Neuron-wide RNA transport combines with netrin-mediated local translation to spatially regulate the synaptic proteome

We first asked whether the most abundant RNA, ribosomal RNA (rRNA) underwent directed targeting from neuronal cell bodies to synaptic sites during synapse formation. Towards this end, we performed fluorescent in situ hybridization (FISH) for 18S and 28S rRNA in cultures (3DIV) containing a bifurcated SN contacting a target L7 MN, with which it formed glutamatergic synapses, and a non-target L11 MN, with which it fasciculated but did not form chemical synapses (Figure 1A and Figure 1—figure supplement 1A). Quantification of 18S (Figure 1B,C) and 28S (Figure 1—figure supplement 1B,C) RNA in the proximal SN branches did not reveal any significant difference in concentration in the branch contacting the L7 target MN and the branch contacting the L11 non-target MN. We also examined the distribution of ribosomes by expressing the Aplysia ribosomal protein S6 tagged with the fluorescent tag dendra2 in bifurcated SNs contacting both target L7 and non-target L11 MNs (Figure 1D). This allowed us to detect the ribosomal protein S6 in distal SNs without any signal from the MN (in the rRNA FISH experiments, we focused on proximal SN processes in order to avoid confounding signal from the MNs). As shown in Figure 1E,F, S6-dendra2 targeted equally to distal SN branches contacting target or non-target MNs. Together, these data indicate that ribosomes are delivered throughout the SN and are not targeted specifically from the soma to branches receiving synaptogenic signals.

Figure 1 with 1 supplement see all

Download asset Open asset

As a candidate localized mRNA, we first examined the transcript encoding the SN-specific neuropeptide sensorin, which not only localizes to distal sensory neurites but also concentrates at synapses. Since sensorin is only expressed in SNs, we could visualize sensorin mRNA in proximal and distal neurites without any background signal from MNs. We have previously shown that the neuritic and synaptic localization of sensorin mRNA are mediated by distinct signals (Meer et al., 2012; Wang et al., 2009) and that sensorin mRNA undergoes localized translation at synapses during 5HT-mediated LTF (Wang et al. 2009). FISH for sensorin mRNA in 3DIV cultures revealed that the sensorin transcript localized equally to both proximal (Figure 2D) and distal (Figure 2B,C) branches of bifurcated sensory neurons contacting L7 or L11 MNs. The pattern of sensorin mRNA in distal branches was distinct, with diffuse distribution of sensorin mRNA in branches contacting non-target MNs and punctate concentrations of sensorin RNA at SN synapses onto target MNs (see Figure 2—figure supplement 1A for quantification of coefficient of variation). However, the same amount of RNA was present in both SN branches. Similar results were observed when we performed FISH analysis on 1DIV cultures (Figure 2—figure supplement 1B,D). These findings are consistent with a lack of directed targeting of sensorin mRNA from soma to branches receiving synaptogenic signals, although there is local redistribution of sensorin mRNA within the neuronal process at synaptic sites.

Figure 2 with 2 supplements see all

Download asset Open asset

We also examined the distribution of the mRNA encoding β-thymosin, which we had previously identified as a localized mRNA in Aplysia SNs (Moccia et al., 2003). Since β-thymosin is expressed in both SN and MN, we examined its localization in proximal SN processes. As shown in Figure 2—figure supplement 2, β-thymosin mRNA was distributed equally well in SN branches contacting L7 target and L11 non-target MNs.

To assay the localization of an activity-dependent transcript, we performed in situ hybridization for EF1α. EF1α was expressed at very low levels in unstimulated neurons and was distributed equally well between L7 target and L11 non-target branches (Figure 3A,B). Its expression was significantly induced following five spaced bath applications of (5HT), as previously described (Giustetto et al., 2003). As shown in Figure 3C,D, we found that equal amounts of 5HT-induced EF1α were delivered to SN branches contacting L7 target and L11 non-target MNs. These results indicate that activity-induced transcripts are not delivered preferentially from the nucleus towards a synaptic target.

Figure 3

Download asset Open asset

We then asked whether local stimulation might regulate the transport of EF1α from the soma towards stimulated synapses. Towards this end, we cultured a bifurcated SN with two target LFS MNs and used local perfusion to deliver five applications of 5HT to the connections made onto one of the MNs, which we have previously shown produces branch-specific LTF (Martin et al., 1997). In control experiments, we locally perfused five applications of artificial seawater (ASW), the vehicle. As shown in Figure 3, FISH for EF1α revealed that local stimulation with 5x5HT-induced expression of EF1α, as evidenced in increase in EF1α signal in the soma (Figure 3G). Strikingly, the stimulated transcript was delivered equally well to both branches (Figure 3F,H). These studies reveal that even following local stimulation, transcriptionally induced mRNAs are delivered throughout the neuronal arbor, to stimulated- and unstimulated-branches.

We next studied the localization of the RNA binding protein Staufen in SNs contacting target and non-target MNs by expressing Aplysia Staufen (ApStaufen) tagged with the fluorescent protein dendra2 in bifurcated sensory neurons. We focused on Staufen because of its previously described role in transporting transcripts into dendrites of mammalian cells (Kiebler et al., 1999; Tang et al., 2001). As shown in Figure 4, we found that ApStaufen-dendra2 was transported equally well to branches contacting target and non-target MNs.

Figure 4 with 1 supplement see all

Download asset Open asset

We also examined the distribution of ApStaufen-dendra2 in SN paired with two target MNs following local stimulation with 5X5HT to induce branch-specific LTF. As shown in Figure 4—figure supplement 1, equal numbers of ApStaufen-dendra2 puncta were present in both SN branches in prior to stimulation and following local perfusion with vehicle (ASW). While local stimulation with 5X5HT increased the number of ApStaufen-dendra2 puncta in the SN processes (suggesting that local stimulation promotes the transport of RNAs out of the soma into the process), the increase was equivalent in stimulated- and unstimulated-SN branches.

Together, our experiments using FISH to analyze rRNAs and sensorin and EF1α mRNAs, and overexpression to analyze S6 ribosomal protein and Staufen, argue that ribosomes, mRNAs, and RNA binding proteins are transported throughout the neuron, without any preferential targeting from the soma to branches receiving synaptic signals. We have previously reported that local translation is spatially restricted to stimulated synapses during synapse-specific long-term facilitation (LTF) of SN-MN synapses (Wang et al 2009); here, we asked whether local translation is also spatially restricted during synapse formation. To do this, we performed immunocytochemistry with anti-sensorin antibodies in sensory-motor cocultures (3DIV, Figure 5). Again, we chose to focus on sensorin because it is only expressed in the SN and thus the signal derives exclusively from the SN. As shown in Figure 5B,C, sensorin protein was present at much higher concentrations in SN branches contacting target MNs than in branches contacting non-target MNs. These data are consistent with sensorin expression being spatially regulated at the level of translational regulation rather than mRNA targeting.

Figure 5 with 1 supplement see all

Download asset Open asset

We previously reported live imaging experiments using a translational reporter consisting of the 5′ and 3′ UTRs of sensorin fused to the photoconvertible fluorescent protein dendra2 to demonstrate synapse-specific translation during synapse-specific LTF of SN-MN synapses (Wang et al., 2009). We now conducted live imaging experiments with this reporter to directly measure local translation (as opposed to somatic translation followed by protein transport into the neurite during synapse formation) (Figure 5D,F). To do this, we expressed the reporter in bifurcated SNs contacting L7 and L11 MNs and removed the SN cell body so that we could specifically monitor translation in neuronal processes. We photoconverted dendra2 in the branches from green to red and used live imaging to detect the appearance of newly translated, green dendra2 signal, which we have previously demonstrated represents new translation (Wang et al., 2009). As shown in Figure 5E,G, we detected a significantly greater increase in green signal in SN processes contacting the L7 target MN than in the SN processes contacting the L11 non-target MN. Together, these experiments indicate that although sensorin mRNA is delivered throughout the SN, it is preferentially translated at sites of synaptic contact.

To determine whether this effect was specific to sensorin mRNA, or whether it applied more generally to global translation, we analyzed the localization of the translation eukaryotic initiation factor4E (eIF4E) and the 4E binding protein (4EBP) in bifurcated sensory neurons contacting L7 and L11 motor neurons (for demonstration of antibody specificity, see Figure 5—figure supplement 1J). As shown in Figure 5B,C and Figure 5—figure supplement 1A,I, both eIF4E and 4EBP proteins distributed equally in SN branches contacting target and non-target MNs. However, when we used phospho-specific antibodies that recognize the activated forms of eIF4E (peIF4E) and 4EBP (p4EBP), we detected significantly higher concentrations in SN branches making synaptic contact with L7 MNs. These findings indicate that translational machinery is distributed throughout the neuron but is preferentially activated at sites receiving a synaptogenic signal. Since phosphorylation of eIF4E and 4EBP activates global translation, our results indicate that translational regulation, rather than mRNA targeting, controls the locally regulated synaptic proteome.

As an additional way of measuring translation, we used the recently developed ribopuromycylation method in which immunofluorescence is used to measure puromycin immobilized on ribosomes by the elongation inhibitor emetine. This metabolic labeling method allows detection of newly translated proteins. As shown in Figure 6, these experiments revealed significantly more protein synthesis in SN varicosities contacting the synaptic L7 target MN than in SN contacts with the non-target L11 MN. These results provide further evidence that synapse formation does not alter RNA targeting within the neuron but rather spatially regulates translation.

Figure 6

Download asset Open asset

Our previous study using translational reporters to visualize local translation of sensorin mRNA during LTF of Aplysia SN-MN synapses (Wang et al., 2009) revealed that serotonin-regulated translation specifically at stimulated synapses and that this regulation required a calcium-dependent trans-synaptic signal from the MN to the SN (Wang et al., 2009). Our current finding that sensorin protein was enriched at synaptic sites as compared to non-synaptic sites (Figure 5B,C) is also consistent with a role for a synaptically localized translational regulation mechanism. These findings suggested that a synaptically generated signal functioned to regulate translation.

We considered the possibility that the guidance factor netrin-1 might serve as such a signal. Netrin-1 is a chemotropic factor known to induce synaptogenesis in Caenorhabditis elegans (Colón-Ramos et al., 2007) and to stimulate translation in growth cones during axon guidance (Campbell and Holt, 2001). Flanagan and colleagues have reported that components of the translational machinery, including ribosomal subunits and translation factors, are tethered at the plasma membrane in neuronal dendrites by binding to the cytoplasmic tail of the netrin receptor Deleted in Colorectal Cancer (DCC), (Tcherkezian et al., 2010). Netrin-1 binding was shown to trigger the release of bound translational components and to thereby promote localized, netrin-1- dependent protein synthesis. Together, these findings suggested that release of netrin-1 from either SN or MN during synapse formation and during synaptic plasticity triggers translation at sites of synaptic but not non-synaptic contacts.

We tested the possibility that netrin-1 regulates the synthesis of localized transcripts by incubating Aplysia sensory-motor neuronal cultures with recombinant Fc-netrin-1. As shown in Figure 7A,C, bath application of netrin-1 (250 ng/ml) for 24 hr increased sensorin immunoreactivity in SN branches contacting the L11 non-target MN, such that there was no longer any branch-specificity of sensorin protein concentration.

Figure 7 with 1 supplement see all

Download asset Open asset

To further examine the effect of netrin-1 on sensorin expression, we cultured SNs with target MNs in the presence or absence of netrin-1 and compared sensorin immunoreactivity. As shown in Figure 7D,E, incubation with netrin-1 significantly increased sensorin immunoreactivity. To differentiate between stimulated local translation and stimulated somatic translation followed by transport into the branches, we cultured bifurcated SN-MN, removed the SN cell body, incubated with netrin-1, and processed the cultures 24 hr later for sensorin immunoreactivity (Figure 7—figure supplement 1A,C). These experiments revealed that netrin-1 increased sensorin concentration equally well in branches contacting both L7 and L11 MNs, consistent with local stimulation of translation. Netrin-1 also increased peIF4E immunoreactivity in SN-LFS cultures in which the SN soma had been removed (Figure 7—figure supplement 1C). The netrin-1-induced increase in sensorin immunoreactivity in intact SN-LFS cocultures was similar to that observed with five spaced applications of serotonin (5HT), which produces LTF of SN-MN synapses and which has been shown to promote sensorin translation (Figure 7E) (Hu et al., 2006; Wang et al., 2009). Moreover, netrin-1-induced translation in intact or soma-lacking SNs was blocked by pre-incubation with the protein synthesis inhibitor anisomycin (10 μM for 24 hr), indicating that the increased immunoreactivity did indeed result from increased translation (Figure 7E and Figure 7—figure supplement 1C).

We considered the possibility that netrin-1 might transform non-synaptic sites to synapses and that the increase in translation occurred as a result of the conversion of non-synaptic to synaptic sites. However, incubation of SN-L11 cocultures or L7-SN-L11 cocultures with netrin-1 did not result in the formation of synapse as shown by the absence of an excitatory post-synaptic potential (EPSP) in the L11 MN after SN stimulation (data not shown). We did find, however, that incubation with netrin-1 significantly increased EPSP amplitude between SN and LFS target MNs (Figure 7F,G). The increase in EPSP amplitude was similar to that seen during LTF induced by five spaced applications of 5HT (data not shown). LTF induced by five applications of 5HT has also been shown to involve significant increases in the number of SN varicosities (sites of synaptic connections) in SN-MN cultures (Glanzman, et al., 1990). We thus measured the effect of netrin-1 on varicosity numbers, and, as shown in Figure 7H, found that it significantly increased the number of SN synaptic varicosities in SN-LFS MN cultures. Together, these findings suggest that the local translation induced by netrin-1 increases synaptic strength and synapse number.

To explore the mechanisms by which netrin-1 regulates translation, we first asked whether it did so by binding to DCC. To test this, we incubated cultured neurons with recombinant netrin-1 in the presence of monoclonal anti-DCC antibodies that have previously been shown to block netrin-1 binding to the DCC receptor (Bennett et al., 1997; Braisted et al., 2000; Manitt et al., 2009). We first showed that a 24 hr incubation of isolated SNs with netrin-1 triggered a significant increase in sensorin immunoreactivity, but that incubation of isolated SNs with anti-DCC (in the absence of netrin-1) did not change sensorin immunoreactivity (Figure 8A,B). While anti-DCC antibodies had no effect on the basal concentration of sensorin in isolated SNs, they significantly decreased basal sensorin immunoreactivity in SN-MN cocultures (Figure 8C,D). Taken together, these findings are consistent with netrin-1 being endogenously released from the post-synaptic MN and binding to DCC on the SN to promote pre-synaptic translation of sensorin.

Figure 8

Download asset Open asset

We next explored the function of netrin-1/DCC signaling during synapse formation. Towards this end, we cultured a SN for 24 hr, then added a MN in the presence or absence of vehicle (ASW), control IgG (non-immune mouse IgG), or function-blocking anti-DCC antibodies, and measured EPSP amplitude 24 hr later. As shown in Figure 8H, EPSP amplitude was significantly lower in cultures incubated with anti-DCC antibodies than in control cultures. This was accompanied by a decrease in sensorin immunoreactivity in the cultures incubated with anti-DCC antibodies (Figure 8C,D). To determine whether netrin-1/DCC also regulated synapse maintenance and/or stability, we cultured SN-MN neurons for 2 days, added anti-DCC antibodies, and then measured EPSP amplitude 24 hr later. As shown in Figure 8E, this also triggered a significant decrease in EPSP amplitude (Figure 8H) and sensorin immunoreactivity (Figure 8D), consistent with netrin-1/DCC signaling contributing to synapse stability/maintenance.

To further examine netrin-1 and DCC expression in Aplysia SN-MN cocultures, we cloned the Aplysia homologs (submitted to GenBank). Aplysia Netrin-1 (ApNetrin-1, accession #KM218335) was 47% identical and 60% conserved with human Netrin-1, and Aplysia DCC (ApDCC, accession # KM218336) was 31% identical and 47% conserved with human DCC. To assay the effect of ApNetrin-1 on synaptic strength, we overexpressed c-terminally tagged dendra2-ApNetrin-1 in MNs (with overexpression of dendra2 as a negative control) and determined the effect on sensorin immunoreactivity (as a proxy for translation) and SN-MN strength. As shown in Figure 9, overexpression of dendra2-tagged ApNetrin-1 in MNs generated a punctate staining pattern throughout the neuronal process. Overexpression of ApNetrin-1 in the MN significantly increased the concentration of sensorin protein in the SN as measured by immunocytochemistry (Figure 9B,C), consistent with a netrin-1-dependent stimulation of translation. Measurement of EPSP amplitude before and 24 hr after overexpression of dendra2-ApNetrin-1, or overexpression of dendra2 as a negative control, revealed that ApNetrin-1 significantly increased synaptic strength between SN and MN (Figure 9E).

Figure 9

Download asset Open asset

Since our previous study had indicated that local translation required a calcium-dependent retrograde signal from the MN (Wang et al., 2009), we next asked whether ApNetrin-1 release from the MN was a calcium-dependent process. To do this, we expressed ApNetrin-1 tagged with dendra2 in the MN, and labeled the MN with Alexa fluor 647 and the SN with ApDCC tagged with mCherry. As shown in Figure 10, despite being expressed in the MN, the majority of the ApNetrin-1-dendra2 signal was found outside of the MN, decorating the processes of the SN, suggesting that ApNetrin-1 was released from the MN and bound to the SN. We note that overexpression of ApDCC-mCherry in the SN increased the ApNetrin-1 signal that decorated the SN (as compared to the signal that was observed in SNs labeled with red Alexa647, as in Figure 9), and thus allowed us to clearly detect binding of green ApNetrin-1 expressed in the MN to the red SN. To determine whether ApNetrin-1 bound to DCC on the SN, we performed these experiments in cultures containing anti-DCC antibodies in the medium (or non-immune IgG as a negative control). As shown in Figure 10—figure supplement 1, incubation with anti-DCC antibodies dramatically decreased the binding of ApNetrin-1 to the SN, as assayed using Pearson's Correlation to monitor the colocalization of green ApNetrin-1-dendra2 and red DCC-mCherry.

Figure 10 with 1 supplement see all

Download asset Open asset

To test whether release from the MN required calcium, we microinjected 50 mM BAPTA into the MN and 3 hr later imaged green ApNetrin-1-dendra2 signal localization. As shown in Figure 10B,D, when calcium was chelated in the MN, green ApNetrin-1-dendra2 signal was sequestered in the MN and did not colocalize with the SN. In contrast, in the absence of BAPTA, the green ApNetrin-1-dendra2 signal decorated the processes of the SN. Together, these findings are consistent with calcium-dependent release of ApNetrin-1 from the MN and binding to the SN.

Studies of mRNA localization in non-neuronal asymmetric cells have revealed that a large number of transcripts localize to specific subcellular compartments. For example, high throughput FISH analyses of over 3000 transcripts in Drosophila embryos indicated that over 70% of mRNAs were expressed in an array of highly specific subcellular patterns (Lécuyer et al., 2007). These findings suggest that targeted localization of mRNA spatially regulates gene expression to essentially define the fate of discrete subcellular compartments. We have previously identified hundreds of mRNAs in the neurites of Aplysia SNs (Moccia et al., 2003), and studies of localized mRNAs in mammalian hippocampal neurons have indicated that hundreds (Eberwine et al., 2002; Poon et al., 2006) to thousands of transcripts (Cajigas et al., 2012) localize to dendrites. While some studies have indicated that a handful of these mRNAs localize to specific subcellular loci within the neuronal process, including sensorin mRNA, which localizes to synapses in Aplysia SNs (Lyles et al., 2006), and arc mRNA, which localizes to stimulated synapses in mammalian dentate granule neurons (Steward et al., 1998), whether or not mRNAs undergo directed targeting from the soma to specific synapses, and whether or not RNA localization defines the fate of specific neuronal compartments, remains an open question.

The Aplysia SN-MN culture system provides an ideal preparation for monitoring mRNA localization and local translation during synapse formation and synaptic plasticity, since a single neuron can be manipulated to form synapses or to undergo transcription- and translation-dependent synaptic strengthening in a branch-specific manner. Using this preparation, we find that the local proteome was regulated at the level of translation rather than by stimulus-induced mRNA targeting from the nucleus. Thus, transcripts, ribosomes, and translational machinery were delivered throughout the neuron, but translation was selectively activated in response to local stimuli such as synapse formation. We note that our studies do not rule out a role for more localized subcellular trafficking of mRNAs within neuronal processes, and/or selective stabilization within processes (as described in Farris et al. (2014)), but they do indicate that synapse formation and synaptic stimulation do not regulate the trafficking from cell soma to process or synapse. Thus, transcripts that are induced by stimuli are delivered throughout the neuron, with their local distribution and translation being regulated by localized cues.

We focused on a role for netrin-1 as the retrograde signal driving translation in the sensory neuron because it has been shown to promote translation in axonal growth cones (Campbell and Holt, 2001), to promote synaptogenesis in mammalian cortical neurons (Goldman et al., 2013), and because Flanagan et al. have reported that the netrin receptor DCC serves as a transmembrane translation regulation complex in dendrites and axons (Tcherkezian et al., 2010). The finding that anti-DCC antibodies decrease sensorin immunoreactivity in SNs paired with MNs, but not in isolated SNs, is consistent with netrin-1 being released from the MN and binding to DCC on the adjacent pre-synaptic SN to promote the translation of sensorin at synapses. Netrin-1 binding to DCC may promote translation by releasing translational machinery from binding DCC cytoplasmic domains within the pre-synaptic terminal, as proposed by Flanagan et al. (2010). Alternatively, netrin-1-induced translation may be downstream of PI3 kinase and/or MAPK pathways, both of which are activated by netrin-1 binding to DCC (Li et al., 2004; Round and Stein, 2007).

We previously reported that serotonin-induced translation of sensorin in SNs required a calcium-dependent retrograde signal from the MN (Wang et al., 2009). In the present study, we show that ApNetrin expressed in MN is released and binds to the SN in a calcium-dependent manner. What the source of calcium is remains an open question. One possibility is that it results from spontaneous release of glutamate from the pre-synaptic SN during synapse formation and/or during synaptic plasticity (Eliot et al., (1994); Jin et al., (2012); Villareal et al., (2007)), although see also Sutton et al. (2006), which indicates that spontaneous release serves to suppress local translation in hippocampal neurons.

While we used recombinant human netrin-1 and function-blocking antibodies against human DCC, we are confident that these reagents reflect the endogenous functions of netrin-1 and DCC in Aplysia. We cloned Aplysia netrin-1 and found that it is 47% identical and 60% similar to human netrin-1 (by comparison, C. elegans netrin-1 is 44% identical and 59% similar to human netrin-1). We also cloned Aplysia DCC and found that it is 31% identical and 47% similar to human DCC (by comparison, C. elegans DCC is 18% identical and 31% similar to human DCC). Moreover, the function-blocking antibody recognizes the extracellular domain, which is the most highly conserved region of the protein (extracellular domain of Aplysia DCC is 39% identical and 55% similar to human DCC). Finally, we show that overexpression of ApNetrin in MNs significantly increases synaptic strength and sensorin immunoreactivity (Figure 9).

We suspect that netrin-1/DCC signaling is one of multiple local cues and signaling pathways that regulate translation in neurons and that translation is not regulated in a digital on/off manner, but rather in a graded manner. This explains, for example, why there is some sensorin translation in SN neurites contacting non-target L11 (Figure 5), and why sensorin protein is expressed in isolated sensory neurons that have been incubated with anti-DCC antibodies (Figure 8A).

Our previously published report that synapse formation alters the localization of sensorin mRNA such that it concentrates at synapses (Lyles et al., 2006; Meer et al., 2012), and that serotonin-regulated translation of a sensorin-based translational reporter was restricted to sites of synaptic contact, suggested that the activity-dependent synaptic localization of mRNA played a critical role in the regulation of its translation. Our current finding that netrin-1 increases translation at synaptic and non-synaptic sites indicates that the precise subcellular localization of the mRNA is secondary to the localization of the netrin-1 signal. Thus, sensorin mRNA does not have to be in the synaptic milieu per se to be translated but rather has to localize to a site in which netrin-1/DCC signaling can occur. We also note that we previously found that there was an increase in the concentration of sensorin mRNA in SN branches contacting target as compared to non-target MNs in more mature (5 DIV) cocultures (Lyles et al., 2006). This suggests that while there is not directed targeting of sensorin mRNA from soma to process during synapse formation, there may be preferential stabilization, over time, of sensorin mRNA at sites of synaptic contact. This finding is consistent with recent studies indicating that regulated RNA degradation functions to localize mRNAs and to limit translation within axons (Colak et al., 2013).

Our study also sheds light on the relative contributions of local translation and somatic translation to the local proteome. Thus, we can compare the amount of netrin-1-induced translation of the sensorin reporter in SN-LFS cocultures with and without the soma (Figure 7—figure supplement 1D). Although one cannot directly compare the total concentration of sensorin in intact SN and isolated SN neurites, our data indicate that netrin-1 is able to induce an equivalent increase in SN neurites in the presence or absence of a SN soma. This suggests that local translation contributes substantially to stimulus-induced changes in the local proteome.

The fact that netrin-1 was able to stimulate global translation at non-synaptic sites as detected by phospho-eIF4E immunocytochemistry (Figure 7—figure supplement 1C) suggests that a large number of transcripts, and not just the two mRNAs we detected by FISH, localize throughout the neuronal arbor. Moreover, we found that every type of RNA that we imaged was delivered from the soma throughout the neuronal arbor, whether during synapse formation or following local stimulation. In contrast, we found that every measure of translation that we employed demonstrated localized translation at synapses and/or stimulated synapses. While our results are consistent with the idea that mRNAs are delivered throughout the neuronal arbor, validation of this idea as a general mechanism will require analysis of a larger pool of specific transcripts.

Taken together, our findings show that a signaling pathway involved in axon guidance also regulates synapse formation by spatially restricting neuronal gene expression. Our findings further underscore an uncoupling of transcriptional and translational regulation in neurons, with stimuli triggering transcription of a population of mRNAs that are delivered throughout the neuron, but with local cues independently regulating their translation. Decentralizing the control of gene expression to individual subcellular compartments in this manner enriches the adaptability and plasticity of the nervous system by enabling each compartment to change its proteome in response to local cues. This mechanism allows all the branches of a neuron to be in a state of readiness to respond to local stimuli by changing their local proteome and hence their synaptic structure and function.

Cultures were prepared from adult 80–100 gram Aplysia (Alacrity, Redondo Beach, CA) for SN-LFS MN cultures or from the abdominal ganglion of juvenile Aplysia (1–4 g, National Aplysia Resource, University of Miami, FL) for L7 and L11 MNs. Animals were anesthetized by injecting 0.35 M MgCl₂ into the body, ganglia were dissected and incubated in protease (10 unit/ml protease XIV, Sigma #P5147) in L15 culture media for 1 hr 50 min to digest the connective tissue sheath. Bifurcated SNs from the pedal–pleural ganglia were paired with L7 and L11 MNs from the abdominal ganglion of juvenile Aplysia or with two LFS MNs from adult animals. Culture methods are described in detail in Zhao et al. (2009).

To induce LTF of SN-MN synapses, we applied five spaced applications of 5HT (10 μM) using either bath or local perfusion. Bath application of 5x5HT consisted of five 5 min applications of 5HT (10 μM) in L15 media with four intervening 20 min washes in L15, for a total stimulation time of 1 hr 45 min. Local perfusion of 5x5HT consisted of local delivery of 5HT (100 μM) to SN-MN synapses with a perfusion pipette; five applications were given at 10 min intervals, with each application consisting of five 5 s pulses given at 10 s intervals. ASW (Artificial Sea Water) was used as a control for all experiments. Cultures were fixed immediately after the last stimulation or at 24 hr after stimulation and processed for immunocytochemistry or FISH.

Confocal images were taken on a Zeiss LSM 700 scanning laser microscope. For FISH and ICC experiments, we used a Picospritzer (World Precision Instrument, Sarasota, FL) to microinject 5 mM Alexa647 (bifurcated SN) or Alexa546 (L7 MN) as volume controls and to clearly differentiate SN from MN neurites at least 3 hr before confocal imaging. Image analysis was done by an observer blind to experimental design.

To investigate the distribution of RNA granules during synapse formation, 200 ng/μl ApStaufen-dendra2 (in pNEX3) together with Alexa647 was microinjected into SNs on DIV 1 and imaged on DIV3. For the RNA granules targeting after local perfusion, images were acquired before and 24 hr after local perfusion.

To measure new translation, we microinjected 200 ng/μl sensorin-dendra2 reporter (in pNEX3, described in Wang et al. (2009)) into SNs on DIV 1. On day 2, we removed the SN cell body and 18 hr later, we photoconverted dendra2 from green to red as a described in Wang et al. (2009), with slight modifications, using an X-cite series120Q excitation light (a 120-watt lamp), delivering two 5 s pulses of UV illumination with a 10 s interval. After imaging, we returned cultures to the 18°C incubator and imaged the green and red signals 24 hr later. To compare new translation in SN branches contacting target or non-target motor neuron, we measured total fluorescence (area (μm2) × mean fluorescence intensity) of both green and red fluorescence using Slidebook5 software. New translation was normalized as the total pixel intensity in the dendra2 green signal to the dendra2 red signal 24 hr after photoconversion.

To test whether ApNet-1 depends on post-synaptic Ca²⁺, 200 ng/μl dendra2-tagged ApNet-1 (green) was microinjected to the LFS MN and 200 ng/μl ApDCC-mCherry (red) was microinjected into the SN at DIV2. 24 hrs later, 50 mM BAPTA in Buffer (1.5 M K-Acetate, 0.5 M KCl, 0.01 M HEPES, pH7.2) was microinjected into the post-synaptic LFS MN. Buffer alone was microinjected as a control. Images were taken pre- and 24 hr after BAPTA or buffer microinjection.

To test the effect of anti-DCC antibodies on the binding of MN-expressed ApNetrin-1-dendra2 to ApDCC-mCherry-expressing SNs, cultures were incubated with anti-DCC antibodies (250 ng/ml) for 24 hr. In control cultures, ASW or non-immune IgG (250 ng/ml) was added for 24 hr. Images were taken 24 hr after incubation.

To measure varicosity number in SN-MN cocultures, a VAMP (synaptobrevin)-mCherry pNEX3 construct was expressed in SNs paired with LFS MNs for 48 hr. The number of varicosities was measured before and 24 hr after incubation with 250 ng/ml netrin-1.

EPSP amplitude was measured as described in Zhao et al. (2003). Briefly, LFS MNs were impaled with a sharp glass electrode (10–15 MΩ) filled with 2 M potassium acetate, and the membrane potential of MNs was held at −80 mV and of SNs at −50 mV. EPSP amplitudes were measured before and 24 hr after 250 ng/ml netrin-1 treatment in L15 media using Axoscope 8.2 and pCLAMP 8 (Axon Instruments, Union City, CA) following intracellular depolarization (1–3 nA for 5 ms) of the SN. Artificial Sea Water (ASW) was used as a vehicle control.

We slightly modified the ribopuromycylation method described in David et al. (2012). Specifically, we cultured Aplysia SNs with L7 target MNs and L11 non-target MNs for 3DIV and then replaced the culture media with L15 media. Cultured neurons were incubated with 200 μM emetine in L15 media for 30 min before adding puromycin. After 30 min, 100 μM puromycin with 200 μM emetin in L15 media was added for 10 min. Cultured neurons were then washed with L15 and immediately fixed with 4% PFA/30% sucrose for 30 min.

FISH was performed as described in Lyles et al. (2006) with slight modifications. Briefly, cultured neurons were fixed with 4% PFA/30% sucrose for 30 min at room temperature. For FISH, after fixation, we hybridized with 40 ng/ml 18S rRNA, 28S rRNA, EF1α RNA, or sensorin RNA probed and imaged after FISH. For β-thymosin RNA FISH, we followed by the protocol as described in ViewRNA ISH processes (Affymetrix, Santa Clara, CA). Image settings were adjusted so that sense 18S rRNA, sense 28S rRNA, sense EF1α RNA, and sense sensorin mRNA probes did not produce any signal and no probes were used as negative control for β-thymosin RNA.

To generate riboprobes, we performed in vitro transcription with T7/T3 polymerase, tagging riboprobes with digoxigenin or biotin and purifying with ProbeQuant G-50 microcolumn (#28-9034-08; GE healthcare, Piscataway, NJ). Before we used riboprobes, we measured their concentration and tested their specificity by dot blotting with anti-digoxigenin-POD (#11207733910; Roche, Indianapolis, IN) or HRP-mouse anti-biotin (#03-3720; Invitrogen, Grand Island, NY). All FISH experiments were performed with sense riboprobes as a control to measure background signals. FISH detection using the TSA amplification system was followed as a described in Lyles et al. (2006).

To investigate netrin-1 mediated translation, 250 ng/ml recombinant Fc human netrin-1 (#ALX-522-455; Enzo life sciences, Farmingdale, NY) in L15 media or functional blocking anti-DCC (#OP45; Milipore, Billerica, MA) was applied to cultures for 24 hr and then washed with L15 media. Cultures were immediately fixed with 4% PFA/30% sucrose for 30 min.

After fixation, ICC was performed as described in Martin et al. (1997). Primary antibodies included custom-made chicken anti-sensorin antibodies (generated against peptide CATRSKNNVPRRFPRARYRVGYMF by Aves Labs, Inc., Tigard, OR), rabbit anti-eIF4E (#9742; Cell Signaling, Beverly, MA), rabbit anti-phospho eIF4E (#9741; Cell Signaling), rabbit anti-4EBP (#9452; Cell Signaling), rabbit anti-phospho 4EBP (#2855; Cell Signaling), and mouse monoclonal anti-puromycin, clone 12D100 (Millipore, #MABE343).

Degenerate primers were used to amplify segments of netrin-1 and DCC from RNA isolated from adult Aplysia CNS. The sequenced fragments were BLAST searched against the following database: http://aplysiagenetools.org/. Identified hits were used to generate primers that were in turn used to amplify full-length Aplysia netrin-1 (GenBank accession #KM218335) and DCC (accession #KM218336) from RNA isolated from CNS. Full-length clones were verified by sequencing.

All results are presented as mean ± SEM. Prism GraphPad was used to perform all statistical analysis. The type of statistical analysis performed is described in each figure legend.

The funder had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank S Haasan for blind image analysis, D Black, JT Braslow, P Chen, TH Ch'ng, VM Ho, GM Martin, K Olofsdotter-Otis, B Uzgil, and L Zipursky for comments on the manuscript, and all members of the Martin lab for helpful discussions. The work was supported by NIH R01NS045324 (to KCM).

1. Received: July 25, 2014
2. Accepted: January 8, 2015
3. Accepted Manuscript published: January 8, 2015 (version 1)
4. Version of Record published: February 4, 2015 (version 2)

© 2015, Kim and Martin

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.