Transcription-translation coupling: Traveling a road under construction

The genetic information stored in DNA is read out in two steps: transcription and translation. The two steps are carried out by two large molecular machines, RNA polymerases (RNAPs) and ribosomes. Both move along a template (DNA or mRNA, respectively) to give a polymeric product (mRNA or protein) that reflects the sequence information of the template. Importantly, the product of the first synthesis step is the template of the second. In eukaryotes, transcription and translation are spatially separated, as the mRNA product of transcription is exported from the nucleus before translation. In contrast, bacteria perform both processes simultaneously. Ribosomes begin to translate an mRNA while the mRNA is still being produced by an RNAP. Thus, ribosomes travel along a track that is not yet finished (Fig. 1).

Figure 1.

Cotranscriptional translation: A ribosome follows an RNAP on the mRNA transcript the RNAP generates. Do the two molecular machines “dock” together to form a supercomplex, the expressome, when the ribosome catches up? If yes, is their coupling direct or mediated by Nus factors? Is the coupling tight or transient and stochastic? To see this figure in color, go online.

Obviously, a ribosome cannot move faster than the RNAP it follows, at least once it has caught up with the RNAP. However, experiments also show that they do not move more slowly, at least not in Escherichia coli. Rather, the speeds of transcription and translation seem to be adjusted to be the same even though they can vary widely for different physiological conditions (1,2).

Cotranscriptional translation appears to have at least two functions: it prevents premature termination of transcription because the ribosome following the RNAP prevents access of termination factors to that RNAP. The following ribosome also can re-activate paused RNAPs by preventing or restricting backtracking and by pushing forward a backtracked RNAP (2).

How do a transcribing RNAP and the ribosome following it coordinate their activity? Do they interact at all beyond being tethered together by the mRNA? These questions have received much attention and there are at least three answers (Fig. 1), all supported by some evidence. Structural studies (in particular cross-linking in living cells, which shows RNAPs and ribosomes in close proximity, and cryoelectron microscopy) suggest that the RNAPs and ribosomes can bind to each other to form an “expressome,” a supercomplex of the two big molecular machines (3,4). However, while close proximity could reflect a direct molecular interaction, it could also be a consequence of a short mRNA tethering the two machines together without specific interactions between them. In addition, there is evidence for RNAP-ribosome interaction mediated by Nus factors, specifically NusG (4,5). Both direct and Nus-mediated interactions could be either tight or transient and stochastic. The third possibility is that there is no physical coupling, but rather a signal related to the physiological state of the cell (e.g., the alarmone ppGpp) controls the speeds of both processes and adjusts both to the cell’s general physiological state. The latter interpretation is supported by experiments in which coupling is disrupted by antibiotics that slow down ribosomes (6). However, this observation does not completely rule out structural coupling, in particular if that coupling is stochastic.

A paper by Li and Chou in this issue of Biophysical Journal (7) addresses precisely this stochastic case with a detailed model for the coupling of transcription and translation. Discrete stochastic models have long been used to describe all types of molecular machines that step along a track with discrete binding sites, including RNAPs and ribosomes (8). Such models describe the motion of ribosomes and RNAPs as a sequence of random events in which the molecular machine makes a discrete step forward (corresponding to one nucleotide or one codon) while simultaneously elongating its polymeric product. These steps occur with a rate that reflects both the internal dynamics of the machine and its interaction with its environment, including other molecular machines. In Li and Chou’s model, the first such stepper, the RNAP, generates the track for the second stepper, the ribosome. The two are therefore described by two positions, n and m, with $n<m$ (Fig. 1). In addition to the simple stepping, internal states of the two machines and transitions between these states account for RNAP pausing, reactivation/pushing of an RNAP by a trailing ribosome, and potential coupling between the two. Coupling is described by a transition to an “associated state” of the two machines, which could be either through direct interaction or Nus mediated: the model of Li and Chou allows the two machines to “dock” to each other with a certain rate when they are within some interaction range. The interaction range also allows some flexibility in the stepping of the coupled machines: they do not need to keep a fixed distance or step in perfect synchrony. However, their distance is constrained to remain below the allowed maximum imposed by this interaction range, at least while the two machines are in the associated state. As a consequence, a ribosome can not only push an RNAP but can also “hold the RNAP back,” and indeed, the model predicts that cotranscriptional translation can speed up or slow down transcription depending on the kinetic parameters. The model can also be extended to include a sequence dependence of stepping and/or pausing in a straightforward way.

The first thing this model makes clear is that the different proposed mechanisms for the coordination of transcription and translation, and specifically for the correlation of the two speeds under different conditions (direct or Nus-mediated interaction, tethering by the mRNA, coordinated regulation by, e.g., ppGpp), are limiting cases of a continuum of models, described by different choices of the rates in a stochastic framework. It is quite possible that different bacterial species or even different sets of genes in the same organism make use of different regions of the parameter space of that model. Indeed, transcription and translation are not coupled in Bacillus subtilis, where RNAP runs away from the ribosome following it (9).

The more difficult question is how the different models (or different kinetic regimes within the stochastic model) can be distinguished experimentally. Here, Li and Chou propose several metrics that can be used to quantify the coupling and that may be used to infer parameters of the model and thereby get insights into whether a direct coupling exists and into its strength and association/dissociation kinetics. So far, the main quantity used to characterize coupling is the delay between completion of an mRNA and completion of the protein after induction of a gene, measured as a bulk average. Li and Chou calculate the distribution of that delay, which would provide more detailed information but is difficult to access experimentally. Single-molecule experiments of cotranscriptional translation would likely be required to determine this distribution. A second parameter they propose, the coupling coefficient, defined as the fraction of RNAPs that are associated with a ribosome at the time they complete the transcript, also requires the simultaneous observation of both machines, that is, the determination of both position variables n and m. An option that may be worth exploring is whether a conditional distribution or a conditional average of the position of the ribosome for a known RNAP position could be determined, e.g., with a sequencing-based method such as ribosome profiling after some clever preselection of the complexes to implement the conditioning.

In addition, Li and Chou study a characteristic that is accessible to more classical experiments, namely the probability of a gap between RNAP and ribosome that exceeds a certain threshold. This quantity can be probed at the bulk level by premature transcription termination because when the ribosome falls behind, the RNAP become accessible to transcription termination factors such as Rho. It should therefore be worth designing experimental conditions that clearly distinguish between different coupling types based on how the premature termination probability depends on some parameter that can be modulated experimentally, e.g., the translation initiation rate or the (local) speed of ribosomes.

Tests that probe the coupling of transcription and translation typically require some way of disrupting the coordination of the two speeds. To rule out coordination by global signaling (such as via ppGpp), it would be ideal if only one gene of interest was perturbed, while the bulk of gene expression proceeded without perturbation. In any case, new studies on transcription-translation coupling can be expected in the next years, and modeling frameworks, such as the one by Li and Chou, can provide some guidance for a quantitative analysis.

Author contributions

S.K. wrote the manuscript.

Editor: Jason Kahn.