Localization and translation of mRNAs within different subcellular domains provides an important mechanism to spatially and temporally introduce new proteins in polarized cells. Neurons make use of this localized protein synthesis during initial growth, regeneration and functional maintenance of their axons. Although the first evidence for protein synthesis in axons dates back to 1960s, improved methodologies, including the ability to isolate axons to purity, highly sensitive RNA detection methods and imaging approaches, have shed new light on the complexity of the transcriptome of the axon and how it is regulated. Moreover, these efforts are now uncovering new roles for locally synthesized proteins in neurological diseases and injury responses. In this Cell Science at a Glance article and the accompanying poster, we provide an overview of how axonal mRNA transport and translation are regulated, and discuss their emerging links to neurological disorders and neural repair.
Neurons are highly polarized cells with cytoplasmic extensions that can extend for millimeters, or even up to meters in large vertebrates. Axons and dendrites constitute the vast majority of the volume and surface area of a neuron, and neurons use localized protein synthesis in these cytoplasmic extensions to spatially and temporally regulate the protein content of these subcellular domains (Jung et al., 2014). Axons provide long-range connections between neurons and their targets that allow the brain, spinal cord and peripheral nerves to communicate. With the known transport rates of proteins, organelles and other macromolecules, the distal axon must respond to environmental stimuli well before anything could be transported there from the cell body. Localized protein synthesis is one way to overcome this distance constraint, but the neuroscience community largely overlooked the possibility of localized mRNA translation in axons until recent years (Box 1). Generating proteins locally within distal axons brings unique advantages for growth, survival and function to these far reaches of the cytoplasm of a neuron, and recent unbiased analyses point to thousands of different mRNAs in axons. In this Cell Science at a Glance article, we aim to summarize new advances in the field and point out where knowledge gaps exist. We focus on the regulation of mRNA transport and translation and the unique functions served from axonally synthesized proteins, highlighting how neuronal health is affected by these mechanisms where loss or gain of function result in disease or altered axon growth capacity. Obviously, it is not possible to cover the entire field in this short article, so we refer the reader to several recent reviews for more detailed summaries (Batista and Hengst, 2016; Costa and Willis, 2017; Kar et al., 2018; Tasdemir-Yilmaz and Segal, 2016).
Early electron microscopy (EM) analyses of rodent brain showed evidence for polysomes at the base of dendritic spines in mature hippocampus (Steward and Levy, 1982), which spurred decades of research focusing on the role of dendritically synthesized proteins in synaptic plasticity (Namjoshi and Raab-Graham, 2017). Those same early EM studies shed doubt on the possibility that axons can synthesize proteins, as no polysomes were found in the axons of the mature hippocampus (Steward and Levy, 1982). Further doubt that translation occurs in axons came from instances of mRNAs having been detected in hypothalamic and olfactory axons, yet ribosomes appeared to be lacking (Denis-Donini et al., 1998; Mohr and Richter, 1992). Thus, it was suggested that the translational machinery and mRNAs are excluded from vertebrate axons (Steward, 1997), even though biochemical evidence already pointed to the possibility of intra-axonal protein synthesis in the 1960s (Koenig, 1965a,b; Koenig, 1967a,b) and, shortly thereafter, EM evidence for ribosomes in axons of the PNS was published (Bunge, 1973; Tennyson, 1970; Yamada et al., 1971).
A series of critical studies taking advantage of the size of the squid giant axon as a tractable model to test for intra-axonal protein synthesis culminated in clear evidence for intra-axonal protein synthesis in those invertebrate neurons (Giuditta et al., 1980, 1986, 1991). Subsequent studies in vertebrates, and then mammals, showed that axons synthesize proteins even in adults (Perry and Fainzilber, 2014; Twiss and van Minnen, 2006). Fueled by recent technical and experimental advances, the field has now moved from phenomenon to understanding the functions served by axonally synthesized proteins in neuronal development, injury responses and disease states. Recent studies are starting to uncover how axonal mRNA transport and translation are regulated, including the use of in vivo systems for analyses of axonal mRNA localization and protein synthesis in mature neurons (Kalinski et al., 2015a; Shigeoka et al., 2016). Recent work has also uncovered how mRNA splice variants contribute to subcellular transcriptomes that undoubtedly include axonal mRNAs (Taliaferro et al., 2016).
How does the neuron know which mRNAs to localize into axons?
Just as RNAs are localized into dendrites and the subcellular regions of non-neuronal cells, mRNA transport into axons is driven by sequences inherent to the RNAs. These sequence motifs have most often been found in 3′ untranslated regions (UTR) of the mRNAs (Andreassi and Riccio, 2009; Gomes et al., 2014), but 5′UTR-localization motifs have also been uncovered (Merianda et al., 2013). Although mRNA protein-coding sequences (CDS) that mediate axonal targeting have not yet been found for axonal mRNAs, CDS motifs have been described in yeast (Kilchert and Spang, 2011), so it is likely that CDS motifs for axonal mRNA targeting will also be uncovered. In another commonality with dendrites and non-neuronal systems, RNA-binding proteins (RBPs) binding to these motifs are necessary for axonal mRNA transport (Khalil et al., 2018; Korsak et al., 2016). Thus, the interaction of an mRNA with RBPs is sequence dependent, but consensus RNA motifs unique to axonal mRNA targeting have yet to be found. Moreover, some axonal localization motifs can also target mRNAs into dendrites (Tiruchinapalli et al., 2003; Vuppalanchi et al., 2010), which could reflect common uses for the encoded proteins in axons and dendrites.
The secondary structures of the RNA motifs are thought to influence their interaction with RBPs (Gomes et al., 2014). The ability to bioinformatically compare secondary structures across RNA species is advancing, so common structural motifs may be discovered soon. Next-generation sequencing of RNAs from RBP immunoprecipitations also holds promise to uncover RBP-recognition motifs that are shared between axonal mRNAs, as seen for recent work with motor axon transcriptomes (Rotem et al., 2017). However, multiple RBPs can bind to the same mRNA and impart different fates to the mRNA. Interactions of mRNAs with RBPs likely begin in the nucleus, either co-transcriptionally or shortly after transcription, and the fate of an mRNA with regard to its subcellular mRNA localization is conferred by the sequential binding of multiple RBPs. Evidence for this is the interaction of β-actin mRNA (ACTB) with zip code binding protein 2 (ZBP2, also called KHSRP and FUBP2) and ZBP1 (also called IGF2BP1 and IMP1) (Pan et al., 2007).
As new knowledge of localization motifs emerges, it will need to be interpreted in the context of multiple protein interactors and different neuron types, as well as for the protein–protein interactions that occur at different sites within the soma and along the axon. For example, recent work shows that γ-actin mRNA localizes into motor axons (Moradi et al., 2017), whereas it appears to be restricted to the soma of sensory and cortical neurons (Bassell et al., 1998; Zheng et al., 2001). In addition, CREB mRNA localizes to embryonic sensory axons, but not axons of sympathetic neurons (Andreassi et al., 2010; Cox et al., 2008). Axonal mRNA populations can similarly change with growth states and as the neuron matures (Gumy et al., 2011; Shigeoka et al., 2016; Taylor et al., 2009). In-depth axonal RNA profiles of different neuronal subtypes will undoubtedly uncover more differences in axonal transcriptomes, physiological states and pathological states, and it is important to keep in mind that differential gene expression as well as combinations of RBPs could drive these.
Mechanism for the regulation of axonal mRNA transport
Just as with proteins and organelles, mRNAs are actively transported into axons by molecular motors. Because of the unified polarity of microtubules in axons, the plus-end-directed kinesin motor proteins are used for long-range anterograde transport in axons on microtubules, whereas myosin motor proteins are used for short-range transport on microfilaments (see poster) (Kalinski et al., 2015b). Dynein motor-dependent retrograde movements have been observed for some axonal RBPs, but it is not clear whether mRNAs are bound to these retrogradely moving RBPs. This could be a mechanism for relocating mRNAs within an axon, for instance in response to extracellular stimuli, or for delivering RBPs back to the cell body for reuse, as has been shown for La protein (van Niekerk et al., 2007).
mRNAs are transported in protein complexes that have been referred to as ‘RNA transport granules’, which are ribonucleoprotein complexes (RNPs) (Kar et al., 2017). Analyses of neural RNP contents show that there are many different RBPs, variable components of translational machinery and proteins that are needed for interaction with motor proteins, as well as many different mRNAs that are components of them (Elvira et al., 2006; Kanai et al., 2004; Krichevsky and Kosik, 2001). Notably, these analyses represent pooled lysates, so the RNPs are those from dendrites, axons and soma. Subsequent work with dendritically localizing RBPs points to there being multiple dendritic RNPs with different contents (Fritzsche et al., 2013; Miller et al., 2009). It is highly likely that axons also contain a diversity of different RNPs, but more sensitive methods will be needed to systematically dissect the makeup of RNPs isolated directly from axons.
Although several RBPs have been shown to localize into axons (see poster), the number is vastly smaller than the thousands of mRNAs that have been detected in axons (Kar et al., 2018). These RBPs likely interact with many different mRNAs, and are analogous to ‘RNA regulons’ (Keene, 2007), but much is still to be learned regarding the breadth and functions of axonal RBPs. Individual axonal mRNAs can also interact with several different RBPs. For example, although ZBP1 is required for axonal localization of β-actin mRNA, heterogeneous nuclear (hn)RNP R, spinal motor neuron (SMN) and HuD (also called ELAVL4) proteins also bind to axonal β-actin mRNA (Glinka et al., 2010; Kim et al., 2015; Rossoll et al., 2003; Tiruchinapalli et al., 2003). The interaction with hnRNP R and ZBP1 contributes to β-actin mRNA localization in motor axons, while ZBP1 or ZBP1 together with HuD may be sufficient for its axonal localization in other neurons (Glinka et al., 2010; Kim et al., 2015; Rossoll et al., 2002). In addition, both HuD and ZBP1 can bind to the AU-rich element (ARE) of Gap43, but it is not clear whether ZBP1 binds the RNA directly or in complex with HuD (Yoo et al., 2013). The binding of some axonal RBPs, such as nucleolin, FUS/TLS, YB-1 (also known as YBX1), Hermes (also known as CD44), TDP-43 (also known as TARDBP) and FMRP, to individual mRNAs has been characterized, but not been tested for interactions with other axonal mRNAs (Antar et al., 2006; Kar et al., 2017; Perry et al., 2016). These RBPs undoubtedly will have multiple mRNA targets among the axonal mRNAs, as was recently shown for SFPQ, which co-assembles with Bclw (also known as Bcl2l2) and lamin B2 (Lb2, also known as Lmnb2) into RNA-transport granules (Cosker et al., 2016). Other axonal RBPs, including TRF2-S (also known as TERF2), ZBP1 and HuD, have been shown to interact with multiple mRNAs that can localize to axons (Bolognani et al., 2010; Jønson et al., 2007; Zhang et al., 2015). However, these interactions can only been inferred based on cross-referencing co-immunoprecipitating mRNAs from brain or non-neuronal cell lysates to axonal transcriptomes, rather than having been validated as true interactions within axons.
Altered axonal RNA transport can impact axon growth, function and survival, and several RBPs linked to neurological diseases have been detected in axons. These include TDP-43 and FUS/TLS, which are mutated in amyotrophic lateral sclerosis (ALS), SMN, whose deficiency causes spinal muscular atrophy (SMA), and SFPQ, whose loss from axons contributes to peripheral neuropathy (Box 2). Analyses of cellular and animal models of ALS, SMA and chemotherapy-induced neuropathy have shown alterations in axonal mRNA localization (Alami et al., 2014; Pease-Raissi et al., 2017; Rotem et al., 2017; Saal et al., 2014), which leads to the intriguing possibility that alterations in axonal protein synthesis can impact disease pathogenesis or progression.
NGF regulates transcription and subsequent axonal transport of Bclw to protect axons of developing PNS neurons from degeneration (Cosker et al., 2016). In mature neurons, the chemotherapeutic agent Paclitaxel decreases the amount of axonal SFPQ–BclW RNPs (Pease-Raissi et al., 2017). Peripheral neuropathy is a significant complication for some chemotherapeutics, and this Paclitaxel-induced axon degeneration is prevented by exogenous peptides targeting the interaction Bclw with the inositol-phosphate-3 receptor (Pease-Raissi et al., 2017).
Neuronal infection with α-Herpes viruses requires retrograde transport of viral particles from distal axons to the nucleus for latent infection (Taylor and Enquist, 2015), and intra-axonal protein synthesis is needed for this retrograde transport (Koyuncu et al., 2013). Although it is not clear which axonal proteins are required, axonal translation of Kpnb1, Dctn1 and Pafahb1 can modify retrograde transport.
Motor neuron disease
Altered axonal mRNA translation is implicated in pathogenesis of motor neuron degeneration. SMA has an onset during childhood, with motor neuron death resulting from loss of SMN. SMN has a well known role in RNP assembly (Khalil et al., 2018), but its axonal localization and the decreased in axonal β-actin mRNA seen upon depletion of SMN suggests that it has other functions (Le et al., 2005; Rossoll et al., 2003; Zhang et al., 2006). Consistent with this, SMN depletion from cultured motor neurons markedly changes axonal mRNA levels, including mRNAs needed for axon growth and synaptic function (Saal et al., 2014). The discovery that mutations in the RBPs TDP-43 and Fus/TLS can be causative for ALS pointed to possible post-transcriptional dysregulation in mature motor neurons. Motor neurons expressing ALS mutants of TDP-43 showed decreased mobility of axonal RNPs and reduced axonal transport of Nefl (Alami et al., 2014). Recent RNA-seq work showed that ALS-causing TDP-43 mutations alter the axonal content of both mRNAs and miRNAs in cultured spinal motor neurons (Rotem et al., 2017).
Work from the Hengst laboratory has surprisingly shown that activation of axonally synthesized proteins can also trigger neuronal degeneration. Stimulation of hippocampal neurons with Aβ1-42 peptide, a causative agent for Alzheimer's disease, triggers translation of Atf4 mRNA in axons with subsequent retrograde transport of ATF4 causing cell death (Baleriola et al., 2014).
Regulation of mRNA dynamics within the axonal compartment
mRNAs are thought to be in a translationally suppressed state during their transport as RNPs (Wells, 2006). For example, ZPB1 undergoes phosphorylation in distal axons, thereby decreasing its RNA-binding affinity and releasing β-actin mRNA for translation (Hüttelmaier et al., 2005). Phosphorylation of ZBP1 is regulated by extracellular stimuli, as sonic hedgehog (SHH) has been shown to increase axonal β-actin mRNA translation in growth cones of the developing spinal cord (Lepelletier et al., 2017). Thus, ZBP1 contributes to both transport and translational regulation of β-actin mRNA, and this function may extend to other mRNA targets of ZBP1.
RBPs can also regulate the stability of their target mRNAs. HuD, which is needed for axonal localization of Gap43 (Yoo et al., 2013) and has been well characterized to stabilize mRNAs (Beckel-Mitchener et al., 2002; Gomes et al., 2017), can also contribute to mRNA translational regulation (Fukao et al., 2009). By contrast, KHSRP can also bind to the ARE of Gap43, but this interaction destabilizes the mRNA (Bird et al., 2013). This suggests that HuD and KHSRP can sometimes compete for binding to the same mRNAs, but with different outcomes (Gardiner et al., 2015). Interestingly, KHSRP binds to nuclear β-actin mRNA and facilitates its interaction with ZBP1 for axonal localization (Pan et al., 2007), emphasizing that RBPs can have different functions in different subcellular compartments.
mRNAs can also compete for binding to the same RBP (see poster). β-actin mRNA and Gap43 compete for binding to ZBP1 for their transport into axons (Donnelly et al., 2011; Yoo et al., 2013). Furthermore, Gap43 and Nrn1 compete for binding to HuD in sensory neurons, but not in CNS neurons that express higher levels of HuD (Gomes et al., 2017). Changes in transcription or stability, which can change the levels of either target mRNAs or RBPs, will thus obviously affect the competition between mRNA and RBPs. Taken together, these observations provide evidence for the dynamic nature and composition of axonal RNPs.
Translation of mRNAs in axons obviously requires ribosomes, tRNAs and translation factors. Stimuli that regulate axonal translation can modify the activity of this translational machinery in axons (Jung et al., 2012). A unique single-molecule imaging approach recently showed that there was an increase in axonal β-actin mRNA translation within 20 s of ligand stimulation (Ströhl et al., 2017), demonstrating that axon terminals can rapidly activate translation in response to extracellular stimuli. This also implies that prior studies that used standard imaging approaches and suggested translation within 5–20 min may have underestimated how quickly axonal translation can occur, including work from our laboratory (Pacheco and Twiss, 2012; Vuppalanchi et al., 2012).
A critical, yet unanswered, question is how translational specificity is driven at the level of individual axonal mRNAs beyond their interactions with RBPs. There is evidence that intracellular Ca2+ levels can impact the specificity for axonal mRNA translation (see poster). Increases in axoplasmic Ca2+ after severing nerve injury (i.e. axotomy) have been implicated in the translation of Kpnb1, RanBP1, Vim (encoding vimentin) and Stat3 (Ben-Yaakov et al., 2012; Hanz et al., 2003; Perlson et al., 2005; Yudin et al., 2008). Additionally, release of Ca2+ from endoplasmic reticulum (ER) stores increases translation of axonal Calr and Grp78 (also known as BiP and Hspa5), but not β-actin mRNA, through phosphorylation of eIF2α (Vuppalanchi et al., 2012). ER Ca2+ release can activate the unfolded protein response (UPR), and UPR-induced translation of Luman/CREB3 in peripheral nervous system (PNS) axons occurs after axotomy (Ying et al., 2014, 2015). Thus, axotomy-induced elevated Ca2+ can drive translation of specific axonal mRNAs.
Translation of mRNAs that are not part of this axotomy response is likely to be regulated by other means, as indicated above for β-actin mRNA. Translation of axonal mRNAs that encode components of the translational machinery brings the potential to impact subsequent translation. For instance, translation of eIF2B2 and eIF4G2 in sympathetic axons contributes to axon growth (Kar et al., 2013). Axonal levels of these two mRNAs are regulated by microRNAs (miRNAs; designated by the prefix miR), which in turn impacts translation of other axonal mRNAs (Kar et al., 2013). miR-182 has also been shown to attenuate axonal synthesis of cofilin 1, with the guidance cue Slit2 attenuating effects of miR-182 (Bellon et al., 2017). mRNAs encoding other translation factors, ribosomal proteins and RBPs feature prominently in motor, sensory and retinal ganglion cell (RGC) axon transcriptomes (Briese et al., 2016; Gumy et al., 2011; Minis et al., 2014; Zivraj et al., 2010), and some miRNAs have been found to be enriched in axons (Natera-Naranjo et al., 2010; Phay et al., 2015; Rotem et al., 2017). Effects of miRNAs, together with local generation of translational machinery, bring the potential to broadly impact translation of many axonal mRNAs.
Ribosomes and mRNAs have long been known to associate with the cytoskeleton (Bassell and Singer, 2001), which could also provide a means to control where proteins are synthesized in axons. The microtubule-binding adenomatous polyposis protein (APC) binds to subsets of localized mRNAs, including axonal mRNAs, and links these mRNAs to microtubules (Mili et al., 2008; Preitner et al., 2014; Villarin et al., 2016). This entails a mechanism to concentrate mRNAs near sites where their protein products are needed, but how this is coordinated at a molecular level is not known. Beyond the cytoskeleton, work from the Flanagan laboratory has shown that ribosome-subunit-bound mRNAs can form a complex with the cytoplasmic domain of deleted in colon cancer (DCC), a component of the netrin receptor (Tcherkezian et al., 2010), thus providing an appealing mechanism to link extracellular stimuli, such as netrin, to the translation of receptor-bound mRNA subsets. It will be of high interest to determine whether this docking mechanism extends to other cell surface receptors and which mRNAs are docked.
Axonally synthesized proteins support axon growth
Given the large distances that can separate the distal axon from its soma, the axon must be able to rapidly respond to extracellular stimuli. This autonomy is perhaps best illustrated by the use of axonally synthesized proteins for growth of developing axons. Axons sense attractant and repellant cues in their environment to navigate their way to the appropriate targets. β-actin mRNA and ribosomes have been detected in growing axons of cultured neurons, and axonal levels of β-actin mRNA increase in response to neurotrophin-3 (NT-3, also known as NTF3) (Bassell et al., 1998; Zhang et al., 1999). Attractant cues have also been shown to activate translation in axons, which is needed for growth cone turning (see poster) (Campbell and Holt, 2001). Moreover, attractant stimuli can trigger movement of mRNAs and the translational machinery to coordinate where proteins are generated within a growth cone (Leung et al., 2006; Yao et al., 2006). Repellant stimuli can also modify mRNA transport and translation in growth cones (Piper et al., 2006, 2005; Walker et al., 2012; Wu et al., 2005). The opposite outcomes with regard to growth direction, which are driven by attractant and repellant cues, are determined in part by which proteins are synthesized. For example, translation of RhoA and cofilin 1 (Cfl1) is increased by repellant cues and that of β-actin mRNA increased by attractant cues (Leung et al., 2006; Piper et al., 2006; Wu et al., 2005). Although many of these observations are from cultured neurons, there is now ample evidence that translation occurs in developing axons in vivo, and that there are specific changes in axonally synthesized protein populations during axonal pathfinding, such as changes in RhoA, β-actin and ErbA2 translation (Brittis et al., 2002; Shigeoka et al., 2016; Walker et al., 2012; Zivraj et al., 2010). A more generalized promotion of axon growth can also be driven by localized synthesis, as exemplified by translation of Par3 (also known as Pard3), TC10 (also known as Rhoq) and Gap43 in developing sensory neurons (Donnelly et al., 2013; Gracias et al., 2014; Hengst et al., 2009).
As noted above, mRNAs of the different actin isoforms show a differential localization in axons of motor neurons compared with other neuron types. β-actin synthesis in sensory and RGC axons has been shown to trigger axon branching (Donnelly et al., 2013; Wong et al., 2017). However, local synthesis of axonal γ-actin rather than β-actin in motor axons is needed for branching, and the authors hypothesized that this difference reflects the higher degree of branching that spinal motor neurons undergo (Moradi et al., 2017). For the RGC axons, this branching occurs in vivo as the axons are approaching their targets to, presumably, facilitate target innervation (Wong et al., 2017). Although these effects are limited to β-actin, it is appealing to speculate that other axonally translated proteins may impart different functions to different neuronal populations.
Axonally synthesized proteins support survival and function of axons
Soma-to-axon signaling has been shown to support survival and function of axons through the delivery of mRNAs and localized translation of new proteins. Developing sensory neurons require nerve growth factor (NGF) for survival, and NGF triggers transcription of the Bclw gene with the subsequent delivery of Bclw mRNA into axons (Cosker et al., 2013). Withdrawal of NGF or preventing axonal translation of Bclw results in axon degeneration (Cosker et al., 2013), and loss of axonal Bclw mRNA has recently been linked to neuropathy (Box 2) (Pease-Raissi et al., 2017). NGF-dependent survival of sympathetic axons has been shown to require intra-axonal translation of myo-inositol monophosphatase-1 (Impa1), and a decrease in Impa1 translation attenuates CREB signaling in these neurons (Andreassi et al., 2010). Translation of Lb2 in RGC axons supports axon survival by maintaining mitochondrial function (Yoon et al., 2012). Further data link intra-axonal translation of other nuclear-encoded mitochondrial proteins to axon function. Specifically, translation of CoxIV and Atp5g1 in sympathetic axons is needed to maintain mitochondrial ATP production in axons, which in turn is required for axon function and survival (Aschrafi et al., 2010, 2008; Kar et al., 2014). ATP is needed for translation, and translation-dependent branching of sensory axons by neurotrophins requires local mitochondrial respiration at incipient branch points where translation occurs (Spillane et al., 2013). The transcriptomes of axons for different neuronal subtypes contain many mRNAs of nuclear-encoded mitochondrial proteins (Kar et al., 2018), so the support of mitochondrial function might be a common role for axonally synthesized proteins.
Dendritically synthesized proteins have a prominent role in synaptic plasticity (Kosik, 2016), and there is emerging evidence that axonally synthesized proteins can support synaptic development and function. Changes in the translated mRNA populations were observed in developing RGC axons as they reach their targets and establish synaptic connections (Shigeoka et al., 2016). Axons of cultured cortical neurons also show shifts in translation in response to glutamate, indicating that neurotransmitters can stimulate axonal protein synthesis, as has been shown for dendritic translation (Hsu et al., 2015). Although it is not clear which proteins are generated in axons in response to glutamate, intra-axonal synthesis of synaptosomal-associated protein 25 (SNAP25), which is needed for synaptic vesicle release, is rapidly increased during formation of presynaptic terminals in hippocampal neurons (Batista et al., 2017). Translation in PNS sensory nerve endings has also been linked to the development of neuropathic pain, and cytokine signaling in distal axons contributes to this (Khoutorsky and Price, 2017). Taken together, studies clearly indicate that axonally synthesized proteins can impact presynaptic function and neural activity.
Axonally synthesized proteins provide a platform for retrograde signaling
Axonally generated proteins provide a platform for the delivery of signals to the soma (see poster). Work in several different types of neurons and under different settings has shown that translation of transcriptional modulators in axons can affect gene expression in the nucleus. NGF-dependent synthesis of CREB in axons promotes the survival of developing sensory neurons, where the retrogradely transported CREB functions as a nuclear transcription factor (Cox et al., 2008). Adult PNS neurons show translation and retrograde transport of the transcription factors Stat3α and PPARγ after axotomy (Ben-Yaakov et al., 2012; Lezana et al., 2016). The chromatin-interacting protein HMGN5 is synthesized in hippocampal axons, and the protein product is transported to the nucleus to modulate gene expression (Moretti et al., 2015). Axonal synthesis of ATF4 is stimulated by Aβ1-42 peptide and retrogradely transported ATF4 can stimulate neuronal apoptosis (Box 2) (Baleriola et al., 2014). A key unsolved issue for these retrogradely transported transcription factors and chromatin modulators is how the proteins are distinguished from those that are synthesized in and reside in the soma.
Translation in axons also provides a means to control retrograde transport. Axonal injury activates translation of importin-β1 mRNA (Kpnb1) in axons (Hanz et al., 2003). By forming a heterodimer with an anterogradely transported importin-α protein, importin-β1 protein links cargo proteins, as the transcription factors mentioned above, to the minus-end-directed microtubule motor protein dynein (Hanz and Fainzilber, 2006). This provides a route for axon-to-nuclear transport, and this axonal synthesis of importin-β1 is needed for subsequent axotomy-induced changes in gene expression that support nerve regeneration (Perry et al., 2012). Interestingly, axonal synthesis of importin-β1 also provides a cell-size-sensing mechanism during development (Perry et al., 2016), emphasizing that different functions can be imparted by the same axonally synthesized protein under different physiological conditions. It will be interesting to see if this results from a qualitative difference in transported cargo or a quantitative difference in how much importin-β1–importin-α3 is delivered under these two different settings.
Recent work also indicates that the dynein-mediated retrograde transport of vesicles can be modified by intra-axonal translation of dynein regulators. Local translation of mRNAs encoding p150Glued and Lis1 (Dctn1 and Pafah1b1, respectively) is required for NGF-induced retrograde transport of signaling endosomes, while only locally synthesized Lis1 is needed for NGF-induced retrograde transport of large vesicles (Villarin et al., 2016). Both Lis1 and p150Glued have stimulatory effects on dynein (Baumbach et al., 2017; Gutierrez et al., 2017; King and Schroer, 2000; Pandey and Smith, 2011) and are involved in the initiation of transport from microtubule plus-ends (Jha et al., 2017; Moughamian and Holzbaur, 2012). Thus, the local synthesis of dynein regulators in distal axons constitutes an ‘on demand’ system to alter retrograde transport. Deletions or mutations of the human Lis1 (PAFAH1B1) gene cause lissencephaly, a brain malformation with severe disruption of cortical development originating from altered migration of neuronal precursors (Bertipaglia et al., 2017). However, Lis1 is also needed in mature neurons, since depletion or knockout of Lis1 in adult sensory neurons disrupts retrograde axonal transport (Hines et al., 2018; Pandey and Smith, 2011), and it is intriguing to speculate that loss of axonally synthesized Lis1 may contribute to this effect in adult neurons.
As we outline above, our knowledge of the axonal transcriptome and its regulation has remarkably advanced our understanding of axonal protein synthesis. Axonally synthesized proteins contribute to growth, function and survival responses within the axon, but also provide an axon-to-soma signaling platform to coordinate neuronal responses culminating in growth, regeneration, survival, or cell death depending on the stimulus. These could have profound effects on development and function of the CNS and PNS. Although we have learned much about the axonal transcriptome and its regulation, there is still a lot to discover, and several pressing issues need to be addressed in future years. First, are there unique needs for axonally synthesized proteins in different neuronal types, as noted for axon branching supported by γ-actin in motor and β-actin in RGC and sensory neurons? Second, many axonally synthesized proteins, such as β-actin and Gap43, are also made in the soma and then transported into axons, and it is not clear which, if any, unique functions the axonally synthesized protein imparts and how. Third, RNA–protein interactions have a critical role in the delivery of mRNAs into axons and translation of mRNAs in axons, but we have only identified a few axonal RBPs and a better understanding of the dynamics of axonal RNPs is needed. Finally, the question of how specificity is achieved in the translation of some mRNAs and not others in response to stimuli needs to be addressed. Reversible RNA methylation, which has recently been implicated in regulating axonal mRNA translation and regeneration (Yu et al., 2017), could bring transcript specificity for translation. With continuingly emerging evidence for the contribution of axonal protein synthesis to neurological disease and regeneration, answering the issues above has the potential to open up new strategies for treatments.
The authors thank Amar N. Kar, Seung Joon Lee, Terika Smith and Priyanka Patel for suggestions and discussions on the topic.
Our work in this area has been supported by grants from the National Institutes of Health (R01-NS041596 and P01-NS055976 to J.L.T.; R01-NS089663 to N.P.-B. and J.L.T.; R01-NS056314 to D.S.S.), National Science Foundation (MCB-1020970), U.S. Department of Defense (W81XWH-13-1-0308), and the Dr Miriam and Sheldon G. Adelson Medical Research Foundation. J.L.T. is the SmartState Chair in Childhood Neurotherapeutics at the University of South Carolina. Deposited in PMC for release after 12 months.