All of the TriMV cDNA clones in the monocistronic constructs with the stable hairpin at the 5′ end of the untranslated region were generated as described previously (Roberts et al., 2015). The monocistronic TriMV firefly luciferase constructs were assembled in the T3 polymerase-driven plasmid, c-myc-T3LUC(pA; Thoma et al., 2004). The TriMV mutated sequences were synthesized as gBlocks^® fragment from IDT-DNA or as GenParts from GenScript and next cloned using the recombination-based cloning kit from GenBuilder from GenScript into the TriMV plasmid cut at HpaI-NcoI sites. The sequence inserted at the very 5′ end of the untranslated region in each clone, to form a stable stem loop with a ΔG > −34 kcal (Kozak, 1989), was CGC GCG CAC GGC CCA AGC TGG GCC GTG CGC GCC.

All RNAs were transcribed in vitro from linearized plasmids or PCR-amplified products using either the T7 MegaScript kit from Ambion or the T3 RNA polymerase from Thermo Fisher. Monocistronic TriMV luciferase constructs and the control firefly luciferase construct with vector sequences were linearized either with SfcI/BfmI to include the poly(A) tail. The Renilla luciferase construct (Promega) using for the in vivo assays was linearized with BamHI to include the poly(A) tail. Reactions were assembled according to the appropriate transcription kit protocol and as previously described (Roberts et al., 2015).

The in vitro translation reactions were performed using the wheat germ extract system kit (Promega, Madison, WI) and as previously described (Roberts et al., 2015). All experiments were performed in triplicate and repeated in at least three independent experiments. For the translation assays in the presence of the anti-sense DNA oligos, luciferase activity was measured for the full-length TriMV 5′ UTR mRNA construct with the strong stem loop at the immediate 5′ end. The translation reaction was set up as previously described (Zeenko and Gallie, 2005) except the addition of magnesium acetate at a final concentration of 0.625 mM to rule out any loss of translation due to magnesium chelation by the DNA oligos. The unmodified DNA oligonucleotides were ordered at IDT-DNA. The sequences were:

Anti YX-AUG: 5′AGGAGAAAAAGAGAAGTA 3′.

Anti-Beta-globin: 5′ CAAAAGCTTGCAGGAAGAGATCCATCTAC 3′.

Anti Y-1: 5′ GAGAATAGTGGAAAAACAGGGA 3′.

Anti Y-2: 5′ AAAGTAAGAAAGTGCCAAGAGAGCAGAAAG 3’.

Oat protoplasts were prepared from an oat cell suspension culture as described (Roberts et al., 2015, 2017b). 1 pmol of each RNA reporter construct was electroporated into approximately 10⁶ cells. To normalize RNA incorporation into cells, 0.1 pmol of capped polyadenylated Renilla luciferase RNA was included in each electroporation. Five hours post-electroporation, the cells were harvested, lysed in 500 μl passive lysis buffer (Dual Luciferase kit, Promega), and centrifuged for 10 min at 15,000 × g. Luciferase activity was measured using 100 μl of the supernatant. 50 μl of the luciferase assay reagent was injected into each sample and read for 10 s. 50 μl of Stop & Glo reagent was then injected and Renilla luciferase activity was measured for six seconds. All experiments were performed in triplicate and repeated in at least three independent experiments.

SHAPE analysis was conducted as described by Liu et al. (2021). In summary, 8 pmol of in vitro RNA previously transcribed as explained above, were diluted to 30 ul of water and denatured at 95°C for 2 min, following a 2 min ice incubation. 7.5 ul of 5X RNA folding buffer (1X folding buffer = 100 mM HEPES pH 8.0, 50 mM MgCl_2, 100 mM NaCl) was added to each sample, and samples were incubated at 30°C for 20 min. Each sample was divided in two tubes. To one of the divided reactions, 3.3 μl of 100 mM NMIA was added, and to the other reaction 3.3 μl of DMSO was added. Samples were incubated at 37°C for 45 min. After modification, the RNA was precipitated with 0.1 volumes of 3 M sodium acetate, 2.5 volumes of cold ethanol, and 0.5 μl of 20 mg/ml glycogen. RNA was refrigerated at −80° C for 30 min and centrifuged for 30 min at 4°C. The pellet was washed twice with cold 70% ethanol and air-dried at room temperature for 5 min, the pellet was reconstituted in 10 μl of 0.5X TE buffer. Integrity of RNA was verified in a 1% TBE agarose gel. For cDNA synthesis, 2 pmol of the control (DMSO), modified (NMIA) RNA, or unmodified RNA (for sequencing ladder) were diluted to 8 μl of 0.5X TE buffer. 1 μl of 10 μM of fluorescently labeled primer (Thermo Fisher, 6FAM was used for DMSO and NMIA samples, and PET was used for sequencing ladders) was added to the DMSO and NMIA reactions. Samples were incubated at 65°C for 5 min and placed in ice for 5 min. After incubation, 8.5 μl of enzyme mix (4 μl 5X SuperScriptIII FS buffer, 1 μl 100 mM DTT, 1 μl 10 mM dNTPs, 2ul H₂O, and 0.5 μl SSIII enzyme) were added (for the sequencing reaction, 2 μl of 1 mM ddATP were also added to the mix). Samples were incubated at 52°C for 1 h and 1 μl of 4 M NaOH was added to the samples to remove the RNA after a 5 min incubation at 95°C. The reaction was neutralized with 2 μl of 2 M HCl and incubated at 95°C for 5 min. Each NMIA and DMSO reactions were combined individually with the sequencing reaction. cDNA was precipitated as described above and resuspended in 20 μl of H₂O. Samples were sent to Genewiz for fragment analysis. Data were analyzed using QuSHAPE software (Karabiber et al., 2013) were peak position were manually adjusted. Three replicates were submitted, and the procedure was repeated three times. After obtaining the reactivity data, secondary structure analysis prediction was performed using the RNAstructure software from the Mathews Lab (Reuter and Mathews, 2010). Primer sequences used: TriMV (798–818) = 5′ TGCTCTCCAGCGGTTCCATCC 3′ and TriMV (834–854) = 5′ GGAACCAGGGCGTATCTCTTC 3′.

5′ UTR RNA fragments were synthesized in vitro by first linearizing the firefly luciferase construct above mentioned with BbsI to remove the luciferase coding region and produce only 5′ UTR segments. The digestion fragment was then in vitro transcribed using the transcription protocol described above. The RNA fragments bear a stable stem loop structure at their 5′ end and were biotinylated at the 3′ end using the Pierce RNA 3′ End Desthiobiotinylation kit from Thermo Fisher (Product No. 20163) and following the manufacturer’s instructions. After biotinylation, RNA was cleaned using the NucleoSpin RNA Clean-up kit from Macherey-Nagel (Product No. 740948.50) and RNA integrity was verified in a 1% agarose gel. 100 μl translation reactions were set up using wheat germ extract system kit (Promega, Madison, WI) with 50 μl of wheat germ extract, 8 μl of amino acid mix, 8 μl of 1 M potassium acetate, 4 μl of 50 mM of GMP-PNP, and 25 pmol of labeled RNA. Reactions were incubated for 10 min at 25°C and UV crosslinked using 254 nm bulbs, for 45–60 s using the auto-crosslink function. After crosslinking, a pull-down of bound proteins was conducted using the Pierce Magnetic RNA-Protein Pull-Down Kit (Product No. 20164) using the manufacturer’s instructions. After incubation with streptavidin beads, 5 washes were preformed prior to the final elution, to remove all unbound proteins and reduce the occurrence of unspecific bounded proteins. An aliquot of the final elution samples was loaded in an SDS-PAGE and revealed with silver staining to verify protein presence (Supplementary Figure 3A). Remaining final elution was sent to the Mass Spectrometry Core Facility at the University of Wisconsin-Madison for Nano Liquid chromatography with tandem mass spectrometry (LC–MS–MS) analysis. Resulting data were used to search against the Triticum aestivum (130,673 entries) UniProt reference proteome (Proteome ID UP000019116) and results were analyzed using the Scaffold software (Proteome Software, Inc). The mass spectrometry analysis was performed on two independent experiments. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2022) partner repository and available with the dataset identifier PXD031443.

Based on the model that an AUG codon, which is preceded by a polypyrimidine CU-tract, can serve as the ribosomal entry site (Jackson and Kaminski, 1995), we examined the TriMV 5′ untranslated region (UTR) primary sequence surrounding the 12 upstream AUG triplets for additional potential ribosomal landing pads. No CU-rich pattern was observed embedded within or in close proximity to any of these triplets. However, we identified two fairly long CU-rich tracts (22 and 39 bases, respectively) at position nts 608–629 (UCCUgUUUUUCCaCUaUUUCUC) and at nts 651–689 (CUUUCUgCUCUCUUggCaCUUUCUUaCUUUCaCaCUCUC), just upstream of the core TriMV Y₁₆X₁₁-AUG₇₄₀ region, which embeds the preferred AUG initiation site at position nts 740–742 (Figure 1A). The two polypyrimidine tracts were spaced from one another and from the core Y₁₆X₁₁-AUG₇₄₀ motif by 21 nucleotides and 23 nucleotides, respectively. Sequence alignments of 12 newly identified TriMV isolates (Redila et al., 2021) revealed the strong sequence conservations of those motifs (Supplementary Figure 1).

We first tested the relevance of these polypyrimidine tracts, which we referred to as Y-1 and Y-2 motifs, respectively, on the IRES-driven translation at the preferred 13th AUG codon. For this purpose, we altered each of the polypyrimidine segments with random sequences in the context of the full-length 739-nts long TriMV 5′ UTR (mY-1: GUUGGAUACCGUUCGACGAGGU; mY-2: GCCGAGACAAAGCAGUAUUGUAAGCUAGUGGAAGCAGUG, respectively; Figure 1A). The TriMV 5′ UTR mutants were tested in the context of a monocistronic luciferase reporter mRNA with a stable hairpin structure at the 5′ end to block 5′ entry of the ribosomes and thus ensure measurements of internal (cap-independent) initiation activity (Roberts et al., 2015; Jaramillo-Mesa et al., 2019). In all of these constructs, the core Y₁₆X₁₁-AUG₇₄₀ motif remained intact with the 13^th AUG triplet corresponding to that of the luciferase reporter gene. We compared the translation efficiency of the mutants to that of the wild-type full-length TriMV 5′ UTR (construct 1-739) and the non-functional TriMV reverse complementary sequence (construct 739-1) in two well-validated and complementary plant translation read-outs: wheat germ extract (in vitro; Figure 1B) and oat protoplasts (in vivo; Figure 1C). Our assay revealed that mutation of the 22-nt long Y-1 sequence (mY-1) abolished translation of the luciferase construct in vitro and in vivo, while the full sequence disruption of the 39-nt long Y-2 stretch (mY-2) dropped translation down to 50% of the wild-type activity in vitro but retained up to 85% of the wild-type level in vivo (Figures 1B,C). Mutants with both auxiliary CU-rich motifs swapped into random sequences (1-739: mY-1 + mY-2 mutant) or with their internal deletion (construct del 605–700) were translationally inactive in vitro and in vivo (Figures 1B,C). These initial observations suggest that these motifs, especially the Y-1 segment, might be necessary for the TriMV IRES-driven translation.

The sequence patterns of the Y-1 and Y-2 polypyrimidine stretches shared a striking resemblance to a Y_nX_m-AUG like motif that is used as a ribosome landing site, (Y_n being a stretch of n C and U bases followed by a spacer X of m random bases), except that they did not embed any AUG triplets in their direct close proximity. We have previously shown that deletion of the region spanning the TriMV Y₁₆X₁₁-AUG₇₄₀ at position nts 709–739 (mutant 1-709) was sufficient to impair IRES-driven translation, revealing the TriMV translation dependency to an YX-AUG motif (Jaramillo-Mesa et al., 2019). The core TriMV Y₁₆X₁₁-AUG₇₄₀ motif consists of a 11-nt X spacer between the CU tract and the preferred AUG codon at position 740. This preferred AUG codon is located 111 bases and 51 bases downstream of each of the identified Y-1 and Y-2 CU-rich tracts, respectively. Based on the model that an AUG codon embedded within a YX-AUG motif can serve as the ribosomal entry site (Jackson and Kaminski, 1995), we set to test whether these auxiliary polypyrimidine segments and their partial resemblance to a conventional Y_nX_m-like motif could functionally reprogram translation. For this purpose, we first truncated the 739-nts long TriMV 5′ UTR to remove the core Y₁₆X₁₁-AUG₇₄₀ motif, positioning the initiation site of the luciferase reporter gene after the Y-1 segment at position nts 648 (mutant 1-648) or after the Y-2 tract at position nts 701 (mutant 1-700; Figure 2A). In these contexts, the AUG triplet was separated by 18 and 11 nucleotides from the closest CU-rich tract, respectively (Figure 2A), resulting in a “Y₂₂X₁₈-AUG₆₄₈” like motif in the mutant 1-648 construct, and a “Y₃₉X₁₁-AUG₇₀₁” like motif in the mutant 1-700 RNA. We compared the translation efficiency of these truncated mutants in vitro to that of the wild-type full-length TriMV 5′ UTR (construct 1-739) and the non-functional reverse complementary sequence (construct 739-1). While the 1-648 deletion mutant was barely functional, with less than 10% translation activity remaining, the 1-700 deletion mutant strongly supported translation to level well above of the wild-type sequence (Figure 2B).

The above observations led us to examine the requirement by which an YX-AUG motif specifies the translational start site. Of particular focus was the TriMV deletion mutant 1-709 in which the 5′ UTR sequence lacks the native Y₁₆X₁₁-AUG₇₄₀ but bears intact Y-1 and Y-2 CU-rich segments with the Y-2 polypyrimidine stretch located 20 bases upstream of the AUG, creating a potential Y₃₉X₂₀-AUG like motif. However, its translation is severely impaired (Jaramillo-Mesa et al., 2019).

We first revisited the importance of the length of the spacer sequence between the CU-tract and the AUG. We compared the translation efficiency of the 1-700 truncated mutant in vitro to that of the wild-type full-length TriMV 5′ UTR (construct 1-739), the non-functional reverse complementary sequence (construct 739-1), and the IRES-impaired deletion mutant (construct 1-709; Jaramillo-Mesa et al., 2019; Figure 3). The 1-700 and 1-709 deletion mutants both lack the native YX-AUG motif and differ in that the AUG codons were positioned 11 and 20 bases downstream of the Y-2 segment, respectively. While 1-709 deletion mutant had 30% of translation activity remaining, the 1-700 deletion mutant strongly supported translation (Figure 3). This observed translation was dependent upon the integrity of both auxiliary CU-rich segments within that region. Altering the 1-700 mutant’s Y-2 tract sequence to random sequences (1-700: mY-2) dropped translation to 30% of wild type, similar to the 1-709 mutant, in line with the requirement of a functional YX-AUG motif for translation; mutating the Y-1 segment (1-700: mY-1) abolished translation. Taken together, the YX-AUG-driven translation is i) dependent upon a “good context spacer length,” likely due to limited scanning ability of the recruited ribosomes as we previously reported for the native TriMV YX-AUG motif (Jaramillo-Mesa et al., 2019) and as reported for the type II EMCV motif (Ohlmann and Jackson, 1999); and ii) dependent on the presence of an additional flanking polypyrimidine track. This is in line with the observed inability of the 1-648 deletion mutant, in which an additional CU-rich segment was missing, to drive translation (Figure 2B).

Secondly, we examined whether the presence of a cryptic AUG could modulate the YX-AUG-driven translation in plants in a similar fashion as reported in both Type I and Type II animal picornavirus IRESes (Hellen et al., 1994; Kaminski et al., 1994). We thus inserted an upstream AUG triplet at position nt 648 (uAUG₆₄₈), 18 nts downstream from the Y-1 motif in the context of the translationally impaired 1-709 deletion mutant (construct 1-709+ uAUG₆₄₈, Figure 4A). This created a chimeric “Y₂₂X₁₈-^aAUG-X₅₈^bAUG” motif that imitated the Type I PV Y₉X₁₈-^aAUG-X₁₅₄-^bAUG motif, in which the first ^aAUG triplet is located 18 nts from the Y motif and separated from a downstream ^bAUG codon by a long spacer sequence, which included the Y2 CU-rich tract (Hellen et al., 1994; Pestova et al., 1994). In the context of the Type I PV YX-AUG motif, the downstream AUG is designated as the initiation site and likely reached by ribosomal scanning (Hellen et al., 1994; Pestova et al., 1994). The inserted uAUG₆₄₈ codon in the TriMV IRES was placed in a similar coding frame and Kozak sequence context of the downstream AUG₇₀₉ codon of the luciferase gene. Remarkably, embedding that AUG right after the Y-1 tract fully restored translation of the deletion mutant 1-709 up to the wild-type level in vitro (constructs 1-709+ uAUG 648, Figure 4B). When we placed the uAUG₆₄₈ codon out of frame of the luciferase initiation site (1-709+ uAUG648 out of frame), translation of the luciferase was maintained up to 70% of the full-length wild-type level in vitro (Figure 4B). This revealed that the inserted uAUG remained cryptic, as intended, and the downstream AUG₇₀₉ remained the targeted start codon. In line with such a model, when we mutated the AUG₇₀₉ codon of the luciferase into a non-initiation AUC site, we abolished translation in the context of the in-frame chimeric YX1-AUG₆₄₈ construct (construct 1-709 + uAUG₆₄₈ + AUG₇₀₉/AUC; Figure 4B) further supporting the cryptic, but essential nature of the inserted AUG in restoring translation of 1-709 deletion mutant. Similar trend in translation was replicated in cells (Figure 4C).

To support the scanning ability of the ribosomal complex that landed on this chimeric YX-AUG₆₄₈ motif to reach the downstream AUG₇₀₉ codon, we extended the spacer sequence separating the two AUG codons from X₆₁ to X₁₅₄ to similar length of that in the PV motif “Y₂₂X₁₈-^aAUG-X₁₅₄^bAUG” (Figures 4A,D). For this purpose, we inserted 93 bases from the native PV spacer sequence. The relative frequency of initiation at this new position of the start codon was 70% of the wild-type level in vitro (Figure 4D), in line with a scanning-mediated ribosomal mechanism.

This observed translation reprogramming was dependent upon the Y-1 and Y-2 segments (Figure 4E). Swapping the Y-1 CU-rich tract with random sequences, and thus eliminating the chimeric YX-AUG motif, abolished the recovered translation (1-709: uAUG 648 + mY-1 construct, Figure 4E). Altering the Y-2 tract sequence to random sequences (1-709 uAUG 648 mY-2) dropped translation to 30% of wild type, similar to the 1-709 mutant, in line with the requirement of an intact YX-AUG motif and an auxiliary CU-rich tract for activity (Figure 4E).

All together our data supported that the assembly of the ribosomal complex and translation initiation can be reprogrammed at each of these upstream CU-rich tracts under specific conditions: the YX-AUG-driven translation is positionally dependent with the critical role of the spacer length separating the preferred AUG and the CU-rich tract, can be modulated by a cryptic AUG codon, and requires a flanking polypyrimidine track for its function.

The above observations led us to hypothesize that the Y-tract regions at upstream positions of the native YX-AUG motif may be the initial recruitment site of the ribosomal complexes onto the TriMV IRES element. To characterize the accessibility of the Y tract regions for such potential interaction, we mapped the overall secondary structure of the 739-nt TriMV 5′ UTR by SHAPE analysis (Figure 5A). The SHAPE-guided structure of the TriMV leader sequence revealed a compact structure with the configuration of the 3′ end of the UTR comprising two defined stem structures, which meet at a three-way junction with the YX-AUG motif located on a downstream unstructured segment. These stems correspond to the hairpin at position nts 469–488 that was previously identified to be crucial for eIF4G recruitment (Roberts et al., 2015, 2017a), and a large bifurcated stem structure at base position nts 514-677 that bears the two auxiliary Y-1 and Y-2 tracts alongside and that appears to protrude out of the 5′ UTR. The Y-motifs are each positioned on bulges and are largely unpaired.

The “free-to-interact” potential of the auxiliary Y-motifs was experimentally tested using anti-sense single-stranded DNA oligonucleotides designed to block the CU-tracts accessibility, in virtue of their sequence complementarity (Figure 5B). To this end, we added DNA oligonucleotides consisting of the sequence complementarity to each of the Y-motifs up to 100 mM to the translation reaction of the full-length wild-type TriMV 1-739 RNA construct. As a control, we included anti-sense DNA oligonucleotides with complementary to the core Y₁₆X₁₁-AUG₇₄₀ motif (anti-YX-AUG DNA), that we previously established to strongly and specifically interfere with TriMV IRES activity (Jaramillo-Mesa et al., 2019). We used unrelated human ß-globin oligo as a negative control (anti-B globin). Our result showed that, similarly to the effect of blocking the Y₁₆X₁₁-AUG₇₄₀ motif for potential binding (Jaramillo-Mesa et al., 2019), translation steadily decreased in the presence of increasing fold molar excess of the anti-sense DNA oligonucleotides targeting each of the auxiliary Y motifs. These observations are in line with the importance of the accessibility of the Y-regions and the YX-AUG motifs to support translation.

To next characterize the role of the auxiliary CU-rich tracts in ribosome recruitment on the TriMV IRES element, we set out to analyze the composition of the translation initiation complex specifically recruited by the Y-1 and Y-2 polypyrimidine tracks onto the viral 5′ UTR. To this end, we used (i) the TriMV 5′ UTR mutant in which the native YX-AUG motif was replaced with unrelated beta-globin sequence to maintain the same length of the 5′ UTR to that of a wild-type sequence. In this UTR construct, the Y1 and Y2 segments were intact (sample 1-739 + mYXAUG),; (ii) the mY-1+ mY-2 TriMV 5′ UTR mutant in which the native YX-AUG motif was intact but both Y1 and Y2 were mutated to random sequences (sample 1-739 + mY1 & mY2), and (iii) the reverse complementary sequence of the TriMV 5′ UTR (sample 739-1), as RNA baits for protein pull-down (Figure 6A). While each of those tested TriMV RNA sequences were functionally impaired for translation, our goal was to gain an insight on the pre-initiation complex that can be recruited by these RNA segments and their present motifs under stringent conditions. Each RNA segment bear at their 5′ end a stable hairpin structure to block 5′ entry of the ribosomes and to allow only internal recruitment of ribosomal complexes onto the RNA, and were biotinylated at their 3′ end. Native ribosomal complexes were let to assemble on those RNAs in the presence of non-hydrolyzable form of GTP, GMP-PNP, in the wheat germ translation reaction and next isolated and eluted through streptavidin beads. The GMP-PNP would arrest translation initiation at the formation of 48S ribosomal complex upon recognition of the preferred AUG and block the joining of the 60S large ribosomal subunit to the complex. The eluted complexes were next analyzed by mass spectrometry. This strategy would determine the enrichment of small and/or large ribosomal subunit proteins and their associated translation factors to each RNA. In particular, we set to compare any ribosomal complex enrichment with the TriMV 5′ UTR sequence in the presence or in the absence of the auxiliary CU-rich tracts (Figure 6B). We first confirmed that no translation was observed with our wild-type TriMV luciferase RNA construct in the presence of the GMP-PP (Supplementary Figure 2). The mass spectrometry analysis of two independent experiments identified a total of 455 host-interacting proteins clusters, including 142 uncharacterized proteins (Supplementary Figure 3 and Supplementary Table 1). We next addressed the composition of the translation initiation complex formed on the translationally impaired TriMV IRES elements that either missed the YX-AUG motif or the auxiliary CU-rich tracts (Figure 6B). The mass spec analysis indicated a strong enrichment of 60S ribosomal related proteins and 40S ribosomal related proteins in the eluate fraction from the mutant RNA that misses the YX-AUG motif but has intact CU-rich tracts (sample 1-739 + mYXAUG; Figure 6B) when compared to the fraction bound to the mutant that misses the CU-rich tracts but bears the YX-AUG motif, and the non-functional 739-1 RNA (sample 1-739 + mY-1 & m-Y2; Figure 6B). Additionally, some translation initiation factors mainly associated to the 40S ribosomal complex, including the eIF3 subunits, were predominantly recovered with the 1-739 + mYXAUG mutant (Supplementary Figure 3B and Supplementary Table 1). All of these observations are in line with a ribosome clustering function of the auxiliary CU-rich tracts.

The potential role of the auxiliary Y-motifs in ribosome recruitment and their ability to functionally act as chimeric YX-AUG motifs led us to test whether we could re-designate the preferred start codon to upstream positions in the context of the full-length TriMV IRES. We thus inserted an uAUG at position 701 in the context of the full-length TriMV UTR and in frame with the downstream luciferase start codon (Figure 7A). In this context, we created a chimeric Y₃₉X₁₁-AUG₇₀₁ motif followed by the original Y₁₆X₁₁-AUG₇₄₀ motif that embeds the luciferase initiation site. Both AUGs are located 11 bases downstream of the closest polypyrimidine segments and in the same Kozak context. We hypothesized that the ribosomal complexes would linearly select the closest AUG codon embedded within the chimeric YX-AUG₇₀₁ motif over the downstream native YX-AUG₇₄₀ motif to initiate translation. We compared the in vitro translation efficiency of this mutant (construct 1-739 + AUG₇₀₁) to that of the wild-type full-length TriMV 5′ UTR (construct 1-739) and the non-functional reverse complementary sequence (construct 739-1). Our result revealed that in the presence of the uAUG₇₀₁ codon, translation of the luciferase was 70% of the wild-type level (construct 1-739 + AUG₇₀₁). However, translation mainly occurred at the downstream native AUG₇₄₀ codon. When the uAUG₇₀₁ codon was followed by a stop codon, translation of the luciferase remained to similar level. Mutating the downstream AUG₇₄₀ of the luciferase gene into an AUC codon abolished translation (Figure 7B). In sum, although the chimeric YX-AUG₇₀₁ motif appears to be in good context for initiation, it was bypassed to favor preferential translation at the downstream native YX-AUG₇₄₀ site, suggesting a non-linear mechanism of YX-AUG recognition and/or a better accessibility of the YX-AUG₇₄₀ site in the context of the native UTR.