A synergistic network of interactions promotes the formation of in vitro processing bodies and protects mRNA against decapping

Abstract

Cellular liquid-liquid phase separation (LLPS) results in the formation of dynamic granules that play an important role in many biological processes. On a molecular level, the clustering of proteins into a confined space results from an indefinite network of intermolecular interactions. Here, we introduce and exploit a novel high-throughput bottom-up approach to study how the interactions between RNA, the Dcp1:Dcp2 mRNA decapping complex and the scaffolding proteins Edc3 and Pdc1 result in the formation of processing bodies. We find that the LLPS boundaries are close to physiological concentrations upon inclusion of multiple proteins and RNA. Within in vitro processing bodies the RNA is protected against endonucleolytic cleavage and the mRNA decapping activity is reduced, which argues for a role of processing bodies in temporary mRNA storage. Interestingly, the intrinsically disordered region (IDR) in the Edc3 protein emerges as a central hub for interactions with both RNA and mRNA decapping factors. In addition, the Edc3 IDR plays a role in the formation of irreversible protein aggregates that are potentially detrimental for cellular homeostasis. In summary, our data reveal insights into the mechanisms that lead to cellular LLPS and into the way this influences enzymatic activity.

Figure 1.

High-throughput LLPS diagrams. (A) Schematic diagram of the proteins used in this study. Lines indicate the intra- and intermolecular interactions that are important for the LLPS process. (B) Comparison of LLPS diagrams for the Dcp1:Dcp2 (y-axis) and Edc3 (x-axis) protein that are constructed based on the visual inspection of microscopy images (left) or direct and quantitative turbidity measurements (right). Both methods provide highly similar phase diagrams, underlining that our high-throughput approach provides rapid and reliable data. The degree of phase separation is indicated in a white (no LLPS), yellow, orange, red (high degree of LLPS) color scale. See also Supplementary Figure S2. (C) The effect of salt on the LLPS process. Salt inhibits LLPS of Dcp1:Dcp2:Edc3. Note the few outliers in the phase diagrams at (e.g. 150 mM salt; 5 μM Edc3 and 30 μM Dcp1:Dcp2) are likely due to the presence of small air bubbles in these conditions. (D) Addition of Pdc1 significantly enhances the LLPS of Dcp1:Dcp2 and Edc3. The effect is due to specific interactions between Pdc1 and Dcp1:Dcp2:Edc3 and is not due to indirect effects of, e.g. molecular crowding (Supplementary Figure S3).

Figure 2.

RNA strongly enhances LLPS. (A) Phase diagrams of Dcp1:Dcp2:Edc3 supplemented with 5 μM RNA of different length. RNA of 30 bases or more significantly moves the phase separation boundaries to lower concentrations, whereas DNA has no effect on the phase diagram. (B) Phase diagrams of Dcp1:Dcp2:Edc3 supplemented with RNA, however, as opposed to panel A, the total amount of nucleotides is kept constant. The phase diagrams show that one 100mer RNA has the same effect as three 30mer RNAs and indicate that the RNA-binding events with Dcp1:Dcp2:Edc3 are fully independent. Shorter RNAs have no influence on the phase separation diagrams, indicating that the minimal length of an RNA that can efficiently be incorporated into the interaction network is around 30 bases. (C) Fluorescence microscopy images of Dcp1:Dcp2:Edc3 that is complemented with fluorescently labeled RNA. The RNA is highly enriched in the droplet phase. Addition of Dcp1:Dcp2 and Edc3 (no preformed Dcp1:Dcp2:Edc3 foci) to RNA (left) results in larger foci as when the RNA is added to Dcp1:Dcp2:Edc3 foci, indicating that RNA is a highly efficient nucleation factor. No fluorescent signal is detected in the droplet phase in the absence of RNA (Supplementary Figure S3B).

Figure 3.

RNA is protected within LLPS foci. (A) RNase cleavage of a 30mer RNA (that contains a single RNaseA cleavage site at position 10) is reduced upon phase separation. This shows that the RNA within in vitro processing bodies is protected against RNase activity. This protection is not due to direct interactions between Edc3, Dcp1 or Dcp2 with RNA as addition of the same amount of the individual Edc3 domains (LSm, IDR and YjeF_N) does not result in RNase protection (and LLPS). The amount of LLPS is indicated on top, where the color scheme of Figure 1B is used. (B) The Dcp2 activity on a short RNA of 20 nt is independent of LLPS. (C) The Dcp2 activity on a longer RNA of 100 nt is significantly reduced upon LLPS. This is not due to direct interactions, as the addition of the individual Edc3 domains has no influence on the decapping activity (see also A). The long RNA is efficiently embedded in the intermolecular interaction network that lead to LLPS (Figure 2), whereas the short RNA is not able to enhance phase separations of Dcp1:Dcp2:Edc3. It should be noted that the measured decapping activity is a weighted average of the activity of Dcp2 inside and outside the in vitro processing bodies, due to the exchange of components between the two phases. The activity of the decapping complex that is purely within the in vitro processing bodies is thus likely even lower than what we observe here.

Figure 4.

The Edc3 IDR specifically interacts with RNA. (A) Phase separation diagrams of Dcp1:Dcp2 and Edc3 in the absence (top) and presence (bottom) of a 30mer RNA. RNA is able to significantly shift the phase separation boundary for the WT Edc3 protein (left). Upon deletion of the IDR the effect of RNA on the phase diagrams is reduced (right). (B) Phase separation diagrams of Edc3 and a 30mer RNA. Edc3 and RNA are sufficient to induce LLPS (left). A version of Edc3 that lacks the IDR is no longer able to undergo phase separations in the presence of RNA only. (C) ¹H-¹⁵N NMR spectra of the Edc3 IDR in the absence (black) and presence (red) of an equimolar amount of RNA of 15 nt. A large number of resonances are significantly weaker or undergo chemical shift perturbations (CSPs), indicating a direct interaction between the RNA and the IDR. A number of assignments are indicated. (D) The extent of the CSPs that are induced by the RNA correlate with the length of the RNA. This indicates that one RNA can interact with multiple IDRs. (E) Plot of the loss of intensity of the NMR signals in the Edc3 IDR upon addition of a 30mer RNA. Three regions in the Edc3 IDR interact with RNA: a region around residue 80, a region around residue 130 and a region around residue 180. (F) ITC binding experiments reveal a μM affinity between a 30mer RNA and the Edc3 IDR. The exact affinity cannot be extracted due to the unknown stoichiometry of the interaction.

Figure 5.

Maturation of in vitro processing bodies. (A) Successive fluorescence microscopy images that were taken of Dcp1:Dcp2:Edc3 phase separated proteins after incubation for 12 h. The proteins formed a gel-like film on the bottom of the well that could be scratched off using a pipette tip. (B) The LLPS (left panels) and maturation (right panels) of the Edc3 protein at different salt concentrations (y-axis) were monitored over time (x-axis). The full-length Edc3 protein (top panel) undergoes phase separation at lower salt concentrations. The formed foci (left) merge and subsequently form a gel-like film (right). Removal of the IDR or the YjeF_N domain in Edc3 inhibited the LLPS and maturation processes, indicating that the interaction between these domains plays an important role in the maturation of Edc3 containing foci. (C) The Edc3 YjeF_N domain interacts directly with two regions in the Edc3 IDR, around residue 100 and around residue 165. Plotted is the loss of intensity of the IDR NMR resonances upon addition of the YjeF_N domain. See also Supplementary Figure S7.