Poly(A) tails are important regulators of mRNA stability and translational efficiency. Cytoplasmic removal of poly(A) tails by 3′→5′ exonucleases (deadenylation) is the rate-limiting step in mRNA degradation. Two exonuclease complexes contribute the majority of the deadenylation activity in eukaryotes: Ccr4–Not and Pan2–Pan3. These can be specifically recruited to mRNA to regulate mRNA stability or translational efficiency, thereby fine-tuning gene expression. In the present review, we discuss the activities and roles of the Pan2–Pan3 deadenylation complex.
Almost every eukaryotic mRNA is synthesized with a poly(A) tail of uniform length: yeast mRNAs have a ~70-nt poly(A) tail, whereas mRNAs in higher eukaryotes contain poly(A) tails of 200–250 nt. Poly(A) tails are added co-transcriptionally and are required for mRNA export from the nucleus . They are bound by PABP [poly(A)-binding protein] (Pab1p in yeast, PABPC1 in mammals). The poly(A) tail (and PABP) protects mRNA 3′ ends from non-specific exonuclease activities . In addition, the polyA tail and PABPs promote translational efficiency. The Pan2–Pan3 and Ccr4–Not multiprotein complexes comprise the main deadenylation activities in eukaryotic cells . Both complexes play key roles in the regulation of poly(A) tail length and gene expression. In the present article, we review the function of Pan2–Pan3, with a focus on the yeast complex.
Discovery of PAN
Early experiments addressing the function of poly(A) tails showed that depletion or inactivation of the essential Pab1p in yeast results in mRNAs with an increased average poly(A) tail length and reduced translation [4,5]. This led to the proposal that poly(A) tail removal and translation initiation are stimulated by Pab1p . Indeed, later work showed that PABPs interact with the eIF4F (eukaryotic translation initiation factor 4F) to stimulate initiation of translation [6,7]. eIF4F also interacts with the 5′-cap and therefore circularizes the mRNA. This ‘closed-loop’ conformation of mRNA provides a quality-control mechanism to prevent translation of partially degraded mRNA or aberrant mRNAs lacking a poly(A) tail. Longer poly(A) tails stimulate translation, whereas deadenylation results in translational repression [8,9].
Although PABPs protect mRNA from degradation by non-specific nucleases, they also appeared to stimulate poly(A)-specific exonucleases. For example, addition of purified Pab1p to extracts made from a Pab1p-deficient yeast strain stimulates shortening of poly(A) tails . Initial searches for the eukaryotic deadenylase PAN [poly(A) nuclease], were performed by fractionating yeast extracts . A 3′→5′ exonuclease activity was identified: this PAN activity could degrade a 3′ poly(A) tail, but not the mRNA body, and was dependent on two proteins: Pan2p and Pan3p [11,12]. Neither of these proteins are essential for viability in yeast, but their disruption results in an increased average poly(A) tail length in vivo . Deadenylation is completely eliminated only when the both CCR4 (one of the two deadenylase subunits of Ccr4–Not) and PAN2 are disrupted.
Pan2 is a 127-kDa protein (Figure 1), first identified in yeast by Boeck et al. , but conserved across all eukaryotes. It contains a nuclease domain of the DEDD/RNase D family of 3′→5′ exonucleases, the same type of domain found in the Caf1/Pop2 exonuclease subunit of the Ccr4–Not complex [13,14]. The exonuclease domain of Pan2 is located at its C-terminus.
Domain structures of Pan2 and Pan3
Pan2 also contains a WD40 repeat domain (at its N-terminus) and a UCH (ubiquitin C-terminal hydrolase) domain (Figure 1). WD40 domains often mediate protein–protein interactions, but an interaction partner for Pan2's WD40 domain has not been identified. Interestingly, the UCH domain is predicted to be inactive due to lack of active-site residues . It is unclear whether the UCH domain binds ubiquitin and what role this plays within the complex. Deletion of the first 332 amino acids of yeast Pan2p does not inhibit complex formation with Pan3p, suggesting that their interaction is mediated through the linker, UCH or exonuclease domain  (Figure 1).
Pan3 is the 76-kDa interacting partner of Pan2  (Figure 1). It contains an N-terminal zinc finger, a pseudokinase domain and a highly conserved C-terminal domain (amino acids 560–679 in yeast Pan3p) of unknown function. The C-terminal domain probably interacts with Pan2, as shown by yeast two-hybrid and pull-down experiments [16,17]. An unstructured linker that connects the zinc finger and pseudokinase domains contains a PAM2 (PABP-interacting motif 2) that binds to the C-terminal domain of PABPs . Yeast two-hybrid experiments also show that the C-terminal two-thirds of yeast Pan3p (amino acids 250–679) can self-interact .
The biochemical activity of the Pan2–Pan3 complex has been characterized extensively in vitro. Early work showed that Pan2–Pan3 binds to poly(U)–and poly(A)–Sepharose, suggesting a high affinity for RNA . Pan2–Pan3 purified from yeast or human cells is dependent on Mg2+, requires a 3′-hydroxy group and the product of the reaction is 5′-AMP [17,19].
Pan2–Pan3 was initially purified from a Pab1p-deficient yeast strain as an exonuclease that requires Pab1p to remove a poly(A) tail from a model mRNA substrate . Similarly, the human complex is also greatly stimulated by PABP . Pan2–Pan3 is specific for poly(A); it is unable to degrade the body of the mRNA. Although some conditions, e.g. addition of spermidine, allow PABP-independent deadenylation, it is generally assumed that PABP is required for recruitment to mRNA in vivo [17,19]. In agreement with this, Pan2–Pan3 does not efficiently remove the final 20–25 nucleotides of the poly(A) tail . PABP binding to poly(A) saturates at approximately 25 A residues per molecule [4,20], so this suggests that Pan2–Pan3 is only active on a poly(A)–PABP complex. In addition, Pab1p mutation in yeast results in longer poly(A) tails, the same phenotype observed in PAN2 and PAN3 deletion strains .
Pan2–Pan3 is predominantly cytoplasmic. However, it is unclear whether it acts solely in the cytoplasm or whether it is also involved in maturation of the poly(A) tail in the nucleus [21–24]. A second PABP found in the nucleus in yeast, Nab2p, may protect poly(A) tails from premature deadenylation .
Roles in mRNA decay
Deadenylation of mRNA is the first and rate-limiting step of mRNA degradation . Following the lag phase due to deadenylation, mRNA degradation proceeds rapidly either by decapping and 5′→3′ degradation by Xrn1, or by the 3′→5′ exonuclease activity of the cytoplasmic exosome [2,26]. Many of the mRNA decay factors localize to mammalian P-bodies (processing bodies), including Pan2 and Pan3 .
A two-phase model of deadenylation has been proposed wherein Pan2–Pan3 removes poly(A) until PABP can no longer bind (12–25 As) . Subsequently, Ccr4–Not removes the remaining As in the ‘terminal deadenylation’ phase, triggering rapid mRNA degradation [28,29]. Indeed, deletion of any single exonuclease subunit (Pan2p, Ccr4p, Caf1p) in yeast results in longer poly(A) tails, but not a severe growth phenotype [11,12], whereas a double PAN2/CCR4 deletion is very detrimental to cell growth with no residual deadenylation . In agreement with this, deletion of CCR4 in yeast impairs terminal deadenylation (removal of the final ~20 As), whereas deletion of PAN2 impairs the removal of the distal portion of a poly(A) tail . Similar conclusions have been made in mammalian cells where both complexes contribute to deadenylation, and overexpression of PAN2 promotes the first phase of deadenylation, but overexpression of CCR4 promotes the second phase [29,30]. These data also suggest functional redundancy between the two complexes.
Different mRNAs have different deadenylation rates, contributing to differences in their relative stabilities in vivo [19,25]. For example, the normally distributive deadenylation activity of Pan2–Pan3 appears to be processive on certain mRNAs (e.g. MFA2) . Interestingly, in trypanosomes, where transcription is polycistronic and mRNA degradation plays a larger role in controlling gene expression, depletion of Pan2 causes stabilization of mRNAs with intermediate half-lives, whereas depletion of Caf1 stabilizes most mRNAs .
Importantly, deadenylation can not only signal mRNA degradation, but also fine-tune gene expression. Deadenylation represses translation, and readenylation by cytoplasmic poly(A) polymerases enhances translation . This control mechanism of deadenylation/readenylation is used to control translation in oocytes, inflammation and neurons .
Recruitment to mRNA and regulation
A major mRNA recruitment mechanism for Pan2–Pan3 is via the Pan3–PABP interaction . However, it is unclear whether Pan2–Pan3 is able to interact with every PABP-bound mRNA, or whether other signals prevent or promote this interaction. Recent studies by Huang et al.  show that phosphorylation of a cluster of serine and threonine residues adjacent to the PAM2 in Pan3 may regulate its interaction with PABP. Ccr4–Not is recruited to mRNA through interactions with proteins that directly bind sequences in mRNA 3′ UTRs, e.g. Puf/Pumilio and TTP (tristetraprolin) proteins [32,34]. For instance, Puf3p promotes mRNA deadenylation by both Pan2–Pan3 and Ccr4–Not, although a direct interaction between Puf3p and Pan2–Pan3 has not been demonstrated . There are few examples of specific recruitment of Pan2–Pan3 to mRNAs.
Both Pan2–Pan3 and Ccr4–Not complexes are recruited to mRNAs via the miRISC (miRNA-induced silencing complex). Specifically Pan3, CNOT1 and PABP are reported to bind to the C-terminal silencing domain of the GW182/TNRC6 subunit of miRISC [36–38] and these interactions are highly conserved . In this context, the deadenylation complexes mediate rapid biphasic deadenylation of miRNA-targeted mRNAs and, as expected, this leads to either translational repression or mRNA decay . Interestingly, there is some evidence that GW182 proteins not only recruit deadenylation complexes, but also may stimulate their deadenylation activity, at least for Ccr4–Not . Overexpression of dominant-negative mutants of Pan2 or Caf1 slows the first and second phases of miRNA-mediated deadenylation respectively . In addition, Ccr4–Not appears to mediate translational repression in response to miRNAs, independent of its deadenylation activity: tethering a catalytically inactive version of the deadenylase subunit Caf1 to a reporter mRNA represses translation, as does tethering other Ccr4–Not components to reporter mRNAs lacking poly(A) tails . It is possible that Ccr4–Not actively inhibits translation, for example through inhibition of the activity of a translation initiation factor . Alternatively, Ccr4–Not could disrupt PABP interaction or the closed-loop conformation of mRNA to inhibit translation initiation indirectly . Further studies are required to clarify this.
Pan3p interacts with the Dun1 kinase, and dun1pan2 and dun1pan3 double mutants are hypersensitive to replication stress . Thus it has been proposed that Dun1 and Pan2–Pan3 function together to regulate Rad5 mRNA. Interestingly, stress inhibits deadenylation (by both Ccr4–Not and Pan2–Pan3), mRNA decay and translation, possibly to allow selective translation of mRNAs specific to the stress response, while retaining other mRNAs to be used upon recovery from the stress [45,46]. The mechanism of stress-induced inhibition of deadenylation is unknown.
Poly(A) tails are required to regulate mRNA stability and efficient translation. Deadenylation is a key step of mRNA degradation, possibly because release of the closed-loop mRNA structure is required to expose the mRNA 5′-cap for decapping . Studies of deadenylation of eukaryotic mRNAs suggest that transcription, mRNA processing, translation and mRNA decay are closely linked . The enzymes involved in each of these key events communicate with each other to create an integrated control mechanism for gene expression. In support of this, genomic methods to monitor mRNA synthesis and decay demonstrate that yeast compensate for a reduction in transcription due to mutants in RNA polymerase II by globally increasing mRNA stability, whereas cells compensate for deletion of Ccr4–Not subunits by decreasing mRNA synthesis . In addition, there is evidence that the Pan2–Pan3 and Ccr4–Not complexes interact in larger assemblies . Future research directed towards understanding deadenylation is required to reveal the molecular mechanisms of the individual components and their larger assemblies.
Note added in proof (received 3 December 2013)
Christie et al.  recently reported the crystal structure of the pseudokinase domain and CTD (C-terminal domain) of Pan3, revealing details of dimerization.
Translation UK 2013: An Independent Meeting held at the University of Kent, U.K., 4–5 July 2013. Organized and Edited by Emma Mead and Rosalyn Masterton (University of Kent, U.K.).
This work was supported by the European Research Council (ERC) [starting grant number 261151 (to L.A.P.)] and by the Medical Research Council [grant number U105192715].