40S ribosomal subunits scan mRNA for the start codon by one-dimensional diffusion

When 40S scanning is rate-limiting, mRNAs with longer 5′ UTRs are expected to be translated less efficiently than mRNAs with shorter 5′ UTRs

To analyze putative experimental outcomes of examining dependence of protein synthesis on 5′ UTR length using GFP reporters, we considered a simple mathematical approximation: The amount of translated protein is inversely proportional to the time needed to synthesize one GFP molecule. The time needed to synthesize one GFP molecule accounts for the duration of 40S scanning, which depends on the length of the 5′ UTR (L), and sum of all other steps of initiation, elongation, and termination, which do not depend on the 5′ UTR length ( in Eqs. 1 and 2). In this simplified approximation, mRNA half-lives and GFP protein degradation rates are assumed to be constant, independent of 5′ UTR length and unaffected by the overall rate of GFP synthesis. The duration of 40S scanning should scale linearly with 5′ UTR length if scanning is driven by a translocase/helicase (Vassilenko et al. 2011): (1) where T = total time, L = length of the 5′ UTR, k_tr = rate of 40S translation along the 5′ UTR, and τ_j = duration of N translation steps other than 40S scanning, including other steps of initiation such as 40S recruitment that are independent of 5′ UTR length.

Alternatively, scanning time should scale as the square of 5′ UTR length if scanning occurs by diffusion (Vassilenko et al. 2011): (2) where D = 40S diffusion coefficient.

The reciprocal relationship between the amount of synthesized protein and 5′ UTR length is expected for both translocation and diffusional mechanism of scanning (Fig. 1). In the experiments reported in this manuscript involving variation in 5′ UTR length, the term in Equations 1 and 2 is expected to be a constant.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 1.

Modeled dependencies of GFP synthesis on the length of the 5′ UTR for the helicase-driven (A, Eq. 1) and 1D diffusion (B, Eq. 2) mechanisms of 40S scanning. Amounts of synthesized GFP are inversely proportional to 5′ UTR scanning time and sum of all other steps of initiation, elongation, and termination, which do not depend on the 5′ UTR length. GFP synthesis is modeled for scenarios when the sum of all other steps of translation, which are independent of 5′ UTR length, are 10-fold faster (> black), equal to (= red), 10-fold slower (< blue), 100-fold slower (<< cyan), or 1000-fold slower (<<< magenta) than the scanning of the 50 nt-long 5′ UTR.

Both translocation and 1D diffusion models predict that if scanning is much faster than the sum of all other steps of protein synthesis, including other steps of initiation, protein synthesis is expected to be nearly independent of 5′ UTR length (Fig. 1). Conversely, in the case of 40S scanning being comparable to or even slower than the sum of all other steps of translation, protein synthesis is expected to show strong dependence on the 5′ UTR length (Fig. 1). Under these conditions, it might be possible to distinguish between translocation and diffusion mechanisms of scanning by fitting the dependence of GFP synthesis on the 5′ UTR length to Equations 1 and 2. However, this may be challenging given similarities in shapes of translocation and 1D curves (Fig. 1).

Another prediction that can be inferred from these simplified models is that a reduction of scanning rate (e.g., via mutation) is expected to more significantly affect translation of mRNAs with longer 5′ UTRs. On the contrary, slowing down any other step of translation is expected to have a stronger relative effect on translation of mRNAs with shorter 5′ UTRs because the scanning-independent steps of translation make larger contributions to the time needed to synthesize one GFP molecule on mRNAs with shorter 5′ UTRs than on mRNAs with longer 5′ UTRs.

Lengthening of the 5′ UTR through addition of UCC/ACCAC inhibitory sequences substantially reduces translation

To experimentally test how the length and secondary structure of the 5′ UTR affect translation and which protein factors are critical for scanning in live cells, we used the RNA-ID reporter system (Fig. 2A; Dean and Grayhack 2012). The RNA-ID reporter provides a robust and sensitive way to examine the effects of cis-regularity mRNA sequences on translation in yeast cells. The RNA-ID DNA construct contains the ORFs of superfolder GFP and the yeast codon-optimized red fluorescent protein variant of mCherry under control of the bidirectional GAL1,10 promoter, with the MET15 or URA3 gene used for selection in yeast, and flanking sequences allowing for the integration into the ADE2 locus of the yeast genome via homologous recombination. GFP levels are measured by GFP fluorescence normalized by the fluorescence of RFP to account for differential activation of the promoter in different cells.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

Lengthening unstructured 5′ UTR enriched with UCC and ACCAC sequences progressively reduces translation of GFP reporter in yeast cells. (A) Schematic depiction of RNA-ID reporter that includes ADE2 loci for integration into the yeast genome, URA3 or MET15 selection marker gene (details in Materials and Methods), Gal 1,10 bidirectional promoter, RFP, and sfGFP ORFs. (B) Design of GFP mRNAs containing unstructured 5′ UTRs (NUS′) of different lengths. The constructs share the same NUS sequence but differ in how much of the NUS sequence is included from the 5′ end. (C) Mean GFP/RFP fluorescence ratios measured in yeast cells, which were transformed with different RNA-ID 5′ UTR NUS′ constructs as indicated. GFP/RFP ratio in GO RNA-ID with 6 nt-long 5′ UTR was set to 1. Spearman rank correlation coefficient R (Virtanen et al. 2020) between length and GFP/RFP ratio was determined to be −0.915 (P < 0.0001). (D) Mean GFP/RFP fluorescence ratios normalized to the ratio of GFP and RFP mRNA levels, which were determined by RT-qPCR. Error bars show standard deviations determined from three biological replicates.

In all RNA-ID constructs, the ORF of superfolder GFP was flanked by the 3′ UTR from yeast RPL41B mRNA (Fig. 2A). RPL41B mRNA is one of the most well-translated mRNAs in yeast cells; the 3′ UTRs of this mRNA were previously precisely identified using rapid amplification of cDNA ends (RACE) (Yu and Warner 2001). To examine the effect of the 5′ UTR length on translation, upstream of the GFP ORF, we introduced intrinsically unstructured sequences ranging from 50 to 500 nt in length in 50 nt increments at the 5′ end of the 5′ UTR while preserving the original Kozak context of the start codon (Fig. 2B; Supplemental Table S1). These nonrepetitive unstructured sequences (NUS) were designed with the previously developed genetic algorithm (Lai et al. 2018). The genetic algorithm evolves a sequence through an iterative sequence alteration/selection process that eliminates all alternative base-pairing interactions, including those pairs that the 5′ UTR nucleotides can form with the entire mRNA sequence including the ORF and 3′ UTR. A single 500 nt NUS was designed, and individual constructs using differing lengths were created by truncation of the 5′ end. Therefore, these constructs share the sequence upstream of the Kozak sequence and the start codon. Importantly, in contrast to CA and CAA repeats commonly used to eliminate RNA secondary structure, NUSs are more complex and more easily propagated in Escherichia coli and yeast cells. Base-pairing probabilities estimated by the RNAstructure software package predict that NUS nucleotides are nearly completely unpaired while most nucleotides in the ORF and 3′ UTR have high (>75%) base-pairing probability (Supplemental Fig. S1).

Resulting RNA-ID constructs (named ⁵⁰NUS′-⁵⁰⁰NUS′ where superscript numbers indicate 5′ UTR length) were transformed into BY4741 haploid strain of Saccharomyces cerevisiae. GFP/RFP ratio for each construct was measured by flow cytometry (Supplemental Fig. S2) and normalized to GFP/RFP levels observed in the RNA-ID GO control strain, in which the GFP mRNA contains a 6 nt-long 5′ UTR (Supplemental Table S1). Extending the 5′ UTR progressively reduced GFP synthesis until it leveled off at the length of 350–400 nt (Fig. 2C). Normalizing GFP/RFP fluorescence by the GFP/RFP ratio of mRNA levels, which were determined by RT-qPCR, indicated that the observed reduction in GFP synthesis was due to a decrease in translational efficiency rather than changes in mRNA levels (Fig. 2D). Translation of GFP mRNAs with 350–500 nt-long unstructured 5′ UTRs was approximately eight times lower on average than the translation of mRNA with a 50 nt-long 5′ UTR. These data showed considerable dependence of translation on the 5′ UTR length, which could be due to either length-dependent or sequence-specific effects. Indeed, further analysis of ⁵⁰NUS′-⁵⁰⁰NUS′ sequences revealed that extending the 5′ UTR from 50 to 500 nt increasingly enriched NUS sequences with C-rich motifs UCC and ACCAC (Supplemental Fig. S3). These C-rich sequences were recently reported to be inhibitory for translation initiation (Niederer et al. 2022).

Order-of-magnitude variations in the length of unstructured 5′ UTR depleted of UCC/ACCAC inhibitory sequences have only modest effects on translation in yeast cells

To test whether observed reduction of protein synthesis in ⁵⁰NUS′-⁵⁰⁰NUS′ constructs was due to the presence of inhibitory C-rich sequences or increased 5′ UTR length, we computationally designed a new set of unstructured 5′ UTRs (⁵⁰NUS1-⁵⁰⁰NUS1) ranging in length from 50 to 500 nt, in which UCC and ACCAC sequence motifs were avoided (Supplemental Table S1). Because sequences adjacent to the 5′ cap may influence 40S recruitment, we extended NUS1 UTRs with 50 nt increments at the 3′ end while keeping the 5′ end constant (Fig. 3A). As expected, RFP fluorescence remained nearly constant between strains transformed with different RNA-ID NUS1 constructs while extending the 5′ UTR in GFP mRNA moderately reduced GFP synthesis (Fig. 3B–D). Transforming ⁵⁰NUS1-⁵⁰⁰NUS1 constructs into the BY4741 and BY4742 strains produced similar results (Supplemental Fig. S4). Translation of ⁵⁰NUS1 was only twofold more efficient than translation of ⁵⁰⁰NUS1 (Fig. 3C–E). Hence, the larger length-dependent reduction of translational efficiency observed in ⁵⁰NUS′-⁵⁰⁰NUS′ constructs seems likely to be due to the presence of UCC/ACCAC inhibitory sequences or, possibly, to effects of 40S recruitment caused by varying the sequence adjacent to the 5′ cap. In any case, with the NUS1 constructs, translation efficiency is relatively independent of the 5′ UTR length.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

Tenfold lengthening of unstructured 5′ UTR moderately reduces translation of GFP reporter in yeast cells. (A) Design of GFP mRNAs containing unstructured 5′ UTRs (NUS1 series) of different lengths. (B) Mean GFP and RFP fluorescence measured in yeast cells, which were transformed with different RNA-ID constructs as indicated. (C) Mean GFP fluorescence in strains expressing RNA-ID mRNA with unstructured 5′ UTRs of different lengths normalized by GFP fluorescence produced in the GO RNA-ID control reporter with a 6 nt-long 5′ UTR. (D) Mean GFP/RFP fluorescence ratios. GFP/RFP ratio in the GO RNA-ID control was set to 1. Spearman rank correlation coefficient R (Virtanen et al. 2020) between length and GFP/RFP ratio was determined to be −0.841 (P < 0.0001). (E) Mean GFP/RFP fluorescence ratios normalized to the ratio of GFP and RFP mRNA levels, which were determined by RT-qPCR. NUS1 and NUS2 are independently designed, divergent, nonrepetitive, unstructured sequences. Error bars show standard deviations determined from three biological replicates.

To further explore how sequence variations in the unstructured 5′ UTR affect translation, we independently designed another 300 nt-long 5′ UTR (³⁰⁰NUS2), in which UCC and ACCAC sequences were also avoided. ³⁰⁰NUS1 and ³⁰⁰NUS2 shared only 57.3% identity in a global pairwise alignment calculated by the EMBOSS Stretcher alignment tool and 58.8% identity in a local alignment calculated by the EMBOSS Water alignment tool (Madeira et al. 2019). Similar sequence identity (mean of 56.9% and standard deviation 2.1%) was found in the local alignment of 100 computationally shuffled (i.e., randomized) ³⁰⁰NUS1 and ³⁰⁰NUS2 sequences. GFP/RFP ratio in ³⁰⁰NUS1 and ³⁰⁰NUS2 were similar (Fig. 3C–E), indicating that as long as inhibitory motifs were eliminated from the design, other sequence variations had minimal effect on translation. We thus focused the rest of our experiments on the ⁵⁰NUS1-⁵⁰⁰NUS1 constructs.

Normalizing GFP/RFP ratio by mRNA levels, which were determined by RT-qPCR, showed that observed changes in GFP synthesis in ⁵⁰NUS1-⁵⁰⁰NUS1 strains were due to modest reductions of translational efficiency rather than changes in mRNA levels (Fig. 3E). It is noteworthy that GFP/RFP ratios (Fig. 3D,E) and GFP levels alone (Fig. 3C) showed similar dependencies on 5′ UTR length. Hence, the transcriptional activity of the Gal1,10 promoter is effectively constant in different cells, and GFP fluorescence alone (without normalizing to RFP levels) can be reliably used to analyze changes in translational efficiency of GFP reporter constructs.

To test whether in yeast cells, ⁵⁰NUS1-⁵⁰⁰NUS1 mRNAs are transcribed from the previously identified transcription start site of the GAL1,10 promoter, we performed 5′ RACE analysis for ⁵⁰NUS1, ³⁰⁰NUS1, and ⁵⁰⁰NUS1 mRNAs. 5′ RACE PCR products were of expected size and sequence (Supplemental Fig. S5), consistent with the predicted transcription start. SDS-PAGE analysis combined with GFP fluorescence imaging revealed a single fluorescent GFP protein band in all ⁵⁰NUS1-⁵⁰⁰NUS1 strains, indicating that only one ORF, which produced fluorescent GFP, was predominately translated (Supplemental Fig. S6). Hence, no alternative near-cognate, in-frame start sites were used for translation initiation in ⁵⁰NUS1-⁵⁰⁰NUS1 GFP mRNAs.

RNA secondary structures in the 5′ UTR strongly inhibit translation

Our experiments with ⁵⁰NUS1-⁵⁰⁰NUS1 reporter constructs showed that extending the 5′ UTR 10-fold over the median length of yeast 5′ UTRs only modestly reduces protein synthesis. Therefore, 40S scanning along an unstructured 5′ UTR seems quite effective and does not hamper translation initiation. We next explored the effects of 5′ UTR secondary structure on translational efficiency and 40S scanning. To that end, we replaced nucleotides 156–199 of ³⁰⁰NUS2 5′ UTR with a GC-rich stem–loop containing 20 bp and a UUAA tetraloop (Fig. 4A). When compared to the ³⁰⁰NUS2 5′ UTR, the ³⁰⁰NUS2 5′ UTR with the cap distal stem–loop (³⁰⁰NUS2-SL) produced ∼150-fold less GFP (Fig. 4B). When adjusted by mRNA levels measured by RT-qPCR, the stem–loop reduced GFP translation by 85-fold (Fig. 4C). Likewise, we replaced 44 nt 150 nt downstream from the 5′ end of ³⁰⁰NUS′ and ⁵⁰⁰NUS′ 5′ UTRs with a cap-distal stem–loop containing 20 bp and a tetraloop. The insertion of this stem–loop into ³⁰⁰NUS′ and ⁵⁰⁰NUS′ 5′ UTRs inhibited GFP synthesis by 30- and 10-fold, respectively (Supplemental Fig. S7). Hence, consistent with many published reports (Pelletier and Sonenberg 1985; Kozak 1986, 1989a; Vega Laso et al. 1993; Babendure et al. 2006; Sen et al. 2015), stable RNA secondary structure strongly inhibits translation, presumably by hindering 40S scanning. These results also indicate that ⁵⁰NUS-⁵⁰⁰NUS (as well as ⁵⁰NUS′-⁵⁰⁰NUS′) mRNA constructs are translated in a cap- and scanning-dependent manner rather than through a noncanonical, cap-independent mechanism (e.g., IRES-dependent initiation).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

RNA secondary structure in the 5′ UTR strongly inhibits translation. (A) Fifty-four nucleotides in the middle of ³⁰⁰NUS2 5′UTR were replaced with a 20 bp-long RNA stem–loop to create ³⁰⁰NUS2 + SL mRNA. Base-pair probabilities determined by the RNAstructure software are shown by color code as indicated. (B,C) Introducing the stem–loop (SL) into the ³⁰⁰NUS2 5′ UTR reduced GFP synthesis. (D–F) A 122 nt-long sequence (¹²²Struct) in the 5′ UTR (gray box in schematics D), which is predicted to fold into multiple stem–loops, inhibits GFP synthesis. (B,E) Mean GFP fluorescence was normalized to that in the GO RNA-ID reporter with a 6 nt-long 5′ UTR. (C,F) Mean GFP/RFP fluorescence ratios normalized to the ratio of GFP and RFP mRNA levels, which were determined by RT-qPCR. GFP/RFP ratio in the GO RNA-ID control was set to 1. Error bars (B,C,E,F) show standard deviations determined from three biological replicates.

Next, instead of introducing a very stable stem–loop into an unstructured 5′ UTR, we replaced the entire 5′ UTR with a 122 nt-long sequence that is derived from the 5′ UTR of human GAPDH mRNA to make the ¹²²Struct construct. The RNAstructure software package predicts that most nucleotides of this sequence have between 50% and 75% probability of base-pairing (Supplemental Figs. S8, S9). Rather than forming one very stable secondary structure, the ¹²²Struct sequence likely folds into a dynamic ensemble of alternative structures. When compared to the ¹⁰⁰NUS1 and ¹⁵⁰NUS1 strains, the ¹²²Struct strain produced five- and sixfold less GFP, respectively (Fig. 4E). The GFP/RFP ratio normalized to mRNA levels in ¹²²Struct strain was two- and threefold lower than that in the ¹⁰⁰NUS1 and ¹⁵⁰NUS1 strains, respectively (Fig. 4F), indicating that a structured 5′ UTR substantially reduces translation relative to an unstructured 5′ UTR of similar length.

Because the ¹²²Struct 5′ UTR is predicted to contain base pairs near the 5′ end (Supplemental Figs. S8, S9), secondary structures of the ¹²²Struct 5′ UTR may inhibit both initial 40S recruitment and following 40S scanning (Babendure et al. 2006). We next aimed to eliminate the effects of secondary structure on 40S recruitment. As the 43S initiation complex covers ∼45 nt of mRNA (Kozak and Shatkin 1977; Brito Querido et al. 2020), we added 50 nt-long unstructured sequences NUS3 or NUS4 upstream of the ¹²²Struct sequence that would allow for unhindered 40S binding. Surprisingly, adding the NUS3 or NUS4 unstructured sequence reduced GFP levels 10- or 14-fold, respectively, relative to GFP levels in the ¹²²Struct 5′ UTR strain (Fig. 4E). GFP/RFP ratios normalized to mRNA levels in NUS3-¹²²Struct and NUS4-¹²²Struct strains were ninefold lower than that in the ¹²²Struct 5′ UTR strain (Fig. 4F). One possible explanation of these results is that 40S binding to the original ¹²²Struct mRNA partially unwound secondary structures of the 5′ UTR and thus aided 40S scanning. 40S recruitment to the unstructured sequence (NUS3 or NUS4) upstream of the ¹²²Struct preserved all secondary structure elements within the ¹²²Struct sequence and thus reduced effectiveness of 40S scanning on NUS3–¹²²Struct and NUS4–¹²²Struct mRNAs. Taken together, experiments with the ³⁰⁰NUS2-SL and ¹²²Struct mRNAs demonstrate that secondary structure elements within the 5′ UTR strongly inhibit protein synthesis by impeding 40S scanning.

Translational helicase eIF4A is not involved in 40S movement along the 5′ UTR

Slowing 40S scanning is expected to more substantially affect mRNAs with longer 5′ UTRs (Fig. 1). We next used mutants to test whether mRNAs with long unstructured 5′ UTRs are particularly sensitive to genetic perturbations of translation factors, which were previously implicated in 40S scanning. DEAD-box helicase eIF4A is a subunit of the 5′ cap recognition complex eIF4F. The helicase activity of eIF4A is thought to be directly involved in ATP-dependent 40S movement along the 5′ UTR (Kozak 1980; Spirin 2009; Brito Querido et al. 2024). In yeast, eIF4A is encoded by two paralogs TIF1 and TIF2. To test involvement of eIF4A in 40S scanning, RNA-ID constructs were transformed into either tif1Δ or tif2Δ deletion strains created in a BY4741 genetic background. RT-qPCR measurements showed an approximately twofold decrease in eIF4A mRNA levels in both tif1Δ and tif2Δ strains when compared to the parent (wild-type) BY4741 strain (Supplemental Fig. S10). Nevertheless, GFP synthesis on the ^50-500NUS1 and ¹²²Struct mRNAs was essentially unaffected by either tif1 or tif2 deletion (Fig. 5A). Likewise, GFP synthesis on the ^50-500NUS′ mRNAs was unchanged in tif1Δ or tif2Δ strains (Supplemental Fig. S11A). This could be due to high expression levels of eIF4A, which is the most abundant factor of translation initiation in yeast cells (Kulak et al. 2014). Consistent with the observation that protein synthesis was unaffected by the deletion of either tif1 or tif2, growth of tif1Δ or tif2Δ strains on YPD plates was similar to the parental, wild-type (BY4741) strain (Supplemental Fig. S12).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 5.

Mutational perturbations of eIF4F subunits similarly reduce translation of GFP NUS1 mRNAs with short and long unstructured 5′ UTRs. GFP fluorescence measured in mutant strains normalized to respective GFP fluorescence measured in the wild-type strain. Error bars indicate standard deviations determined from three biological replicates. (A) eIF4A-encoding tif1, tif2 or eIF4B-encoding tif3 were deleted in BY4741 background as indicated. (B) Either wild-type (TIF1) or temperature-sensitive (tif1-A79V) alleles of the eIF4A-encoding tif1 were expressed in the strain lacking both chromosomal copies of tif1 and tif2. (C) An amino acid substitution L614F, which reduces eIF4G binding to eIF4A, was introduced into tif4631 in the background of tif4632 deletion to create the temperature-sensitive tif4632Δtif4631-L614F strain. (D) An amino acid substitution M570K, which reduces eIF4G binding to eIF4A, was introduced into tif4632 in the background of a tif4631 deletion to create temperature-sensitive tif4631Δtif4632-M570K strain. BY4741 strain is considered as wild type in A, C, and D. In experiments with temperature-sensitive strains tif2Δtif1-A79V and tif4632Δtif4631-L614F, the Gal 1,10 promoter was induced concurrently with raising the incubation temperature from 30°C to 37°C. Respective wild-type control strains received identical treatment.

Because eIF4A is essential for yeast viability, we could not delete both TIF1 and TIF2 genes simultaneously. Instead, we integrated RNA-ID constructs into the strain lacking both TIF1 and TIF2 genes and expressing the temperature-sensitive allele tif1-A79V (Sen et al. 2015). A79V substitution is thought to weaken eIF4A binding to RNA (Schmid and Linder 1991; Linder and Jankowsky 2011; Sen et al. 2015) and severely hamper yeast growth at 37°C (Supplemental Fig. S13). To examine the effects of the temperature-sensitive A79V mutation in eIF4A, GFP expression was induced by the addition of galactose concurrently with switching from 30°C to 37°C. In the absence of galactose, negligible levels of GFP were accumulated, indicating tight transcriptional control of the Gal 1,10 promoter (Supplemental Fig. S14). When compared at 37°C to the strain in which the wild-type TIF1 allele was expressed, expressing tif1-V79A decreased GFP levels fourfold in all ^50-500NUS1 constructs irrespective of 5′ UTR length (Fig. 5B). Similar uniform reduction of GFP synthesis was observed for all the ^50-500NUS′ mRNAs (Supplemental Fig. S11B). Translation on the ¹²²Struct mRNA containing a 122 nt-long structured 5′ UTR was reduced sixfold, that is, fairly similar to translation of mRNAs with unstructured 5′ UTRs (Fig. 5B). This result is consistent with published data indicating that eIF4A stimulates translation of mRNAs regardless of the structural complexity of their 5′ UTR (Yourik et al. 2017).

Helicase activity of eIF4A is stimulated by another initiation factor, eIF4B, which is encoded by a single-copy gene TIF3. We next deleted the nonessential gene TIF3 in a BY4741 background. In the tif3Δ strain, translation of GFP mRNAs with 50–200 nt-long unstructured 5′ UTRs was reduced twofold, while translation of GFP mRNAs with 300–500 nt-long unstructured 5′ UTRs was unaffected (Fig. 5A). Similar results were obtained with the ^50-500NUS′ reporters (Supplemental Fig. 11A). Higher sensitivity of mRNAs with shorter 5′ UTRs to TIF3 deletion may indicate involvement of this protein in the 40S recruitment step of translation initiation (Eqs. 1 and 2). Remarkably, the ¹²²Struct mRNA, containing a 122 nt-long structured 5′ UTR, was far more sensitive to eIF4B deletion than mRNAs with unstructured 5′ UTRs as GFP levels decreased sixfold in the tif3Δ strain. This observation agrees with ribosome profiling data showing that eIF4B stimulates translation of mRNAs with structured 5′ UTRs (Sen et al. 2016).

eIF4A helicase and ATPase activities are stimulated by interaction with another subunit of the eIF4F complex, eIF4G (Rogers et al. 2002; Schutz et al. 2008; Rajagopal et al. 2012). In yeast, eIF4G is encoded by two paralogs, TIF4631 and TIF4632. When we deleted the TIF4632 gene, translation of all RNA-ID constructs was unaffected (Fig. 5C). We next introduced an amino acid substitution L614F into tif4631 in the presence of tif4632Δ. The L614F mutation weakens the interaction of eIF4G with eIF4A and produces a temperature-sensitive phenotype, which can be suppressed by overexpression of eIF4A or eIF4B (Supplemental Fig. S13; Park et al. 2013). When measured at 37°C, the L614F mutation in tif4631 reduced GFP synthesis of the ⁵⁰NUS1, ⁵⁰⁰NUS1, and ¹²²Struct mRNAs by four-, seven- and fourfold, respectively (Fig. 5C). Relative to tif4632Δ strain, GFP levels in tif4632Δ tif4631-L614F strain transformed with either the ⁵⁰NUS′ or ⁵⁰⁰NUS′ RNA-ID reporter were equally reduced by threefold (Supplemental Fig. 11C).

Because mRNAs with longer NUS1 5′ UTRs were slightly more sensitive to tif4631-L614F mutation than mRNAs with shorter 5′ UTRs, we also made reciprocal mutational perturbations in tif4632 and tif4631 to further examine involvement of eIF4A–eIF4G interaction in scanning. In that, we deleted tif4631 and introduced an amino acid substitution M570K, which weakens the interaction of eIF4A with eIF4G (Park et al. 2013), into tif4632. Deletion of tif4631 lowered GFP levels fivefold irrespective of 5′ UTR length in both the ^50-500NUS1 and ^50-500NUS′ reporters (Fig. 5D; Supplemental Fig. S11D). Mutation M570K in tif4632 further reduced GFP levels 1.5- to twofold in all 5′ UTR NUS1 and NUS′ reporters (Fig. 5D; Supplemental Fig. S11D). Relative to the tif4631Δ strain, translation of the ¹²²Struct mRNA decreased by 1.2-fold in the tif4631Δ tif4632-M570K strain (Fig. 5D).

Taken together, our results indicate that amino acid substitutions disrupting the eIF4A–eIF4G interaction similarly affected GFP reporters with short and long 5′ UTRs. Hence, the eIF4F complex as well as eIF4A alone are unlikely to be responsible for 40S movement along the 5′ UTR.

Loss-of-function mutations in eIF3g and eIF3i moderately affect translation of mRNAs with short and long unstructured 5′ UTRs

We next probed whether initiation factors eIF3g and eIF3i, which are encoded by yeast genes TIF35 and TIF34, respectively, affect 40S scanning. These two subunits of multimeric initiation factor 3 bind near the mRNA entry site of the 40S subunit, that is, the “front” side of the small subunit in respect to the direction of scanning. eIF3g and eIF3i were shown to stimulate ATPase activity of eIF4A (Yourik et al. 2017). In a recent cryo-EM reconstruction of the 48S initiation complex, two eIF4A molecules were visualized. One was bound at the mRNA entry site; the other was bound to the eIF4G•eIF4E•5′cap near the “rear” side of the small subunit (Brito Querido et al. 2024). eIF3i was seen making extensive interactions with eIF4A bound to the mRNA entry site, providing further evidence for possible eIF3i involvement in helicase and translocase activities of the 43S initiation complex.

In yeast, triple alanine substitution K194A/L235A/F237A of conserved residues in the RRM of eIF3g (tif35-KLF) or a single-point mutation Q258R in the WD40 repeat 6 of eIF3i (tif34-Q258R) was shown to produce growth defects (Cuchalova et al. 2010). These mutations also affected the efficiency of reinitiation on the GCN4 mRNA, which contains four short uORFs upstream of the main ORF (Cuchalova et al. 2010). Based on these observations, it was hypothesized that eIF3g and eIF3i may be involved in 40S scanning (Cuchalova et al. 2010). We re-created these eIF3g and eIF3i mutations in the BY4741 genetic background (Supplemental Fig. S13). When compared to BY4741, in the temperature-sensitive tif35-KLF strain, GFP expression decreased two- to 2.5-fold in all ^50-500NUS1 RNA-ID constructs irrespective of 5′ UTR length (Fig. 6A), showing no evidence of eIF3g involvement in 40S scanning. Translation on the ¹²²Struct mRNA was reduced 1.5-fold, similar to translation of mRNAs with unstructured 5′ UTRs (Fig. 6A).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 6.

Mutations in eIF3g and eIF3i similarly reduce translation of GFP NUS1 mRNAs with short and long unstructured 5′ UTRs. Amino acid substitutions were introduced in eIF3g-encoding tif35 (A) or eIF3i-encoding tif34 (B) in the BY4741 genetic background. GFP fluorescence measured in mutant strains was normalized to the respective GFP fluorescence measured in the wild-type (BY4741) strain. Error bars indicate standard deviations determined from three biological replicates. In experiments with the temperature-sensitive tif35-KLF strain, the Gal 1,10 promoter was induced concurrently with raising the incubation temperature from 30°C to 37°C for both tif35-KLF and wild-type cells.

Amino acid substitution Q258R in eIF3i affected translation of GFP mRNAs with shorter unstructured 5′ UTRs somewhat more significantly than translation of mRNA with longer 5′ UTRs: GFP expression on the ⁵⁰NUS1 and ⁵⁰⁰NUS1 mRNAs decreased four and two times, respectively (Fig. 6B). In contrast to the moderate reduction of GFP synthesis on mRNAs with unstructured 5′ UTRs, Q258R substitution in eIF3i diminished translation of the ¹²²Struct mRNA 15-fold (Fig. 6B). Hence, eIF3i may be involved in the unwinding of secondary structure in 5′ UTRs rather than 40S movement along mRNA itself.

Translational helicases Ded1 and Slh1 are not involved in 40S movement along the 5′ UTR

We next examined involvement of two translational helicases Ded1 and Slh1 in 40S movement along the 5′ UTR. Ded1 (yeast ortholog of mammalian DDX3X) is a DEAD box helicase that is implicated in unwinding stable stem–loops in 5′ UTRs during translation initiation (Sharma and Jankowsky 2014; Sen et al. 2015). We transformed RNA-ID constructs into a yeast strain containing the cold-sensitive allele ded1-120 (Supplemental Fig. S13), which encodes Ded1 bearing amino acid substitutions G108D and G494D (Sen et al. 2015). These mutations are thought to impair ATP binding or hydrolysis (Linder and Jankowsky 2011). GFP expression in the cold-sensitive ded1-120 strain (Supplemental Fig. S13) was induced by the addition of galactose simultaneously with the temperature shift from 30°C to 18°C. When compared to GFP expression in the BY4741 strain at 18°C, translation of all reporter mRNAs with unstructured 5′ UTRs in ded1-120 strain was reduced (Fig. 7A). Contrary to the idea of Ded1 involvement in 40S translocation during scanning (Berthelot et al. 2004), mRNAs with shorter unstructured 5′ UTRs were more significantly affected as GFP expression on ⁵⁰NUS1 and ⁵⁰⁰NUS1 mRNAs decreased seven- and threefold, respectively (Fig. 7A). Similar results were obtained with ^50-500NUS′ reporters (Supplemental Fig. S11E).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 7.

Translational helicases Ded1 and Slh1 are not essential for 40S scanning. GFP fluorescence measured in mutant strains normalized to respective GFP fluorescence in the wild-type (BY4741) strain. Error bars indicate standard deviations determined from three biological replicates. (A) NUS1 RNA-ID constructs were transformed into a ded1-120 strain containing cold-sensitive allele ded1-120, which encodes the mutant form of Ded1 helicase bearing the amino acid substitutions G108D and G494D. (B) NUS1 RNA-ID constructs were transformed into the slh1Δ or asc1Δ strains, helicase Slh1 and ribosomal protein Asc1, respectively. In A, GFP synthesis was induced concurrently with the temperature shift from 30°C to 18°C.

Consistent with the importance of Ded1-mediated unwinding of mRNA secondary structure (Sharma and Jankowsky 2014; Sen et al. 2015; Gupta et al. 2018), translation of the ¹²²Struct mRNA containing a structured 5′ UTR diminished nearly 20-fold in ded1-120 cells (Fig. 7A). Taken together, our data support the model suggesting that Ded1 is critical for melting secondary structures but not the actual movement of the 40S along the 5′ UTR.

Recently, another translational helicase, ski2-like ASCC3, was linked to 40S scanning, as ASCC3 knockdown lowered translational efficiency and 40S occupancy along the 5′ UTR in mammalian cells (Kito et al. 2023). ASCC3 and its yeast ortholog Slh1 are known to disassemble stalled 80S ribosomes in the ribosome-associated quality control (RQC) pathway (Filbeck et al. 2022). Noteworthy, ASCC3 was shown to be a 3′-to-5′ translocase, that is, it moves in the direction opposite to scanning 40S subunit (Dango et al. 2011; Best et al. 2023). Nevertheless, we tested whether yeast Slh1 may be responsible for the 40S movement along the 5′ UTR by deleting the single nonessential gene encoding this helicase. Because Slh1 interacts with nonessential protein Asc1 on the small ribosomal subunit (Best et al. 2023), we also expressed GFP reporters in asc1Δ strain. Relative to BY4741 wild-type cells, GFP expression of the ⁵⁰NUS1 construct increased 1.4- and twofold in slh1Δ and asc1Δ strains, respectively (Fig. 7B). Neither genetic perturbation substantially affected GFP expression with long 5′ UTRs. Deletions of slh1 and asc1 similarly affected the ^50-500NUS′ reporters (Supplemental Fig. 11F). Hence, our data provide no supporting evidence for Slh1 involvement in 40S scanning.