A roadmap for ribosomal RNA folding and assembly during transcription

Abstract

Ribonucleoprotein (RNP) assembly typically begins during transcription when folding of the newly synthesized RNA is coupled with the recruitment of RNA-binding proteins (RBPs). Upon binding, the proteins induce structural rearrangements in the RNA that are crucial for the next steps of assembly. Focusing primarily on bacterial ribosome assembly, we discuss recent work showing that early RNA-protein interactions are more dynamic than previously supposed, and remain so, until sufficient proteins are recruited to each transcript to consolidate an entire domain of the RNP. We also review studies showing that stable assembly of an RNP competes against modification and processing of the RNA. Finally, we discuss how transcription sets the timeline for competing and cooperative RNA-RBP interactions that determine the fate of the nascent RNA. How this dance is coordinated is the focus of this review.

The fate of an RNA begins at birth

Many RNAs fold into stable structures that allow them to perform essential biochemical processes, such as gene regulation, splicing, and protein synthesis. A classic example is the bacterial 16S ribosomal RNA (rRNA) that adopts a unique structure and binds 20 or more ribosomal proteins (RPs) to form the 30S ribosomal subunit that participates in protein synthesis. Early electron micrographs of actively transcribed ribosomal DNA revealed the precursor rRNA associating with proteins as it was elongated by RNA polymerase (RNAP) [1], establishing that rRNA folding and assembly is directly coupled with its synthesis.

The advent of transcriptome-wide structure probing experiments has spotlighted the potential for any RNA to adopt structures and form ribonucleoproteins (RNPs; see Glossary) during their transcription that can influence gene expression [2,3]. For instance, precursor messenger RNAs (mRNAs) are not only subject to co-transcriptional capping, splicing, and polyadenylation [4,5], but they are also packaged with nuclear export factors to generate a mRNP that is protected from nucleases [6–8]. Certain long non-coding RNAs such as Xist and NEAT-1 regulate genes near their site of transcription, hinting that these RNAs begin interacting with their binding partners as they are synthesized [9–11].

Because RNAs start to fold and recruit RNA-binding proteins (RBPs) at their sites of transcription, one may ask how and to what extent these early folding events determine the ultimate fate of each transcript. Although ribosomes are more stable than most other RNPs in the cell, ribosome assembly illustrates principles that help us understand the formation (and dissolution) of more dynamic RNPs. For example, RPs mainly recognize the structure of the rRNA [12], and thus proper folding of the rRNA during transcription determines the likelihood of recruiting RPs to their correct binding sites. At the same time, RP binding frequently alters the rRNA structure after the two components come together [13], so that RP binding can itself change the folding path of the rRNA (Figure 1). Here, we first review how initial RNA folding events can influence RBP interactions, specifically focusing on co-transcriptional bacterial rRNA folding and ribosome assembly. We then describe how dynamic and often transient initial rRNA-RBP interactions can drive ribosome assembly and expand these observations to broader RNPs. Finally, we discuss the influence of transcription rate and auxiliary factors on these assembly steps.

Figure 1. RNA folding kinetics during transcription.

A. During transcription, stem loops are predicted to form as soon as the RNA emerges from the RNA polymerase (RNAP) exit channel. In a classic study of E. coli tRNA^fMet, the anticodon stem-loop formed in 80 μs, whereas the acceptor stem, which involves bases much further apart in the primary sequence of the tRNA, closed in 2 ms [88]. Tertiary folding of the tRNA is delayed because it requires all of the stem-loops to be transcribed [89]. Once complete, the tertiary structure can form in 5-10 ms [88]. B. Single-molecule optical trap experiments showed that SRP RNA hops between metastable base paired states during transcription [23]. The metastable hairpins are replaced by native long-range base pairs after the entire RNA is synthesized. C. Co-transcriptional SHAPE-Seq and single-molecule vectorial folding demonstrated that the ZTP/ZMP riboswitch forms an unstable tertiary interaction that exchanges with a more stable alternative terminator stem-loop if ZMP fails to bind before the alternative hairpin is completely elongated [27,90]. D. RP S15 stably binds a three-helix junction in the 16S rRNA only after the helix junction is fully transcribed and able to fold [91].

Folding is an unavoidable hazard during RNA synthesis

Nascent transcripts begin to fold as they emerge from the elongating RNAP, because the timescale for RNA base pairing is much shorter than the timescale for transcription (Figure 1; Box 1). Consequently, interactions between nucleotides near the 5’ end of the RNA form seconds or minutes earlier than interactions with downstream nucleotides [14,15]. It has long been thought that sequential RNA folding during transcription streamlines RNP assembly by committing the RNA to a specific subset of possible structures. In reality, transcription makes folding more variable, because sequential folding favors short-range base pairs over long-range interactions that dictate the RNA’s overall structure. In the absence of these long-range interactions, bases that initially pair during transcription may need to reorganize to achieve the final structure [16,17].

Box 1: Kinetics of RNA folding.

The primary sequence of an RNA encodes Watson-Crick (WC) base pairs that define its secondary structure. Once formed, the double helices may align and pack together in a tertiary structure that is defined by non-WC interactions such as triplexes, kissing loops, and pseudoknots.

In physiological salts, RNA hairpins are often stable and able to form in 10-100 μs [77]. The hairpin closing time depends on the size of the loop and the likelihood of correctly establishing the first base pair (see Figure 1). RNA tertiary structure is typically less stable than secondary structure and may form in ~1 ms to 1 s [72].

Many RNA sequences encode more than one secondary structure with similar stabilities. As a result, non-native structures kinetically compete with native folding [78,79]. RNAs that have evolved to form a single compact structure encode tertiary interactions that stabilize the proper base pairs, but sometimes trap the RNA in a misfolded state [77]. Random sequences also base pair, but their secondary structures are extended and inhomogeneous [80] because they lack long-range tertiary constraints.

The importance of sequential folding was demonstrated in early studies on phage MS2 [18], and on ribozymes that become kinetically trapped in misfolded structures when upstream bases pair improperly during transcription [14]. By circularly permuting the ribozyme sequences, so that different base pairs are transcribed first, it was shown that the 5′ to 3′ order of folding favors particular folding intermediates [19,20]. This is because formation of a stable upstream hairpin during transcription can be enough to commit an RNA to a particular folding pattern. Indeed, Mahen et al found that once formed, RNA hairpins ≥10 bp long (~3 kcal/mol) are sufficiently stable to resist pairing with downstream nucleotides in yeast cells [21]. Single-molecule vectorial folding experiments showed that even the small twister ribozyme is more likely to misfold on its maiden folding attempt than on subsequent attempts [22].

Nevertheless, folding during transcription is not simply “first come, first served”. Most RNA hairpins are shorter than 10 bp, and stochastic reorganization of weakly paired segments as the RNA elongates can lead to “topological frustration” that takes many minutes to resolve. For example, a recent single-molecule RNA transcription study found that a local non-native hairpin near the 5’ end of the signal recognition particle (SRP) RNA is later resolved in favor of native long-range base pairs once the entire RNA was transcribed (Figure 1) [23]. At intermediate transcript lengths, the authors observed frequent hopping between alternative structures, illustrating frustration within the sequential folding path. Thus, the inevitable initial folding of RNA during transcription is not a boon, but a hazard to which nature must adapt.

Conformational switching is an evolved trait

Countless non-coding RNAs require long-range interactions spanning hundreds of nucleotides. How do such RNAs fold correctly if these long-range interactions are likely to mispair during transcription? Key insights have come from studies of riboswitches, which regulate transcription termination in response to ligand binding [24]. A number of chemical footprinting, computational, and single-molecule methods have been recently deployed to understand how riboswitches fold into structures that enable read-through or termination during transcription (Supplemental Table 1). Ensemble and single-molecule experiments on the flavin mononucleotide and purine riboswitches showed that the forward rates of riboswitch folding and ligand binding, rather than thermodynamic stability, were crucial for committing the transcript to one or the other conformation before RNAP passed by the downstream termination site [22,25,26].

High-throughput structure-probing demonstrated that the ZTP riboswitch forms a metastable tertiary structure during transcription that is solidified by binding of the ZMP ligand [27] (Figure 1). This metastable tertiary structure is crucial for the switch mechanism, because it prepares the RNA to recognize ZMP and because it enables rapid conformational exchange via a ‘toehold’ mechanism. In general, transcriptional riboswitches seem to avoid being trapped in misfolded structures by encoding metastable structures that are designed to switch conformation as the RNA is elongated [24].

Variable RNA folding during transcription delays protein recruitment

Folding strategies that rely only on interactions within the RNA itself become increasingly less efficient for long transcripts due to the sheer number of possible different structures that may form. However, misfolding of long RNAs can be circumvented by RBPs, which may capture native interactions or lower the kinetic barriers to RNA refolding [28–30]. Moreover, assembly of a multi-body RNP affords greater control over the timing of molecular interactions during RNA elongation. These considerations raise the question of how quickly an elongating transcript can recruit its partner RBPs, and whether RNA-protein binding during transcription makes RNP assembly more efficient.

Bacterial 30S ribosome assembly involves the stable association of 21 RPs with the 16S rRNA, as well as temporary interactions with dozens of assembly factors (AFs), nucleotide modifying enzymes, and processing enzymes that interact with the rRNA at different stages of assembly. The RPs bind to the rRNA in a cooperative hierarchy that ultimately produces stable ribosomal subunits (Box 2). Only recently, however, has it been possible to examine how RPs bind during transcription.

Box 2: Hierarchy of Ribosome Assembly.

Historically, the mechanism of bacterial 30S ribosome assembly has been studied via in vitro reconstitution of native 16S rRNA with purified ribosomal proteins (RPs). Classical assembly mapping experiments defined a hierarchy of protein addition [33,81,82]. In this hierarchy, primary RPs uS4, uS7, uS8, bS15, bS17, and bS20 bind directly to the rRNA, whereas the remaining 14 secondary and tertiary assembly proteins only join after another protein has joined the complex. This hierarchy arises from conformational changes in the rRNA that are induced by protein binding [45], rather than from direct protein-protein interactions (Figure I). The resulting cooperativity of assembly ensures the completion of 30S and 50S subunits [33].

The order of primary and secondary protein binding is not strict, but can follow different (parallel) paths along a free energy landscape defined by the protein and RNA interactions [53]. This landscape was initially thought to arise from the thermodynamic stability of the protein-RNA contacts [29]. Later, pulse-chase mass spectrometry experiments showed that the kinetics of protein addition also follows the assembly map [53,75], indicating that primary proteins pre-organize the binding sites of later assembly proteins (Figure I).

If assembly proceeds through parallel pathways, do these pathways originate at a single point? Time-resolved hydroxyl radical footprinting experiments established that 30S assembly can nucleate from different rRNA domains concurrently when the rRNA is pre-folded [13], indicating that various helical junctions can interact with proteins independently. Next, interactions between helix junctions consolidate the major 30S domains [83–86]. Although each 30S domain can start to assemble independently of the others, the 5′ domain folds more rapidly than the 3’ domain under many conditions [13,34], suggesting that the 5′ to 3′ polarity of assembly during transcription is encoded in the rRNA. Modular assembly paths have been also proposed for the bacterial 50S subunit [87] and the yeast small and large subunit processomes [55]. After domains are formed, interdomain interactions are established, leading to formation of the native particle.

Figure I. Hierarchy of ribosome assembly.

Primary RPs pre-organize rRNA helix junctions to promote binding of secondary and tertiary proteins [45]. In this example, primary proteins S4 and S20 can associate with the rRNA independently and in any order, while secondary protein S16 can only form a stable complex after both S4 and S20 have stably bound the rRNA and preorganized both helix junctions [81,82].

One obstacle to studying the assembly path during transcription is the need to monitor RNA synthesis, folding, and protein binding simultaneously. Two single-molecule approaches [31,32] were recently developed to overcome this challenge (Figure 2). We developed single-molecule colocalization co-transcriptional assembly (smCoCoA) to record binding of Escherichia coli RP uS4 to full-length pre-16S rRNA during and after transcription (Figure 2C). Duss and co-workers developed a similar single-molecule method to measure colocalization of E. coli RPs uS7 and uS3 with a fragment of the 16S rRNA comprising its 3’ domain (Figure 2D). An advantage of single-molecule assays is that protein binding can be correlated with transcription, even when individual polymerases move at different speeds.

Figure 2. Studying co-transcriptional assembly of ribosomes at the single-molecule level.

A. 16S secondary structure highlighting the domains recognized by ribosomal proteins (RPs) uS4 (magenta), uS7 (blue), uS9 (orange). uS4 and uS7 bind helix junctions that are closed by long-range helix 3 (green) and helix 28 (red), respectively. B. Structure of the 30S ribosome (PDB: 2AVY) showing long-range interactions in the binding sites for uS4, uS7, and uS9. C. Single-molecule colocalization co-transcriptional assembly (smCoCoA) utilizes total internal reflection fluorescence microscopy (TIRFm) to measure protein binding during transcription. Protein induced fluorescence enhancement (PIFE) marks the moment when RNA polymerase (RNAP) passes over a Cy3 fluorophore at the end of the pre-16S DNA template. Colocalization of fluorophore-tagged RPs with the nascent rRNA reports on uS4 binding [31]. D. A zero-mode waveguide instrument monitors transcription by measuring the change in fluorescence intensity owing to movement of the fluorescent DNA towards the bottom of the waveguide. Single-molecular Förster resonance energy yransfer (smFRET) detects specific binding of uS7 to the RNA, which was labeled near h28 by a complementary Cy3-labeled DNA oligomer [32].

While proteins uS4 and uS7 stably bind with the native 16S rRNA, and nucleate assembly of the 16S 5′-central domains and 3′ major domains, respectively [33], surprisingly, neither protein was able to stably bind the elongating rRNA during transcription [31,32]. Instead, the proteins repeatedly encountered the transcript but dissociated after 0.1–0.2 s, only occasionally forming more stable complexes. The likelihood of stable binding was much greater for short transcripts containing a minimal uS4 or uS7 binding site, suggesting that uS4 and uS7 bind the full-length rRNA poorly because longer rRNAs initially misfold during transcription. For uS4, this was confirmed by allowing the newly transcribed pre-16S rRNA to equilibrate to a more native structure before the protein was added [31]. After this equilibration, stable uS4 binding was more common, whereas transient binding was less frequent.

If uS4 and uS7 struggle to recognize their rRNA binding sites owing to variable folding of transcripts, how can ribosome assembly nucleate during transcription at all? Although uS4 and uS7 were unable to stably bind the nascent rRNA on their own, the likelihood of forming specific and longer-lived complexes during transcription increased dramatically when nearby RPs, such as uS12 or uS19, were also present [31,32]. This enhancement was specific, as non-RPs had no effect. An analysis of the uS4 binding kinetics showed that additional RPs did not change the lifetimes of the complexes (uS4 dissociation), but rather increased the probability of specific uS4 binding, presumably binding of additional proteins favors a more native rRNA structure. Thus, when a transcript can interact with multiple RPs, the forward kinetics of assembly suddenly becomes more efficient.

Long-range RNA interactions fall into the trap

One may ask whether certain protein binding sites are more vulnerable to variable folding than others. Interestingly, the binding sites of uS4 and uS7 both straddle long-range 16S helices 3 and 28 that close the 5′ and 3′ major domains, respectively (Figure 2A, B). This raised the possibility that stable binding is prevented by mispairing of these long-range interactions during transcription. Duss and coworkers tested this idea using single-molecule Förster Resonance Energy Transfer (smFRET) and found that base pairing of helix 28 was hindered in the full 3’ domain rRNA compared to shorter rRNA fragments [32]. Furthermore, formation of helix 28 correlated with uS7 binding, demonstrating that the inability of uS7 to bind longer transcripts could be partially attributed to malformation of helix 28.

The two sides of helix 28 are separated by another domain of rRNA structure that is recognized by protein uS9. This insertion lengthens the delay in the synthesis of the 3′ half of helix 28 and increases opportunities for misfolding. Indeed, Duss et al found that transcripts lacking the uS9 domain were more likely to fold into a structure competent to bind uS7, consistent with the idea that sequence separation is important [32].

Long-range tertiary interactions may also exacerbate misfolding of the 16S 3′ domain. In the native 30S ribosome, the uS9 domain docks against the uS7 binding site. Although helix junctions within the uS7 and uS9 binding sites begin to fold independently [13,34], their interface cannot form until the 3′ side of helix 28 has been transcribed. If the uS9 domain misfolds or interacts too stably with uS7 before helix 28 is properly formed, the rRNA may be trapped in non-native structures that are incompetent for complete assembly of the 16S 3′ domain. This possibility was supported by in-cell footprinting of 30S ribosomes showing that this region is improperly structured in particles lacking uS9 [35] or RimM AF that acts on the 3′ domain [36].

Unstable complexes drive assembly

If most RPs sample the rRNA dynamically as it first starts to fold during transcription, and if too-stable local interactions can trap the rRNA in unproductive intermediates, this implies that transient non-native complexes may help kick-start ribosome assembly. How can this be? Many RPs protect their binding sites in stages [13], suggesting that the protein and RNA co-fold into the native complex after they come together. For example, detailed studies of uS4 binding showed that the initial encounter complexes sampled many different conformations [37]. Even more surprising was that most stable uS4 binding events initially passed through an on-path non-native intermediate, demonstrating that such complexes can lead to the desired RNP [37]. Although low-affinity, non-native interactions have often been disregarded as unproductive [38], the prevalence of such protein binding events during transcription suggests that they may contribute more to assembly than previously thought.

Transient binding and on-pathway non-native intermediates may be common to many RNP assembly mechanisms. For example, non-native intermediates are thought to contribute to binding of HIV Rev protein to the REV response element (RRE) in the HIV genome, in which binding of multiple Rev proteins promotes stable RNP formation [39]. During co-transcriptional splicing of the yeast mitochondrial b15 group I intron, fast non-specific binding of the protein Cbp2 hastened the formation of slow, specific RNP interactions that accelerate intron splicing [40]. Finally, it was recently shown that the mouse protein Dazl only resides on a particular mRNA for 1–2 s but nonetheless influences its translation, highlighting that transient RNA-protein binding can trigger changes in gene expression [41].

Short-lived interactions avoid sequestering RPs on misfolded or defective transcripts, while also allowing the rRNA to refold. In particular, RP binding can alleviate mispairing at helix junctions or misorientation of the rRNA helices relative to one another [37,42,43]. One possibility is that helix junctions in different parts of the rRNA must reorient at the same time to establish a larger stable tertiary structure (Box 2). In this scenario, only when a sufficient number of correct RNA-protein interactions has been achieved is the newly transcribed rRNA committed to forming a 30S complex. Therefore, by intensifying the cooperativity between RPs, unstable or variable rRNA folding during transcription may paradoxically improve the quality of assembly. A mechanism that relies on dynamic RP sampling of the rRNA during transcription may be especially advantageous in the crowded environment of the cell, in which RNAs are bombarded by proteins from the moment of birth.

An open question is whether multiple RPs must reside on a transcript at the same time to favor productive uS4 or uS7 recruitment (Figure 3). RPs bound to different sites in the rRNA are known to communicate through allosteric changes in the RNA structure, which is a basis for the hierarchical assembly map (Box 2) [44,45]. Based on the binding lifetimes for uS4 binding to pre-16S rRNA in single-molecule experiments, and assuming that all RPs behaved similarly to uS4, we estimated that two or more RPs can load within minutes if they bind cooperatively but can take hours if they bind independently [31]. In cells, co-residency may be increased by condensation of RPs at the sites of transcription. Under favorable conditions, it is conceivable that multiple proteins could quickly coalesce on the rRNA during transcription, committing it to assembly.

Figure 3. Models for how transient protein binding influences RNA folding and ribonucleoprotein (RNP) assembly during transcription.

Initial RNA structures are assumed to be dynamic and heterogeneous (left) before they assemble into the final RNP (right). Top pathway, two or more proteins residing on the same transcript may cooperate to stabilize a larger domain of RNA structure. Bottom pathway, transient binding of one protein leaves the rRNA in a conformation that is competent to bind a second protein.

An additional possibility is that transient protein binding switches the elongating rRNA into a new ensemble of conformations that are not typically sampled by the naive RNA (Figure 3). If the kinetic energy barrier between the two RNA conformations is ≥12 times greater than thermal energy (k_BT), the new conformation could persist for ≥0.1 s after the protein dissociates. At present, it is not known whether such transient binding is sufficient to influence the next binding event and bias the RNA folding pathway during transcription toward more native-like structures.

Linking transcription elongation with RNP assembly

If proper assembly of an RNP depends on metastable structures, some transcripts may only be competent to recruit partner RBPs within a ‘window of opportunity’ that is established by the timeline of RNA synthesis (Figure 4, Key Figure). Evidence for the importance of transcription speed comes from the inability to assemble functional 50S subunits when T7 RNAP was used to transcribe the 23S rRNA in E. coli cells at 37 °C [46]. This defect was suppressed by growing cells at a lower temperature that also decreased the T7 transcription rate. Similarly, the elongation rate is tightly coupled with rRNA folding in yeast, as slower Pol I elongation has been shown to disrupt rRNA processing and ribosome assembly (reviewed in [47]). Whereas Duss and colleagues found that slower elongation by E. coli RNAP made folding of the 16S 3’ domain less efficient in their single-molecule assay [32], we found that modulating the transcription elongation rate did not affect uS4 recruitment to the 16S 5’ domain; indicating that transcription speed may influence the folding of certain rRNA regions more than others [31].

Figure 4, Key Figure. Transcription defines the timeline for ribosome folding, assembly, modification, and processing.

RNAs begin to fold as they are synthesized by RNA polymerase. As the RNA elongates, ribosomal proteins (RPs; blue) alter the structure and dynamics of the elongating transcript. Assembly factors (AFs) and nucleotide modification enzymes, such as pseudouridine synthases (Ψase; green), compete with RPs for access to the RNA while it folds and elongates. Multiple RPs bind the RNA transiently, remodeling its structure, such as by orienting helix junctions. RP’s such as uS4 and uS7 must co-fold with their binding sites to stably join the complex. Once sufficient long-range interactions have accumulated, the entire RNP domain is able to assemble.

Although the average transcription speed defines the time window for rRNA folding and assembly, transcriptional pausing can specifically alter the folding pathway of the transcript [14,16,48]. On the one hand, a paused transcription elongation complex may impair folding by sequestering one side of a long-range helix [49]. On the other hand, pausing can allow upstream nucleotides more time to refold and stably recruit an RP before the rest of the rRNA domain is transcribed. Recently, stable secondary structure in the nascent rRNA was proposed to speed up transcription by preventing RNAP backtracking [50]. This observation raises the question of whether faster transcription improves RNA folding, or efficient folding improves transcription, or both.

In addition to the mechanics of transcription, factors that associate with the elongating RNAP have been shown to influence the process of ribosome assembly (Figure 4). NusA, a component of the bacterial rrn transcription anti-termination complex (rrnTAC), has been implicated in folding of the pre-rRNA by anchoring the 5′ end of the rRNA to RNAP until the 5′ end can base pair with downstream residues to form the long-range duplex needed for RNase III processing [51,52]. The rrnTAC may also directly facilitate assembly by reducing RNA misfolding. A recent cryo-electron microscopy (cryo-EM) structure of the E. coli rrnTAC revealed that S4 extends a channel of positively charged residues leading away from the RNA exit site [52], raising the possibility that the rrnTAC helps the transcript fold as it emerges from the RNAP active site.

Assembly factors and modification enzymes

Ribosome assembly involves many auxiliary proteins, such as AFs, modifying enzymes, and processing enzymes (Figure 4). These factors often open or distort the rRNA structure and time the establishment of long-range interactions in the immature subunits [36]. While most bacterial AFs function after transcription and initial processing of the pre-rRNA by RNase III [53], many yeast and human AFs are core components of large processomes that assemble on the pre-rRNA during transcription (reviewed in [54,55]). The small subunit (SSU) processome encases the pre-18S rRNA within a protective shell of protein that holds each 18S domain apart, presumably allowing each domain to assemble separately before interacting with the others.

RNA modifications are known to help stabilize RNA structures and the rRNA is one of the most heavily modified RNAs in the cell [56], yet the function of individual modifications is still not clear. Many of the RNA modification enzymes are thought to interact with the rRNA during transcription since their binding sites become inaccessible after the rRNA has folded and assembled into a stable RNP (Figure 4) [57,58]. Therefore, the modification machinery likely competes with RPs for access to the newly synthesized pre-rRNA.

An interesting example is modification of 16S nucleotide 516 by the pseudouridine synthase RsuA. This nucleotide lies within the uS4 binding site, and thus must compete with binding of protein uS4 to the same region [57,59]. Indeed, purified RsuA fails to efficiently modify full-length 16S compared to truncated RNAs indicating that RNA folding inhibits modification [60]. Delayed co-transcriptional folding of the uS4 binding site during transcription may create an opportunity for RsuA_modification prior to stable uS4 binding.

In yeast, many rRNA modifications are directed by small nucleolar RNAs (snoRNAs) that base pair with the rRNA [61,62]. Therefore, snoRNA binding directly competes with intramolecular rRNA folding during transcription. Recent processome structures indicate that snoRNAs likely bind the rRNA very early in assembly, consistent with the idea that snoRNAs control the timing of certain rRNA folding events by preventing inappropriate interactions from trapping the rRNA in a misfolded conformation [63,64].

In addition to specific AFs and targeted modification enzymes, general RBPs with chaperone activity may be important for remodeling trapped rRNA intermediates, particularly at low growth temperatures (Figure 4). The bacterial hexameric RNA chaperone Hfq has been proposed to bind to the 16S rRNA 5’ leader and promote a conformational change that is crucial for co-transcriptional processing of the pre-rRNA [65]. There are also four RNA helicases in E. coli that play a role in ribosome assembly and maturation. The E. coli helicase SrmB may function during transcription since it associates near the 5’ end of the 23S rRNA forming an RNP with RPs L4 and L22 [66]. RNA helicases may act generally to remodel a variety of misfolded rRNA intermediates since deletion of SrmB can be rescued by overexpression of another RNA helicase, CsdA (Cold shock domain A) [67] and recent work showed that CsdA can unwind a number of different RNA substrates [68]. Many more RNA helicases are required for ribosome assembly in yeast and metazoans suggesting that active remodeling of the rRNA is critical for efficient assembly [69–71].

Concluding Remarks

Details of how ribosome assembly is coupled with transcription and co-transcriptional RNA folding are beginning to emerge, offering lessons for the assembly of other RBPs. Since RNA folding occurs concomitantly with synthesis and is inherently heterogenous, proteins must recognize and bind to an evolving set of RNA structures during the early stages of assembly. It is becoming clear that RNAs can adopt a range of different structures during transcription, particularly when the final structure is defined by many long-range interactions, as is the case for rRNA. Frequent, transient protein binding to partially folded or even incorrectly folded rRNA during transcription may shape subsequent folding steps, contributing to fast and accurate assembly of the final ribosomal subunit.

Further work is required to understand the physical basis for how proteins respond to evolving RNA structure during RNP formation, which will be aided by single RNA footprinting and multi-color single molecule fluorescence microscopy (see Outstanding Questions). It will be particularly important to understand whether other classes of RBPs form transient complexes during transcription, and the extent of cooperative RBP binding to eukaryotic pol II transcripts and long non-coding RNAs. Examining how RNPs are formed during transcription is also important for examining how proteins interact with any non-coding RNA, particularly with respect to drug design aimed at inhibiting particular protein-RNA interactions that lead to disease.

Outstanding Questions.

• How is ribonucleoprotein (RNP) assembly nucleated during transcription if complexes formed during transcription are short-lived?

• How does transient protein binding affect the folding pathway of elongating RNAs? Do multiple proteins need to reside on the RNA at the same time to commit the RNA to assembly?

• Is a condensed phase required to nucleate ribosome assembly during transcription, particularly if initial interactions are short-lived?

• How much RNA structure and how many proteins must bind to commit a transcript to assembly and avoid turnover?

• How does binding of other proteins, such as assembly factors, chaperones, helicases, RNA polymerase (RNAP)-associated factors, and RNA modification enzymes, compete with binding of ribosomal proteins and affect nucleation of RNP assembly?

• How well do the models proposed for transient protein binding in bacteria hold up in other organisms or for other types of RNPs? Do yeast ribosomal proteins also bind the elongating pre-rRNA transiently at first?

• Do the small nucleolar RNPs (snoRNPs) bind transiently during co-transcriptional assembly?

• Can the dynamic nature of initial RNP assembly be harnessed for targeted drug design?

Highlights.

Nascent RNAs form variable structures as they are transcribed, creating an unavoidable hazard for ribonucleoprotein (RNP) assembly

To form the correct structures during synthesis, strategies have emerged such as metastable RNA switches, modular domain folding, and protein-guided folding

Proteins primarily form transient, non-native complexes with elongating RNAs

Transient protein binding may be sufficient to guide RNA folding and initiate RNP assembly during transcription

Other transcription-coupled processes, such as RNA modification, binding of assembly factors, and processing by nucleases, kinetically compete with RNA folding and the assembly of stable RNP domains.

Glossary

Assembly Factor (AF)

A protein that assists the assembly or maturation of ribosomal subunits but is not a component of the mature ribosome.

Backtracking

A type of long-lived transcriptional pause involving reverse translocation of the RNA and DNA through a secondary channel within RNAP [49]. Backtracked polymerases can be reactivated through RNA cleavage with the help of transcription factors such as GreA/GreB in bacteria and TFIIS in yeast [72].

Helix junction

Intersection of two or more RNA double helices connected by 0 to 7 ‘unpaired’ nucleotides. Helix junctions can be grouped into sequence families that adopt preferred helix orientations based on tertiary interactions in the joining segments.

Non-coding RNAs

refers to any RNA that does not encode for a protein. Non-coding RNAs carry out specific functions, which may depend on folding into a complex 3-dimensional structure [73].

Processome

large protein complexes that assemble on rRNA in eukaryotes during transcription. They help fold, modify, and process immature small and large ribosomal subunits [63].

Protein Induced Fluorescence Enhancement (PIFE)

A phenomena that is observed when a protein come in close proximity (1-10 nm) to a fluorescent dye causing a sudden increase in fluorescence intensity owing to a change in the molecular environment of the fluorescent dye [74].

Pulse-chase mass spectrometry

An isotope labeling method for measuring the stable association of proteins with ribosomal complexes [75].

Ribonucleoprotein (RNP)

A stable complex formed between at least one RNA and one protein.

Ribosome Assembly

The process of assembling a ribosomal subunit from its RNA and protein components [33].

The large and small ribosomal subunits assemble separately before coming together to carry out protein synthesis.

Small nucleolar RNA (snoRNA)

small non-coding RNAs in eukaryotes that assemble with proteins to guide specific functions such as RNA modification and processing [62]. SnoRNPs are essential for ribosome assembly and maturation in eukaryotes.

Time-resolved Hydroxyl radical footprinting

A RNA structure probing technique that utilizes hydroxyl radicals to cleave the solvent exposed backbone of RNA [36].

Total internal reflection fluorescence microscopy (TIRFm)

A type of fluorescence microscopy that utilizes an evanescent wave to only illuminate molecules tethered to the slide surface (~100 nm into solution), while freely diffusing fluorescent molecules remain dark.

Transcription Elongation Complex

An RNA polymerase (RNAP), transcript and DNA template in the elongation phase of transcription, plus any associated transcription elongation factors.

Zero-mode waveguide (ZMW)

A microscopy tool that illuminates the sample within a nanometer-sized well [76]. ZMW can be used with high concentrations of labeled proteins, which is not possible using conventional TIRFm.