 800 μL of a 50% Ni2+–nitrilotriacetic acid suspension (Qiagen, Courtaboeuf, France) to a BioRad Polyprep chromatography column. Washes and imidazole elutions were performed according to the manufacturer’s instructions. The eIF4A eluate was then dialyzed and applied to an FPLC Hi-TrapQ column (GE healthcare Europe, Vilizy-Villacoublay, France).
In vitro transcription and translation
Globin UTR, UTR–AUG2 and AUG1–AUG2 constructs linearized with EcoRI were used as DNA templates for in vitro RNA transcription with T7 RNA polymerase (Promega France, Charbonnières, France), as previously described . For synthesis of capped transcripts, the GTP concentration was reduced to 0.32 mm, and m7GpppG cap analog (New England BioLabs France, Evry, France) was added at a concentration of 1.28 mm. Capped and uncapped mRNAs were translated for 30 min at 30 °C in a 10-μL reaction volume in the presence of 50% volume nuclease-treated RRL or untreated RRL (Promega France, Charbonnières, France), 75 mm KCl, 0.5 mm MgCl2, 20 μm each amino acid (except for methionine), and 0.25 mCi·mL−1 [35S]methionine. Translation products were then separated by 12% SDS/PAGE. The gel was dried and subjected to autoradiography for 12 h by the use of Kodak Biomax films (Fisher Scientific, Illkirch, France), and quantified with a Molecular Dynamics PhosphoImager.
The in vitro-synthesized L protease was prepared in the reticulocyte lysate from translation of the pMM1 clone, as previously described . Cleavage of eIF4GI by L protease was analyzed by western blotting as previously described . L protease (0.125–0.5 μL), hippuristanol (0.15–0.6 μm) and the recombinant eIF4AWT and eIF4AR362Q proteins (0.25 and 0.5 μg), were incubated for 10 min in the RRL before RNA addition.
Sucrose gradient density analysis
Ribosomal complexes were assembled on 32P-labeled globin UTR or HIV-1 leaderless (AUG1–AUG2) mRNA. Nuclease-treated RRL was pretreated with hippuristanol (10 μm) and cycloheximide (1 mg·mL−1) for 10 min at 30 °C. One picomole of RNA was then incubated for 10 min at 30 °C in the presence of 0.5 mm magnesium acetate, 75 mm potassium acetate, 20 μm amino acid, and 8 U of RNAsin (Promega France, Charbonnières, France). Reactions were stopped on ice, layered over 10–50% sucrose gradients (25 mm Tris, pH 7.6, 6 mm MgCl2, 75 mm KCl) and sedimented by ultracentrifugation at 26 9794 g in an SW40 Ti rotor for 3 h at 4 °C. Fractions were collected, blotted onto a Hybond N+ membrane (Amersham France, Courtaboeuf, France), exposed, and scanned. The amount of RNA in each fraction was determined, and expressed as the percentage of total counts.
We thank Dr T. Pestova and Dr C. Hellen (SUNY Downstate Medical Center, USA) for the gift of the eIF4AWT and eIF4AR362Q expression constructs, Dr J. Pelletier (McGill University, Canada) for the generous gift of hippuristanol, and Dr S. Morley (University of Sussex, UK) for antibody against eIF4GI. We also thank R. Zwizwai for proofreading the manuscript. Work in the laboratory of T. Ohlmann and B. Sargueil was supported by grants from the ANRS, SIDACTION and a ‘contrat d’interface’ between INSERM and the Virobiotec, Biological Ressources Center, Hospices Civils de Lyon. S. de Breyne and N. Chamond were recipients of an ANRS fellowship.