This study has shown that cytoplasmic polyadenylation is required for growth cone collapse in response to Sema3A, suggesting that cytoplasmic polyadenylation may regulate guidance-cue-induced local translation. The mRNA of a well-known regulator of cytoplasmic polyadenylation, CPEB1, is present at low levels in RGCs, but CPEB1 protein is not detected in the retina, and knockdown of CPEB1 function does not obviously affect retinal axon guidance. However, UV cross-linking revealed the presence of other CPE-binding proteins in the retina, and dominant-negative inhibition of CPE-binding in RGCs causes axon outgrowth defects. These results indicate that cytoplasmic polyadenylation and CPE-mediated regulation of mRNAs are important for RGC axon development.
Cordycepin is a specific inhibitor for mRNA polyadenylation. Although it inhibits RNA synthesis , Sema3A-induced growth cone collapse is known to be transcription-independent . Cordycepin can also inhibit deadenylation; an unidentified 3'-5' deadenylase in HeLa cell extracts cannot act on mRNAs capped at the 3' end by a cordycepin residue . However, this is not an entirely off-target effect, as it still addresses cytoplasmic control of poly(A) tail length, and cordycepin could only affect this deadenylase if it is added to the poly(A) tail by polyadenylation. The adenosine control argues against action by cordycepin on adenosine receptors and rules out the possible inhibition of adenylate cyclase by cordycepin, because adenosine and cordycepin inhibit adenylate cyclase activity equally . Finally, our finding that cordycepin slightly reduces but does not abolish the Sema3A-induced increase in translation suggests that cordycepin leaves Sema3A signal transduction pathways relatively intact. Consistent with this, cordycepin does not inhibit Xenopus oocyte maturation induced by injection of c-Mos, the synthesis of which occurs early in maturation , implying that cordycepin does not affect kinases downstream of c-Mos in oocyte maturation, such as mitogen-activated protein kinase (MAPK) and p90Rsk .
We have attempted to detect Sema3A-induced polyadenylation of candidate axonal mRNAs, such as RhoA , which contains a CPE in the 3'UTR in Xenopus and is polyadenylated in maturing Xenopus oocytes (data not shown), but have been hampered by the extremely low amounts of RNA obtainable from isolated Xenopus retinal axons, which are not adequate to obtain reliable results with existing PCR-based poly(A) tail assays [67, 68]. In addition, it is difficult to say a priori which mRNAs might be polyadenylated, as only a handful of mRNAs are known to undergo guidance cue-induced translation in axons . Nevertheless, our cordycepin results imply that axonal mRNAs are polyadenylated in response to Sema3A; future studies using more sensitive, unbiased assays may identify these mRNAs.
The contrast between clear, though weak, detection of CPEB1 mRNA in RGCs and the inability to detect CPEB1 protein in the retina suggests that CPEB1 protein might be present at very low levels. Indeed, in one study, the CPEB1 remaining in oocytes after maturation-associated degradation was undetectable under normal loading and exposure conditions, even though CPEB1 levels in mature oocytes have been variously reported as 25% or 3–5% of CPEB1 levels in immature oocytes [26, 69]. Given that CPEB1 was detected in oocytes in this study, we infer from the amount of CPEB1 in oocytes  and the amounts of protein loaded in our western blots (see Materials and methods) that the inability to detect CPEB1 in retinas suggests that the relative abundance of CPEB1 protein is about 500 times lower in the retina than in oocytes. This is not implausible, as CPEB1 mRNA levels are also much higher in oocytes than in the embryonic retina (data not shown), and in cDNA libraries collected from Xenopus tropicalis, CPEB1 is an abundant clone in eggs (0.045% of all clones, compared to 0.023% for ATP synthase, γ-subunit) whereas it is not detected in cDNA libraries between gastrulation and stage 45 .
Thus, the inability to detect CPEB1 protein leaves open the possibility that RGCs contain a small amount of CPEB1 protein. However, radiolabeled CPE-containing RNA was bound by proteins that could not be detected by anti-CPEB1 western blot or immunoprecipitation, suggesting that even if a small amount of CPEB1 is present, its function in regulating CPE-containing mRNAs may be taken over by other proteins whose identities are unknown. The other members of the CPEB family, CPEB2–4, are expressed in embryonic eyes (Figure 2F), and although they are not fully cloned in Xenopus laevis, human CPEB2 has a predicted molecular weight of 62 kDa, suggesting that the approximately 60 kDa CPE-binding band could be CPEB2. CPEB2–4 have been reported not to bind to one copy of CPE sequence (UUUUAAU) , but this may not be true for all mRNAs if the CPE is located in a loop structure. Future work on Xenopus CPEB2 awaits cloning of the gene and generation of an anti-XCPEB2 antibody. In addition, KSRP (K-homology splicing regulatory protein; known as VgRBP71 in Xenopus and FUBP2, MARTA1, or ZBP2 in other species) binds CPE sequences in mice (Y-SH and JDR, unpublished observations) and regulates the localization of β-actin mRNA in neurons . Although VgRBP71 is 71 kDa , its rat homolog MARTA1 (molecular weight 74 kDa) was originally identified as a 90 kDa protein binding to the 3'UTR of microtubule-associated protein 2 mRNA in UV cross-linking assays [73, 74], suggesting that the approximately 95 kDa CPE-binding band in Figure 4 could be VgRBP71. Future studies may identify the CPE-binding proteins in the Xenopus retina.
The effect of CPEB1-AA suggests that these non-CPEB1 CPE-binding proteins, or at least regulation of CPE-containing mRNAs, are important for RGC axon development. This regulation may occur via mRNA localization, translational repression, translational activation, or all three, as occurs with CPEB1 [18, 30]. CPEB1-AA would displace native CPE-binding proteins, thereby causing mis-regulation of CPE-containing mRNAs (Figure 5A–C). Indeed, expression of dominant-negative CPEB1 in Purkinje cells causes defects in cerebellar long-term depression and motor learning, while elimination of endogenous CPEB1 does not [33, 35], suggesting that non-CPEB1 CPE-binding proteins are also involved in synaptic plasticity.
Our finding that CPE-mediated mRNA regulation is important for axon outgrowth is consistent with other studies demonstrating roles for post-transcriptional regulation in axon formation and extension. For example, regulation of neurofilament-M mRNA by heterogeneous nuclear ribonucleoprotein (hnRNP) K is required for axon outgrowth in Xenopus, although hnRNP K is unlikely to be a CPE-binding protein, as it binds to poly(C) sequences . In addition, translational regulation of the neuronal polarity regulator SAD kinase (also known as BR serine/threonine kinase 2 or brain-selective kinase 2) by the mammalian target of rapamycin (mTOR) pathway controls axon formation . It is likely that coordinated regulation of many mRNAs by multiple RNA-binding proteins is required for the complex program of axon extension.
Given that axon extension and growth cone collapse are in some ways opposite phenomena, the effect of dominant-negative CPEB1 on axon extension seems opposed to the requirement for cytoplasmic polyadenylation in growth cone collapse. These can be reconciled by noting that the CPE-binding proteins displaced by CPEB1-AA may not necessarily regulate cytoplasmic polyadenylation, or may regulate the polyadenylation of only a subset of mRNAs that are polyadenylated upon Sema3A stimulation. It would be interesting to directly test the connection between cytoplasmic polyadenylation and the retinal CPE-binding proteins by asking whether cordycepin inhibits axon outgrowth as CPEB1-AA does, or if CPEB1-AA inhibits growth cone collapse as cordycepin does. However, the former experiment would be difficult to interpret given the inhibition of transcription by cordycepin over the timescales required to study neurite outgrowth, while the latter experiment is precluded by the lack of CPEB1-AA-positive axons growing out of transfected retinal explants.
Even if CPE-binding proteins do indeed regulate cytoplasmic polyadenylation, the apparent contradiction described above can be resolved by noting that axon extension and Sema3A-induced collapse occur at different time points of RGC axon development; the effect of CPEB1-AA on axon outgrowth is observed early and most likely includes an effect on neurite initiation (Figure 5D), whereas Sema3A is more effective at collapsing old growth cones (cultured after stage 35/36) than young growth cones . Thus, CPE-mediated mRNA regulation and cytoplasmic polyadenylation may have different roles at different developmental stages. Alternatively, just as protein synthesis is required for both attractive and repulsive responses, CPE-mediated mRNA regulation and cytoplasmic polyadenylation may be involved in both attractive and repulsive responses; future work may examine this possibility.
The conclusion that non-CPEB1 CPE-binding proteins, which may or may not regulate cytoplasmic polyadenylation, are involved in RGC axon outgrowth leaves open the question of how cytoplasmic polyadenylation is regulated. It is not necessarily surprising that different mechanisms would regulate cytoplasmic polyadenylation in oocytes and embryos. For example, even though maternal mRNAs are silenced in immature oocytes from stage I to stage VI, PARN is not expressed until stage III [78, 79], suggesting that other mechanisms not involving PARN must deadenylate and silence maternal mRNAs in early immature oocytes. In addition, in early Drosophila embryos, regulated translation of germ plasm mRNAs is correlated with their poly(A) tail length, but seems to be independent of the Drosophila CPEB homolog ORB . Similarly, in Xenopus early embryogenesis, cytoplasmic polyadenylation of mRNAs such as activin receptor is mediated by U-rich sequences (U12–27) similar to, but distinct from, the CPE bound by CPEB1 (U4-6A1-2U) during oocyte maturation [81–83]. These U-rich sequences are bound by ElrA (elav-like ribonucleoprotein A) [84, 85], suggesting that ElrA mediates cytoplasmic polyadenylation, although this has not been directly demonstrated. In addition, although ElrA is unlikely to be one of the CPE-binding proteins in Figure 4, as its molecular weight is 36 kDa and it does not bind the cyclin B1 3'UTR, it can bind to the CPE bound by CPEB1 (U4-6A1-2U) in some mRNAs such as cyclin E1 . ElrA is expressed in Xenopus throughout development , making it a potential regulator of some CPE-containing mRNAs and cytoplasmic polyadenylation in the retina.
In addition to ElrA, a role in regulation of the poly(A) tail length of target mRNAs has been described for other proteins, Musashi  and Pumilio , as well as the micro-RNA let-7 . Although Musashi (also known as nervous system-specific RNP protein, or Nrp-1) is not expressed in Xenopus differentiated RGCs , we have detected Pumilio and miRNAs in RGCs (F van Horck and M-L Baudet, unpublished observations). Pumilio and let-7 repress target mRNAs by stimulating deadenylation, as CPEB1 does in immature oocytes. If these or other factors repress and deadenylate mRNAs in unstimulated growth cones, Sema3A stimulation might cause them to release their target mRNAs, allowing them to be polyadenylated by default, which would explain why cordycepin prevents Sema3A-induced collapse. Future studies may determine whether these RNA-binding proteins, micro-RNAs, CPE-binding proteins, or other mechanisms regulate cytoplasmic polyadenylation in RGC axons, aided by the identification and 3'UTR sequence analysis of mRNAs that are polyadenylated upon guidance cue stimulation.