Intrinsically unstructured sequences in the mRNA 3′ UTR reduce the ability of poly(A) tail to enhance translation

Graphical Abstract

The 5′ cap and 3′ poly(A) tail of mRNA are known to synergistically stimulate translation initiation via the formation of the cap•eIF4E•eIF4G•PABP•poly(A) complex. Most mRNA sequences have an intrinsic propensity to fold into extensive intramolecular secondary structures that result in short end-to-end distances. The inherent compactness of mRNAs might stabilize the cap•eIF4E•eIF4G•PABP•poly(A) complex and enhance cap-poly(A) translational synergy. Here, we test this hypothesis by introducing intrinsically unstructured sequences into the 5′ or 3′ UTRs of model mRNAs. We found that the introduction of unstructured sequences into the 3′ UTR, but not the 5′ UTR, decreases mRNA translation in cell-free wheat germ and yeast extracts without affecting mRNA stability. The observed reduction in protein synthesis results from the diminished ability of the poly(A) tail to stimulate translation. These results suggest that base pair formation by the 3′ UTR enhances the cap-poly(A) synergy in translation initiation.

INTRODUCTION

Interactions of translation factors with the 5′ and 3′ ends of mRNA play critical roles in the initiation of protein synthesis in eukaryotes. Canonical eukaryotic translation initiation involves the recruitment of the small (40S) ribosomal subunit to the 5′ end of mRNA. Next, the 40S is thought to scan the mRNA in search of the AUG codon most proximal to the 5′ end of mRNA where protein synthesis begins upon joining of the large ribosomal subunit [1]. The recruitment of the small ribosomal subunit to the 5′ end depends on the presence of a 5′ 7-methyl-guanosine cap, which is recognized by initiation factor eIF4E. eIF4E also interacts with the initiation factor eIF4G, which aids in the recruitment of the small ribosomal subunit bound with a number of other initiation factors [1].

The poly(A) tail at the 3′ end of mRNA was shown to stimulate translation initiation [2, 3]. Translation efficiencies of four mRNA substrates, namely, mRNA with cap and poly(A) tail, mRNA with a cap but without poly(A) tail, mRNA with poly(A) tail but without cap, and mRNA without either cap or poly(A) tail, were compared in experiments performed in vivo and in cell extracts [4]. In these experiments, poly(A) stimulated translation of capped mRNAs to a larger degree than translation of uncapped mRNAs [5, 6]. This phenomenon was described as “synergy” between the 5′ cap and 3′ poly(A) tail[5].

The cap-poly(A) tail synergy is believed to be mediated by the binding of cap-binding factor eIF4E and poly(A) binding protein (PABP) to different parts of eIF4G [7-9]. The eIF4E•eIF4G•PABP complex is thought to “circularize” the mRNA (making a "closed loop”) [10-12]. The interaction between eIF4E, eIF4G and PABP is conserved from yeast to humans and is thought to play an important role in the initiation of protein synthesis in eukaryotes. However, mounting evidence suggests that mRNAs vary in the degree to which their translation depends on the presence of eIF4G-PABP interactions [4, 13-17]. Mutations that disrupt eIF4G-PABP interaction in yeast and mammalian cells have little or no effect on overall translation levels and cell viability [18-21]. Recently published data also suggest that the closed-loop state of mRNA is a transient rather than stable configuration of actively translated mRNAs [20, 22]. Furthermore, evolutionary conservation of the closed-loop structure is somewhat puzzling because of the significant entropic cost expected for protein-mediated mRNA circularization. Although formation of eIF4E•eIF4G•PABP complex has been shown to stabilize eIF4E binding to the 5′ cap [23], the role of eIF4G-PABP-mediated mRNA circularization in translation initiation is obscure. It is unclear to what extent mRNA sequence and secondary structure influence cap-poly(A) translational synergy.

mRNA secondary structure might have a profound effect on the formation of the cap•eIF4E•eIF4G•PABP•poly(A) complex and mRNA circularization. Recent computational and FRET studies demonstrated that the ends of RNA molecules are brought in the proximity of a few nanometers regardless of RNA length and sequence because of extensive intramolecular basepairing interactions [24-28]. Formation of multiple stem-loops renders mRNA much more compact than the random coil polymer and thus brings mRNA ends close to each other. Importantly, this universal property of RNA sequences does not depend on the presence of evolutionary conserved basepairs [25]. Only RNAs, which are depleted of guanosines and have low sequence complexity, do not form compact structures with short end-to-end distances [25].

Colocalization of mRNA 5′ and 3′ ends has been observed in cells in which eIF4G-PABP interaction was disrupted by mutagenesis [20]. Thus, even in the absence of the eIF4E•eIF4G•PABP complex, the 5′ and 3′ ends of RNA may be within a few nanometers of each other, i.e. over an order of magnitude closer than the end-to-end distance expected for an average length mRNA in a random coil conformation. PABP binding to the poly(A) tail can facilitate the recruitment of the cap-binding protein complex, eIF4E•eIF4G, to the 5′ end of the mRNA since the 5′ and 3′ ends of mRNA are intrinsically close. Hence, the eIF4G-PABP interaction may have emerged throughout evolution to exploit the intrinsic closeness of mRNA ends [17, 25].

To examine the role of inherent mRNA compactness on translation initiation, we replaced either the 5′ or 3′ UTRs in model mRNAs with intrinsically unstructured sequences. Our results in cell-free wheat germ extract, yeast extract, and in rabbit reticulocyte lysate (RRL) suggest that base pairs formed by the 3′ UTR, but not the 5′ UTR, stimulate translation initiation and enhance cap-poly(A) tail synergy.

RESULTS

Introduction of unstructured sequences into the 3′ UTR of GAPDH mRNA decreases mRNA translational efficiency

To test how intrinsic mRNA compactness affects translation initiation, we followed translation of one isoform of human GAPDH mRNA in wheat germ extract (WGE). GAPDH mRNA is highly translated in human cells [29, 30]. Our FRET experiments demonstrated that the 5′ end of 5′ UTR and the 3′ end of the 3′ UTR of GAPDH mRNA folded in the absence of protein factors are just 6 nm apart in spite of the sequence being 1327 nucleotides in length [25]. WGE is a commercially available highly active translation system that is commonly used in studies of eukaryotic translation. Furthermore, another advantage of using WGE is that the analysis of translation in cell extracts is not complicated by changes in mRNA transcription, splicing, processing and mRNA transport from the nucleus into the cytoplasm. In contrast to another popular in vitro translation system, rabbit reticulocyte lysate (RRL), translation in WGE is stimulated by the presence of both 5′ cap and 3′ poly(A) tail (Suppl. Fig. 1a).

To test whether GAPDH mRNA folds into compact structures with short end-to-end distances in WGE, we labeled the 5′ and 3′ ends of wild-type GAPDH mRNAs, which lacked the 5′ cap and poly(A) tail, with donor (Cy3) and acceptor (Cy5) fluorescent dyes, respectively. Labeled GAPDH mRNA was desalted and stored in water. When FRET was measured in the absence of any cations, negligible levels of energy transfer between fluorophores attached to mRNA ends was observed, indicating that the mRNA is in unfolded conformations (Fig. 1a). By contrast, a 0.75 ± 0.02 FRET value was observed after GAPDH mRNA was incubated for 10 minutes in WGE that lacked amino acids needed for translation (Fig. 1a). Thus, in WGE, GAPDH mRNA folds into compact structures with an average end-to-end distance of ~2.3 nm. The end-to-end distance measured in WGE is shorter than the 6 nm end-to-end distance previously measured for GAPDH mRNA in the presence of 1 mM MgCl₂ and 100 mM KCl [25]. This difference might be due to molecular crowding in WGE that promotes mRNA folding and compaction [31, 32]. Cell lysate might also affect photophysical properties of fluorophores, resulting in an apparent increase in FRET.

Figure 1. Folding of GAPDH mRNA and kinetics of GAPDH translation in WGE.

(a) FRET values measured in Cy3/Cy5-labeled wild-type GAPDH mRNA folded either in WGE or in water lacking any cations. Each FRET value represents the mean ± one SEM of three independent experiments. (b) Phosphoimager scan of 15% SDS-PAGE showing [³⁵S] Met-labeled GAPDH protein (indicated by arrow) translated in WGE programed with wild-type GAPDH mRNA for 15, 30, 45, 60, 75, 90, 105, and 120 min. (c) ImageJ quantification of GAPDH synthesis in (b) normalized by GAPDH amount accumulated in 120 min. The black line is a single exponential fit.

To disrupt the basepairing interactions that render GAPDH mRNA compact, we replaced 106 nt in either the 5′ or 3′ UTR of GAPDH mRNA with CA repeats, which have low basepairing potential (Fig. 2a). Basepairing probability estimates using the RNAstructure software package show that most bases of the wild-type sequence of GAPDH 5′ and 3′ UTRs have high (>75%) basepairing probability (Fig. 3). By contrast, CA repeats are predicted to be nearly completely unpaired while the rest of the GAPDH mRNA remains mainly basepaired (Fig. 3). These calculations are consistent with our previous computational and FRET experiments showing that CA repeats disrupt mRNA intramolecular secondary structure and dramatically increase mRNA end-to-end distance [25].

Figure 2. Design of human GAPDH (a) and GFP (b) mRNA variants with unstructured UTRs.

GAPDH and GFP ORFs are colored orange and yellow, respectively. Unstructured segments in mRNA sequences are colored red. 5′ CA and 3′ CA are CA repeats. 5′ NUS1, 3′ NUS2, 3′ NUS3 and 3′ NUS4 are non-repetitive, unstructured sequences. 3′ S1, 3′ S2, 3′ S3 and 3′ S4 (blue) are shuffled sequences, starting from the wild-type sequence.

Figure 3.

The nucleotide base pairing probability of wild-type GAPDH mRNA and designed sequences. The base pair probability was estimated by a partition function calculation. The Y-axis is the probability of base pairing. The X-axis represents the nucleotide position in the mRNA. The 5′ UTR region is labeled with red. The 3′ UTR is labeled with grey. 5′CA and 5′NUS1 introduced an additional 7 nucleotides in the 5′ UTR, hence the highlighted regions are shifted in those sequences relative to the others.

To further test the role of mRNA compactness in translation initiation, we also replaced either the 5′ or 3′ UTR of GAPDH mRNA with non-repetitive unstructured sequences (NUS), which were designed using the previously described genetic algorithm, to create 5′ NUS1 GAPDH and 3′ NUS2 GAPDH mRNA variants, respectively (Fig. 2a) [25]. The genetic algorithm evolves the original sequence to reduce the basepairing probabilities in a specified region and therefore results in less compact structures. Instead of disrupting a single set of basepairing interactions, NUS sequences are selected to eliminate all alternative helixes formed within a UTR and between the 5′ and 3′ UTRs in the ensemble of structures of the original wild-type sequence. Importantly, in contrast to CA and CAA repeats, which are commonly introduced to disrupt RNA secondary structure, NUS are more complex sequences that are easily maintained and propagated in E. coli cells, thus enabling preparation of templates for in vitro transcription. Noteworthy, 5′ NUS1 and 3′ NUS2 sequences were independently designed. NUS1 and NUS2 share no sequence identity in global pairwise alignment and only 60.8% of sequence identity in a local alignment as calculated by the EMBOSS Water alignment tool [33]. Similar sequence identity (mean of 57.4% and standard deviation of 3.8%) was found in local alignment of NUS2 with computationally shuffled NUS2 sequences. Thus, the NUS1 and NUS2 sequences shared as much sequence identity as two random sequences with similar nucleotide content (i.e. number of A, C, G and U). Both 5′ and 3′ NUS sequences were computationally estimated to dramatically diminish basepairing in respective UTRs (Fig. 3). Our previous FRET measurements corroborated these computational predictions by showing that the introduction of the NUS2 sequence into the 3′ UTR of GAPDH mRNA indeed diminished energy transfer between mRNA ends [25].

The wild-type, 5′ CA, 5′ NUS1, 3′CA and 3′ NUS2 variants of GAPDH mRNA were transcribed with a 30 nucleotide-long poly(A) tail and capped. mRNAs were pre-incubated for 10 minutes in WGE, which lacked amino acids, to enable mRNA folding. Translation was initiated by the addition of 20 amino acids including [³⁵S]-methionine. Regardless of which mRNA was used, accumulation of [³⁵S]-methionine-labeled GAPDH protein, quantified using SDS-PAGE, leveled off after 60-75 minutes of translation (Fig. 1b-c, Suppl. Fig. 1b). Observing plateau in 30 minutes to an hour of translation is typical for protein synthesis in cell lysates and is due to depletion and degradation of ATP, GTP amino acids and components of ATP regeneration systems like phosphoenolpyruvate [34].

The introduction of an unstructured sequence into the 5′ UTR of GAPDH mRNA led to a 1.5-fold increase in GAPDH protein levels synthesized in 1 hour of translation in WGE (Fig. 4a). In contrast to experiments with 5′ CA and 5′ NUS1 GAPDH mRNA, translation of partially unstructured 3′ CA and 3′ NUS2 mRNA produced 2-fold less of full-length [³⁵S]-methionine-labeled GAPDH protein than translation of the wild-type mRNA sequence (Fig. 4a). To discern whether the observed decrease in GAPDH translation primarily resulted from alterations in the 3′ UTR primary structure or its secondary structure, we replaced 106 nucleotides in the 3′ UTR of GAPDH mRNA with a shuffled sequence (i.e. a sequence with the same number of A, C, G and U nucleotides as the wild-type in this region, but in random order) (Fig. 2a). An estimation of base pairing probabilities showed that, while introduction of shuffled sequences altered patterns of basepairing interaction, the average basepairing probability throughout the 3′ UTR remained largely unaffected (Fig. 3). Hence, GAPDH mRNAs containing shuffled sequences in their 3′ UTRs are expected to fold into compact structures similar to the wild-type GAPDH mRNA. Using Förster resonance energy transfer (FRET), we previously found that end-to-end distances in in vitro-folded wild-type and 3′ S1 (Shuffled 1) GAPDH mRNAs are indeed similar [25]. Remarkably, the translation efficiencies of 3′ S1 (Shuffled 1) and wild-type GAPDH mRNAs were also similar (Fig. 4a) while translation of 3′ S2 mRNA was decreased by just 10%. Therefore, the introduction of NUS into the 3′ UTR likely inhibited translation by disrupting the secondary structure of mRNA rather than by elimination of a specific sequence that enhances translation of the wild-type GAPDH mRNA.

Figure 4. The introduction of unstructured sequences into the 3′ UTR of GAPDH mRNA decreases mRNA translational efficiency in WGE and diminishes stimulation of translation by poly(A) tail.

(a) Relative GAPDH levels synthesized in 1 hour in WGE programmed with different mRNA variants normalized to GAPDH amount synthesized from wild-type GAPDH mRNA. Asterisks indicate that translation of a given mRNA variant is different from translation of the wild-type mRNA, as p-values determined by the two-tailed Student t-test were below 0.05.

(b) Poly(A)-mediated stimulation of translation shown as a ratio of GAPDH levels synthesized from capped, polyadenylated (+cap/+poly(A)) vs capped, deadenylated (+cap/−poly(A)) mRNA. Asterisks indicate that poly(A)-mediated stimulation of translation of a given mRNA variant is significantly lower than that of the wild-type mRNA, as p-values determined by the one-tailed Student t-test were below 0.05.

(c) Changes in mRNA levels, 2^−ΔCt, during 1-hour incubation in WGE that are measured by RT-qPCR. (a-c) Bar graphs show mean values, error bars show standard error of the mean (SEM) determined from triplicate experiments.

While experiments with secondary structure disruptions in the 3′ UTR support the idea that mRNA compactness stimulates translation, lack of translational inhibition by unstructured sequences in the 5′ UTR seemingly contradicts our original hypothesis. As discussed later, secondary structures in the 5′ UTR may have a pleotropic effect on translation initiation affecting both the efficiency of 40S scanning and the stability of the eIF4E•eIF4G•PABP complex. Alternatively, compactness of the 3′ UTR alone might stimulate cap-poly(A) synergy. Hence, we proceeded to study the effects of unstructured sequences introduced into the 3′ UTR.

Unstructured sequences in the 3′ UTR of GAPDH mRNA diminish the ability of poly(A) tail to stimulate translation

Why does introduction of unstructured sequences in the 3′ UTR decrease translational efficiency? Unstructured sequences in the 3′ UTR may destabilize the cap•eIF4E•eIF4G•PABP•poly(A) complex and thus diminish cap-poly(A) tail synergy. To test how the presence of the unstructured sequences in the 3′ UTR affects the ability of poly(A) tail to stimulate translation, for each GAPDH mRNA variant, we examined translation of two mRNA isoforms: mRNA with cap and poly(A) tail (+/+) versus mRNA with a cap but without poly(A) tail (+/−) (Fig. 4b). Quantitative analysis of poly(A) effect on translation of uncapped GAPDH mRNA was hampered by low translation efficiency of uncapped mRNA isoforms (Suppl. Fig. 1a).

Relative to isoforms lacking poly(A) tail, the presence of poly(A) tail stimulated translation of wild-type, 3′ S1 (Shuffled 1) and 3′ S2 (Shuffled 2) mRNAs by 4.5 to 5.4 fold (Fig. 4b, Table 1). In 5′ CA mRNA, poly(A) tail stimulated translation by 6.3 fold suggesting that unstructured sequences in 5′ UTR do not affect cap-poly(A) synergy. By contrast, poly(A) tail enhanced translation of 3′ CA and 3′ NUS2 mRNAs by only 2 fold (Fig. 4b, Table 1). Hence, the reduced translation efficiency of capped and polyadenylated 3′ CA and 3′ NUS2 mRNAs (Fig. 4a) relative to wild-type mRNA was mostly due to weakened poly(A)-mediated stimulation of translation. To further scrutinize this interpretation, instead of replacing only the last 106 terminal nucleotides of the 3′ UTR with unstructured sequences (as done in 3′ CA and 3′ NUS2 mRNAs), we replaced almost the entire 3′ UTR of GAPDH mRNA with non-repetitive unstructured sequence NUS3 (202 nucleotides out of 210 nucleotides of the original 3′ UTR). Rendering the entire 3′ UTR of GAPDH mRNA unstructured (Fig. 3) essentially abrogated the ability of poly(A) tail to stimulate translation of capped 3′ NUS3 mRNA as ratio of GAPDH levels translated from +cap+poly(A) vs +cap−poly(A) mRNAs decreased from 5.4 (wild-type mRNA) to 1.2 (3′ NUS3 mRNA).

Table 1.

Poly(A)-mediated stimulation of translation in mRNAs with unstructured 3′UTRs is significantly lower than that in mRNAs with structured 3′UTRs.

mRNA	Ratio +cap+poly(A)/ +cap−poly(A), mean value + SEM	p values determined relative to WT mRNA
WT GAPDH	5.4 ± 0.5
5′ CA GAPDH	6.3 ± 1.7	0.31
3′ S1 GAPDH	5.3 ± 0.7	0.46
3′ S2 GAPDH	4.5 ± 0.5	0.14
3′ CA GAPDH	2.0 ± 0.5	3×10−3
3′ NUS2 GAPDH	2.2 ± 0.2	2×10−4
3′ NUS3 GAPDH	1.2 ±0.2	1×10−4
WTGFP	3.6 ± 0.5
3′ S3 GFP	3.3 ± 0.2	0.3
3′ S4 GFP	3.8 ± 0.2	0.4
3′ NUS4 GFP	2.1 ± 0.3	0.01
WT GFP (in yeast extr.)	3.4 ± 1.1
3′ NUS4 GFP (in yeast extr.)	1.0 ± 0.3	0.04

In mRNAs with unstructured 3′UTRs, poly(A)-mediated stimulation of translation, i.e. ratio of protein levels synthesized from capped, polyadenylated (+cap/+poly(A)) vs capped, deadenylated (+cap/−poly(A)) mRNA, is lower than that measured for the wild-type mRNA, as p-values determined by the one-tailed Student t-test were below α (0.05). GAPDH and GFP mRNA variants were translated in WGE unless indicated otherwise. GFP mRNA variants were also translated in yeast cell extract.

To examine how the introduction of unstructured or shuffled sequences into GAPDH mRNA 5′ and 3′ UTRs affect mRNA stability and capping efficiency, we used RT-qPCR, 3′ RACE and 5′ capping assay. Our results indicate that GAPDH mRNA variants were capped with similar efficiency (Suppl. Fig. 2). RT-qPCR showed no signs of substantial mRNA degradation in WGE during the time course of translation for any of studied mRNA variants (Fig. 4c). No difference in mRNA stability between poly-adenylated and non-adenylated GAPDH mRNA variants was detected (Fig. 4c). Finally, 3′ RACE showed no evidence of mRNA deadenylation in WGE during translation in either wild-type or 3′ NUS2 mRNA capped, polyadenylated variants (Suppl. Fig. 3). Hence, changes in GAPDH synthesis levels produced by the introduction of intrinsically unstructured sequences into the 5′ and 3′ UTRs of GAPDH mRNA (Fig. 4a-b) are not due to differences in mRNA stability but are rather indicative of differences in translational efficiencies.

An intrinsically unstructured sequence in the 3′ UTR of GFP reporter reduces GFP synthesis and cap-poly(A) synergy

Next, we tested whether unstructured sequences in the 3′ UTR inhibit translation only in the context of GAPDH mRNA or unstructured sequences in 3′ UTR inhibit translation of other mRNAs. To that end, we constructed a reporter mRNA in which the superfolder (sf)GFP open reading frame (ORF) was flanked by a 19 nucleotide-long 5′ UTR and 206 nucleotide-long 3′ UTR from yeast RPL41B mRNA, the gene that encodes yeast ribosome protein L41B (Fig. 2b). Computation estimates that this RPL41B-UTR_ (“wild-type”) GFP mRNA forms extensive secondary structures throughout the entire sequence (Fig. 5). Replacing the last 110 nucleotides of the 3′ UTR of the RPL41B-UTR_GFP mRNA with unstructured sequence NUS4 designed by the genetic algorithm is predicted to diminish the basepairing probabilities of nucleotides in the 3′ UTR (Fig. 5). Noteworthy, in local alignment performed using EMBOSS Water algorithm, NUS4 shared only 55.2% and 58.1% identity with NUS1 and NUS2, respectively. In addition, we created two other GFP mRNA variants by replacing 110 nucleotides at the 3′ end of 3′ UTR of the original RPL41B-UTR_GFP mRNA with shuffled sequences (Fig. 2b). Shuffled sequences S3 and S4 are estimated by computation to form extensive secondary structures (Fig. 5).

Figure 5.

The per nucleotide base pairing probability of wild-type GFP mRNA and designed sequences. The base pair probability was estimated with a partition function calculation. The Y-axis is the probability of base pairing. The X-axis represents the nucleotide position in the mRNA. The 3′ UTR was labeled with grey.

Similar to another fast-folding GFP variant, EGFP, sfGFP used in our experiments matures and becomes fluorescent in live cells 2-3 faster than wild-type GFP [35]. The maturation lifetime of EGFP during cell-free expression was reported to be 3-8 minutes [36, 37]. Thus, we can reliably follow accumulation of synthesized sfGFP protein during translation by fluorescence with a <8-minute lag at the beginning of the translation reaction (Fig. 6a).

Figure 6. An unstructured sequence in the 3′ UTR of GFP mRNA reduces the synergy between a cap and a poly(A) tail in WGE.

(a) Fluorescence (relative fluorescence units, RFU) of GFP synthesized from capped and polyadenylated “wild-type” (black), 3′ S3 (blue), 3′ S4 (magenta) and 3′NUS4 (red) RPL41B-UTR_GFP mRNAs as a function of time. (b) Poly(A)-mediated stimulation of translation shown as a ratio of GFP levels synthesized from capped, polyadenylated (+cap/+poly(A)) vs capped, deadenylated (+cap/−poly(A)) mRNA. (c) Changes in mRNA levels, 2^−ΔCt, during 1-hour incubation in WGE that are measured by RT-qPCR. (d) The degree of synergy between the cap and poly(A) tail was calculated according to the formula Synergy= $\frac{[translationof+cap+poly(A)mRNA]\times [translationof-cap-poly(A)mRAN]}{[translationof-cap+poly(A)mRNA]\times [translationof+cap-poly(A)mRNA]}$. For each GFP mRNA variant, four modifications of mRNA were translated for 1 hour in WGE: without cap or poly(A)₃₀ (−/−), without cap and with poly(A)₃₀ (−/+), with cap and without poly(A)₃₀ (+/−), with both cap and poly(A)₃₀ (+/+). Dots (a) and bar graphs (b-d) show mean values, error bars (a-d) show standard error of the mean (SEM) determined from triplicate experiments. Asterisks indicate that poly(A)-mediated stimulation of translation (b) or synergy (d) are significantly lower than those measured for the wild-type mRNA, as p-values determined by the one-tailed Student t-test were below 0.05.

Capped GFP mRNA variants transcribed with a 30 nucleotide-long poly(A) tail were translated in WGE. Similar to GAPDH translation in WGE, accumulation of GFP leveled off after 75 minutes of translation. The introduction of NUS into the RPL41B-UTR_GFP mRNA resulted in a two-fold decrease in GFP synthesis (Fig. 6a). In contrast to NUS-containing mRNA, 3′ S3 (Shuffled 3) GFP and 3′ S4 (Shuffled 4) GFP mRNAs translated as well as the original RPL41B-UTR_GFP mRNA (Fig. 6a). Hence, NUS sequence inhibits translation by disrupting mRNA secondary structure rather than by perturbing the original sequence of the 3′ UTR.

Relative to isoforms lacking poly(A) tail, the presence of poly(A) tail stimulated translation of the original RPL41B-UTR (“wild-type”), 3′ S3 (Shuffled 3) and 3′ S4 (Shuffled 4) GFP mRNAs by 3.3 to 3.8 fold (Fig. 6b, Table 1). By contrast, poly(A) tail enhanced translation of 3′ NUS4 GFP mRNAs by only 2.1 fold (Fig. 6b, Table 1). Hence, consistent with data obtained for GAPDH mRNA variants, an unstructured sequence in the 3′ UTR reduced the ability of poly(A) tail to stimulate translation of GFP reporter mRNA.

RT-qPCR experiments revealed no substantial degradation of +cap +poly(A) and +cap −poly(A) tail modifications of “wild-type” and 3′ NUS4 GFP mRNA variants during translation in WGE (Fig. 6c). Minor variations in mRNA stability (i.e. slight decrease in +cap +poly(A) “wild-type” GFP mRNA levels) do not account for observed differences in the amounts of synthesized GFP. 3′ RACE experiments show no sign of poly(A) tail shortening during translation in WGE for either RPL41B-UTR_GFP or 3′ NUS3 GFP mRNAs (Suppl. Fig. 4). Therefore, the introduction of an unstructured sequence into the 3′ UTR of GFP mRNA reduces GFP synthesis by hampering translation efficiency without affecting mRNA stability.

Sensitivity of fluorescent GFP detection enabled quantitative comparison of the poly(A) tail effect on translation of capped versus uncapped GFP mRNA variants. To that end, for each GFP mRNA variant, we examined translation of four mRNA isoforms: mRNA with cap and poly(A) tail (+/+), mRNA with a cap but without poly(A) tail (+/−), mRNA with poly(A) tail but without cap (−/+), and mRNA without either cap or poly(A) tail (−/−). Cap-poly(A) synergy was determined from the ratio of levels to which a poly(A) tail stimulates expression of capped versus uncapped mRNA according to the formula: Synergy=$\frac{[translationof+cap+poly(A)mRNA]\times [translationof-cap-poly(A)mRAN]}{[translationof+cap-poly(A)mRNA]\times [translationof-cap+poly(A)mRNA]}$ [5].Synergy values larger than 1 indicate synergy between the 5′ cap and 3′ poly(A) tail. In other words, synergy values above 1 show that poly(A) tail stimulates translation of capped mRNA more than translation of uncapped mRNA. Cap-poly(A) synergy values for 3′ S3 (Shuffled 3) and 3′ S4 (Shuffled 4) GFP mRNAs were not significantly different from the 2.5 synergy value observed with the RPL41B-UTR “wild-type” GFP mRNA (Fig. 6d, Suppl. Fig. 5, Table 2). By contrast, introduction of NUS4 into the 3′ UTR led to a 1.8-fold decrease in cap-poly(A) synergy when compared to cap-poly(A) tail synergy in the original RPL41B-UTR_GFP mRNA (Fig. 6d, Suppl. Fig. 5, Table 2). These data provide evidence that unstructured sequences in the 3′ UTR reduce cap-poly(A) translational synergy.

Table 2.

Cap-poly(A) synergy in GFP mRNA with unstructured 3′ UTR is significantly lower than that in mRNAs with structured 3′ UTRs.

mRNA	Synergy, mean ± SEM	p values determined relative to WT mRNA
WT GFP	2.5 ± 0.5
3′ S3 GFP	1.8 ± 0.2	0.18
3′ S4 GFP	2.4 ± 0.3	0.47
3′ NUS4 GFP	1.4 ± 0.2	0.03

In GFP mRNA with unstructured 3′UTR (3′ NUS4), cap-poly(A) synergy, Synergy= $\frac{[translationof+cap+poly(A)mRNA]\times [translationof-cap-poly(A)mRNA]}{[translationof+cap-poly(A)mRNA]\times [translationof-cap+poly(A)mRNA]}$, is lower than that measured for the wild-type mRNA, as p-value determined by the one-tailed Student t-test was below α (0.05). GFP mRNA variants were translated in WGE.

Unstructured sequences in the 3′ UTR diminish the ability of poly(A) tail to stimulate translation in yeast extract

To test whether unstructured sequences in the 3′ UTR inhibit translation in other translation systems that are different from WGE, we prepared cell extracts from Saccharomyces cerevisiae. Proteins synthesis in yeast cell extract showed strong dependence on the 5′ cap as translation of uncapped “wild-type” GFP mRNA was barely detectable (Fig. 7a). Relative to isoform lacking poly(A) tail, the presence of poly(A) tail produced 3.4-fold stimulation of cap-dependent translation of “wild-type” GFP mRNA (Fig. 7a-b). Unstructured sequence in the 3′ UTR abrogated the ability of poly(A) tail to stimulate translation of capped 3′ NUS4 GFP mRNA as ratio of GFP levels translated from +cap+poly(A) vs +cap−poly(A) mRNAs decreased from 3.4 (“wild-type” GFP mRNA) to 1.0 (3′ NUS4 GFP mRNA). Importantly, RT-qPCR data indicate that introduction of unstructured sequence in the 3′ UTR did not decrease mRNA stability of both de- and polyadenylated mRNA isoforms (Fig. 7c). In fact, stability of “wild-type” +cap−poly(A) GFP mRNA was somewhat lower than stability of 3′ NUS4 GFP mRNA isoforms. Hence, variations in mRNA levels do not account for the observed abrogation of poly(A) tail ability to stimulate translation of capped 3′ NUS4 GFP mRNA. Hence, in yeast cell extract, unstructured 3′ UTR diminished the ability of poly(A) tail to stimulate translation without reducing mRNA stability.

Figure 7. An unstructured sequence in the 3′ UTR of RPL41B-UTR_GFP mRNA decreases mRNA translational efficiency in yeast extract.

(a) Fluorescence (relative fluorescence units, RFU) of GFP synthesized from capped and polyadenylated “wild-type” (black), capped and deadenylated “wild-type” (blue), uncapped and polyadenylated “wild-type” (orange), capped and polyadenylated 3′ NUS4 (red) and capped and deadenylated 3′ NUS4 (cyan) RPL41B-UTR_GFP mRNAs as a function of time. (b) Poly(A)-mediated stimulation of translation shown as a ratio of GFP levels synthesized from capped, polyadenylated (+cap/+poly(A)) vs capped, deadenylated (+cap/−poly(A)) mRNA after 1-hour incubation in yeast extract. (c) Changes in mRNA levels, 2^−ΔCt, during 1.5-hour incubation in yeast extract that are measured by RT-qPCR. Dots (a) and bar graphs (b-c) show mean values, error bars (a-c) show standard error of the mean (SEM) determined from triplicate experiments. Asterisk indicates that poly(A)-mediated stimulation of translation of 3′ NUS4 mRNA is significantly lower than that of the wild-type mRNA, as p-values determined by the one-tailed Student t-test were below 0.05.

mRNAs with structured and unstructured 3' UTRs are translated with similar efficiencies in rabbit reticulocyte lysate (RRL).

We next examined how the presence of an unstructured sequence in the 3′ UTR affects translation in cell extracts with weak or no translation stimulation by the poly(A) tail. Translation in rabbit reticulocyte lysate (RRL) is known to exhibit low cap- and poly(A) dependence [38]. By adding 10-fold excess of uncapped and deadenylated firefly luciferase mRNA to RRL, we created conditions where translation was appreciably stimulated by the 5′ cap but remained weakly dependent of the presence of the 3′ poly(A) tail. The 5′ cap produced nearly 10-fold stimulation of GAPDH synthesis while the 3′ poly(A) tail enhanced translation of wild-type capped GAPDH mRNA by less than 2 fold. Under these conditions of diminished translation stimulation by the poly(A) tail, translation efficiencies of wild-type and 3′ NUS3 GAPDH mRNAs were similar (Fig. 8a, Suppl. Fig. 6). Stabilities of capped wild-type and 3′ NUS3 GAPDH mRNAs were also similar (Fig. 8b). These data indicate that unstructured sequences in the 3′ UTR lessen translation efficiency by reducing the ability of poly(A) to simulate cap-dependent translation, but have no effect on translation when translation stimulation by the poly(A) tail is nearly absent.

Figure 8. GAPDH mRNAs with structured and unstructured 3' UTRs in RRL are translated in RRL with similar efficiencies.

(a) Relative GAPDH levels synthesized in 1.5 hour in RRL programmed with wild-type and 3′NUS3 mRNA variants normalized to GAPDH amount synthesized from wild-type capped, polyadenylated (+cap/+poly(A) GAPDH mRNA. (b) Changes in mRNA levels, 2^−ΔCt, during 1.5-hour incubation in RRL that are measured by RT-qPCR. (a-b) Bar graphs show mean values, error bars show standard error of the mean (SEM) determined from triplicate experiments.

DISCUSSION

We aimed to test the hypothesis that the intrinsic compactness of mRNA may facilitate translational synergy between the 5′ cap and 3′ poly(A) tail by stabilizing the cap•eIF4E•eIF4G•PABP•poly(A) complex. Consistent with this hypothesis, we observed that unstructured sequences in the 3′ UTR inhibit translation and diminish the ability of the poly(A) tail to stimulate cap-dependent translation without changing mRNA stability in both WGE and yeast cell extract. Previous publications suggested that increasing the length of the 3′ UTR diminishes stimulatory effect of poly(A) tail on translation while introducing a stem-loop into the 3′ UTR does not affect cap-poly(A) synergy [39-41]. However, secondary structures of the 3′ UTR and the entire mRNA were not considered or analyzed in those studies, which hampers the comparison with our data.

In contrast to unstructured sequences in the 3′ UTR, the introduction of NUS and CA repeats into the 5′UTR of GAPDH mRNA did not inhibit GAPDH synthesis or the ability of poly(A) tail to stimulate cap-dependent translation. The secondary structure in the 5′ UTR might play a dual role in translation initiation. On the one hand, basepairing interactions between the 5′ and 3′ UTRs might stimulate protein synthesis by stabilizing the cap•eIF4E•eIF4G•PABP•poly(A) complex. On the other hand, stable stem-loop structures in the 5′ UTR were shown to inhibit ribosome scanning of the 5′ UTR during translation initiation [42, 43], suggesting that secondary structure in the 5′ UTR hampers protein synthesis. Furthermore, upon abrogation of secondary structure in the 5′ UTR, entropic destabilization of cap•eIF4E•eIF4G•PABP•poly(A) complex might be offset or even exceeded by stabilization of eIF4G binding to the 5′ UTR. Possibly, due to superposition of these effects, unstructured sequences in 5′ UTRs do not inhibit but rather stimulate translation. Alternatively, as discussed below, rather than the closeness of mRNA ends, compactness of the 3′ UTR alone might facilitate translation re-initiation mediated by the cap•eIF4E•eIF4G•PABP•poly(A) complex.

Why do unstructured sequences in the 3′ UTR diminish cap-poly(A) synergy? It has been reported that unstructured sequences at the 3′ end of 3′ UTR inhibit protein expression by hampering mRNA 3′ end processing and polyadenylation [44]. However, since our experiments were performed in cell-free extracts with mRNAs containing poly(A)₃₀ added through in vitro transcription, reduced translation observed in mRNAs with unstructured 3′ UTRs is not due to defective polyadenylation. It is unlikely that the introduction of unstructured sequences into the 3′ UTR of GAPDH and GFP mRNAs inhibits translation by replacing translation-enhancing sequences in the original 3′ UTR or by recruiting specific inhibitory factors. Indeed, randomized sequences introduced into the 3′ UTR of GAPDH and GFP mRNAs did not substantially affect synthesis of GAPDH and GFP, respectively. Additionally, NUS2, NUS3, NUS4 and CA repeats similarly inhibited translation but shared little sequence identity. The observed decrease in cap-poly(A) synergy is unlikely due to sequestration of PABP or eIF4G by unstructured sequences in the 3′ UTR as they do not inhibit translation when they are introduced into the 5′ UTR (Fig. 4). Still, we cannot exclude the possibility that unstructured sequences in the 3′ UTR may hamper translation in WGE by recruiting endogenous single-strand RNA binding proteins that have low sequence specificity.

It is more likely that unstructured sequences in the 3′ UTR hamper translation by increasing the entropic cost for the cap•eIF4E•eIF4G•PABP•poly(A) complex formation. Indeed, unstructured sequences in the 3′ UTR specifically diminished the ability of poly(A) tail to stimulate translation and cap-poly(A) synergy (Fig. 4, 6). Furthermore, increasing the length of the 3′ UTR unstructured sequence from 106 nucleotides (3′ NUS2 and 3′ CA mRNAs) to 202 nucleotides (3′ NUS3) reduced poly(A)-induced stimulation of translation by 2 fold, essentially abrogating the effect of the poly(A) presence (Fig. 4b). This result is consistent with the idea that extending the length of the 3′ UTR unstructured sequence increases the entropic cost for mRNA circularization during the cap•eIF4E•eIF4G•PABP•poly(A) complex formation.

Another model, which explains different translational effects of intrinsically unstructured sequences in the 5′ and 3′ UTRs, is that compactness of the 3′ UTR is important for the re-initiation of the 40S subunit after termination of protein synthesis. Although scarce, experimental data suggest that after translation termination, ribosomes can re-initiate protein synthesis without disassociating from the mRNA in the process facilitated by the poly(A) tail [12, 45, 46]. This re-initiation mechanism was hypothesized to require mRNA circularization [45] and thus might underlie cap-poly(A) synergy mediated by the cap•eIF4E•eIF4G•PABP•poly(A) complex formation [10]. Unwinding of ORFs by translating ribosomes make mRNAs less compact [20, 22] but should not drastically reduce secondary structure in the 3′ UTR. Compactness of the 3′ UTR might facilitate one-dimensional diffusion of the 40S subunit along the 3′ UTR from the stop codon to the 5′ cap of mRNA bound to eIF4E•eIF4G•PABP•poly(A) complex. Possibly due to expanded distance between the stop codon and the cap•eIF4E•eIF4G•PABP•poly(A) complex, unstructured sequences in the 3′ UTR (but not the 5′ UTR) diminish cap-poly(A) synergy.

Our study has its own limitations. For example, levels of cap-poly(A) synergy observed in our experiments performed in WGE and yeast cell extracts are more modest than those observed in situ translation systems such as tobacco protoplasts or Chinese hamster ovary cells [5]. Substantial variations in cap-poly(A) synergy between different translation systems were previously reported in the literature. Furthermore, different cell extract preparations from the same organism can substantially differ in cap-poly(A) synergy. For example, depending on the presence/absence of endogenous (competing) mRNAs and concentrations of ribosomes, cap-poly(A) synergy in rabbit reticulocyte lysates varies from strong to absent [23, 38]. Variations in concentration and stoichiometry of ribosomes, initiation factors and mRNAs likely underlie disparities in cap-poly(A) synergy between different translation systems. Nevertheless, WGE, which is commercially available and easy to use, and yeast cell extracts sufficiently recapitulates poly(A)-induced stimulation of translation (Fig. 4, 6, 7, Supplementary Fig. 1a). 30 nucleotide-long poly(A) tail of reporter mRNAs used in our work is significantly shorter than the length of the tail added during transcription termination, which is thought to vary between ~90 (yeast) and ~250 (animals) adenosines [47]. However, new sequencing methods revealed that median lengths of poly(A) are much shorter and range between 30 (yeast) to 50-100 (human) adenosines [48]. Importantly, a 30 nucleotide-long poly(A) tail accommodates at least one PABP molecule [49-51]. Furthermore, short (30-60 nucleotide-long) poly(A) tails are found in highly translated mRNAs [52]. Hence, translation of our reporter mRNAs in WGE should provide biologically relevant observations.

Our results offer new insights into the cap-poly(A) synergy. The effect of the 3′ UTR secondary structure on the cap-poly(A) synergy might indicate that translation could be regulated through changes in mRNA compactness. However, natural mRNA sequences have a nearly universal propensity to fold into extensive secondary structures, which render mRNA compact [25]. Therefore, it is unlikely that variations in the 3′ UTR secondary structures is a major factor underlying differences in translational efficiency between distinct genes. Nevertheless, importance of mRNA compactness for cap-poly(A) synergy may explain why the closed-loop state of mRNA is a transient rather than stable configuration of actively translated mRNAs [20, 22]. Ribosomes and other RNA helicases might keep actively translated mRNA predominantly unstructured and thus destabilize the cap•eIF4E•eIF4G•PABP•poly(A) complex. By contrast, translationally silent mRNAs are likely free to fold into compact structures. Hence, the cap•eIF4E•eIF4G•PABP•poly(A) complex might stably form only in translationally silent mRNAs and thus play an important role in translational (re)-activation of these mRNAs.

MATERIALS AND METHODS

Design of mRNA variants.

Non-repetitive unstructured sequences (NUS) sequences were designed using the orega program (optimizing RNA ends with a genetic algorithm) of the RNAstructure software package (https://rna.urmc.rochester.edu/RNAstructure.html) as previously described [25]. Base pair probabilities were estimated using the partition function in RNAstructure [53]. Construction of pSP64A vectors (Promega) encoding GAPDH and GFP variants via Gibson assembly was done as previously described [25]; constructs encoding wild-type, 3′CA, 3′S1, 3′S2 and 3′ NUS2 GAPDH mRNA variants were previously made [25]. A GFP reporter containing UTRs from yeast RPL41B gene was synthesized by GenScript and subsequently cloned into pSP64A vector. Sequences of all GAPDH and GFP mRNA variants used in this work are provided in Supplementary Materials.

mRNA preparation.

mRNAs were transcribed either with or without a 30-nucleotide long poly(A) tail by T7 RNA polymerase-catalyzed run-off in vitro transcription using DNA templates linearized by either EcoRI or SacI, respectively. 10 μg of linearized plasmid DNA was added to a 1 mL transcription reaction mixture containing 80 mM HEPES-KOH pH 7.5, 2 mM spermidine, 30 mM dithiothreitol, 25 mM NaCl, 8 mM MgCl₂, 0.8 mM NTPs and 2 μM T7 polymerase and incubated at 37 °C for 4-6 hours. The synthesized RNA was precipitated with 0.3 M sodium acetate, pH 5.3 and 2.5 volumes of ethanol. After ethanol precipitation, mRNAs were re-suspended in 100 μl ddH₂O, layered on the top of 5-20% sucrose gradients in 10 mM Hepes-HCl, pH 7.5, 100 mM KCl, and 1 mM EDTA and centrifuged in a SW-41 rotor (Beckman Coulter) for 12 hours at 37,000 rpm at 4 °C. After sucrose gradient fractionation, mRNA was precipitated by the addition of two and half volumes of ethanol.

mRNA was capped using vaccinia virus enzyme (NEB). 10 μg of in vitro transcribed mRNA was capped at 37°C for 30 min in a reaction buffer containing 50 mM Tris-HCl, pH 7.8, 1.25 mM MgCl₂, 6 mM KCl, 2.5 mM DTT, 1 mM GTP, 100 μM S-adenosylmethionine, 125 nM capping enzyme (NEB) and 1 U/μL RNase Inhibitor (NEB). To analyze the capping efficiency, capping reaction was carried out in the presence of 1 mM GTP and 2 μCi/μl α-[³²P]-GTP (Perkin Elmer). mRNA was desalted twice using 1 mL G-25 spin-columns equilibrated with water. Capping efficiency was examined by 5% denaturing acrylamide gel, followed by phosphorimager analysis.

Ensemble FRET measurements.

The 5′ and 3′ ends of the wild-type GAPDH mRNA were labeled with Cy3 and Cy5 fluorophores, respectively, as previously described [25]. Ensemble FRET data were acquired and analyzed using the ratioA method as previously described [25].

In vitro translation in wheat germ extract (WGE).

GAPDH mRNA (1.2 μg) was pre-incubated in 50% (v/v) micrococcal nuclease-treated WGE (Promega) in the presence of 140 mM potassium acetate and 0.8 U/μl murine RNase inhibitor (NEB) for 10 minutes at 25 °C. Translation was initiated by the addition of an amino acids mixture lacking methionine (0.08 mM) and [³⁵S] methionine (0.5 μM) in a final volume of 20 μl. Translations were performed for 60 min at 25 °C. Translation reactions were analyzed by 15% (w/v) SDS-polyacrylamide gel electrophoresis followed by phosphoimaging. For time course measurements of GFP synthesis in WGE, in vitro transcribed GFP mRNAs were translated in WGE as described above except all amino acids were unlabeled. The reactions (60 μl) were transferred to a 384-well plate and GFP fluorescence signals were recorded by a microplate reader at room temperature for 2 hours (λ_ex = 485 nm; λ_em = 510 nm).

In vitro translation in rabbit reticulocyte lysate (RRL).

GAPDH mRNA (1.2 μg) was pre-incubated in 70% (v/v) micrococcal nuclease-treated WGE (Promega) in the presence of 0.8 U/μl murine RNase inhibitor (NEB) and 12 μg of uncapped firefly luciferase mRNA for 5 minutes at 25 °C. Translation was initiated by the addition of an amino acids mixture lacking methionine (0.02 mM) and [³⁵S] methionine (0.5 μM) in a final volume of 20 μl. Translations were performed for 90 min at 30 °C. Translation reactions were analyzed by 12% (w/v) SDS-polyacrylamide gel electrophoresis followed by phosphoimaging.

Yeast cell extract preparation.

Yeast cell extract preparation was performed as previously described [54]. Saccharomyces cerevisiae cells (MBS strain [MATa ade2-1 his3-11, 15 leu2-3, 112 trp1-1 ura3-1 can1-100 (rho+) L-o, M-o]; gift from Vasili Hauryliuk) were grown at 30 °C in YPD medium until they reached OD₆₀₀ = 3 to 4. Then cells were washed once with cold mQ water, twice with buffer C (20 mM HEPES pH 7.5, 100 mM potassium acetate, 2 mM magnesium acetate, 2 mM DTT, 0.5 mM PMSF) and pelleted at 5000 g for 10 minutes at 4 °C. The cells were resuspended in 1/10 of the volume of the wet pellet weight with buffer C supplemented with protease inhibitor cocktail. Small frozen cell pellets were generated by freezing the cell suspension in liquid nitrogen and were stored at −80 °C. Typically, 10 g of cells were obtained from 3 L culture. Then the pellets were crushed with a pre chilled ceramic mortar and pestle in liquid nitrogen until a fine powder was obtained and were thawed on ice for 3 to 5 hours. Afterwards the lysate was transferred to a 50 ml falcon tube and was centrifuged twice at 29,242 g for 15 minutes. The obtained sample was run through the Zeba Desalt Spin column (Pierce) preequilibrated with buffer C. The absorbance of the obtained cell extract in 260 nm was 200 to 300 OD/ml. The cell extract was stored at −80 °C until further use.

In vitro translation in yeast extracts.

Following buffers were used: 1 x energy mix – 1 mM ATP, 0.25 mM GTP and 25 mM creatine phosphate, freshly made; 1 x common buffer – 10 mM Hepes-KOH pH 7.5 and 1 mM DTT; 1 x variable buffer – 2.05 mM Mg(OAc)₂ and 100 mM KOAc. First, the extract was treated with 23 gel U/μl micrococcal nuclease (NEB) supplemented with 0.5 mM CaCl₂ for 10 min at 25 °C. The reaction was stopped with 2 mM EGTA pH 7.5. Then, the pre-mix containing 5 x energy mix, 0.6 mg/ml creatine kinase, 5 x common buffer and 5 x variable buffer was prepared. 20 μl reaction with 1 x pre-mix, 0.08 mM amino acids, 0.8 U/μl RNase inhibitor, treated extract (A₂₆₀ = 100 OD/ml) and 2 μg GFP mRNA was measured at 24 °C on the plate reader (Spark Tecan) every 128 sec for 1 hour.

RT-qPCR assay.

For quantitative RT-PCR (qRT-PCR), total RNA translated in WGE was extracted with phenol:chloroform followed by ethanol precipitation. cDNA was prepared with the Superscript III First Strand Synthesis System for RT-PCR (Invitrogen) using random primers. 1 μg of mRNA extracted from cell lysate and 0.5 μg random primer (Invitrogen) were incubated together at 65°C for 5 minutes and then placed on ice for at least 1 minute. Next, RNA was incubated in 1x First Strand Buffer (Invitrogen) containing 10 mM DTT, 40 U/ul RNAase OUT (Invitrogen)), and 200 U of Superscript II reverse transcriptase (Invitrogen) at 25°C for 10 minutes. cDNAs were synthesized at 42°C for 50 min; RNA was then degraded by incubating the reaction at 90°C for 10 minutes. 40 cycles of qRT-PCR were performed using the SYBR Green PCR Master Mix (Applied Biosystem) and Applied Biosystems^™ StepOne^™ Real-Time PCR System. Each qPCR cycle consisted of DNA melting at 95°C for 15 s and primer annealing/template extension at 60°C for 1 min. The first cycle consisted of template extension at 50°C for 2 minutes followed by DNA melting at 95°C for 10 min. Changes in mRNA levels after 1-hour translation were quantified relative to amounts of mRNA extracted from cell lysate before translation as 2^−ΔCt from respective Ct (cycle threshold) values where ΔCt = Ct_{after_transiation} – Ct_{before_transiation}. Sequences of qPCR primers are provided in Supplementary Table 1. For GAPDH mRNA, four reactions with distinct pairs of primers, which spanned GAPDH mRNA sequence from 5′ to 3′ end, were used to examine mRNA integrity. Because no differences between these four PCR reactions were observed in each mRNA variant, results of four reactions were combined to calculate 2^−ΔCt. Only one pair of primers was used to examine GFP mRNA levels.

3′ RACE assay.

Analysis of poly(A) tail length was performed using the Poly(A) Tail-Length Assay Kit (Thermo Fisher Scientific) following the manufacturer′s instructions. Briefly, 500 ng of RNA was used in a G/I tailing reaction in which poly(A) polymerase adds a limited number of guanosine and inosine residues to the 3′ end of poly(A)-containing mRNAs. After reverse transcription using the newly added G/I sites as priming sites, cDNAs of interest were probed by PCR performed with a gene-specific forward primer designed upstream of the polyadenylation site and a universal reverse primer that includes the poly(A) tail of the gene of interest. For GAPDH mRNAs, primer GAPDH_seqF3 (5′ AGAGCACAAGAGGAAGAGAGA 3′) was used as the gene-specific forward primer; for GFP mRNAs, primer GFP_F_CB0721 (5′ TATACATCACGGCAGACAAACA 3′) was used as the gene-specific forward primer.

• The role of mRNA secondary structure in cap-poly(A) translational synergy is unclear.

• We show that unstructured sequences in the 3’ UTR decrease mRNA translation.

• Unstructured sequences in the 3’ UTR diminish ability of the poly(A) tail to stimulate translation.

• Base pair formation by the 3’ UTR enhances cap-poly(A) synergy in translation.