Regulation of the multisubunit CCR4-NOT deadenylase in the initiation of mRNA degradation

Abstract

The conserved CCR4-NOT complex initiates the decay of mRNAs by catalyzing the shortening of their poly(A) tails in a process known as deadenylation. Recent studies have provided mechanistic insights into the action and regulation of this molecular machine. The two catalytic enzymatic subunits of the complex hydrolyze polyadenosine RNA. Notably, the non-catalytic subunits substantially enhance the complex’s affinity and sequence selectivity for polyadenosine by directly contacting the RNA. An additional regulatory mechanism is the active recruitment of the CCR4-NOT to transcripts targeted for decay by RNA-binding proteins that recognize motifs or sequences residing predominantly in untranslated regions. This targeting and strict control of the mRNA deadenylation process emerges as a crucial nexus during post-transcriptional regulation of gene expression.

Introduction

Over the last decades, the view of a relatively passive role of mRNAs as mere information carriers has changed due to the increasing appreciation that mRNA itself has a central role in the specific and tight control of gene expression on the level of individual transcripts. The mRNA life cycle is highly regulated, from transcription in the nucleus to translation by the ribosomes in the cytosol. Removing mRNAs from the cellular pool endows the cell with a fast and efficient means to regulate their expression. mRNA decay mechanisms are diverse but follow a limited number of enzymatic pathways. The exonucleolytic removal of the stretch of up to 200–300 adenosines at the 3′-end, known as the poly(A) tail, initiates the canonical cytosolic decay process [1]. The two principal deadenylation enzyme complexes, PAN2-PAN3 and CCR4-NOT, are recruited and activated on specific transcripts, leading to decay [2]. The poly(A)-binding protein (PABP) that coats the poly(A) tail stimulates the PAN2-PAN3 complex, leading to more efficient activity on longer tails that are coated by multiple PABP molecules [3,4]. The CCR4-NOT complex, however, was thought to prefer shorter tails devoid of PABP. These insights coalesced into a two-stage or biphasic mechanistic model in which PAN2-PAN3 initiates deadenylation while the CCR4-NOT completes it [5,6].

In somatic cells, deadenylated mRNA species are quickly degraded and do not accumulate in the cell. This observation led to the view that deadenylation is the step that controls the overall rate of the mRNA decay process [2,7]. However, the mechanisms regulating mRNA deadenylation’s speed and efficiency remained poorly understood. The degradation machinery presented such a challenge for the biochemical and structural study because mRNA decay complexes, in contrast to splicing and translation machinery, cannot be readily isolated in a compositionally homogeneous form and in quantities suited for in vitro mechanistic studies. Technological developments in the recombinant production of multisubunit complexes have enabled the in vitro studies of these highly dynamic molecular machines. These studies paint a picture of two very different complexes responsible for poly(A) tail shortening of mRNAs. The PAN2-PAN3 complex can be recruited to microRNA targets [8] and is generally stimulated by the presence of PABP on the poly(A) tail [4]. However, while PAN2-PAN3 preferentially trims long poly(A) tails and has limited effects on the transcriptome, the CCR4-NOT complex is the principal cytoplasmic deadenylase in human cells [9]. Furthermore, CCR4-NOT is a significant hub for integrating many specific translational repression mechanisms that funnel into the central mRNA decay pathway, including microRNAs.

This review discusses the mechanisms that allow CCR4-NOT to regulate gene expression in a transcript-specific manner. Recent studies described how different RNA-associated proteins guide the CCR4-NOT complex and physically tether it to specific transcripts to silence them. These studies revealed deadenylation to be a precise and highly regulated process. We also present a model of allosteric regulation of CCR4-NOT via its subunits with multiple RNA binding sites distributed along the extended surface of the complex. Finally, we propose that this regulation may provide an additional means of conferring target specificity for highly tuned and exquisitely precise regulation of mRNA expression.

CCR4-NOT is a modular and highly dynamic molecular machine

Two catalytic subunits comprise the twin ‘engines’ of the CCR4-NOT complex: the DEDD-type exonuclease CAF1 and the EEP-type deadenylase CCR4 [10,11], which form the central nuclease module of the complex [12]. Due to their functional importance and biochemical tractability, they were the first CCR4-NOT subunits to be structurally characterized in isolation [13–16]. Although structural work revealed that the leucine-rich repeat (LRR) domain of CCR4 mediates the interaction with CAF1 within the nuclease module [17,18], it also became apparent that the two nuclease domains containing the active sites are positioned very flexibly with respect to each other in the absence of bound substrate RNA [19,20].

The nuclease module docks onto the central MIF4G domain of the ~2,400-residue NOT1 subunit, which serves as the ‘scaffold’ of the entire CCR4-NOT assembly (Fig. 1) [12,17]. NOT1 consists of an array of helical repeat domains [17,18,21–25], which typically function as protein-protein interaction domains. Likewise, the other CCR4-NOT subunits assemble on NOT1 in several discrete modules. At the N-terminal end of NOT1, the tetratricopeptide repeat (TPR)-containing protein NOT10 and its partner NOT11 assemble as a heterodimer on the extended HEAT repeat-containing domain of NOT1 [12,17]. Further along NOT1, a short helical domain [17,26] is next to the central MIF4G domain that anchors the nuclease module [17,18]. Immediately C-terminal to the nuclease module NOT1 contains a coiled-coil domain that folds only in the presence of its partner, the armadillo repeat-containing subunit CAF40/NOT9 [21,22]. A MIF4G-like domain of unknown function [23] separates the CAF40-binding region from the very C-terminal hairpin repeat-containing domain that functions as the docking site for NOT2 and NOT3, which collectively comprise the NOT module [24,25] (Fig. 1). Biochemical and structural work hinted that these modules might function somewhat autonomously within the CCR4-NOT because they fold and assemble independently. Attempts to produce high-resolution structural information of the complete yeast [27,28] and human [29] CCR4-NOT have proven unsuccessful due to its high intrinsic flexibility. Given its highly flexible, modular architecture, we speculate that CCR4-NOT may not adopt a single or even a series of discrete conformations without the RNA substrate or other stabilizing factors.

Figure 1. The architecture of the mammalian CCR4-NOT complex.

A schematic view of the domain organization of NOT1 and bound subunits, together with the structures and structural models of the characterized modules and subassemblies. The experimental structures shown here are the N-terminal part of S. cerevisiae NOT1 (PDB 4B8B) [17], the human CAF40 module (PDB 4CRU) [21], the C. thermophilum NOT1 MIF4G-C (PDB 6H3Z) [23] and the human NOT module (PDB 4C0D) [24]. The model of the human nuclease module was prepared by superposing the NOT1-MIF4G-CAF1 dimer (PDB 4GMJ) [18] as well as the CCR4 LRR and the CCR4 EEP domains, respectively (both PDB 7AX1) [19] onto the S. cerevisiae nuclease module (PDB 4B8C) [17]. No experimental structural information was reported for the NOT10–11 module. AlphaFold2 structure prediction suggests that NOT10 (UniProt ID Q9H9A5) and NOT11 (UniProt ID Q9UKZ1) are composed primarily of helical repeats. NOT10 and NOT11 do not exist in S. cerevisiae and other ascomycetes; all other subunits are conserved throughout eukaryotes.

CCR4-NOT is an extended binding platform for RNA-binding factors

Many mRNAs carry sequence motifs (often embedded in their 3′ untranslated regions) recognized by RNA-binding protein factors that then recruit CCR4-NOT by physically tethering it to the mRNA, resulting in transcript-specific deadenylation. Several proteins interact with the nucleases themselves, such as SMG5/7 in nonsense-mediated mRNA decay, which stimulates the degradation of transcripts with premature stop codons [30], and BTG2/Tob, which recruits the complex to PABP-coated tails [16]. However, most other RNA-associated proteins recruit CCR4-NOT via non-enzymatic modules (Fig. 2), thus rationalizing the functional relevance of some of the non-catalytic subassemblies in CCR4-NOT. For instance, the protein DND1 interacts with the N-terminal region of NOT1 [31], and tristetraprolin (TTP) binds to the HEAT-repeat NOT1 domain positioned C-terminally to the NOT10–11 module (in addition to another contact with CAF40) [26]. However, the main docking sites for RNA-binding factors are in the C-terminal half of NOT1. The NOT module is used as a docking platform by proteins of the Nanos family [32,33], Roquin [34], NOT4 [35,36], Bicaudal-C [37], HELZ [38] and Pumilio [39], and the small CAF40 subunit is another hotspot for interactions (Fig. 2) [21,22,34,35,40–42].

Figure 2. Transcript-specific recruitment of CCR4-NOT by RNA-binding proteins.

A) A gallery of co-structures of CCR4-NOT subunits or subassemblies with interaction partners. CAF40 is a hotspot of interactions and possesses tandem tryptophan (W) binding pockets on its convex side that are used by GW182 proteins and TTP (PDB 4CRV) [21,42], and a peptide-binding pocket on the convex side which recognizes Bag-of-marbles (Bam; PDB 5ONA) [40], metazoan NOT4 (PDB 6HOM) [35], Roquin (PDB 5LSW) [34] and RNF219 [41]. Tristetraprolin (TTP) binds the small helical domain N-terminal of the central MIF4G (PDB 4J8S) [26]. Tob folds into a globular domain associated with CAF1 (PDB 2D5R) [16]. The NOT module is contacted and recruited by Drosophila Nanos (PDB 5FU7) [32], Hs Nanos (PDB 4CQO) [33], and yeast NOT4 [36].

B) A schematic depiction of recruitment of the CCR4-NOT complex by RNA-associated proteins. Recruiting factors interact with target transcripts indirectly (GW182, mediated by miRNA-AGO complex; Tob/BTG, mediated by PABP) or directly (most likely all other displayed proteins) and then contact CCR4-NOT on one or multiple surfaces. RNA-binding domains are typically globular folds, but interactions with CCR4-NOT are mediated mainly by short peptides; a notable exception is Tob which binds CAF1 using a folded domain.

It is striking that all structurally characterized CCR4-NOT—RBP interactions, except for BTG2/Tob, involve short linear motifs, i.e., sequence motifs of ~10–30 residues, which are more conserved compared to the surrounding low-complexity or unstructured protein regions [43]. These interacting regions may even be as simple as two aromatic side chains binding to twin hydrophobic pockets, as in the case of the GW182 proteins involved in microRNA-mediated mRNA silencing [21,22] or TTP (Fig. 2A) [42]. These sequence motifs recognize specific binding sites on the CCR4-NOT complex where highly conserved residues tend to cluster. In many cases, such motifs appear to fold into secondary structure elements upon binding [40]. This mode of interaction is present in many signaling networks, where interactions must be highly specific and readily reversible [43]. Furthermore, these motifs can be quickly acquired, changed, and even lost entirely during evolution, as shown in the example of the protein Nanos [32,33].

In essence, the CCR4-NOT complex is one rather sizeable binding platform recognized by many factors that guide it to its respective targets. The apparent advantage of recruiting a general decay factor is that any transcript containing a specific regulatory element can be regulated independently of all other transcripts. Furthermore, this regulation can be closely controlled by rapidly altering the availability of the recruitment factors, for example, in response to external stimuli necessitating a change in the gene expression program of the cell. For instance, posttranslational modifications regulate the interaction of RNF219 and Tob2 proteins with CCR4-NOT [41,44].

Combinatorial effects determine CCR4-NOT recruitment efficiency

In the cytosol, bulk mRNAs are several orders of magnitude more abundant than CCR4-NOT. In this environment, different transcripts likely compete with each other for the recruitment of the complex, and local concentrations of both transcripts and their bound CCR4-NOT recruitment factors may influence the outcome of CCR4-NOT recruitment and deadenylation/degradation efficiency. Intriguingly, there are cases where short linear motifs from different recruitment factors bind to the same surface on CCR4-NOT, suggesting the possibility of competitive recruitment (Fig. 3A). The most striking example of this binding exclusivity is the concave side of CAF40. Evolutionarily independent yet highly similar peptides of at least four different CCR4-NOT interactors target this crucial region (Fig. 2A) [34,35,40,41]. There is also biochemical evidence that peptides of CCR4-NOT interactors do indeed compete for binding to the CAF40 subunit [40]. On the opposite side of CAF40, the two tryptophan-binding pockets can bind GW182 proteins and TTP [21,22,42], further supporting the plausibility of mutually exclusive interactions.

Figure 3. Mechanisms of CCR4-NOT complex activation.

A) Competition of transcripts for CCR4-NOT recruitment. Some binding sites on the CCR4-NOT surface are not unique but are shared by several RNA-binding proteins. Consequently, those proteins might, in principle, be able to compete for recruitment of CCR4-NOT to their bound transcripts. One example is the peptide-binding site on the concave side of CAF40 (A) which is used by evolutionarily unrelated peptides of proteins Bag-of-marbles (Bam) [40], Roquin [34], and metazoan NOT4 [35]. The tandem W-binding sites on the convex side of CAF40 (B) can accommodate Trp residues from GW182 proteins [21,22] as well as TTP [42].

B) Additive effects might enhance CCR4-NOT recruitment efficiency. Many transcripts contain several or multiple sequence motifs and miRNA-target sites and thus could be bound simultaneously by multiple CCR4-NOT recruiters. These proteins could, in turn, recruit one CCR4-NOT complex by multiple interactions, as shown here by the example of a hypothetical transcript containing one miRNA-binding site and a stem-loop structure that Roquin binds, leading to a tighter interaction and more efficient recruitment than in the situation when only one recruitment factor would be present. Alternatively, several RNA-associated proteins bound on one mRNA could recruit several CCR4-NOT complexes.

C) Substrate RNA binding may enhance CCR4-NOT activity. Several RNA-binding interfaces have been identified or proposed on different parts of CCR4-NOT, i.e., (1) the nuclease module [19], (2) CAF40 [48], (3) the NOT module [25], and (4) the NOT10–11 module [29]. Binding of the substrate transcript might enhance the overall affinity of CCR4-NOT, improving deadenylation efficiency and processivity, stabilizing a favorable conformation of the complex, position the 3′-end of the substrate close to the active sites, and terminate the exonuclease activity at the end of the poly(A) tail.

D) Model of general and specific mRNA deadenylation. Newly synthesized mRNAs possess long poly(A) tails coated by copies of PABP. In general, bulk mRNA decay, PAN2-PAN3, is recruited to the transcript by directly binding PABP, and the 3’-end is positioned optimally to reach the active site [4]. At this stage, CCR4-NOT can already be recruited by BTG/Tob (not shown) bound to PABP but cannot yet outcompete PAN2-PAN3. Once the penultimate PABP dissociates, PAN2-PAN3 loses affinity for the transcript. At the same time, CCR4-NOT can remove the terminal PABP, associate with the naked substrate, and distributively shorten the remaining tail. This pathway can be circumvented by RNA-binding proteins recruiting CCR4-NOT to specific transcripts to elicit fast and efficient displacement of PABP and subsequent processive deadenylation. In both cases, deadenylation is followed by rapid 5′-decapping and exonucleolytic decay of the mRNA body.

Some RNA-associated proteins contact CCR4-NOT not with just one but via multiple motifs, and this appears to be a cooperative phenomenon. For example, several proteins, including Roquin and NOT4, interact with both CAF40 and the NOT module [34,35], while HELZ and Drosophila Nanos use multiple motifs to bind the NOT module (Fig. 2) [32,38]. In addition, GW182 proteins contain multiple tryptophan residues that bind to the tandem pockets in CAF40 [21,22] and additional sites, including the NOT module (our unpublished data), and TTP uses the CAF40 tryptophan pockets and also contacts NOT1 with another peptide [26,42]. Furthermore, many mRNAs contain multiple recognition sequences for different RNA-binding factors. These could, in principle, recruit multiple CCR4-NOT complexes or cooperate to recruit a single CCR4-NOT complex more efficiently. These speculations are yet to be corroborated by evidence, but both mechanisms would result in a more robust and focused deadenylation with subsequent decay (Fig. 3B).

Cooperative RNA substrate binding stimulates CCR4-NOT activity

Apart from specific recruitment by protein factors, another layer of regulation of CCR4-NOT stems from binding to the mRNA substrate itself (Fig. 3C). The recent reconstitution of recombinant CCR4-NOT from yeast [45] and human proteins [29] allows direct and systematic measurement of deadenylation rates in a completely controllable and manipulable system [46]. Furthermore, the Passmore laboratory demonstrated that the complete fission yeast CCR4-NOT is highly specific towards poly(A) sequences and more active than the isolated nuclease heterodimer [45]. This data indicates that the non-catalytic part of the complex somehow contributes to the deadenylation activity of the nucleases.

We recently found that the human CCR4-NOT is also strikingly more active than the isolated nuclease heterodimer, which suggests that the regulation of nuclease activity by the non-catalytic part is a conserved mechanism [29]. The CAF40 module contributes to accelerated deadenylation [47], and we could attribute similar effects to the NOT and NOT10–11 modules in vitro [29]. Since the nuclease, CAF40, NOT10–11, and NOT modules all bind RNA directly [19,25,29,48], we proposed a model in which multiple RNA-binding surfaces bind and stabilize the RNA substrate within the complex (Fig. 3C) [29].

These observations suggest that direct binding to the RNA substrate may optimize CCR4-NOT deadenylation efficiency. One consequence of this interaction may be the stabilization of a favorable conformation and optimal substrate orientation for enzymatic hydrolysis. It is especially noteworthy that the CAF40 module appears to serve an important stimulatory role, as evident by the inhibition of deadenylation by the Bam and RNF219 proteins that bind CAF40 directly [29,41]. Because CAF40 is in the immediate proximity of the nuclease module on the NOT1 scaffold, we speculate that RNA bound to CAF40 may be optimally and stably oriented toward the active sites of the CCR4 and CAF1 nucleases. Furthermore, the binding sites on the NOT module and CAF40 are selective for the binding of non-poly(A) sequences [25,48], which may serve as a mechanism to terminate processive RNA hydrolysis once the poly(A) tail reaches a critical length [29,45]. While relatively weak affinity for poly(A) may permit CCR4-NOT to ‘slide’ along the shortened tail, the complex may, in turn, associate with non-A sequences with higher affinity resulting in deceleration and even stalling. We propose that efficient deadenylation likely requires recruitment of CCR4-NOT by one or several external factors and substrate binding by the non-catalytic parts of the complex.

The venerable biphasic model of deadenylation in which the PAN2-PAN3 would initially trim mRNA poly(A) tails and then hand over to the CCR4-NOT [5] is due for an update. The yeast CCR4 can efficiently displace PABP from poly(A) tails, suggesting that CCR4-NOT may indeed functionally substitute for PAN2-PAN3 (Fig. 3D) [49,50]. CCR4-NOT can also indirectly interact with long PABP-coated poly(A) tails via BTG/Tob [51]. Trimming and PABP removal by PAN2-PAN3 could be bypassed altogether in a scenario where RNA-associated factors recruit CCR4-NOT and physically tether it to their target transcripts (Fig. 3D). Another layer of complexity to this mechanism is that the mRNA substrate and recruiting factors may compete for binding to the CCR4-NOT, such as in CAF40, where RNA- and peptide-binding interfaces overlap [29,34,41,52].

Concluding remarks and outlook

Emerging structural, biochemical, and functional evidence points to very tight control of CCR4-NOT deadenylation machinery on many levels. However, we are still only just beginning to glimpse the exquisite regulation of this essential molecular machine. Since recombinant, purified full-length CCR4-NOT [29,45] and many additional factors necessary for targeted mRNA deadenylation are now available for biochemical and structural study; we are finally in a position to dissect these mechanisms with systematic precision. In addition, it will be fascinating to test the hypotheses regarding the combinatorial effects of multiple RNA-binding proteins and RNA-binding sites on CCR4-NOT recruitment and targeted deadenylation.

Further, it will be imperative to look beyond CCR4-NOT itself and assess its role in conjunction with other factors in mRNA decay, such as the PAN2-PAN3 complex and the downstream decay factors. A substantial body of new functional and structural data revealed a complex interaction network of CCR4-NOT involving the decapping-associated helicase DDX6 [21,22] and the exonuclease XRN1 [53]. In addition, the discovery that the coding sequence of the transcript can influence deadenylation suggests very close mechanistic links with translation [54]. To understand how these factors integrate, we envision further structures of multi-factor complexes to glean the mechanism and an in vitro recapitulation of complete mRNA decay pathways in the test tube. This exciting prospect is increasingly within our reach.

Highlights:

• The CCR4-NOT complex is a highly regulated molecular machine

• CCR4-NOT is recruited to specific mRNAs by RNA-binding proteins

• RNA-binding surfaces on several CCR4-NOT modules promote activity

• Combinatorial effects allow precise deadenylation control on a single-transcript level