Single-nucleotide control of tRNA folding cooperativity under near-cellular conditions

Significance

RNA structures control many essential functions of life, including gene expression, protein synthesis, and development. To perform these functions, RNAs often adopt complex tertiary folds. Here, we study the thermodynamics of late-transcriptional folding of eukaryotic and prokaryotic tRNAs. In near-cellular conditions, tRNAs gain folding cooperativity only when nearly all nucleotides in the acceptor stem are transcribed. Furthermore, the native 5′ and 3′ extensions in the precursor do not affect the folding pathway. Therefore, tRNA folding cooperativity is controlled by a single nucleotide under near-cellular conditions.

Abstract

RNA folding is often studied by renaturing full-length RNA in vitro and tracking folding transitions. However, the intracellular transcript folds as it emerges from the RNA polymerase. Here, we investigate the folding pathways and stability of numerous late-transcriptional intermediates of yeast and Escherichia coli transfer RNAs (tRNAs). Transfer RNA is a highly regulated functional RNA that undergoes multiple steps of posttranscriptional processing and is found in very different lengths during its lifetime in the cell. The precursor transcript is extended on both the 5′ and 3′ ends of the cloverleaf core, and these extensions get trimmed before addition of the 3′-CCA and aminoacylation. We studied the thermodynamics and structures of the precursor tRNA and of late-transcriptional intermediates of the cloverleaf structure. We examined RNA folding at both the secondary and tertiary structural levels using multiple biochemical and biophysical approaches. Our findings suggest that perhaps nature has selected for a single-base addition to control folding to the functional 3D structure. In near-cellular conditions, yeast tRNA^Phe and E. coli tRNA^Ala transcripts fold in a single, cooperative transition only when nearly all of the nucleotides in the cloverleaf are transcribed by indirectly enhancing folding cooperativity. Furthermore, native extensions on the 5′ and 3′ ends do not interfere with cooperative core folding. This highly controlled cooperative folding has implications for recognition of tRNA by processing and modification enzymes and quality control of tRNA in cells.

RNAs adopt unique 3D structures that allow them to perform essential functions in the cell, including catalysis, protein synthesis, and gene regulation. RNA folds in a hierarchical manner, wherein secondary structure forms before higher-order tertiary structure (1). In classical in vitro solution conditions of high salt, typically 1 M Na⁺ (2) or >10 mM Mg²⁺ (3), transcripts can get trapped in very stable misfolded structures that can persist for minutes to hours (4). However, recent studies (5, 6) have shown that in near-cytoplasmic salt of just ∼140 mM K⁺ and 0.5 to 2.0 mM free Mg²⁺ (7–10), functional RNAs fold in a two-state, cooperative manner with less population of intermediates over a smoother landscape.

We chose to investigate the native folding of transfer RNA (tRNA), as it is one of the most prevalent RNAs in cells. The native structure consists of four stems—the acceptor stem, D stem, anticodon stem, and TΨC stem—and has extensive tertiary interactions between the D-stem loop and TΨC-stem loop, with the acceptor-stem loop and anticodon-stem loop not engaged in any tertiary interactions (Fig. 1A). Transfer RNAs are transcribed as precursor molecules with 5′-leader and 3′-trailer sequences that are removed in the nucleus by RNase P at the 5′ end and multiple nucleases at the 3′ end (11). In yeast, 3′-CCA is added posttranscriptionally by nucleotidyl transferase. During the posttranscriptional processing, tRNA is modified and then matured tRNA is aminoacylated. Extensive folding studies on tRNA have been performed on the fully processed cloverleaf sequence (5, 6, 12, 13); however, to our knowledge no studies on tRNA late-transcriptional intermediates have been reported. Proper folding of the cloverleaf as well as the 5′ and 3′ extensions during transcription is important for 5′- and 3′-end processing, posttranscriptional modifications, and biological function (14, 15).

Fig. 1.

Single-nucleotide addition induces folding cooperativity in yeast tRNA^Phe intermediates in near-cellular conditions. (A) Structures of yeast tRNA^Phe and transcriptional intermediates (−11YF, −7YF, −5YF, −4YF, −3YF), which are truncated at the indicated positions. Orange dashed lines depict native tertiary interactions. (B–E) Thermal denaturation of FL yeast tRNA^Phe and transcriptional intermediates in buffer or near-cellular conditions (rows), with 0.5 or 2.0 mM free Mg²⁺ (columns). Colors and symbols for each construct are indicated.

Herein, we analyze effects of several key components of the cytoplasm on late-transcriptional yeast and Escherichia coli tRNA intermediates. Specifically, we investigate the influences of cellular ionic conditions, crowding, amino acid-chelated Mg²⁺ (aaCM), and combinations. Furthermore, the effects of the 5′-leader and 3′-trailer sequences on the folding pathway are investigated. Cooperative folding of full-length tRNA transcripts in cytoplasm mimics has been previously reported and attributed to destabilization of secondary structure and stabilization of tertiary structure (6); the molecular origin of these effects may be due to interaction of the RNA with small molecules, but needs further investigation. Thus, we hypothesized that intermediates of late-transcriptional folding pathways would also be destabilized in cytoplasm mimics, since intermediates have weak secondary structure under these conditions and might not form all tertiary contacts. Additionally, it has been reported that the 5′ stems of RNA regions forming long-range interactions are involved in nonnative pairings until the 3′-pairing region is transcribed (16). Therefore, we hypothesized that noncooperative folding would be observed in intermediates with short 3′ ends on the acceptor stem. Consistent with these notions, we observe folding cooperativity herein only under near-cellular conditions and only in intermediates with complete and strong base pairing in the acceptor stem. We also observe that the cloverleaf region in constructs with the precursor 5′ and 3′ extensions also folds in a two-state manner, with no change in cloverleaf structure. Thus, perhaps nature may have selected for the cloverleaf structure to fold in a two-state manner when the acceptor stem can form, and it does so without any apparent interference from flanking nucleotides.

Results

We previously reported that secondary structure is destabilized in crowded cytoplasm mimics (6), and recently cooperative folding of some RNAs was reported in vivo (16). Here, we investigated effects of such conditions on the thermodynamics of late-transcriptional intermediates, which involves folding as the nascent transcript emerges from the RNA polymerase. Late-transcriptional intermediates of yeast tRNA^Phe (YF) with truncations on the 3′ end of 11 nt (−11YF) to 3 nt (−3YF) were prepared (Fig. 1A). We hypothesized that −11YF lacks proper tertiary interactions because the secondary structure is weak without formation of the acceptor stem, which facilitates long-range interactions; therefore, −11YF is a good model for nonnative folding. Intermediates −7YF, −5YF, −4YF, and −3YF were chosen because tertiary structure has the potential to form as the length and stability of the acceptor stem increase, and two-state folding to the native state might occur with increasing length of the nascent transcript. Folding of these intermediates and of full-length (FL) tRNA^Phe was probed in buffer, as well as in solution conditions that mimic different aspects of the cytoplasm using the following three near-cellular conditions: 1) crowding (20% polyethylene glycol [PEG] 8000), 2) amino acids chelated to 4.1 or 14.0 mM Mg²⁺, and 3) a combination of crowding and aaCM, all in a background of either 0.5 or 2.0 mM free Mg²⁺, respectively, where “free” means Mg²⁺ that is fully coordinated by water molecules, which mimic eukaryotic and prokaryotic divalent conditions, respectively.

Single-Nucleotide Addition Induces Cooperative Folding of Yeast tRNA^Phe in Eukaryotic Near-Cellular Conditions.

We first examined folding of intermediates under eukaryotic-like ionic conditions of 0.5 mM free Mg²⁺. Intermediates without complete transcription of the acceptor stem −11YF and −7YF exhibit highly noncooperative folding in buffer as well as all three near-cellular solution conditions (Fig. 1 B and C and SI Appendix, Fig. S1 A and B). In particular, the melting curves show multiple transitions over a broad temperature range. These observations suggest that these late-transcriptional intermediates are adopting multiple conformations and/or unfolding in a highly multistate manner.

We then turned to longer intermediates that might fold into native-like structures, specifically late-transcriptional intermediates, −5YF, −4YF, and −3YF. In the presence of buffer alone, without cellular additives, broad melting transitions with multiple peaks and a low-magnitude ΔH are observed (Fig. 1B and SI Appendix, Table S1). These observations indicate multistate folding of these late-transcriptional intermediates. Notably, −4YF and −3YF unfold at 54.5 and 57.4 °C, which are lower temperatures than FL, indicating lower stability (SI Appendix, Table S1). We next tested the folding of these intermediates in near-cellular conditions of 20% crowding and aaCM. In cells, there is estimated to be between 20 and 40% molecular crowding that can affect RNA folding (5, 6). While there is 0.5 mM free Mg²⁺ in eukaryotic cells, there is also an additional ∼4 mM Mg²⁺ weakly chelated to amino acids, which can enhance RNA folding and catalysis (17, 18) and therefore can affect the stability of late-transcriptional intermediates. In near-cellular conditions, cooperative folding is observed in intermediates −4YF and −3YF, with both constructs exhibiting just a single transition, similar to FL tRNA^Phe (Fig. 1C). The T_Ms cluster in the narrow range between 64.7 and 67.4 °C, increasing with acceptor-stem length (SI Appendix, Table S1). Given the hierarchical nature of RNA folding, this observation supports two-state, cooperative folding in which secondary structure melts out concomitant with tertiary structure, and secondary structure influences thermostability (17). Surprisingly, the intermediate with one less nucleotide (−5YF) folds in a highly noncooperative manner (Fig. 1C), suggesting that the addition of a single nucleotide (from −5YF to −4YF) results in a large increase in stability, leading to a large increase in cooperativity.

The extent of cooperativity can be measured by the value of ∆H, where a large negative value indicates a cooperative folding transition (19). We compare the cooperativity of intermediates −4YF and −3YF with YF by the ratio of their ∆H values, ∆H_I/∆H_FL (5, 6). In buffer, ∆H_I/∆H_FL is <1, indicating differing extents of partial to noncooperative folding of late-transcriptional intermediates (SI Appendix, Table S1). In the near-cellular condition of crowding with aaCM, the ∆H values for −4YF, −3YF, and YF are similar to each other, around −110 kcal/mol, with ∆H_I/∆H_FL close to 1.0 (SI Appendix, Table S1). Thus, only under near-cellular conditions, late-transcriptional intermediates fold in a highly cooperative manner, similar to YF.

To determine if cooperativity is driven by crowding, aaCM, or a combination of the two, we tested the folding in each solution individually in a background of 0.5 mM Mg²⁺. In the presence of 20% PEG8000 alone, YF unfolds in a single transition with a more negative ∆H of −54.8 kcal/mol compared with −47.4 kcal/mol in buffer (SI Appendix, Fig. S1A and Table S1). Strikingly, late-transcriptional intermediates −4YF and −3YF are not stabilized by these same crowding conditions, as they still unfold in a multistate manner, with ∆H values of just −42.4 kcal/mol for −4YF and −22.8 kcal/mol for −3YF (SI Appendix, Table S1). Observing that crowding dramatically destabilizes late-transcriptional intermediates compared with YF, we next probed folding of the intermediates in aaCM alone. In the presence of aaCM, there is a large change in the folding transitions of the intermediates. There is now striking similarity between the melting curves of −4YF, −3YF, and FL (SI Appendix, Fig. S1B). All three curves have the same minor low-temperature transition at ∼45 °C and the high-temperature transition T_Ms are similar, over a narrow temperature range of 62.8 to 65.6 °C for −4YF to YF (SI Appendix, Table S1). The low-temperature transition, which is the same for all constructs, is attributed to the tertiary structure unfolding, and the high-temperature transition, which increases with construct length, is attributed to secondary structure. These unfolding assignments are based on the hierarchical nature of RNA folding, as well as prior assignments of tRNA folding transitions (13).

Similar to 0.5 mM Mg²⁺, the late-transcriptional intermediates −11YF (only buffer conditions), −7YF, and −5YF display noncooperative behavior in 2 mM free Mg²⁺ conditions (Fig. 1 D and E and SI Appendix, Fig. S1 C and D). Therefore, we again focus on the late-transcriptional intermediates −3YF and −4YF. In buffer with 2.0 mM free Mg²⁺, YF folds in a two-state manner, and −3YF and −4YF late-transcriptional intermediates fold in a three-state manner (Fig. 1D). Specifically, there is a single unfolding transition for YF at 65.1 °C but two distinct transitions for −3YF and −4YF composed of broad transitions at 37.8 and 39.4 °C and sharper transitions at 63.9 and 62.2 °C, respectively (SI Appendix, Table S1). As described above, the low-temperature transition is attributed to tertiary-structure unfolding and the high-temperature transition is attributed to secondary-structure unfolding, consistent with the hierarchical nature of RNA folding. Upon the addition of 20% PEG8000, YF and late intermediates fold in a two-state manner, with T_Ms of 67.9, 69.5, and 69.0 °C for −4YF, −3YF, and YF, respectively (SI Appendix, Fig. S1C and Table S1). Furthermore, with aaCM as well as the combined PEG with aaCM condition, very similar unfolding transitions of intermediates and FL are observed (Fig. 1E and SI Appendix, Fig. S1D). In both of these cellular-like conditions, a single transition is observed for −3YF, −4YF, and YF, with T_Ms clustered between 68.9 and 71.7 °C. The magnitude of ∆H_folding is slightly larger for all constructs in aaCM with a crowder than in aaCM alone, which is −126 and −107 kcal/mol, respectively, for YF (SI Appendix, Table S1). Notably, the ∆H_I/∆H_FL for −3YF and −4YF is near-unity in aaCM and in aaCM with a crowder, indicating that these constructs are unfolding just as cooperatively as full-length but only in near-cellular solutions.

Late-Transcriptional Intermediates of tRNA from Other Classes and Other Species Gain Folding Cooperativity in Near-Cellular Conditions.

To determine if single-nucleotide control of folding cooperativity is a general property for tRNA, we studied folding of E. coli tRNA^Phe (EF) (SI Appendix, Fig. S2), as well as of yeast and E. coli tRNA^Ala (YA and EA, respectively) (SI Appendix, Fig. S3 A and C). Unexpectedly, EF unfolds noncooperatively in all conditions (SI Appendix, Fig. S2B). RNA concentration-dependent thermal denaturation overlaps with each other (SI Appendix, Fig. S2C, Inset), suggesting that the noncooperative behavior does not arise from tRNA dimerization. We then probed the secondary structure of EF. Temperature-dependent in-line probing (ILP) shows unusually high stability in the anticodon stem, marked by no degradation even at high temperature, and low stability in other regions (SI Appendix, Fig. S2D). This is likely because of the successive four GC base pairs in the anticodon stem, which might prevent the RNA from folding cooperativity. Comparison of ILP reactivity at 35 °C between EF and YF shows that some regions in EF have higher ILP reactivity than yeast tRNA^Phe (SI Appendix, Fig. S4 A and B). For example, the variable loop in EF has higher reactivity. This is consistent with the crystal structure of the EF transcript, where U45 does not participate in the triplet base pairing in the tRNA core and is positioned toward solvent (SI Appendix, Fig. S4C) (20). The TΨC stem in EF also has higher reactivity, suggesting that the stem forms a different structure from the YF, which is consistent with the recent work by Kothe and coworkers where folding of the EF transcript was probed by aminoacylation and it was found that the E. coli tRNA^Phe transcript tends to have misfolded intermediates if there is no assistance of the RNA chaperone of the tRNA modification enzyme (TruB) which recognizes the TΨC-arm region (21). Thus, while the global structure of EF is likely to be similar to YF, some specific local regions might also affect the folding cooperativity of EF.

In contrast to EF, both YA and EA unfold cooperatively in the near-cellular conditions (SI Appendix, Fig. S3 B and D), similar to yeast tRNA^Phe. We then tested late-transcriptional intermediates of EA (Fig. 2). In buffer with 0.5 or 2 mM Mg²⁺, all intermediates unfold cooperatively with slightly higher values of ∆H to FL EA, suggesting that folding cooperativity of late-transcriptional intermediates of EA is comparable or higher than the FL EA (Fig. 2 B and C and SI Appendix, Table S2). Therefore, the acceptor stem is not required for folding cooperativity in buffer, and the single-stranded 5′-acceptor nucleotides do not interfere with other regions in nonphysiological buffer conditions. In the near-cellular conditions with 20% PEG8000 and aaCM, EA unfolds cooperatively and the late-transcriptional intermediate (−4EA) unfolds more cooperatively than the other intermediates, which have additional peaks showing lack of folding cooperativity (Fig. 2 D and E). This result supports our observation from yeast tRNA^Phe in near-cellular conditions.

Fig. 2.

Single-nucleotide addition induces folding cooperativity in E. coli tRNA^Ala intermediates under near-prokaryotic cellular conditions. (A) Secondary structure of tRNA^Ala intermediates. Positions of terminal nucleotides for each intermediate are boxed. (B–E) Thermal denaturation of FL E. coli tRNA^Ala and transcriptional intermediates in buffer or near-cellular conditions (rows), with 0.5 or 2.0 mM free Mg²⁺ (columns). Colors and symbols for each construct are indicated.

Cooperative Folding Is Dependent upon the Strength of the Acceptor Stem.

In the previous sections, we discussed how single-nucleotide addition, especially the final nucleotide in the acceptor stem, can induce folding cooperativity. We propose two hypotheses for this: Folding cooperativity arises from structural rearrangements in the intermediates on the path to the final folded state, or folding cooperativity is influenced by the strength of the acceptor stem.

To test the first hypothesis, changes in structure of the folded tRNAs were monitored by native gel electrophoresis, in-line probing, and small-angle X-ray scattering (SAXS) (SI Appendix, Figs. S5–S8 and Table S3). On native gels containing near-cellular concentrations of Mg²⁺, no change in compaction is observed in the late-transcriptional intermediates under all four solution conditions tested (SI Appendix, Fig. S5 A and B), suggesting no global change in structure. The SAXS data further support this (SI Appendix, Fig. S6 and Table S3), where the R_g and D_max values, measures of size and aspect ratio, are similar among the late-transcriptional intermediates and FL tRNA in buffer and near-cellular conditions. The ILP profiles of full-length and intermediates in buffer are also very similar, indicating that native final structure is adopted for all (SI Appendix, Fig. S5C). Furthermore, the ILP signal is similar for each RNA construct in dilute and near-cellular solution conditions, again providing no indication of change in structure (SI Appendix, Fig. S5 D–F). Therefore, it appears that late intermediates do not misfold globally or locally. It should be mentioned that SAXS and ILP do not allow detection of very minor populations of misfolded tRNAs, and our data support that such minor misfolded intermediates are less than the detection limit.

To address the latter hypothesis that folding cooperativity is influenced by the strength of the acceptor stem, we made two mutations to the acceptor stem of yeast tRNA^Phe intermediate constructs (Fig. 3A). In the background of intermediate −5YF, a U69C mutant was made to change a G•U wobble into a GC base pair, with the hypothesis that intermediate −5YF U69C would fold more cooperatively. In the opposite sense, in the background of intermediate −4YF, a C72U mutant was made to change a GC base pair into a G•U wobble, with the hypothesis that −4YF C72U would fold in a multistate manner. Indeed, we observe that with the U69C mutation, the −5YF construct folds in a more cooperative manner in buffer and near-cellular conditions (Fig. 3B). In contrast, the C72U mutation in the −4YF intermediate results in multistate folding of this intermediate, that was two-state without the mutation (Fig. 3C). In sum, we conclude that onset of cooperativity during late-transcriptional folding is favored by the strength and length of the closing stem, which plays an indirect role in determining folding cooperativity (i.e., there is no direct contact between the acceptor stem and the remainder of the tRNA structure).

Fig. 3.

Mutations in the acceptor stem of yeast tRNA^Phe modulate cooperativity. (A) Mutations made to the acceptor stem in truncated intermediates −5YF and −4YF. (B and C) Thermal denaturation of yeast tRNA^Phe mutants (B) −5YF U69C and (C) −4YF C72U in buffer (open circles) and 20% PEG8000 with aaCM in the background of 2.0 mM free Mg²⁺ (filled circles). The native sequences are shown in black.

We noticed that yeast tRNA^Phe, E. coli tRNA^Ala, and yeast tRNA^Ala have a G•U wobble in their acceptor stems. Varani and McClain pointed out that G•U wobbles have strong negative potential in their major grooves, which can attract metal ions (22). Inspection of the crystal structure of yeast tRNA^Phe (Protein Data Bank ID code 1EHZ) revealed an Mg²⁺ ion in the major groove of the acceptor stem near this G•U wobble. Although this ion does not bridge to the remainder of the tRNA, it still could enhance cooperativity of folding indirectly because, as described above, strength of the acceptor stem helps determine cooperativity, and binding of an Mg²⁺ ion should enhance stability of a stem. The acceptor-stem Mg²⁺ ion is not essential for cooperative folding, however, as evidenced by cooperative unfolding of −5YF U69C, which has no G•U wobble in its stem (Fig. 3B), consistent with an indirect role in the wild type as mentioned above.

5′ and 3′ Extensions Do Not Affect tRNA Core Folding Cooperativity.

In cells, tRNAs are transcribed with extensions off the 5′ and 3′ ends, and these have the potential to interact with the cloverleaf core and alter the folding and structure of the tRNA core. In yeast, tRNA^Phe is transcribed with an intron that forms a stable hairpin structure as an extension of the anticodon loop, and the intron does not interfere with the secondary or tertiary structure of the tRNA core (23). Therefore, we studied the intron-less transcript with native flanking sequences. We tested the effects of the folding of the tRNA core fused to the native 5′ leader, together referred to as the “5′-leader” construct, and of the tRNA core fused to the native 5′-leader and 3′-trailer sequences, referred to as the “precursor” construct (Fig. 4A). In buffer with 2.0 mM Mg²⁺, both of these constructs show two distinct transitions with very broad unfolding transitions that occur over ∼50 °C (Fig. 4B). The same behavior is observed in the near-cellular conditions tested above of crowding and aaCM (Fig. 4C).

Fig. 4.

Thermal denaturation supports apparent noncooperative folding of precursor yeast tRNA^Phe under near-cellular conditions. (A) Sequence of 5′-leader and precursor tRNA. The 5′-leader construct contains the 5′ leader and tRNA core, and the precursor construct contains the 5′ leader, tRNA core, and 3′-trailer sequences. (B and C) Thermal denaturation curves of 5′-leader (pink) and precursor (purple) sequences in (B) buffer and (C) 20% PEG8000 with aaCM in the background of 2.0 mM free Mg²⁺.

We hypothesized that perhaps the tRNA core was still folding in a two-state manner and that the broad transitions observed could be attributed to base unstacking in the extended regions. To uncouple changes in the 5′ and 3′ extensions from the core, we used the more detailed method of in-line probing, which reports structure at the nucleotide level rather than globally like UV-detected thermal denaturation (SI Appendix, Figs. S9 and S10). We first consider the ILP profiles at a single, physiologically relevant temperature. At 35 °C in buffer and near-cellular conditions, the ILP profiles of both the 5′-leader and precursor constructs show high signal in the 5′ and 3′ extensions, and native ILP patterns of alternating strong and weak signal in the tRNA core. Thus, in both of these solution conditions, the tRNA core is adopting the native secondary structure, while the extensions off the 5′ and 3′ ends are unfolded and not interfering with the core (SI Appendix, Fig. S11).

To gain insight into the folding pathway, the core tRNA portion of the ILP signal from the above constructs was analyzed at 12 temperatures between 35 and 75 °C. The signal of the nucleotides in each stem was globally fit as a set of melting curves to obtain a single T_M and single ΔH of folding for the entire core (Fig. 5 and SI Appendix, Fig. S12 and Table S4). We first consider the 5′-leader construct. Global fitting for a single unfolding transition was performed on all base-paired nucleotides in the 5′-leader construct in buffer and near-cellular conditions, and the fits are excellent with χ² values of 2.2 and 0.9, respectively, and errors of the T_M and ΔH less than 16% (Fig. 5 and SI Appendix, Table S4). The quality of the fits supports the conclusion that in these conditions the tRNA core is unfolding in a two-state manner. The T_M values are 59.9 and 65.5 °C and ΔH values are −129 and −146 kcal/mol in buffer and near-cellular conditions, respectively. The enthalpy value in near-cellular conditions is similar to the theoretical fully cooperative ΔH of −150.0 kcal/mol, as calculated using nearest-neighbor parameters, and close to the experimental tRNA core ΔH of −126 kcal/mol measured in near-cellular conditions by thermal denaturation (SI Appendix, Table S1). Thus, at the T_M, the secondary and tertiary structures appear to unfold together. It remains possible that at lower temperatures the secondary structure forms independent of tertiary structure, even under secondary structure-weakening physiological conditions as supported by experiments on tRNA fragments under cellular-like conditions (6).

Fig. 5.

Variable-temperature ILP supports cooperative folding of the tRNA core in the presence of the 5′-leader flanking sequence. Global fitting of variable-temperature in-line probing signal of the 5′-leader construct in buffer and physiological conditions (columns). Global fitting for a single unfolding transition was performed on all double-stranded regions (rows) simultaneously in tRNA in (A–D) buffer and (E–H) 20% PEG8000 with aaCM to obtain a single T_M and ∆H of folding for each condition, which can be found in SI Appendix, Table S4. All samples contain 2.0 mM free Mg²⁺. Global fits (lines) and data points are shown for each stem in tRNA.

Next, we turned to the precursor construct. As with the 5′-leader construct, high-quality global fits on the precursor construct in buffer and near-cellular conditions were obtained, supporting cooperative tRNA core folding when both flanking regions are present (SI Appendix, Fig. S12 and Table S4). Global fits of the core tRNA ILP portion of data provide reliable T_M values of 58.5 and 65.2 °C, in agreement with those for the 5′-leader construct, further supporting the conclusion that flanking regions do not contribute to core folding. In addition, the ΔH value for folding of the precursor construct in buffer was −136 kcal/mol, in agreement with that for the 5′-leader construct. Overall, these results suggest that the extended regions of tRNA do not affect the final cloverleaf structure nor the cooperativity of folding.

Discussion

RNA performs many essential functions in the cell, including catalysis, small-molecule binding, and gene regulation. To function optimally, RNAs need to fold into the proper 3D structures without a significant population of misfolded states. Proper folding to the native state is especially important for tRNA. Misfolding of tRNA is implicated in stalling of the ribosome, incorrect reading of the anticodon, and human disease (15, 24). In cells, RNAs fold during transcription, and the functional region of RNAs is often transcribed with flanking nucleotides on the 5′ and 3′ ends. To prevent misfolds, nature can select for sequences that will not adopt stable nonnative structures during late-transcriptional folding (Fig. 6).

Fig. 6.

Model for the late steps of tRNA folding. Without the 3′ portion of the acceptor stem (−11 to −5), the structures form native-like structures but are weak. As the 3′ portion of the acceptor stem is transcribed, the TΨC loop docks into the D loop, beginning with the −4 nucleotide. The 5′-leader and 3′-trailer sequences do not interfere with folding of the cloverleaf and are single-stranded. SL, stem loop.

Typically, RNA folding is studied on renatured transcripts; however, in cells, RNA folds as it emerges from the polymerase. Cotranscriptional folding is influenced by polymerase pausing and rates of folding vs. transcription. Yeast RNA polymerase is known to pause more frequently and for longer durations on AU-rich sequences (25). The 3′ trailer of tRNA^Phe is highly AU-rich with 14/21 (67% AU) bases being A or U. It is therefore plausible that the polymerase pauses on the 3′ trailer, which is a similar length to the polymerase exit tunnel, allowing the tRNA core to fold.

In this study, we examined the effects of near-cellular conditions on the folding, stability, and structure of tRNA late-transcriptional intermediates. This was explored in traditional buffer and near-cellular conditions and, based on our previous work (6), we predicted that late-transcriptional folding intermediates would be destabilized in near-cellular conditions. We used a near-cellular condition that included PEG8000 and aaCM. These conditions are not unique, however, as other crowders facilitate RNA function, as do other small molecule-chelated Mg²⁺ species. For instance, we showed similar RNA cooperativity in crowders of various PEGs, dextrans, and a Ficoll (5), and facilitation of hammerhead and DNAzyme activity in NDP-chelated Mg²⁺ (26). These observations enhance the generality of the findings to different and more complex cellular-like conditions.

We tested late-transcriptional intermediates of −11 to −3 nt of the 3′ end of yeast tRNA^Phe. In buffer with near-cellular concentrations of Mg²⁺, the intermediates that can partially form in the acceptor stem show multiple peaks in melting curves, indicating that there are multiple populated states with varying stability (Fig. 1 B and D). This type of folding is nonfavorable because stable misfolds populate, leading to very rugged folding to the native state. Remarkably, in near-cellular conditions, intermediate −5YF folds in a multistate manner but upon the addition of a single nucleotide the −4YF intermediate folds cooperatively (Fig. 1 C and E). This folding behavior of the late-transcriptional intermediates is also observed in E. coli tRNA^Ala in the near-cellular condition (Fig. 2). In the buffer condition, the ∆H values for late-transcriptional intermediates of EA suggested that they are more cooperative than the FL EA. In addition, even in the near-prokaryotic cellular condition, −4EA has a very minor peak at 50 °C (Fig. 2E). Such broader/small transition peaks of FL EA suggest that the tRNA has a small fraction of folding intermediates which are due to the acceptor stem. Furthermore, the nucleotide position in which folding cooperativity is induced can be modulated by changing the strength of the acceptor stem (Fig. 3). Folding cooperativity can therefore be induced in shorter transcripts with stronger acceptor-stem base pairing. Such indirect contributions to folding cooperativity have been explored previously in our laboratory using model functional RNAs and triplexes (6, 27).

Functional RNAs are often studied in the minimal-length construct needed to function; however, in vivo, these constructs are flanked by sequences on the 5′ and 3′ ends. In some cases, these flanking regions have been shown to modulate function (28). tRNAs are transcribed with a 5′-leader and a 3′-trailer sequence, and the effects of these sequences on the cloverleaf structure folding pathway are largely unexplored (5, 6, 12). Our experimental data suggest that nature has selected for 5′-leader and 3′-trailer sequences to be highly unstructured and to not interact with the fully folded core tRNA structure (SI Appendix, Figs. S9–S11). Additionally, with and without flanking sequences, the cloverleaf folds in a two-state, highly cooperative manner (Fig. 5 and SI Appendix, Fig. S12), suggesting that nature selected for flanking sequences that would not change the tRNA folding pathway. Flanking sequences have been previously reported to influence ribozyme function both positively and negatively through formation of self-structures and ribozyme-inhibiting structures (29–31). What is remarkable about tRNA is that in the context of the full-length transcript, the flanking structures do not appear to form any structure at all, either self-structure or structure with the tRNA core.

Maturation of tRNAs is a highly controlled process. In yeast, it involves cleavage of the 5′ leader with RNase P, while trimming of the 3′ end is thought to be carried out by tRNase Z and other endonucleases (32). The CCA motif is then added to the 3′ end with tRNA nucleotidyl transferase and the tRNA is transported out of the nucleus. Our data suggest that the cloverleaf is fully folded with the 5′ and 3′ extensions, which may help the processing enzymes recognize fully transcribed RNA. Indeed, cross-linking experiments and crystallographic analysis suggest that RNase P recognizes the folded state of the tRNA, making contacts with the acceptor stem, as well as the D- and TΨC-stem loops, where tertiary contacts are concentrated (14, 33). Less is known about processing of the 3′ end, but our data suggest that correct structure formation of the cloverleaf and no structure of the 3′ trailer could be necessary for processing.

After the 5′ and 3′ ends have been processed, there is a single dangling A on the 3′ end, which we call −3 (11). This transcript is recognized by nucleotidyl transferase, which adds the CCA motif onto the 3′ end. Similar to RNase P, nucleotidyl transferase recognizes the tertiary interactions between the D and TΨC loops, so the cloverleaf has to be folded into the native tertiary structure (34), which we find occurs with −4 and is maintained in −3 (Fig. 1). This folding property leads tRNA processing enzymes and modification enzymes to recognize nascent tRNA efficiently in the cell. While our data on E. coli tRNA^Phe (SI Appendix, Figs. S2 and S4) suggest that some FL tRNAs do not fold fully cooperatively, there are many processing and modification enzymes as well as RNA chaperones that bind to tRNA/pre-tRNA in cells, and these proteins can assist in folding of local structure of such tRNAs, which would decrease the fraction of misfolded tRNAs, as some literature suggests (21, 35, 36). Overall, our data suggest that there is inherent folding regulation in the sequence of tRNA, where the addition of a single nucleotide on late-transcriptional intermediates dramatically improves tRNA folding behavior.

Conclusions

In cells, RNAs fold as the growing transcript emerges from the RNA polymerase. Recent studies have shed light on cotranscriptional folding of riboswitches and revealed that riboswitch folding pathways are highly dependent on the presence of ligand during transcription (37, 38). However, not as much is known about the late-transcriptional folding of nonligand-dependent functional RNAs. Here we investigated the stabilities of tRNA late-transcriptional intermediates in near-cellular conditions. Our data reveal that in near-cellular conditions, most late-transcriptional intermediates are unstable and form weak structures. The first stable intermediate forms when all of the secondary-structure interactions can be made. Perhaps nature selects for this type of late-transcriptional folding to avoid populations of misfolded states.

Many functional RNAs have 5′ and 3′ ends in close proximity, similar to tRNA. These include several classes of riboswitches and ribozymes (39, 40). This long-range interaction could assure that native tertiary structure only forms once the entire RNA is transcribed. This could provide a mechanism for avoiding misprocessing and misfolding of functional RNAs.

Materials and Methods

Chemicals were purchased commercially. RNA constructs were prepared by T7 transcription from a hemiduplex or full-duplex template, gel-purified, and buffer-exchanged. Thermal denaturation experiments were performed on an OLIS spectrometer with data collected every 0.5 °C. In-line probing and native PAGE experiments were carried out using 5′-³²P–labeled RNAs. In the SAXS experiments, in-line size-exclusion SAXS and hand-loading SAXS were used for buffer conditions and PEG conditions, respectively. Please refer to SI Appendix for a more detailed description of materials and methods.

Data Availability Statement.

All data for all figures, supporting figures, and tables are available at https://scholarsphere.psu.edu/collections/h44558f98h.