The mechanism of mRNA cap recognition

Abstract

During translation initiation, mRNA molecules must be identified and activated for loading into a ribosome^1-3. In this rate-limiting step, the heterotrimeric protein eukaryotic initiation factor eIF4F must recognize and productively interact with the 7-methylguanosine cap at the 5' end of the mRNA and subsequently activate the message^1-3. Despite its fundamental, regulatory role in gene expression, the molecular events underlying cap recognition and mRNA activation remain unclear³. Here we generate a single-molecule fluorescence imaging system to examine the dynamics with which eIF4F discriminates productive and non-productive locations on full-length, native mRNA molecules. At the single-molecule level, we observe stochastic sampling of eIF4F along the length of the mRNA and identify allosteric communication between the eIF4F subunits that ultimately drive cap-recognition and subsequent activation of the message. Our experiments uncover functions for each subunit of eIF4F and we conclude by presenting a model for mRNA activation that precisely defines the composition of the activated message. This model provides a general framework for understanding how mRNA molecules may be discriminated from one another and how other RNA-binding proteins may control the efficiency of translation initiation.

Extended Data Figure 1.. Purification, labeling, and validation of biochemical components.

(a) SDS-PAGE gel showing all of the purified eIFs required to perform mRNA recruitment gel-shift assays. The inset shows a red fluorescence scan of the outlined portion of the SDS-PAGE gel demonstrating that the factors are fluorescently labeled. n=1 experiment. (b) Standard mRNA recruitment gel-shift assay showing recruitment of mRNA onto a stable 48S-PIC in the presence of eIF4F. As expected, diminished recruitment is observed in the absence of eIF4B or eIF4E. n=1 experiment. (c) Titration-regime, anisotropy-based RNA binding experiments performed with either 500nM or 50nM FAM-labeled, 51-nucleotide rpl41a fragment. These experiments show that anisotropy continues to increase beyond a 1:1 molar ratio of mRNA:eIF4G, confirming multiple eIF4G molecules bind the short RNA, as previously observed. Each titration-regime experiment was performed once. Gel source data can be found in Supplementary Figure 1.

Extended Data Figure 2.. Validation of smFRET between Cy5-eIF4G:E and Cy3-mRNA.

(a) Hypothetical intensities/FRET trajectories showing what is expected to happen if Cy5-eIF4G:E binds at a cap-proximal location on the capped mRNA construct. Upon binding, the green fluorescence intensity decreases while the red intensity increases for the duration of the binding event, leading to an increase in E_FRET while Cy5-eIF4G:E is bound near the donor fluorophore. (b) Shows similar hypothetical trajectories, but in these trajectories Cy5-eIF4G:E binds at a cap-distal location far away from the donor fluorophore. Because the acceptor and donor fluorophores are not near each other, no changes in intensity or E_FRET are detected before photobleaching. (c) Shows an example experimental trajectory for Cy5-eIF4G:E binding to the short, capped 24-nucleotide 5’ UTR of rpl41a. Because the RNA is short, we expect only a single binding site. At 5nM, we observe transient binding, but at 50nM (d), both the association and dissociation kinetics increase, indicating that eIF4G is able to facilitate its own dissociation. (e) Quantification of the concentration dependent pre-steady-state association rate and population weighted average dissociation rate of eIF4G:E from the capped, 24 nucleotide construct, which indicates facilitated dissociation is occurring. Facilitated dissociation is cartooned in (f) whereby the release of one of RNA binding domains of eIF4G, also called microdissociation, opens up a binding location for a new molecule of eIF4G, which then facilitates the release of the first eIF4G molecule. (g) 2D histograms showing the concentration dependent exchange of Cy5-eIF4G:E bound to the capped, 24-nucleotide construct when increasing concentrations of unlabeled eIF4G:E are injected into the flowcell and quantification to the right. PBL denotes photobleaching limited dissociation rates. n=2 biological experiments for the experiments in (e) and (g); data are presented as means and error bars correspond to standard deviations.

Extended Data Figure 3.. Cy5-eIF4G:E facilitates its own dissociation on full-length rpl41a mRNA.

(a) Survival-time analysis showing that the steady-state dissociation (left) of Cy5-eIF4G:E on the capped mRNA construct is biphasic, whereas the pre-steady-state dissociation rate (right) is monophasic, indicating that steady-state dissociation events are a mixture of facilitated and non-facilitated dissociation events. (b) Comparison of the rate constants derived from steady-state and pre-steady-state experiments indicate facilitated dissociation by eIF4G. Thus, pre-steady-state experiments are more reflective of the inherent dissociation rate of eIF4G:E. Comparison of the pre-steady-state dissociation rates recorded with and without shuttering of the excitation light reveals that dissociation of eIF4G:E is limited by photobleaching. The steady-state dissociation rates plotted are population weighted averages of the two phases shown in (a). n=2 biological experiments experiment for each condition. (c) The ability of unlabeled eIF4G:E to exchange with pre-associated Cy5-labeled eIF4G:E on the capped full-length rpl41a mRNA construct is tested by injecting 5 nM or 400 nM unlabeled eIF4G:E into the flowcell after allowing Cy5-eIF4G:E to bind. *The 2D histogram showing the ejection of free eIF4G:E from the flowcell is repeated from Figures 1 and Figures 3 in the main text for comparison. n=2 biological experiments for each condition. (d) 2D histograms showing the photobleaching limited dissociation of Cy5-eIF4G:E from an uncapped fragment of rpl41a encompassing the first cap-distal position on the mRNA (left) and the photobleaching limited dissociation from a second internal labeling position on full-length mRNA after ejection of all free Cy5-eIF4G:E with buffer. (e) A summary of the association and dissociation kinetics for the 24-nucleotide 5’ UTR fragment, internal fragment around the cap-distal fluorophore, and a second cap-distal location. PBL denotes photobleaching limited dissociation rates. All data in this figure are presented as means and all error bars correspond to standard deviations.

Extended Data Figure 4.. eIF4G:E can “hop” along an mRNA.

Example trajectories taken from pre-steady-state ejection experiments on the capped construct where Cy5-eIF4G:E dissociates from the local position on the mRNA, causing E_FRET to drop to 0, but later recovers, despite the all free eIF4G:E being removed from the flowcell, when the same, or another eIF4G:E bound elsewhere on the mRNA, microdissociates and then “hops” to the local position near the donor fluorphore on the mRNA. These example trajectories are shown for both continuous illumination (a) and excitation light shuttering (b) experiments. (c) The fraction of trajectories which contain hopping events is quantified for the capped and uncapped mRNA constructs as a function of ejection with buffer, ejection with buffer supplemented with eIF4A+B+ATP, and shuttering interval. Comparison of the hopping events to a control experiment performed on the capped, 5’ UTR of rpl41a which is too small to contain a second eIF4G binding site with which eIF4G could hop to and from reveals that hopping is prevalent on rpl41a. n=2 biological experiments for each condition. Data are presented as means and error bars correspond to standard deviations.

Extended Data Figure 5.. eIF4G:E-Cy5 facilitates its own dissociation.

(a) A bar graph showing the population-weighted average, steady-state dissociation rate of eIF4G:E compared to the pre-steady-state dissociation rate. Similar to Cy5-eIF4G/G:E, free, unbound, eIF4G:E-Cy5 is able to exchange with bound eIF4G:E-Cy5 leading to an increased apparent dissociation rate in steady-state experiments. This serves as additional evidence that eIF4G and eIF4E function as a complex. n=2 biological experiments for each condition. Data are presented as means and error bars correspond to standard deviations. (b-d) The equivalent ejection experiments shown in Figure 4 but with the acceptor fluorophore on eIF4E instead of eIF4G. Similar to with Cy5-eIF4G:E, eIF4G:E-Cy5 is stripped from every position except from the cap. This demonstrates that eIF4G:E is stripped as a complex during the search for the cap. PBL denotes photobleaching limited dissociation rates.

Extended Data Figure 6.. Association and dissociation kinetics of Cy5-eIF4G:E in the presence of eIF4A and eIF4B.

(a) The pre-steady-state association rates calculated from a Cy5-eIF4G:E concentration series, also plotted in Figure 1, and the pre-steady-state association rates calculated from a single Cy5-eIF4G:E concentration point of 3 nM in the presence of saturating eIF4A+ATP or eIF4A+eIF4B+ATP. eIF4A and eIF4B do not appear to alter the association kinetics. (b) The equivalent steady-state association kinetics for the experiments described in (a). (c) The steady-state dissociation kinetics calculated from the steady-state portion of the injection experiments for eIF4G:E, eIF4G:E+eIF4A+ATP, and eIF4G:E+eIF4A+eIF4B+ATP. eIF4A and eIF4B enhance the steady-state dissociation rate on every construct except the capped construct in the presence of eIF4E. (d) The equivalent pre-steady-state experiment for those described in (c). eIF4A destabilizes eIF4G/G:E from each mRNA construct in an ATP-dependent manner except in the presence of both eIF4E and the cap. A red line at the highest tick mark on the Y-axis of (b) and (c) denotes the change in axis scale between the two graphs—highlighting that the steady-state dissociation rate far exceeds the pre-steady-state dissociation rate due to facilitated dissociation. (e-f) Similar experiments were performed as described in (a) and (c) except 2 μM eIF4A•ATP was allowed to bind the mRNA prior to the injection of the eIF4F complex. *The data denoted corresponds to the same corresponding data for eIF4G:E+A plotted in (a-d) to serve as a direct comparison with and without pre-equilibration of eIF4A on the mRNA. For (a) – (f) n=2 biological experiments for each experimental condition; data are presented as means and error bars correspond to standard deviations.

Extended Data Figure 7.. ATP binding, but not hydrolysis, or eIF4B is required for stripping eIF4G from mRNA.

To further validate that the only thing preventing the stripping activity of eIF4A is recognition of the cap by eIF4E, we utilized a mutant of eIF4G:E_W104A in which one of the critical tryptophan residues involved in cap recognition is mutated to alanine, thus diminishing the affinity of eIF4E to the cap. eIF4A was effectively able to remove eIF4G from capped positions when cap-recognition was perturbed with these mutations or by competing for cap binding with 200 μM cap analog m⁷GpppA (b). eIF4A does not require ATP hydrolysis (c) or eIF4B (d-g) but does require ATP binding for its strippase activity (d-g). PBL denotes photobleaching limited off rates.

Extended Data Figure 8.. Direct interaction between eIF4G and eIF4A drives the strippase activity of eIF4A and facilitates mRNA recruitment.

(a) 2D histograms illustrating the concentration dependence of stripping eIF4G:E from uncapped mRNA as a function of eIF4A concentration performed at 17 °C and acquired with a 200ms exposure time, quantified in (b). The half-maximal stripping velocity is achieved between 100–200 nM eIF4A whereas the K_D of eIF4A to the 51mer 5’ fragment of rpl41a in the absence of eIF4G is ~10 μM as measured by fluorescence anisotropy. n=2 biological experiments for each concentration. (c) This illustrates that eIF4A must bind eIF4G to facilitate stripping. n=3 biological experiments. (d) 48S mRNA recruitment reaction time course (top) and quantification (bottom) show that eIF4A is essential for in vitro mRNA recruitment, and that perturbation of cap-recognition by either removal of eIF4E from the reaction or mutation of eIF4E_W104A to perturb cap binding results in a rate defect which can be compensated for by increasing concentration of eIF4G. n=3 biological experiments for each mRNA recruitment condition. Gel source data can be found in Supplementary Figure 1. All quantified data in this figure are presented as means, and error bars correspond to standard deviations.

Figure 1.. eIF4G/G:E stably associates with mRNA independent of its binding position or the cap.

(a) Cartoon schematic of the full-length, Cy3-labeled rpl41a mRNA constructs used in this work. (b) Cartoon representation of the stopped flow injection and ejection pre-steady state experiments used to measure k_on and k_off. (c) An example trace taken from a Cy5-eIF4G:E injection experiment is displayed. The trace can be divided into two portions: the first binding event in the trajectory encompasses the pre-steady-state portion, whereas the later portions of the trajectory display steady-state binding and dissociation events until bleaching of the donor fluorophore. On the right, a surface contour plot showing the aggregate behavior of all the trajectories in an injection experiment is displayed. (d) The equivalent example trace and surface contour plot for ejection experiments are shown. Since all of the excess Cy5-eIF4G:E has been removed from the flowcell, only the pre-steady-state behavior can be inferred from ejection experiments. (e) The association rate of Cy5-eIF4G/G:E is linear with respect to concentration and depends more on the presence of eIF4E than on the cap. n=2 biological experiments for every condition. (f) Quantitation of the pre-steady state association and dissociation rates for Cy5-eIF4G/G:E, demonstrating that eIF4G uniformly binds to mRNA irrespective of the cap or its binding location. n=2 experiments for every condition. For the plots in © and (f), data are presented as mean values, while error bars correspond to standard deviations. For the association rate bar plot in (f), the bar coordinate corresponds to the fit shown in © and the error is the error of the fit.

Figure 2.. eIF4E functions in concert with eIF4G irrespective of the cap.

(a) Example trajectories of Cy5-eIF4E binding to the capped mRNA construct in the absence (left) or presence (right) of unlabeled eIF4G. In the absence of eIF4G, Cy5-eIF4E only transiently samples the cap, whereas it stably binds in the presence of eIF4G. (b) Surface contour plots post-synchronized to the steady-state transitions of eIF4E out of the bound state: Time = 0 marks the time at which eIF4E dissociates from the vicinity of the donor fluorophore, such that the negative time axis indicates how long Cy5-eIF4E remains bound before dissociation and the positive time axis shows the delay time between binding events. (c) Quantitation of the association and dissociation rates of Cy5-eIF4E and eIF4G:E-Cy5 from the three mRNA constructs. The association rates for the capped, uncapped, and internal constructs are determined from experiments performed at 5 nm eIF4G:E-Cy5, whereas the capped –4G association rate is derived from the slope of the apparent association rates of Cy5-eIF4E performed across three concentrations. The nearly identical rates derived from equivalent experiments performed with Cy5-eIF4G, plotted in Figure 1, are shown as purple dots to aid the comparison between Cy5-eIF4E and Cy5-eIF4G. n=2 biological experiments for each experimental condition. Because the capped –4G rates were determined using a concentration series and this condition does not facilitate its own dissociation, the dataset for each concentration point was used to calculate the dissociation rate (n=6). Data are presented as mean values whereas error bars correspond to standard deviations, except for the association rate of capped –4G, where the value is calculated from the fit of the association rate concentration series and the error bar corresponds to the error of the fit.

Figure 3.. eIF4A strips eIF4G:E from RNA.

Surface contour plots showing ejection experiments where excess, unbound Cy5-eIF4G:E in the flowcell is replaced with buffer lacking Cy5-eIF4G:E or replaced with buffer lacking Cy5-eIF4G:E but supplemented with eIF4A, eIF4B, and ATP (right). Each experiment was recorded at both 10 frames per second (fps) and 1/3 fps to obtain kinetic information in both the seconds (black) and minutes (gray) timescales. These experiments were performed on the (a) capped, (b) uncapped, (c) capped RNA without eIF4E, and (d) internal mRNA constructs. Each experiment begins with Cy5-eIF4G:E bound to the RNA. E_FRET starts at a high value and remains stable until photobleaching in the absence of eIF4A, eIF4B and ATP. However, in the presence of eIF4A, eIF4B and ATP, E_FRET rapidly decays within ~3 seconds (guiding line). This indicates that eIF4A strips eIF4G:E from RNA. eIF4E recognition of the cap (a) is able to resist this stripping. (e) Pre-steady-state off rates were calculated from the individual trajectories of the experiments performed in (a)-(d). These rates show eIF4G is destabilized by 30–40 fold in the presence of eIF4A, eIF4B, and ATP, but eIF4G is resistant to this destabilization in the presence of both eIF4E and the cap. n=2 biological experiments for each experimental condition. Data are presented as means and error bars correspond to standard deviations.

Figure 4.. The mechanism of mRNA activation.

Our experiments lead us to the following new model of mRNA activation: In the first step, eIF4G:E nonspecifically samples and binds to mRNA. At the second step, the pathway bifurcates into productive (green) or nonproductive (red) pathways. In the non-productive fork (2b), eIF4G:E binds far away from the cap, preventing eIF4E from associating with the cap, allowing eIF4A to recycle the eIF4F complex by either facilitating its intramolecular hopping to another place on the same mRNA molecule (3b) or completely removing it from the RNA (3c). In the productive fork, eIF4G:E binds at the 5’ UTR near the cap, and the eIF4E-cap interaction allosterically signals to eIF4A, through eIF4G, not to strip eIF4G. This stalled eIF4F complex at the cap constitutes the previously elusive activated mRNP, in which we speculate eIF4A may be held in an inactive conformation which can be reactivated by an eIF4B containing 43S-PIC prior to loading the mRNA onto the ribosome (3a). Notably, because eIF4A is present at high concentrations in the cell, this model is agnostic to which step eIF4A binds eIF4G:E, and due to its modest affinity, we speculate it is dynamically sampling the complex throughout the cycle.