5’UTR G-quadruplex structure enhances translation in size dependent manner

Translation initiation in bacteria is frequently regulated by various structures in the 5’ untranslated region (5’UTR). Previously, we demonstrated that G-quadruplex (G4) formation in non-template DNA enhances transcription. In this study, we aimed to explore how G4 formation in mRNA (RG4) at 5’UTR impacts translation using a T7-based in vitro translation system and in E. coli. We showed that RG4 strongly promotes translation efficiency in a size-dependent manner. Additionally, inserting a hairpin upstream of the RG4 further enhances translation efficiency, reaching up to a 12-fold increase. We found that the RG4-dependent effect is not due to increased ribosome affinity, ribosome binding site accessibility, or mRNA stability. We proposed a physical barrier model in which bulky structures in 5’UTR prevent ribosome dislodging and thereby increase the translation output. This study provides biophysical insights into the regulatory role of 5’UTR structures in bacterial translation, highlighting their potential applications in tuning gene expression.


Introduction
Gene expression is a tightly regulated process to ensure e cient utilization of resources and adaptation to changing environments.This regulation occurs at various levels, including transcription, translation, and the level of mRNA and protein 1,2,3,4,5,6,7 .In bacteria, the absence of a nuclear membrane necessitates rapid post-transcriptional regulations to enable quick responses to the environmental stimuli 8,9,10,11 .Untranslated regions (UTR) of RNA have emerged as key players in regulating translation initiation by presenting noncanonical structures to translational machinery or by recruiting proteins and enzymes that recognize RNA sequences, modi cations, or structures 12,13 .The 5' untranslated region (5'UTR) of bacterial mRNA serves multiple functions that are critical for gene regulation and protein synthesis.First, it typically contains a conserved AG-rich Shine-Dalgarno (SD) sequence, located a few nucleotides upstream of the translation start site (TSS).The SD sequence basepairs with the 16S ribosomal RNA (rRNA) to guide the binding of the small ribosomal subunit, providing a well-de ned mechanism for initiating translation 14 .Second, bacterial 5'UTRs often harbor cis-acting regulatory elements, such as upstream open reading frames (uORFs), which stall the ribosome and control the access to downstream TSS.Additionally, bacterial 5'UTRs can serve as a platform for RNAbinding proteins and small RNAs that regulate translation e ciency.For example, small RNA coupled with an RNA binding protein Hfq can bind a 5'UTR to stimulate translation initiation or trigger mRNA degradation 10,15,16 .Furthermore, secondary structures within 5'UTR play a critical role in regulating RNA stability and translation e ciency.Speci cally, co-transcriptionally folded structures can in uence translation initiation and the rate of translation.Previous research on bacterial 5'UTR primarily focused on the ribosome binding site (RBS), which includes the SD sequence and a short range (10-20 nt) upstream and downstream of the SD region 14,17 .Several studies showed that secondary structures, such as pseudoknots and hairpin stem-loop that form across the SD sequence can inhibit translation by preventing ribosome binding 18,19 .Notably, temperature sensitive hairpins in the 5'UTR of E. coli can regulate translation by masking or unmasking RBS or start codon (AUG) to turn on or off translation initiation, respectively 20 .Similarly, riboswitches control translation through changes in mRNA conformation upon ligand binding, enabling rapid responses to environmental cues 21,22 .Speci c sequence elements, such as purine-rich regions or G-quadruplexes, can also affect translation in a context-dependent manner.Previous studies demonstrated that an RNA G-quadruplex (RG4) structure located near the SD sequence inhibits the base pairing between 16S rRNA and the mRNA 23,24 .However, it remains unclear how and to what extent the 5'UTR structure impacts bacterial gene regulation.
Our previous research has uncovered the role of potential G-quadruplex sequence (PQS) in non-template DNA in promoting transcription through co-transcriptional formation of R-loop and G4 structure 25 .In this context, the transcribed mRNA bears a G4 structure at the 5' end, which prompted us to investigate whether such G4 structures in RNA modulates translation outcome.G-quadruplexes form in single stranded DNA or RNA that harbors repetitive runs of guanines interspersed with non-guanine, loop sequences.Four guanine bases come together in a coplanar arrangement to form a tetrad, which stacks in multiple layers.The size and stability of the structure depend on the composition and the length of the loops 26 .Increasing evidence suggests that RG4 structures are involved in translation regulation in eukaryotes, often blocking translation initiation when present in the 5'UTR 27,28,29,30,31 .Some RG4 structures have also been shown to function as Internal Ribosome Entry Sites (IRES), stimulating translation independent of a start site 32,33,34 .Furthermore, computational studies have identi ed numerous PQS across prokaryotic species positioned non-randomly in non-coding RNA segment that precedes mRNA, indicating an evolutionarily conserved function of 5'UTR RG4 in bacterial genome 35,36,37 .While some studies demonstrate reduced translation by the 5'UTR RG4, it is not clear if different sequence composition and hence structure of RG4 would drive a different type and level of translational regulation.
In this study, we investigated the role of 5'UTR RG4 structures in E. coli translation.We inserted a series of PQSs in non-template DNA, upstream of a GFP reporter gene to allow for the formation of RG4 at 5'UTR.
Using the T7 expression system, we measured in vitro transcription and subsequent translation in realtime, which were used to calculate the translation e ciency.We found that the presence of RG4 in the 5'UTR led to enhanced translation both in vitro and in E. coli.Longer loops within RG4 resulted in higher translation yield.Moreover, insertion of a hairpin upstream of an RG4 further increased translation.Taken together, we demonstrate that the translation enhancement scales with the size of the 5'UTR structures.
We propose a mechanism by which the 5'UTR structures act as a physical barrier that may prevent dissociation of ribosome from the mRNA and thereby promote translation.

RNA G4 increases translation e ciency
To quantify the translation e ciency, we set up an in vitro translation assay which contains reagents for the T7 RNAP for transcription and E. coli translation system (Fig. 1a) 25,38 .We prepared DNA construct with T7 promoter followed by the ribosome binding site (RBS), a uorescent reporter which encodes superfold GFP (sfGFP), and a transcription terminator sequence.We used our previously established protocol to measure the real-time transcription; DNA molecular beacon becomes uorescent when annealed to a transcribed RNA which bears a complementary sequence (Fig. 1b) 39 .In parallel, the intensity of sfGFP was obtained and plotted as a translation read-out (Fig. 1c).Hence, the real-time transcription and translation activities can be simultaneously measured by collecting intensities of the molecular beacon and GFP over time using a plate reader.The GFP signal is expected to rise after that of the molecular beacon because of the time delay between transcription and translation and the maturation time required for the sfGFP folding 40 .Based on the simultaneous measurement, we can calculate the translational e ciency for each reaction by normalizing the translation signal (sfGFP) by the transcription signal (molecular beacon).
To examine the effect of 5'UTR RNA G-quadruplex (RG4) on translation, we inserted a potential Gquadruplex forming sequences (PQS) in between the T7 promoter and the RBS such that the PQS is 44 bp downstream of T7 promoter and 43 bp upstream from the RBS.The PQS was inserted into either a template (T) or a non-template (NT) strand for comparison.Hence, the PQS insertion in NT is expected to produce the G4-bearing transcript which can fold into RG4 while the PQS in T, and the scrambled control (C) sequence will not fold into RG4 (Fig. 2a).In agreement with our previous study, the PQS-NT led to approximately 30% higher transcription e ciency than in PQS-T (Fig. 2b) 25 .The initiation rate of transcription was quanti ed by taking the linear increase of uorescence intensity in the initial phase of each curve (Fig. 2c).Surprisingly, the translation reporter, sfGFP signal revealed that the PQS-NT induced over four-fold higher protein product compared to the control (Fig. 2d).The translation e ciency (TE) was calculated by dividing the initiation rate of sfGFP signal by the transcription initiation rate for each condition.(Fig. 2e).The 4-fold difference observed in translation cannot be explained by the 30% difference in transcription between the NT and T, suggesting an additional mechanism that promotes translation post-transcriptionally.Due to the PQS orientations, we expect that the mRNA from NT, but not T or control forms an RG4 structure at 5'UTR.Thus, we tested for the RG4 formation by applying N-methyl mesoporphyrin IX (NMM) in the transcription reaction 41 .NMM is a G4 ligand that exhibits induced uorescence upon binding G4 42 .As expected, the NMM uorescence displayed a prominent increase over the transcription time in NT, but not in T and C (Fig. 2f), indicating a progressive and robust formation of RG4 exclusively in NT condition.The selective NMM uorescence for NT suggests that RG4 is likely responsible for the enhanced translation since the three constructs differ only by the PQS region.In addition, the half-life obtained from the uorescence increase re ected the order of events i.e., transcription signal increased rst, followed by the NMM intensity re ecting the RG4 formation, and the GFP intensity (Fig. 2g).This further supports that the RG4 is responsible for the enhanced translation.

Bulkiness of RG4 drives translational enhancement.
PQS can vary in its sequence composition, which gives rise to diverse conformations of varying stability, and bulkiness 41,43 .Based on the result obtained above, we asked if different PQS sequences produce various levels of translational enhancement.To focus on the effect of G4, we removed the sequence upstream of PQS and inserted a series of PQS with varying loop length without changing the guanine triplets (Fig. 3a).Since RNA G4 primarily folds in parallel conformation in which all the guanine strands run in the same orientation 44 , we envision that the loop sequences will protrude out from the central tetrad core.The NMM based assay revealed that transcription of all PQS-NT sequences produced RG4 (Supplementary Fig. 1a-c).Despite varying levels of NMM signal acquired for different RG4s, single molecule FRET assay displayed that both small (short looped) and large (long looped) RG4 form a stable G4 structure without structural dynamics (Supplementary Fig. 1d, e).All PQS-NT which result in RG4 containing RNA consistently led to higher translation product than its counterpart PQS-T construct (Fig. 3b).Surprisingly, PQS-NT sequences led to 4-fold higher translation than the control (Fig. 3c), strongly re ecting the role of RG4 in promoting translation.Next, to examine the relationship between the RG4 sequence and the translation level, we plotted the translation e ciency against the total loop length of each RG4.Strikingly, the loop length of RG4 is highly correlated to the translation e ciency with a correlation coe cient of r = 0.89, indicating that the longer loop length which likely represents higher bulkiness of individual RG4 structure drives translation enhancement (Fig. 3d).Overall, our data demonstrate that all RG4 structures elevate translation and the bulkiness of RG4 accentuates the translational enhancement.
Hairpin and RG4 structures synergistically promote translation.
By comparing the results presented in Fig. 3C and Fig. 2, we noticed that despite the same PQS sequence, the translation enhancement was higher in Fig. 2. Upon close examination, we recognized that there is a 10 bp hairpin forming sequence located upstream of PQS in the construct used in Fig. 2, but not in Fig. 3C.This observation led us to test if an additional 5'UTR structure can further enhance translation.
To investigate the effect of two tandem structures, we divided the 5'UTR into four segments (Fig. 4a and Supplementary table 1); upstream of RG4 (1), RG4 (2), downstream of RG4 (3), and RBS to start codon (4).We applied an RNA structure prediction tool (UNAfolds 45 ) to calculate the folding energies of all positions except position 2, because RG4 folding cannot be accurately predicted by currently available tools.The folding energy, ΔG estimated for the positions 3, 4, and 3 + 4 were − 5.6, 0.9, and − 11.8 kcal/mol, respectively, indicating weakly folded state of the downstream sequence (Supplementary table 1).To avoid interference with the ribosome binding, we decided to vary sequence located upstream of PQS.We note that the position 1 sequence used in Fig. 3 has a low folding energy (ΔG = -3.1 kcal/mol) (see Supplementary table 1), which most likely stays unfolded at 37 ℃; we named the construct 0 hp henceforth.To study the impact of hairpin in modulating translation, we introduced hairpins of 4 bp, 8 bp, 10 bp and 12 bp stem length which are expected to have folding energy (ΔG) of -5.9, -10.4,-15.9, and − 21.4 kcal/mol, respectively (Fig. 4a).We inserted each hairpin at position 1, upstream of the RG4 sequences.Following the trend seen in 0 hp cases (Fig. 3), the translational e ciency for each hairpin group exhibited an RG4 size dependence (Fig. 4b and c), indicating that the RG4 structures remained in different hairpin constructs.In addition, the translation was further enhanced as a function of the hairpin length (Fig. 4b).For example, the translation of cMyc increased from 1.4 to 5 folds for 0 hp and 12 hp, respectively and that of 199 increased from 3.3 to 8.5 folds for 0 hp and 12 hp, respectively.However, none of the hairpins enhanced translation by itself in the absence of RG4 (the control of each hairpin), suggesting that single hairpin structure cannot drive the translation enhancement independently.Next, we tested tandem hairpins by placing 6, 10, 14 and 17 hp (folding energies ΔG: -6.5, -21.0, -31.3, -42.1 kcal/mol, respectively) in addition to the 10 hp (Supplementary Fig. 2).We found that the two hairpins enhance translation only to a level of a single RG4 (cMyc) regardless of the folding energy (Supplementary table 1).This suggests that the enhancement was induced by the structure rather than the folding energy and that the RG4 is more potent than the hairpins in promoting translation.Indeed, the RG4 loop length-dependent translational e ciency is exhibited for all hairpin variants (Fig. 4c).We compiled and projected all the results to a 2-D heatmap which presents a distinct trend that translation strength is highly correlated with both the hairpin stem size (vertical axis) and G4 loop lengths (horizontal axis) (Fig. 4d).Taken together, hairpin contributes to enhanced translation only when present with RG4.This nding raises an intriguing question about the underlying mechanism by which 5'UTR structure upstream of RBS enhances translation.
What is the mechanism that enables 5'UTR structures to elevate translation in a size dependent manner?We reasoned that the 5'UTR structure may have an impact on either the ribosome or the mRNA.To de ne the mechanism, we set out to test four hypotheses: i) RG4 increases the ribosome binding a nity (Fig. 5a); ii) RG4 improves the accessibility of RBS to ribosome (Fig. 5e); iii) RG4 increases the mRNA lifetime (Fig. 6a); iv) RG4 stabilizes the ribosome-bound state (Fig. 6c).

RG4 does not attract ribosome
We tested the hypothesis that RG4 enhances translation by increasing the a nity to ribosomes, perhaps by acting like an IRES (internal ribosome entry site) in viral translation 46 .This hypothesis posits that the mRNA containing RG4 structure (based on PQS-NT) will recruit more ribosomes than the mRNA without RG4 (C, PQS-T) (Fig. 5a), resulting in higher translation.To test this hypothesis, we performed a competition assay in which RG4 bearing competitor RNA was applied to the translation reaction in molar excess.The competitor RNA constructs included a negative control, polyU 40 nt (Fig. 5b-1), an RG4 alone (Fig. 5b-2), a single strand (ss) RNA with a hairpin but without RG4 (Fig. 5b-3), an RG4 anked by the neighboring sequence found in mRNA (Fig. 5b-4), a ssRNA with RBS (Fig. <link rid=" g5">5</link>b-5) and a ssRNA with both RG4 and RBS (Fig. 5b-6).If our hypothesis is correct, we expect to see reduced translation in conditions 2, 4, and 6, all of which contain RG4.We con rmed that there was no signi cant difference in transcription rate among the six conditions, indicating that the competitor RNAs did not affect the overall transcription (Fig. 5c).The translation result revealed that only the RBS containing RNA (5 and 6) lowered the translation signi cantly while the other conditions, (1-4) did not (Fig. 5d), suggesting that the RBS containing competitor, but not the RG4 bearing strands competed for the ribosome binding, thus lowering the translation.Therefore, we show that the RG4 does not increase a nity toward ribosome.

RG4 does not increase RBS accessibility
Next, we hypothesized that the formation of RG4 increases RBS accessibility to ribosome by preventing RBS from folding into an inaccessible secondary structure (Fig. 5e).To examine the accessibility of the RBS, we applied a molar excess of a molecular beacon that bears sequence complementary to RBS (Fig. 5f).The beacon uoresces upon hybridizing to the RBS; hence the intensity of the beacon represents the accessibility of the RBS.We tested the accessibility in two ways.First, we performed titration of puri ed transcript of NT, C and T to a xed concentration of molecular beacon (400 nM).The dissociation constant, K d was 367 (± 25), 364 (± 23), 370 (± 21) nM for NT, C and T respectively, indicating negligible difference in RBS accessibility among the three constructs (Fig. 5g).Second, we applied the molecular beacon before and after heating up the transcript from NT, C and T to test if the heat induced unfolding will increase the RBS accessibility.Despite the overall increase, the similar intensities of molecular beacon among NT, C and T indicated that RBS is accessible regardless of RG4 (Fig. 5h).Hence, both assays corroborate to re ect that RBS is fully accessible in all three constructs.Therefore, the RG4 unlikely acts via making the RBS accessible to ribosome loading.

RG4 does not increase mRNA lifetime
Next, we hypothesized that RG4 increases the lifetime of the mRNA by stabilizing the mRNA (Fig. 6a) since secondary structures on RNA can increase RNA lifetime by preventing RNA degradation 13 .We performed RT-PCR to compare the mRNA from NT, C and T after three hours of translation.The highly similar mRNA levels tested by two different sets of primers in NT, C and T provide robust evidence that RG4 does not play a role in stabilizing mRNA (Fig. 6b).

RG4 may stabilize ribosome bound to mRNA
The negative results obtained in the rst three hypotheses strongly suggest that the effect of RG4 in translational regulation must occur after the ribosome loads on the mRNA.This raises a possibility that RG4 may promote translation by stabilizing the ribosome bound to mRNA, perhaps by preventing the ribosome from dislodging (Fig. 6c).To test the hypothesis, we applied an RNA helicase, DHX36 (or RHAU) to unfold the RG4 structure during the translation reaction (Fig. 6d).DHX36 is a well-studied RG4-speci c helicase which should effectively remove the RG4 structure formed in mRNA 47,48,49,50,51 .Previously, we reported an ATP dependent repetitive unwinding mechanism by which DHX36 unfolds RG4 using singlemolecule FRET 52 .In agreement with our previous nding, DHX36 displayed a strong a nity to RG4 and unfolded the structure even at sub nanomolar concentration (Supplementary Fig. 3).Strikingly, when applied to the translation reaction, DHX36 reduced translation in a dose-dependent manner (Fig. 6e) without impacting the transcription (Fig. 6f and 6g), strongly suggesting that RG4 structure is responsible for the increased translation.We propose that the RG4 acts as a physical blockade which stabilizes the ribosome bound state by preventing ribosome from dislodging from the mRNA.
Translation enhancement pattern is observed in E. coli.
Next, we asked if the hairpin and RG4 mediated translation enhancement also operates in E. coli.Unlike the cell-free translation system which only contains essential reagents for transcription and translation, cellular environment is enriched with other proteins, including helicases, RNA binding proteins and RNases that can modulate the gene expression process and thus change the translation enhancement effect by the 5'UTR structures 53,54 .To test this, we prepared a dual-color uorescence plasmid reporter.The T7 promoter-GFP was built with the 5'UTR structures for an experimental readout whereas the T7 promoter-mCherry was constructed without 5'UTR elements to serve as an internal control (Fig. 7a).The GFP expression was normalized against the mCherry expression to obtain the relative translation yield for various 5'UTR sequences (Supplementary Fig. 4a-c).The GFP expression was con rmed and visualized by uorescence imaging (Fig. 7b).Later, we quanti ed the GFP and mCherry expression by acquiring realtime uorescence and absorbance (A 600 ) which were recorded simultaneously by the plate reader.The GFP intensity displayed the same orientation dependence of NT > C > T as we observed in vitro (Fig. 7c), while the mCherry intensity remained similar (Fig. 7d), suggesting that the same RG4 dependent translation enhancement occurs in E. coli cells.By using RT-PCR analysis, we con rmed that the difference is not based on the mRNA expression (supplementary Fig. 4d).We cloned six sets of plasmids with varying PQS sequences inserted either in NT or T and quanti ed the translation e ciency.The result re ects the same pattern as before; PQS-NT produces higher GFP signal than PQS-T and the translation further increases when bulkier PQS is inserted in NT (Fig. 7e).Taken together, our data indicates that the 5'UTR structure dependent translational enhancement exists both in vitro and in cells.

Discussion
Here we applied a real-time transcription-translation coupled assay (Fig. 1) to demonstrate the impact of RG4 at 5'UTR in promoting translation (Fig. 2).This enhancement is highly correlated with the total loop length i.e the size of the RG4 (Fig. 3) and such effect is further accentuated when a hairpin structure is added in tandem (Fig. 4).We demonstrate that the RG4 mediated translation enhancement is not due to elevated a nity to ribosome, increased accessibility of RBS (Fig. 5), or improved stability of the mRNA (Fig. 6 top).The RG4 structure is the key to promoting the translation as the helicase induced unwinding completely abolished the effect (Fig. 6 bottom).We propose that RG4 serves as a physical blockade that prevents the ribosome from falling off the mRNA and thereby directing it toward the protein synthesis.In addition, we demonstrate that the same mechanism operates in E. coli cells (Fig. 7), suggesting its potential application for controllable gene expression in E. coli.
We nd that RG4 with longer loops induces higher translational enhancement (Fig. 3).While longer loops can enlarge the overall size of the RG4, they can weaken the folded state of the G4, based on the studies done for DNA G4 i.e. longer loop lengths lead to less stable folding of DNA-G4 due to lower folding energy 41,43 .We tested whether the folded state of RG4 varies between the short loop (111) and the longer loop (199) RG4 via smFRET assay.Surprisingly, both RNAs showed a steady high FRET state, indicating a stably folded G4 structure (Supplementary Fig. 1d, e).This evidence suggests that the G4 folding in RNA is inherently more stable than the G4 in DNA.To further weaken the RG4, we mutated the middle guanine to adenine to destabilize the core of RG4.Strikingly, the mutation constructs still showed a similar enhancement effect as the original RG4 (Supplementary Fig. 5), re ecting that the mutated RNA can still fold into a structure that can increase translational output.Taken the data together with the dualhairpin result (Supplementary Fig. 2), translational enhancement up to 4-6 folds increase occurs regardless of the structure, but the strongest effect requires both stable folding and large size of the structures, for example, 177, 555 and 199 with a hairpin with a long stem.
The effect of RG4 depends on its location.We found that translation was abolished when PQS was positioned at 10 bp upstream of the SD sequence (data not shown), likely by blocking ribosome subunit association to the RBS.This agrees with Holder and Hartig's previous work which demonstrated that inserting G-quadruplex 20 bp upstream of the start codon decreased translation e ciency without changing transcription.Again, G4 likely inhibits the interaction of the 16S ribosomal RNA with the SD region 23,24 .Therefore, we inserted G4 at 46 bp upstream to SD to prevent such structural inhibition.Whether the level of enhancement relies on the distance between RG4 and RBS, or RG4 and other structures warrants future study.
Our observation may be partially explained by the "standby model" 55,56 , in which an upstream hairpin provides a temporary position for the ribosomal 30S subunit to stay, waiting for the unfolding of the SD sequence.According to this model, the rate-determining step becomes the recruitment of ribosome subunit, which is slower than the waiting time for unfolding.In our case, however the extended sequence adjacent to SD (position 3 in Fig. 4a) is predicted to have a folding energy of -5.6 kcal/mol, which falls within the energy proposed by the standby model (less than − 10 kcal/mol) 55 .Nevertheless, the current version of the standby model only accounts for a hairpin structure near the SD sequence, which cannot be extended to the effect of the RG4 structure positioned far from the RBS.Therefore, we propose that the RG4 structure with or without the hairpin may play a role of a blockade in preventing ribosome from falling off the mRNA.
We also considered that the translation enhancement may result from the coupled transcription and translation in E. coli where RNA synthesis by RNAP facilitates the recruitment of ribosomal subunits and initiate translation before the transcription is terminated i.e. co-transcriptional translation (CTT) 57,58 .However, in our system, CTT is unlikely because T7 RNAP transcription rate (220 ~ 230 nt/s) 59,60 is signi cantly higher than E. coli ribosome translation rate (42-51 nt/s) 61 .That is, ribosomes will be loaded onto a nascent RNA post RNA synthesis rather than being coupled to transcription.In order to examine the effect in the presence of CTT, we also cloned the same 5'UTR sequence and GFP gene to an E. coli promoter P L−LacO system.Surprisingly, we still observed a huge increase of GFP signal in nontemplate construct than in template (Supplementary Fig. 6), suggesting the structural effect may still function in regular E. coli gene under CTT.However, the mechanism in a pure E. coli system should be studied more systematically in the future.
To summarize, we have demonstrated a size-dependent 5'UTR structure effect in translation enhancement in T7 and E. coli system.We focused primarily on RNA G-quadruplex and hairpin structures.
Although we examined varying sizes of both the RG4 and hairpin, our study opens a wide window of opportunity for future studies, for example, investigating the role of a pseudoknot structure positioned either upstream or downstream of RG4 or in between RG4 and RBS.Furthermore, we want to note that in all our experiments, PQS insertion into template (PQS-T) which produces C-rich transcript, strongly suppressed the translation e ciency both in vitro and in vivo experiments.This suggests an opposite function of C-rich RNA in down-regulating the gene expression.In conclusion, our study provides a new mechanism and function of 5'UTR mRNA in bacterial translation and provides a novel cloning scheme for tunable gene expression system.Declarations overnight product was rstly digested by DNase I (0.1 U/µL) in DNase reaction buffer (10 mM Tris-HCl pH 7.6, 2.5 mM MgCl 2 , 0.5 mM CaCl 2 ) at 37℃ for 30 minutes.Later, the reaction was quenched by adding 1 µL 0.5 M EDTA, followed by inactivating at 75℃ for 10 minutes.RNA was puri ed by Monarch RNA CleanUp kit (NEB).

In Vitro Translation Assay
Ensemble in vitro translation assay was conducted by PURExpress In Vitro Translation kit (NEB) and performed by TECAN Spart plate reader at 37℃.Each reaction (25 µL) was premixed with 55 ng linear DNA (4 nM), 400 nM molecular beacon (see Supplementary Table 2), RNase inhibitor murine (0.8 unit/ µL), and 10 µL Solution A. To measure the RG4 formation, NMM was added in premixed solution at nal concentration of 1 µM.The reaction was initiated by adding 7.5 µL Solution B and loaded on a 384-well plate (white and transparent bottom, Thermo Scienti c).The Cy3, GFP, and NMM were excited at λ ex 545, 485, and 393 nm and detected at λ em 570, 510, and 610 nm, respectively.Both excitation and emission were assigned with 10 nm slit size.The initiation transcription rate was quanti ed from the linear part (10 to 25 minutes) of the Cy3 intensity curve.The translation rate was quanti ed from the linear part (25 to 50 minutes) of the GFP intensity curve.Each rate was normalized to the transcription rate and translation rate of 10 hp control DNA construct, respectively.The normalized translation e ciency was calculated by dividing the normalized translation rate to the normalized transcription rate.The half-life was de ned by the time that the intensity reached 50 percent of the plateau.

In vitro Transcription NMM Assay
Ensemble in vitro transcription for real-time NMM measurement was performed by TECAN Spark plate reader at 37℃.Each sample was prepared with 1 nM linear DNA template in transcription buffer (40 mM   Tris-HCl pH 8.3, 50 mM KCl, 6 mM MgCl 2 , 2 mM spermidine, 1 mM dithiothreitol), RNase inhibitor murine (0.4 unit/µL), T7 RNA polymerase (1.25 unit/µL), and 1 mM NMM.The reaction was initiated by adding NTP mix for a nal concentration of 1 mM.Each reaction (100 µL) was loaded on 96-well transparent plate (Thermo Scienti c).The data was collected at λ ex 393 nm (slit size 10 nm) and λ em 610 nm (slit size 10 nm).For emission spectrum, the data was collected at λ ex 393 nm (slit size 10 nm) and λ em 580-650 nm (slit size 10 nm).

RNA G4 Competition Assay
RNA competition assay was modi ed from in vitro translation assay by adding RNA competitors.PolyU 40 and cMyc RG4 were ordered from IDT.Other RNAs were synthesized from the 5'UTR of 10 hp control and 10hp cMyc DNA.The DNA templates were PCR-ampli ed by T7 promoter primer, short-length primer 1, and long-length primer 2 (see Supplementary Table 2).RNA puri cation protocol was described in RNA Preparation.The reaction was performed with 55 ng of 10 hp cMyc-NT DNA and 5 µM of competitor RNA by plate reader at 37℃.

RBS Probe Binding Assay and Accessibility Assay
A molecular beacon complementary to the ribosome binding site (RBS) was ordered from IDT and labeled by Cy5 and quencher at each end (see Supplementary Table 2).The Cy5 intensity of beacon was measured at λ ex 640nm and λ em 665nm with 10 nm slit size.The binding assay was conducted by incubating 400 nM beacon and RNAs of 10hp control, T, and NT at 37℃.RNA puri cation protocol was described in RNA Preparation.The data points were collected by titrating RNA concentration in a 2-fold series dilution from 1.75 µM to 24 nM.The K d was tted to the binding curve by OriginPro.For accessibility assay, each sample (10 µg of RNA) was mixed with 400 nM of the RBS beacon and incubated with or without heating treatment.The heated samples were incubated at 80℃ for 5 minutes and cooled by 10℃ intervals for 5 minutes until 37℃.After incubation, intensities of all the samples were measured by TECAN plate reader at 37℃.

DHX36 puri cation and titration experiment
The E. coli strand with DHX36 plasmid was made in the lab and described in a previous publication 52 .The E. coli was inoculated in TB medium and grew overnight.Next day, the culture was diluted to OD 600 of 0.01 and grew till OD 600 of 0.6 at 37℃ and induced protein expression with 1 mM IPTG at 14℃ for overnight.The puri cation protocol followed previous publications 52 .Protein concentration was quanti ed by standard BSA (NEB) calibration curve by SDS-PAGE, and the aliquots of protein samples were stored in -80℃.For DHX36 titration assay, the protein was diluted by TNM buffer (10 mM Tris-HCl, pH 8.0; 50 mM NaCl; 5 mM MgCl 2 ) to avoid the change of reaction volume and buffer condition.1 µL of each titrated DHX36 and additional 1 mM ATP were premixed with solution A, and the reaction was initiated by adding solution B. The experimental protocol is described in the In Vitro Translation Assay.

E. Coli dual uorescence assay
PQS-contained dual uorescence plasmids, described in DNA Preparation, were transformed into BL21(DE3) E. coli, and grew in LB medium at 37°C overnight.The cultures were diluted to OD 600 of 0.01 and grew at 37°C by TECAN plate reader with a 24-well transparent plate (1 mL for each culture).OD 600 measurements were performed every 10 min, followed by an orbital shaking mode (215 rpm) for a 10 min interval.The gene expression was induced by 1 mM IPTG at OD 600 of 0.4.After the induction, the protein expression was monitored by an auto-loop measurement of 10 min shaking, OD 600 , GFP (λ ex 485/ λ em 510, slit 10 nm), and mCherry (λ ex 585/ λ em 610, slit 10 nm) for 10 hours.The translation e ciency was de ned by the ratio of GFP to mCherry and normalized to real-time OD 600 (Supplementary Fig. 4).The ratio of 210 min after induction represented the maximal e ciency of each strain.The cultures after 210min induction were collected and extracted mRNA for RT-qPCR.The pellet was treated with 20 mg/mL lysozyme, and the mRNA was extracted by using Qiagen RNeasy Kits.The primers of qPCR were listed in Supplementary Table 2.
smFRET Assay The PQS RNA oligonucleotides (see Supplementary Table 2) for smFRET were purchased from IDT with terminal amine modi cation for Cy3 labeling.The 18-mer RNA primer for immobilizing PQS RNA on slides was purchased from IDT and later labeled Cy5.The labeling protocol was described in a previous publication.RNA samples were annealed in TE buffer (10 mM Tris-HCl and 1 mM EDTA, pH 8) at the ratio 1:1.The mixtures were heated at 80°C for 5 min and slowly cooled to room temperature (1°C per min).The single molecule assays were performed by using a home-built prism-type total internal re ection uorescence microscope (TIRFM) at room temperature (23.0 ± 1.0°C) 25,62 .RNA sample (10 nM) was diluted to 25 pM and immobilized on a PEG-coated quartz slide by neutravidin (0.05 mg/mL).The reaction buffer (50 mM Tris-HCl, pH 7.5; 50 mM NaCl; 50 mM KCl; 5 mM MgCl2; 5% glycerol; 80 units RNase inhibitor murine) with an oxygen scavenging system (1 mg/mL glucose oxidase, 0.8% v/v glucose, ~ 10 mM Trolox, and 0.03 mg/mL catalase).The smFRET experiment was performed by two solid-state lasers, 532 nm and 640 nm lasers.Each measurement was recorded with a 100 ms time resolution by smCamera software and analyzed with Interactive Data Language (IDL).The outputs were processed with custom MATLAB script to generate trajectories and FRET histograms.The details of smFRET data process were described in previous publications 62,63 .For DHX36 assay, protein (1 nM) or ATP (1 mM) was premixed into reaction buffer, and a ow system was applied to study real-time binding and unwinding events 63 .

Statistics Analysis
Data shown in Fig. 2-7 were obtained from individual and independent experiments.All the numbers were calculated and presented in value ± SEM.The statistics tests were calculated by two-sided paired or unpaired t-test, depending on the data.The average numbers, SEM, and statistics P-values were reported in Supplementary Table 3.    b, Competitor RNA used in translation reaction.The RG4 is PQS cMyc (Fig. 3a), and the hairpin is 10 hp (Fig. 4a).RNA sequences are provided in Supplementary Table

2
. c, Transcription rate of PQS-NT with addition of competitor RNA.Data are presented as mean ± SEM of independent experiments (n = 3).Shown in c represents the signi cance between PQS-NT and addition of competitor, where *P < 0.05 (twosided paired t-test).d, Normalized GFP intensity.Additions of RNA 1-4 show no signi cant difference from PQS-NT while RNA containing the RBS (5 and 6) reduces the GFP production, suggesting successful competition requires RBS not RG4.e-g Hypothesis two: RG4 enhances the accessibility of RBS to ribosome.e, Cartoon depicting the release of RBS from other secondary structure by RG4.f, Schematic for examining the accessibility of the RBS.The mRNA is either heated to remove all secondary structure or not heated.A Cy5-labled molecular beacon complementary to the RBS is applied to determine accessibility.g, Binding curves of the RBS molecular beacon to RNAs.The data points are collected by titrating RNA concentration in a 2-fold series dilution from 1.75 µM to 24 nM and represents the average number from independent experiments (n = 5).The dissociation constants, K d , are 367 (± 25), 364 (± 23), 370 (± 21) nM for NT (blue rectangle), Control (gray circle), and T (purple diamond), respectively, showing a negligible difference in binding a nity (two-sided paired t-test).h, Fluorescence intensities of molecular beacon with and without heating.There is no signi cant difference among the control, T, and NT with or without heating.Data are presented as mean ± SEM of independent experiments (n = 3).NS: nonsigni cant (two-sided paired t-test).