This Message Will Self-Destruct: NMD Regulates Axon Guidance

Abstract

The navigation of axons to their final destination can involve a sequence of steps requiring different sets of guidance receptors. In this issue of Cell, Colak et al. show that regulated intra-axonal protein synthesis coupled to nonsense-mediated mRNA decay controls a switch in Robo3.2 expression critical for navigation.

RNA localization and local translation provide ways to direct the synthesis of proteins with spatial and temporal precision: at subcellular locations where they are needed, or in response to the timing of extracellular cues (Jung et al., 2012). While many studies have looked at mechanisms that switch on local translation, less is known about mechanisms that may subsequently halt translation and thereby limit the amount of protein produced. Jaffrey and colleagues (Colak et al., 2013) now show that nonsense-mediated mRNA decay (NMD) provides a way to limit the translation of a localized mRNA, in a mechanism that operates with temporal and spatial specificity at a specific location within the cell.

NMD was originally identified as a mechanism for quality control that eliminates aberrant transcripts with a premature stop codon. In most normal mRNAs, stop codons are located only in the last exon. Therefore, during the pioneer round of translation the ribosome reaches the stop codon only after it has displaced all the exon junction complexes (EJCs) which are deposited at exon boundaries during splicing. When mutation introduces a stop codon upstream of the last exon, at least one EJC will be left on the mRNA when the ribosome reaches the stop codon. This recruits the NMD machinery and leads to RNA degradation. Although NMD has primarily been studied as a surveillance mechanism, many normal transcripts have introns downstream of the stop codon, and roles have begun to emerge for NMD in regulating normal biological processes (Kervestin and Jacobson, 2012).

Neurons, being highly polarized cells, provide a well suited cell type to study subcellularly localized events. Indeed, axons can survive being severed from the cell body, providing an unambiguous test for events localized to the axon. Spinal commissural neurons are a particularly well studied model for axon guidance (Figure 1A)(Dickson and Zou, 2010). The axonal growth cone is initially attracted toward the midline floor plate. However, the floor plate is an intermediate target, so the axon must undergo a drastic switch in responsiveness after midline crossing: it is now repelled by the floor plate, and gains responsiveness to cues that guide the next step of its journey toward the brain. This switch in responsiveness involves a profound change in the set of axonal guidance receptors expressed. One mechanism for this switch is provided by the observation of mRNA sequence-specific upregulation of translation within commissural growth cones upon crossing the floor plate (Brittis et al., 2002).

Figure 1. NMD in commissural axon guidance.

(A) Guidance of commissural axons across the midline floor plate involves a switch from attraction to repulsion. Robo3.2 is upregulated upon crossing. (B) Robo3.2 is a repellent receptor for the midline cue Slit. In NMD-deficient Upf2 knockout neurons, Robo3.2 level is increased, and axon trajectories are further from the midline. (C) Model for midline activation of mRNA translation coupled to NMD. The NMD pathway inhibits further translation and promotes mRNA degradation.

Colak et al. focused on the receptor Robo3 in commissural axons (Figure 1A, B). Previous work by Tessier-Lavigne and colleagues showed that Robo3 exists in two alternatively spliced isoforms, Robo3.1 and Robo3.2, which act sequentially and with opposing roles (Chen et al., 2008). Robo3.1 is on pre-crossing axons and is essential for them to reach the floor plate, blocking the activity of other Robos which act as receptors for the midline repellent Slit. When axons cross the midline, they abruptly upregulate the Robo3.2 isoform, a receptor that contributes to repulsion from the midline. The switch from Robo3.1 to Robo3.2 does not seem to result from regulation of alternative splicing, since both mRNAs coexist in the cell throughout the process. However, the Robo3.2 mRNA differs from Robo3.1 by retention of an intron that contains a stop codon, making it a predicted NMD target (Black and Zipursky, 2008) and giving it extra 3'UTR sequences.

The study by Colak et al. can be divided in two parts. The first provides compelling evidence for local axonal synthesis of Robo3.2 (Figure 1C). The evidence for this comes from experiments on intact or severed axons from spinal cord explants. Robo3.2 mRNA was shown to be present in axons, and was translated locally in response to a signal provided by floor plate explants or conditioned medium. In contrast, the Robo3.1 mRNA was not detected in axons, suggesting that Robo3.1 protein is synthesized in the cell body. In addition to demonstrating regulated axonal synthesis of Robo3.2, it is interesting to note that these results suggest that the short retained Robo3.2 intron contains a localization element for transport into axons.

The second part of the study reveals a novel axonal role for NMD. The upregulation of Robo3.2 translation by floor plate signals was accompanied by mRNA downregulation, a hallmark of NMD. Further experiments used conditional knockout mice with the key NMD protein Upf2 selectively removed from spinal commissural neurons. Explanted neurons or severed axons from these mice showed increased Robo3.2 protein upregulation in response to floor plate signals, as well as increased mRNA levels. Moreover, in vivo axon tracing revealed abnormal post-crossing axon guidance, including trajectories located further than normal from the midline, consistent with increased repulsion from the floor plate (Figure 1B). This fits well with previous studies showing that the distance of the trajectory from the floor plate is determined by Robo levels (Reeber et al., 2008; Jaworski et al., 2010). Together, these results provide strong evidence for a role of NMD in intra-axonal Robo3.2 regulation and axon guidance.

What could be potential advantages of this novel mechanism involving coupling of local axonal translation with NMD? It is often suggested that a key advantage of local protein synthesis is amplification: allowing a single transported mRNA to produce many protein molecules. However, this notion is challenged by the present study, since NMD limits the amount of protein produced, potentially to a single round of translation per mRNA. Rather, the effect of NMD here can be conceptualized like the self-destructing tape in a spy movie: a pre-programmed mechanism that destroys the message as soon as the tape reaches its end. More formally, in this model translation is inhibited by a feedback loop triggered by the process of translation itself (Figure 1C, inset).

How, then, might such a mechanism be useful in the context of midline axon guidance? A possible advantage of a rapid negative feedback loop immediately following translational activation is that it could potentially produce a sharp and transient spike in protein synthesis, leading to a rapid step-like increase in protein concentration. A sharp increase in repellent receptor level might help lock commissural axons on the post-crossing side and prevent them from re-crossing. Additionally, if the post-crossing protein concentration steps up to a stable plateau, rather than continuing to rise due to ongoing synthesis, this might help each axon to stably select a consistent trajectory, potentially pre-specified by the amount of mRNA loaded from the cell body.

In addition to commissural neurons, NMD components were observed by Colak et al. in growth cones of other neuron types, suggesting wider roles in axon guidance. Moreover, NMD was previously shown to regulate levels of the synaptic protein Arc (Giorgi et al., 2007), suggesting that NMD might be localized to individual synapses; an idea that seems all the more plausible in view of the evidence from Colak et al. for localized NMD. Beside the novel insights this study provides, it also raises interesting questions. What is the molecular identity of the floor plate signals that regulate translation? What intra-axonal translation regulatory machinery is involved? Does the NMD step follow inevitably from translation activation, or could it be regulated independently? Finding all the answers may not be easy, but hopefully it will not be mission: impossible.