Dcp2 is thoroughly conserved in eukaryotes and remains the best studied eukaryotic decapping enzyme that is active preferentially on long m⁷G capped RNA substrates (van Dijk et al., 2002; Piccirillo et al., 2003; Cohen et al., 2004) (Figure 1A). It is also active in vitro on TMG-capped mRNAs, albeit with lesser efficiency than on m⁷G-capped mRNAs (Cohen et al., 2004). Dcp2 specifically cleaves the alpha-beta pyrophosphate bond of capped mRNAs, yielding m⁷GDP and 5′-monophosphorylated RNA (Lykke-Andersen, 2002; van Dijk et al., 2002; Wang et al., 2002). An invariant glutamic acid (S. cerevisiae E153, S. pombe E147) in the Nudix motif serves as a general base for catalysis and its mutation is sufficient to fully impair decapping activity in vitro and in vivo (Dunckley and Parker, 1999; Aglietti et al., 2013). An invariant lysine (S. cerevisiae K135, S. pombe K129) that functions as a general acid stabilizes the leaving group and is critical for RNA binding (Aglietti et al., 2013). The functional assignment of these residues is supported in the structure of the product (m⁷GDP)-bound Dcp2 where both residues are positioned near the beta-phosphate group of m⁷GDP (Charenton et al., 2016; Wurm et al., 2017). In the tertiary structure, four additional glutamic acid residues of the Nudix domain coordinate divalent cations (Mg²⁺ or Mn²⁺) that are required for catalysis (Aglietti et al., 2013). The Nudix motif is followed in C-terminus by a linker and a partially conserved Box B motif that is enriched in positively charged residues which forms an RNA-binding channel along with residues on the Nudix hydrolase domain (Wang et al., 2002; Deshmukh et al., 2008). In S. cerevisiae, the Box B motif is directly followed by a stretch of residues that is rich in hydrophobic amino acids or with long aliphatic side chains, which interact with Edc3 (Harigaya et al., 2010; Charenton et al., 2016). Meanwhile in S. pombe, binding to Edc3 mainly involves a region slightly downstream within the C-terminal intrinsically disordered region (IDR) (Fromm et al., 2012; Charenton et al., 2016), discussed below.

In Dcp2, the Nudix domain is preceded by a short and flexible hinge connected to a structured region often called the N-terminal Regulatory Domain (She et al., 2008; Floor et al., 2010; Charenton et al., 2016; Wurm et al., 2017; Mugridge et al., 2018b). This domain contains a highly conserved stretch of residues identified as Box A (Wang et al., 2002). Box A ensures the specificity of Dcp2 and its removal leads to aberrant production of m⁷GMP in addition to m⁷GDP in vitro (Piccirillo et al., 2003). Most importantly, the N-terminal regulatory domain harbors a set of conserved residues that interact with the cap which together with the Nudix hydrolase domain, form the composite active site of Dcp2 (Floor et al., 2010; Charenton et al., 2016; Wurm et al., 2017; Mugridge et al., 2018b). Keys to Dcp2 activity are an invariant “gatekeeper” tryptophan (S. pombe W43, S. cerevisiae W50) and a glutamic acid (S. pombe D47, S. cerevisiae D54) that are positioned on top of a loop called the 190s loop of the Nudix domain (Floor et al., 2010; Aglietti et al., 2013; Charenton et al., 2016; Wurm et al., 2017; Mugridge et al., 2018b). Together, these residues sandwich the cap when Dcp2 adopts an active conformation (Floor et al., 2010; Aglietti et al., 2013; Charenton et al., 2016; Wurm et al., 2017; Mugridge et al., 2018b). Other conserved residues are also located in the vicinity and interact with the phosphate groups of the cap (Charenton et al., 2016). Without a substrate, the cap-binding tryptophan (S. pombe W43) is blocked by a base-stacking interaction with an aromatic residue in the Nudix domain (S. pombe Y220, S. cerevisiae Y222) (Mugridge et al., 2016). This interaction is liberated upon substrate interaction, enabling both residues to engage with the cap (Mugridge et al., 2016). This aromatic residue also recognizes the first nucleotide of the capped mRNAs with a preference towards purines over pyrimidines (Mugridge et al., 2018b). In addition to containing residues that form the active site, the N-terminal regulatory domain of Dcp2 also interacts with Dcp1 via hydrophobic interactions (She et al., 2008). Interestingly, the N-terminal part of Drosophila and C. elegans Dcp2 also feature long intrinsically disordered regions (IDRs) (Figure 1A) but their significance for decapping catalysis and regulation is currently unknown.

The presence of a disordered C-terminal IDR is a common feature of eukaryotic Dcp2 proteins. The functions of the C-terminal IDR have been most extensively studied in yeast, where it serves both positive and negative regulatory functions through distinct elements. The positive regulatory elements are encoded as leucine-rich sequences commonly referred to as Helical Leucine-rich Motifs or HLMs (Gaudon et al., 1999). In S. cerevisiae, at least 10 HLMs were identified, first defined as short sequences bearing a core LLXΦL motif where Φ denotes any hydrophobic residue (Gaudon et al., 1999). In vitro pull-down assays demonstrated that each of these HLMs, except for HLMs 1 and 9, can directly interact with the decapping activator PatL1 (Pat1) (Charenton et al., 2017). In vivo deletions further suggested that HLMs 2–6 (equivalent to HLMs 1-5 in the He et al. study) contributed the bulk of PatL1 (Pat1) binding (He et al., 2021). Intriguingly, a strain bearing the deletion of HLMs 2–9 (equivalent to all HLMs in the He et al. study) still decays endogenous decapping targets as efficiently as wild-type (He et al., 2021). This could possibly be explained by a functional compensation by HLM1 and the Edc3 binding motif. In S. pombe, 7 putative HLMs have been identified (Fromm et al., 2012; Fromm et al., 2014). S. pombe HLMs can promote decapping activity in vitro by recruiting several decapping activators such as PatL1 and Edc3 with different affinities (Fromm et al., 2014; Lobel et al., 2019). Curiously, and unlike in S. cerevisiae and S. pombe, the C-terminal IDR of Drosophila, C. elegans and human Dcp2 lack any obvious HLMs. Human Dcp2 instead encodes a structured motif at the extreme C-terminal region, which recruits the decapping activator Edc4 (Chang et al., 2014). The rapid divergence of IDR and HLM sequences in Dcp2 C-terminus across species raises important questions on their possible functional redundancy or compensation by decapping activators.

In addition to HLMs, the C-terminal IDR of S. cerevisiae and S. pombe Dcp2 also contains negative cis regulatory elements. An inhibitory element, often referred to as Inhibitory Motif, was first reported by the Jacobson lab in a complementation experiment revealing that certain truncated fragments of Dcp2 were constitutively active and could bypass the requirement for Edc3 (He and Jacobson, 2015). Systematic deletions of the C-terminal IDR further mapped the inhibitory motif to a 25-amino acid region enriched in prolines and phenylalanines (He and Jacobson, 2015). More recently, the Gross lab identified two inhibitory motifs in S. pombe Dcp2, one similar to the S. cerevisiae inhibitory motif and the other that is exclusively found in closely related Schizosaccharomyces species (Paquette et al., 2018). Their structural implications remain unclear due to lack of crystal structures of these motifs in the context of the full-length active or inactive Dcp2. Nonetheless, at least one of the inhibitory motifs in S. pombe could interact with a Dcp2 fragment containing the N-terminal regulatory and the Nudix hydrolase domains (Paquette et al., 2018), supporting a possibility that they may block the active site.

Whether metazoan Dcp2 harbors such inhibitory motifs remains to be confirmed. However, the C-terminal 60 amino acids of human Dcp2 are subjected to ubiquitination and subsequent proteasomal degradation of Dcp2 (Erickson et al., 2015). Thus, across a wide range of species, Dcp2 is subjected to both positive and negative regulation by its C-terminus through diverse, and possibly diverging, mechanisms.

Dcp1 is the regulatory subunit of the Dcp1/2 holoenzyme (She et al., 2008). In S. cerevisiae and S. pombe, Dcp1 is essential for decapping activation and strains wherein Dcp1 is deleted exhibit severely impaired decapping activity (Beelman et al., 1996; Sakuno et al., 2004). Dcp1 interacts directly with Dcp2 in S. cerevisiae and S. pombe (She et al., 2008) and strongly potentiates Dcp2 activity in vitro (Steiger et al., 2003). In contrast, human Dcp1 only interacts with Dcp2 with low affinity, but their interaction is stabilized by Edc4 (Chang et al., 2014). This three-way interaction is required for decapping in vitro and mutations that specifically impair the interface between Dcp1 and Dcp2 fails to rescue the degradation of a reporter mRNA (Chang et al., 2014). The requirement for Edc4 likely explains earlier observations that recombinant Dcp1 from C. elegans, Drosophila and human cells are insufficient to promote catalysis by Dcp2 in vitro (van Dijk et al., 2002; Cohen et al., 2004; Lin et al., 2008). Thus, Dcp1 promotes Dcp2 activity in yeast and metazoans, but their functional architecture seems to diverge and involve additional cofactors in metazoans (Chang et al., 2014).

Dcp1 is primarily characterized by an EVH1 domain (Callebaut, 2002) that mediates the interaction with the N-terminal regulatory domain of Dcp2 (She et al., 2004; She et al., 2008) (Figure 1B). In Drosophila and possibly in other species, the EVH1 domain of Dcp1 also interacts with proline-rich sequences in the exonuclease Xrn1, thus physically coupling decapping and decay (Braun et al., 2012). In S. cerevisiae and S. pombe, Dcp1 is solely composed of the EVH1 domain (Callebaut, 2002). Interestingly, metazoan Dcp1 proteins also contain a longer IDR extension, a difference that is mirrored by the shortening of the Dcp2 C-terminal IDR. Furthermore, the IDR of Dcp1 in Drosophila and humans encodes an HLM that can interact with Edc3, further suggesting that some of the functions of Dcp2 IDR have been transferred over to Dcp1 in metazoans (Tritschler et al., 2009a; Fromm et al., 2012). The extreme C-terminus of metazoan Dcp1 further encodes a structured region that enables trimerization and is required for Dcp2 to interact with Edc4 (Tritschler et al., 2009b). However, the molecular basis for the interface between metazoan Dcp2 and the trimeric form of Dcp1 remains to be determined structurally.

Edc1 and Edc2 are encoded by loci that were identified as suppressors of Dcp1/2 mutations in S. cerevisiae (Dunckley et al., 2001). They encode small, intrinsically disordered proteins sharing 42% amino acid similarity (Dunckley et al., 2001), that enhance decapping in vitro by 140- and 40-fold, respectively (Borja et al., 2011), and can bind RNA (Schwartz et al., 2003). The decapping enhancement function of Edc1 was attributed to two of its motifs, namely an LPXP motif which interacts with Dcp1 EVH1 domain (Borja et al., 2011; Valkov et al., 2016; Wurm et al., 2016), and a YAG activation motif that binds in a groove between the Dcp2 N-terminal regulatory and Nudix hydrolase domains, thereby stabilizing the cap-binding pocket (Wurm et al., 2017; Mugridge et al., 2018b) (Figure 2A). The latter contributes to an increase in the affinity of the Dcp1/2 holocomplex towards RNA, which in turn enhances catalysis (Wurm et al., 2016; Wurm et al., 2017). In contrast, S. cerevisiae Edc2 lacks the YAG motif and how it promotes decapping is not well understood (Figure 2B). Both mammalian homologs of Edc1/2, PNRC1 and PNRC2, contain the LPXP and YAG motifs and are likely to activate Dcp1/2 via similar mechanisms. Curiously, PNRC1 and PNRC2 seem to have specialized in distinct RNA decapping functions. On one hand, human PNRC1 is predominantly nuclear and can recruit Dcp1/2 complex to the nucleolus, where this interaction was proposed to effect decapping of the U3 and U8 snoRNAs (Gaviraghi et al., 2018). On the other hand, PNRC2 is cytoplasmic and localizes to P bodies (Cho et al., 2009), but it forms a complex and functions with a different set of decapping activators, including Upf1 and Smg5, which are involved in the nonsense-mediated decay (NMD) pathway (Lai et al., 2012; Cho et al., 2013). Any link between PNRC1 and PNRC2 with other decapping activators, and whether they are involved more generally in mRNA decapping pathways outside of NMD, remains to be determined.

Edc3 was implicated in decapping through several independent large-scale proteomic surveys in S. cerevisiae, which detected interactions with decapping proteins Dcp1, Dcp2, PatL1 (Pat1) and the Lsm proteins (Fromont-Racine et al., 2000; Uetz et al., 2000; Gavin et al., 2002). Its function in decapping activation was first demonstrated in S. cerevisiae where deletion of Edc3 exacerbates the decapping impairment caused by hypomorphic Dcp1 or Dcp2 alleles (Kshirsagar and Parker, 2004), and recombinant Edc3 clearly potentiates decapping by Dcp1/2 complex in vitro (Harigaya et al., 2010; Nissan et al., 2010; Fromm et al., 2012). In human cells, Edc3 also localizes to P bodies and promotes decapping by Dcp2 (Fenger-Gron et al., 2005).

Edc3 (Figure 2C) encodes an N-terminal LSm (Sm-like) domain which, unlike the canonical Sm motif, neither multimerizes into Sm rings nor binds RNA, but rather accommodates different protein-protein interactions (Tritschler et al., 2007). This domain interacts directly with a region immediately upstream of the first HLM of Dcp2 in K. lactis and S. cerevisiae (Charenton et al., 2016), with the HLMs themselves in S. pombe, or with the Dcp1 HLM in metazoans (Fromm et al., 2012; Fromm et al., 2014). Downstream of the LSm domain of Edc3 is a long IDR that harbors a conserved FDF (phenylalanine-aspartic acid-phenylalanine) motif and interacts with Ddx6 (Tritschler et al., 2009a). Lastly, the C-terminal end of Edc3 is characterized by a YjeF-N domain with a Rossman fold topology that facilitates Edc3 homodimerization (Ling et al., 2008) and in turn promotes phase separation in vitro (Tibble et al., 2021). In vivo, Edc3 appears to act redundantly with another LSm domain-containing decapping activator, Lsm14 (S. cerevisiae Scd6, detailed below), as decapping is only impaired when both proteins are deleted simultaneously (Decourty et al., 2008).

Interestingly, S. cerevisiae Edc3 can promote the degradation of two specific transcripts in an atypical and deadenylation-independent manner. Firstly, a short motif on Edc3 that is conserved among Saccharomycetaceae species can interact with the ribosomal protein Rps28, and together bind to a hairpin in the 3′-untranslated region of the rps28b mRNA to promote its degradation (Badis et al., 2004; Kolesnikova et al., 2013; He et al., 2014). This Edc3-mediated degradation of rps28b requires translation and thus presents intriguing autoregulation of ribosomal protein levels that directly involves decapping cofactors (He et al., 2014). Secondly, Edc3 promotes the degradation of nuclear export factor YRA1 pre-mRNA, which uniquely evades NMD in the cytoplasm (Dong et al., 2007). The mechanism likely involves direct recruitment of Edc3 to cis elements in the YRA1 intron following translational repression (Dong et al., 2010).

Edc4 was identified as a component of P bodies in human cells, where it co-localizes with Dcp1/2 (Fenger-Gron et al., 2005; Yu et al., 2005). It was also captured in a screen for miRNA-mediated silencing components in Drosophila S2 cells (Eulalio et al., 2007b). Edc4 (Figure 2D) encodes an N-terminal WD40 domain that arranges as a circularized seven-bladed beta-propeller and facilitates interaction with Dcp1 trimers (Chang et al., 2014). This domain is successively followed by a serine-rich linker and a C-terminal alpha-helical hairpin repeat similar to those found in ARM and HEAT-repeat proteins (Jinek et al., 2008). In humans, the proximal C-terminus of Edc4 promotes Edc4 oligomerization, as well as interacts directly and simultaneously with Dcp2 and Xrn1 (Chang et al., 2014). These interactions are mediated by short Edc4-binding motifs in Dcp2 and Xrn1 (Braun et al., 2012; Chang et al., 2014). Interestingly, the Edc4-binding motif in Dcp2 is not conserved in Drosophila and C. elegans (Chang et al., 2014), and thus how Edc4 interacts with Dcp2 in these organisms remain unclear. In humans, Edc4 interaction with Xrn1 also alleviates the inhibition of deadenylation caused by Xrn1-mediated sequestration of Caf1, one of the catalytic deadenylation subunit (Chang et al., 2019). Therefore, Edc4 could indirectly enhance decapping by promoting deadenylation, and more directly by scaffolding decapping and decay enzymes. Recently, Edc4 was involved in the inhibition of a novel mRNA decay pathway initiated by the endonuclease MARF1 (Brothers et al., 2020). Edc4 can thus serve both as an enhancer of decapping and as a repressor of mRNA decay pathways.

Edc4 does not have any clear ortholog in S. cerevisiae, but the S. pombe Pdc1 protein exhibits commonalities with Edc4 (Wang et al., 2013; Fromm et al., 2014). Pdc1 bears a WD40-repeat and a distal C-terminus that folds into helical repeats found in Drosophila Edc4 despite sharing only 17% of sequence identity (Fromm et al., 2014). Pdc1 also interacts with the LSm domain of Edc3 through at least three HLMs encoded in its N-terminus (Fromm et al., 2014), and co-localizes with other decapping factors to P bodies (Wang et al., 2013). Whether and how Pdc1 directly activates the Dcp1/2 holoenzyme in S. pombe remains to be determined. A structural comparison of Edc4 and Pdc1, alone or in interaction with Dcp1/2, could likely prove insightful.

Pby1 is a decapping activator in yeast that is related to the human tubulin tyrosine ligase (TTL) protein (Sweet et al., 2007; Rao and Parker, 2017; Charenton et al., 2020). Pby1 involvement in mRNA decapping and decay regulation was suggested by its association with P bodies through large-scale interactome studies (Gavin et al., 2006; Decourty et al., 2008). This was confirmed by fluorescence imaging of S. cerevisiae where Pby1 co-localizes with Dcp2 in P bodies (Sweet et al., 2007). Pby1 is composed of an N-terminal domain that resembles the SurE phosphatase family (Iwasaki and Miki, 2007) and a C-terminal domain consisting of an ATP-grasp fold (Fawaz et al., 2011) (Figure 2E). Interestingly, the structure of the Pby1-Dcp1-Dcp2-Edc3 complex of S. cerevisiae revealed that several conserved residues in the Pby1 C-terminal domain interact with the Nudix domain of Dcp2 (Charenton et al., 2020). Moreover, point mutations that impair this interaction disperse Pby1 from P bodies to the cytoplasm (Charenton et al., 2020). In vivo, the decapping-promoting function of Pby1 is inferred from the observation that Pby1 overexpression restores the growth defect phenotype of decapping mutants in S. cerevisiae, namely a double deletion of PatL1 and Ddx6 (Dhh1) and a triple deletion of PatL1, Edc3 and Lsm14 (Scd6) (Charenton et al., 2020). This effect is critically contingent upon Pby1 interaction with Dcp2 (Charenton et al., 2020).

How Pby1 precisely impacts Dcp2 activity remains an important question to address. The catalytic activity of the ATP-grasp domain is dispensable for function in vivo, and in vitro decapping assays did not show a direct enhancement in decapping activity upon addition of recombinant Pby1, possibly due to a missing co-factor (Charenton et al., 2020). Furthermore, Pby1 deletion did not significantly alter the stability of reporter mRNAs (Sweet et al., 2007). One possibility could be that Pby1 functions predominantly in certain cellular states or when other decapping activators are involved, as suggested by its genetic interactions (Charenton et al., 2020). The latter is also in line with the observation that Pby1 could drive Dcp2 localization into P bodies in the absence of Edc3 and Lsm14 in S. cerevisiae (Rao and Parker, 2017). Assaying the decapping-promoting function of Pby1 in mutant backgrounds might provide further mechanistic and regulatory insights. Lastly, whether metazoan orthologs of Pby1 are involved in decapping, or whether other metazoan proteins might serve functions that are orthologous to Pby1 would be interesting avenues of investigation.

PatL1 (Pat1) was linked to decapping through a genetic screen for S. cerevisiae mutants that fail to degrade an unstable MFA2pG reporter mRNA (Hatfield et al., 1996), and independently in a suppressor screen that renders a Pab1 deletion viable (Boeck et al., 1998; Bonnerot et al., 2000). Lsm1, which is part of the cytoplasmic Lsm1-7 complex (Sharif and Conti, 2013; Wu et al., 2014), was also captured in the latter screen, and as with PatL1 and Dcp2, its deletion strongly inhibits decapping (Bouveret et al., 2000; Tharun et al., 2000). Furthermore, strains bearing deletions of Lsm1 or PatL1 have very similar transcriptome profiles (He et al., 2018). The decapping and decay role for metazoan PatL1 was robustly established by the observation that tethering of PatL1 leads to destabilization of reporter mRNAs both in Drosophila and human cells (Haas et al., 2010; Ozgur et al., 2010).

Structurally, the PatL1 protein can be divided into four functional regions: N-terminal, proline-rich, Mid and C-terminal domains (Haas et al., 2010) (Figure 2F). The N-terminal domain is a predicted disordered region and harbors an FDF motif (equivalent to the DW motif in metazoan PatL1) that binds directly to Ddx6 (Dhh1) (Haas et al., 2010; Sharif et al., 2013). This domain is dispensable for decapping activity and for S. cerevisiae growth (Pilkington and Parker, 2008), likely due to redundancy of decapping mechanisms that converge on Ddx6 (Sharif et al., 2013). The proline-rich region of PatL1 possibly interacts with the EVH1 domain of Dcp1 (Braun et al., 2012), which may provide a direct link to the decapping holoenzyme. The Mid and C-terminal domains are conserved across species and fold into a helical organization (Braun et al., 2010; Fourati et al., 2014; Wu et al., 2014). The Mid domain co-immunoprecipitates with the CCR4-NOT deadenylase complex (Haas et al., 2010), but whether this involves direct binding remains to be determined. A highly conserved motif in the N-terminal portion of the C domain, along with some residues in the Mid domain, provides a bipartite module for binding to the Lsm1-7 complex (Braun et al., 2010; Fourati et al., 2014). This interaction promotes decapping by stabilizing the binding of Lsm1-7 to RNA (Chowdhury et al., 2007; Lobel et al., 2019). In yeast, the other end (C-terminal) of the PatL1 C-terminal domain consists of a yeast-specific motif that interacts directly with the HLMs of Dcp2 and Xrn1 (Charenton et al., 2017; Lobel et al., 2019). This interaction mirrors the function of metazoan Edc4 (Chang et al., 2014) in coupling decapping with exonucleolytic degradation. PatL1 interaction with Dcp2 was proposed to alleviate decapping autoinhibition, possibly by altering the conformation of the C-terminal IDR of Dcp2 (Charenton et al., 2017; Lobel et al., 2019). Point mutations that impair the HLM-binding sites on PatL1 significantly impair decapping in vitro and in vivo (Charenton et al., 2017). Together, the Mid and C-terminal domains are sufficient for PatL1 function in vivo (Pilkington and Parker, 2008).

In addition to its decapping and decay scaffolding functions, the affinity and preference of the PatL1/Lsm1-7 complex for deadenylated RNAs provide a rationale for the functional linkage of the decapping machinery with deadenylated mRNAs. In vitro, a PatL1/Lsm1-7 octamer binds the 3’ end of mRNAs with greater affinity towards oligoadenylated (<10 terminal adenines) than polyadenylated mRNAs, and an even greater affinity towards RNA bearing a stretch of ∼6 uracils (Chowdhury et al., 2007). Although the Lsm1-7 complex seems to prefer U-rich sequences on its own, PatL1 broadens its affinity towards more A-rich sequences (Lobel et al., 2019; Lobel and Gross, 2020). This agrees with RNA-seq in human cells where PatL1 depletion preferentially stabilizes AU-rich mRNAs (Vindry et al., 2017).

In contrast to a role in decapping activation, a PatL1/Lsm1-7 complex has also been involved in protecting deadenylated mRNAs from degradation (He and Parker, 2001) and in keeping transcripts in a translationally repressed state in S. cerevisiae (Coller and Parker, 2005) and during early oocyte development in X. laevis (Marnef et al., 2010). Under hyperosmotic stress, the PatL1/Lsm1-7 complex is recruited to stress-induced transcripts to repress their translation without impacting mRNA stability, and deletion of PatL1 or Lsm1 reactivates the translation of these transcripts (Garre et al., 2018). Evidence supports a similar function in metazoans. In Drosophila, PatL1 associates with the Ddx6 (Me31B) and GIGYF proteins to repress translation (Peter et al., 2019). The same interaction between PatL1 (PATR-1) and GIGYF (GYF-1) has also been reported in C. elegans, where GYF-1 represses translation of some developmental miRNA targets (Mayya et al., 2021). Thus, promotion of decapping and decay or translational repression by PatL1 seems to depend on its specific interactions and on environmental and developmental contexts.

Lastly, the functions of PatL1 and Lsm1-7 proteins in mammals extend beyond their involvement in mRNA silencing. For example, while human PatL1 is predominantly (∼80%) cytoplasmic, the nuclear PatL1 pool associates with Lsm2-8, U6 snRNA and SART3 as part of the U6 snRNP in Cajal body (Vindry et al., 2017). This complex promotes exon cassette inclusion, with knockdown of PatL1 in cells leading to changes in about 180 alternative splicing events with weak splice donor sites (Vindry et al., 2017).

4E-T (Cup) is a metazoan-specific protein that has dual roles in translational repression and decapping & decay by virtue of its direct interactions with translational repressors and decapping co-factors. Most of its sequence is unstructured and poorly conserved, but 4E-T features two eIF4E-binding sites (a canonical YXXXXLϕ and a non-canonical 4E-binding motif) in the N-terminus (Dostie et al., 2000; Nelson et al., 2004), and a conserved 4E-T (Cup) Homology Domain (Kamenska et al., 2014) (Figure 2G). The two eIF4E-binding motifs of 4E-T can also interact with another cap-binding protein, the translational repressor 4EHP (Kubacka et al., 2013). Two separate motifs in the 4E-T Mid and C-terminal regions also interact with Ddx6 and Lsm14 (Nishimura et al., 2015), which in turn can interact with other decapping activators (Brandmann et al., 2018). The importance of 4E-T in mRNA decapping and decay is supported by the observation that knockdown of 4E-T stabilizes Tristetrapolin (TTP)- and miRNA-mediated decay reporter mRNAs in HeLa cells (Ferraiuolo et al., 2005; Nishimura et al., 2015). Mechanistically, this decapping and decay enhancement is thought to function through the competitive release of the 5′-cap from the translation initiation complex (eIF4E, 4G, 4A), rendering the cap accessible to decapping complex (Ferraiuolo et al., 2005; Nishimura et al., 2015).

In contrast with its role in decapping enhancement, substantial evidence indicates that 4E-T can also protect target mRNAs from decapping and decay. In Drosophila, in situ hybridization and qRT-PCR results indicated that targeted mRNAs are destabilized upon depletion of 4E-T (Cup), or upon mutation in its eIF4E-binding motif (Broyer et al., 2017). In Drosophila S2 cells, 4E-T (Cup)-bound mRNAs are deadenylated and capped, but are destabilized when the eIF4E-binding motif of 4E-T (Cup) is mutated (Igreja and Izaurralde, 2011). Curiously, transfection of a construct encoding only the Mid or C domain leads to deadenylation, decapping and decay of Cup-bound mRNAs, suggesting that the N-terminus of 4E-T inhibits decapping and decay (Igreja and Izaurralde, 2011). Similarly, in human HEK293 cells, tethering of 4E-T also resulted in the stabilization of reporter mRNAs bearing an AU-rich element or miRNA-binding sites (Räsch et al., 2020). This result stands at odds with earlier tethering experiments (Ferraiuolo et al., 2005; Nishimura et al., 2015). Mechanistically, 4E-T could repress mRNA translation by increasing the affinity of 4EHP towards the cap (Chapat et al., 2017), which in turn could block Dcp1/2 from accessing the cap.

A possible explanation for the alternative fates of 4E-T–bound mRNAs could lie in the availability and binding affinities of eIF4E and 4EHP to 4E-T. Perhaps in certain cell types or under specific conditions where 4EHP is abundant or more readily interacts with 4E-T, protection from decapping could be favored for 4E-T–bound mRNAs. It is also possible that this affinity for the cap can be modulated through conformational changes due to other interactions with 4E-T. Better structural insight on 4E-T and its interacting partners and precise quantification of the affinity of the 4E-T/4EHP complex in its native niche will likely be required to shed light on the role of 4E-T in mRNA translational repression, decapping and decay.

The Lsm14 (Rap55) protein was first identified as a component of mRNP complexes in oocytes and early embryos of the amphibians Pleurodeles waltl and X. laevis (Lieb et al., 1998), and later as an essential component of Dcp1 and Edc4-containing P bodies in human cells (Bloch et al., 2005; Tanaka et al., 2006; Yang et al., 2006). The S. cerevisiae ortholog of Lsm14 (Scd6) was implicated in decapping through a screen for growth defects under a null allele of Edc3 (Decourty et al., 2008).

Lsm14 (Figure 2H) is related to Edc3 in sharing an N-terminal LSm domain (Albrecht and Lengauer, 2004). In S. cerevisiae Lsm14 (Scd6), this domain is necessary and sufficient to promote the decapping of a reporter mRNA (Zeidan et al., 2018) and interacts with the HLMs of Dcp2 (Fromm et al., 2012). The orthologous domain interacts with the EVH1 domain of Dcp1 in metazoans (Tritschler et al., 2008). In contrast with Edc3, which ends with a structured YjeF-N dimerization domain, the C-terminal end of Lsm14 consists of a long IDR that harbors an FDF motif and RGG repeats that vary in numbers in different species (Marnef et al., 2009). Metazoan Lsm14 features two clusters of RGG repeats that sandwich the FDF motif, while the yeast ortholog (Scd6) has only one RGG repeat located downstream of the FDF motif (Lieb et al., 1998; Marnef et al., 2009). RGG motifs are important for Lsm14 to localize to P bodies, but their contribution seems to vary between cell types or experimental conditions (Tanaka et al., 2006; Yang et al., 2006; Matsumoto et al., 2012). For example, both RGG clusters are necessary and sufficient in human Hep-2 cell lines (Yang et al., 2006) but seem to have a lesser impact in HeLa cells (Tanaka et al., 2006; Matsumoto et al., 2012). Although the precise mechanism of Dcp1/2 enhancement by Lsm14 is currently poorly understood, Lsm14 (Scd6) and Edc3 in S. cerevisiae maintain a functional pool of Dcp2 in the cytoplasm, as their combined deletion results in nuclear retention of inactive Dcp2 (Tishinov and Spang, 2021).

In addition to decapping enhancement, Lsm14 also mediates translational repression when tethered to reporter mRNAs in X. laevis oocytes (Tanaka et al., 2006) and S. cerevisiae (Zeidan et al., 2018). In S. cerevisiae, this function is attributed to the interaction between the RGG motifs on Lsm14 with eIF4G within the eIF4F complex, thereby repressing the assembly of 48S initiation complex (Rajyaguru et al., 2012). In Drosophila and human Lsm14, the LSm domain recruits 4E-T (Cup), and a bipartite phenylalanine-rich motif (FDF and TFG) in the Lsm14 IDR segment interacts with Ddx6 (Me31B) (Tritschler et al., 2008; Brandmann et al., 2018). The three proteins (Lsm14, Ddx6 and 4E-T) form a highly conserved translational repressor complex (Tritschler et al., 2008; Brandmann et al., 2018). In Drosophila and C. elegans early embryonic development, this complex represses a subset of maternal mRNAs (Boag et al., 2008; Wang et al., 2017), and mutants of the C. elegans Lsm14 ortholog (CAR-1) die early in embryogenesis due to failed cytokinesis (Audhya et al., 2005).

Since Lsm14 utilizes the LSm domain to interact with both Dcp1 and 4E-T (Cup) (Tritschler et al., 2008), the capacity of Lsm14 to induce decapping or translational inhibition might be a consequence of the relative affinities and cellular availabilities of Dcp1/2, 4E-T and Ddx6. This hypothesis could be tested through detailed quantification of their expression in different cellular and developmental contexts.

Ddx6 (Dhh1/Me31B) is a DEAD-box helicase that functionally intersects with deadenylation, decapping activators, and translational repression across multiple eukaryotic species (Weston and Sommerville, 2006). Ddx6 (Figure 2I) and its orthologs are characterized by a core helicase composed of two RecA-like domains separated by a short linker (Ostareck et al., 2014). Like other DEAD box helicases, the Ddx6 active site is formed when the two RecA domains come into close interaction in the presence of ATP and RNA (Ostareck et al., 2014). Unlike other DEAD box helicases, Ddx6 has a weak ATPase activity and needs additional factors to be activated (Dutta et al., 2011). More specifically, direct binding of the MIF4G domain of CNOT1, a scaffold subunit of the CCR4-NOT deadenylase complex, to the C-terminal RecA domain of Ddx6 (Dhh1) changes the active site conformation and activates ATP hydrolysis (Maillet and Collart, 2002; Mathys et al., 2014; Mugler et al., 2016). Mutation of the CNOT1/Ddx6 binding interface de-represses a reporter mRNA bearing let-7 miRNA binding sites; hence ATPase activity is important for silencing through Ddx6 (Dhh1) (Mathys et al., 2014; Rouya et al., 2014).

Ddx6 can promote decapping and decay through its extensive interactions with decapping activators. The C-terminal RecA domain of Ddx6 can bind to Edc3, PatL1, Lsm14 and 4E-T using overlapping binding interfaces (Tritschler et al., 2008; Sharif et al., 2013; Ozgur et al., 2015). The possibility of competitive binding to Ddx6 suggests that it may split the partner proteins into distinct complex populations, enacting either translational repression (in the case of PatL1, Lsm14, or 4E-T) or decapping (in the case of Edc3, PatL1, Lsm14, or 4E-T) on deadenylated mRNAs. Since ATP hydrolysis by DEAD box helicases leads to reduced RNA binding (Hondele et al., 2019), activation of the Ddx6 ATPase by CNOT1 (Not1) may release the deadenylated mRNA to be accessible for the recruitment of other Ddx6-associated decapping activators.

Interestingly, the interaction between Ddx6 and the CCR4-NOT scaffold subunit is thus far the only direct physical link between the deadenylation and decapping machineries. This is a critical concept, as mRNA deadenylation is a typical prelude to decapping and decay (Muhlrad et al., 1994; Garneau et al., 2007). Deadenylation is triggered by a wide variety of elements in the 3′-untranslated region and their associated factors (Mayya and Duchaine, 2019). The coupling of deadenylation, decapping and decay is seen in a variety of decay pathways including silencing through the miRNA Induced Silencing Complex, and by RNA-binding proteins such as Pumilio (Van Etten et al., 2012) and AU-rich associated Tristetrapolin (Sandler et al., 2011). These factors directly recruit the deadenylase complex through CNOT1 (Braun et al., 2011; Fabian et al., 2013; Enwerem et al., 2021), but among the RNA binding proteins above, only Tristetrapolin is known to directly interact with Dcp2 thus far (Maciej et al., 2021). Since Ddx6 is a highly abundant protein in cells (Ghaemmaghami et al., 2003; Beck et al., 2011) and it directly interacts with CNOT1, it could serve as a versatile platform to couple deadenylation with decapping and decay machineries in a broad variety of regulatory pathways.

A multitude of RNA-seq, CLIP-seq and ribosome profiling studies led to diverging conclusions on whether Ddx6-bound transcripts are committed to degradation or sequestered and stored in a translationally repressed state (Radhakrishnan et al., 2016; Courel et al., 2019). In S. cerevisiae, Ddx6 (Dhh1) preferentially binds and promotes the degradation of transcripts with suboptimal codons (Radhakrishnan et al., 2016). Conversely, in HEK293 cells Ddx6 represses the translation of suboptimal codon-containing transcripts that are AU-rich without affecting their stability, but promotes the degradation of GC-rich transcripts (Courel et al., 2019). Both during human embryonic stem cell differentiation (Freimer et al., 2018) and in C. elegans oogenesis (Boag et al., 2008), Ddx6 predominantly enables translational inhibition without affecting mRNA stability. The mode of silencing by Ddx6 can also change in the course of development; in early Drosophila embryo, Ddx6 (Me31B) binding is initially associated with reduced translational efficiency at 0–1 h post egg-laying, without impacting mRNA stability, but is correlated with reduced mRNA stability at later timepoints (1–3 h) (Wang et al., 2017). This transition is attributed to the reduced availabilities of Ddx6 (Me31B)’s main interacting partners 4E-T (Cup) and Lsm14 (Trailer Hitch) in the later developmental stages (Zavortink et al., 2020). The different conclusions drawn from these diverse studies suggest that the fate of Ddx6-bound mRNAs could vary depending on intrinsic mRNA features, but also, as with the other decapping activators discussed above, through availability and competitive interactions with decapping or translational factors.

Several observations support the possibility that P bodies represent active sites for mRNA decapping and decay. Firstly, abolishing decapping or decay activities via genetic depletion of Dcp2 or Xrn1 enlarges P bodies in both S. cerevisiae and Drosophila (Sheth and Parker, 2003). Secondly, kinetic measurements of S. pombe Dcp2 activity in vitro demonstrated that Edc3 alleviates Dcp2 autoinhibition and promotes its activity in LLPS condensates (Tibble et al., 2021). Similarly, deadenylation of reporter transcripts is enhanced in phase-separated granules in vitro (Sheu-Gruttadauria and MacRae, 2018). A simplistic view is that the convergence of deadenylation, decapping and decay factors inside P bodies might ensure a coupling between deadenylation and decay and enhance their kinetics, thereby leading to robust mRNA turnover.

In contrast with this model, several observations suggest that decapping and decay can, and do also occur outside of P bodies. Firstly, mutations that abolish visible Edc4- and 4E-T (Trailer Hitch)-positive P bodies in Drosophila S2 cells had no impact on the degradation of reporter mRNAs (Eulalio et al., 2007a). Secondly, Edc1 can activate decapping in the cytoplasm as well as in LLPS condensates in vitro (Tibble et al., 2021). Thirdly, quantification of cytoplasmic versus P body–localized pool of decapping proteins in S. cerevisiae suggests that only a small fraction (∼10%) of decapping activators localizes to P bodies, with the exception of Dcp2 for which ∼30% of the total pool is concentrated in P bodies (Xing et al., 2020). Therefore, while P bodies may be a possible site for enhanced decay, several lines of independent evidence indicate that they are not required and certainly not the sole site for mRNA decapping and decay. It is possible that P bodies can serve to sequester specific sets of mRNAs away from the cytoplasm, to be degraded upon regulated cellular or environmental cues. In support of this, comparative RNA-seq of P bodies purified from S. cerevisiae revealed that different sets of mRNAs are enriched in P bodies under different stress conditions (Wang et al., 2018). Metabolic labelling showed that some of these P body–localized transcripts are destabilized, although this study could not conclusively infer whether their destabilization occurs inside or outside of P bodies (Wang et al., 2018). Since components of the 3′-to-5′ decay pathway are depleted from P bodies (Hubstenberger et al., 2017; Youn et al., 2018), it is also possible that P body localization may sort specific mRNAs toward a particular decay route.

Although the aforementioned evidence supports a function for P bodies in RNA decapping and decay, substantial evidence instead indicates that they can serve as a storage compartment for translationally repressed mRNAs. It was noticed early on that P bodies increase in size upon inhibition of translation and disassemble upon translational reactivation, in both yeast and Drosophila cells (Brengues et al., 2005; Eulalio et al., 2007a). More recently, ribosome profiling and RNA-seq of purified Lsm14-positive P bodies from HEK293 cells indicate an enrichment of transcripts that are poorly translated (Hubstenberger et al., 2017). Furthermore, a detailed single-molecule resolution kinetic study in HEK293 cells did not detect any difference in mRNA decay rate inside or outside of Ddx6-containing P bodies (Wilbertz et al., 2019). The lack of observed differences in mRNA decay rates may be in part due to the selective purification of Ddx6- or Lsm14-containing P bodies in the latter two studies. As discussed above, in some conditions, both Ddx6 and Lsm14 can repress translation without affecting mRNA stability. It will be interesting to compare this observation with the mRNA profile of purified P bodies obtained from selective enrichment of Dcp2 or Edc1-4.

Notwithstanding the above, the diversity of proteins and interactions upon which P bodies are scaffolded could reflect a functional heterogeneity across P body foci, with some favoring or dedicated to translational repression and others biochemically geared towards decapping and decay. From a functional standpoint, a few reports suggested diverging roles for P bodies in distinct cellular states. On one hand, P bodies promote differentiation of pluripotent stem cells, as well as neural and intestinal progenitors in human cells (Di Stefano et al., 2019). Conversely, they maintain the pluripotency of mesenchymal stem cells in human cells (Di Stefano et al., 2019) and Drosophila (Buddika et al., 2022). It stands to reason that intrinsic features of targeted mRNAs, interactions with RNA-binding proteins, or spatiotemporal expression of P body components may alter P body composition and function. Detailing the heterogeneity of P bodies across cell types, but also within a defined cell state, may prove useful in resolving the apparent functional divergence and specificity of the roles for these pervasive foci.