1. Developmental Biology
  2. Neuroscience
Download icon

Receptor-specific interactome as a hub for rapid cue-induced selective translation in axons

Research Article
  • Cited 2
  • Views 1,751
  • Annotations
Cite this article as: eLife 2019;8:e48718 doi: 10.7554/eLife.48718

Abstract

Extrinsic cues trigger the local translation of specific mRNAs in growing axons via cell surface receptors. The coupling of ribosomes to receptors has been proposed as a mechanism linking signals to local translation but it is not known how broadly this mechanism operates, nor whether it can selectively regulate mRNA translation. We report that receptor-ribosome coupling is employed by multiple guidance cue receptors and this interaction is mRNA-dependent. We find that different receptors associate with distinct sets of mRNAs and RNA-binding proteins. Cue stimulation of growing Xenopus retinal ganglion cell axons induces rapid dissociation of ribosomes from receptors and the selective translation of receptor-specific mRNAs. Further, we show that receptor-ribosome dissociation and cue-induced selective translation are inhibited by co-exposure to translation-repressive cues, suggesting a novel mode of signal integration. Our findings reveal receptor-specific interactomes and suggest a generalizable model for cue-selective control of the local proteome.

Introduction

mRNA localization and local translation are major determinants of the local proteome (Zappulo et al., 2017). This seems particularly important for morphologically complex cells such as neurons, where the axonal sub-compartment and its growing tip, the growth cone, often far away from the cell body, rapidly perform specialized functions (Holt and Schuman, 2013). During neuronal wiring, specific interactions between extrinsic cues and receptors mediate guidance of axons to their proper target area and axon branching in this area (Stoeckli, 2018; Manitt et al., 2009; Marshak et al., 2007; Cioni et al., 2013). The rapid axonal responses to several guidance cues require local protein synthesis (Jung et al., 2012; Campbell and Holt, 2001). For example, attractive guidance cues, such as Netrin-1, trigger axonal translation of mRNAs encoding proteins that facilitate actin assembly, whereas repulsive cues trigger the local synthesis of cytoskeletal proteins involved in actin disassembly (Leung et al., 2006; Wu et al., 2005; Piper et al., 2006). This cue-specific mode of translation enables growth cones to steer differentially – towards or away – from the source of such cues (Lin and Holt, 2007; Lin and Holt, 2008). Unbiased detection of newly synthesized proteins in the axon compartment has revealed further complexity showing that different guidance cues stimulate the regulation of distinct signature sets of >100 axonal nascent proteins within just 5 min, many of which are not cytoskeletal-related (Leung et al., 2006; Yao et al., 2006; Wu et al., 2005; Cagnetta et al., 2018; Cioni et al., 2018). Several mechanisms are known to control different aspects of axonal translation, including microRNA regulation (Bellon et al., 2017), mRNA modification (Yu et al., 2018), modulation of the phosphorylation of eukaryotic initiation factors (Cagnetta et al., 2019), RNA-binding protein (RBP) phosphorylation (Sasaki et al., 2010; Lepelletier et al., 2017; Hüttelmaier et al., 2005) and receptor-ribosome coupling (Tcherkezian et al., 2010). The latter is a particularly direct and attractive mechanism to link cue-specific signalling to differential mRNA translation. However, this mechanism has been shown only for the Netrin-1 receptor, deleted in colorectal cancer (DCC), in commissural axon growth cones and HEK293 cells (Tcherkezian et al., 2010). It is unknown whether receptor-ribosome coupling is a widespread mechanism used by different receptors and in different cell types, and whether it regulates selective local translation.

Here, we show in the axonal growth cones of retinal ganglion cells (RGCs) that receptor-ribosome coupling is used by several different guidance receptors known to trigger local protein synthesis (DCC, Neuropilin-1 and Robo2, but not EphB2), indicative of a common mechanism. Interestingly, the receptor-ribosome interaction is mRNA-dependent and immunoprecipitation (IP) reveals that distinct receptors associate with specific RNA-binding proteins (RBPs) and subsets of mRNAs. Upon cue-stimulation, ribosomes dissociate from their receptors within 2 min and receptor-specific mRNAs are selectively translated. We also find that co-stimulation with EphrinA1 blocks the Netrin-1-induced DCC receptor-ribosome dissociation and selective translation in axons, suggesting a new regulatory mechanism for integrating different signals. Together, this study provides evidence that receptor-ribosome coupling is a common mechanism across different receptors and cell types, and suggests that receptor-specific interactomes act as a hub to regulate the localized and selective cue-induced mRNA translation.

Results

Multiple guidance cue receptors interact with ribosomes

In retinal axons, Netrin-1 and Sema3A mediate growth cone steering and branching (Campbell and Holt, 2001; Manitt et al., 2009; Campbell et al., 2001). Specifically, the rapid chemotropic responses to Netrin-1 and Sema3A are mediated, at least in part, by local translation (Campbell and Holt, 2001). The Netrin-1 receptor, DCC, was previously reported to associate with ribosomes in spinal commissural axon growth cones (Tcherkezian et al., 2010). We first asked whether the interaction of DCC with ribosomes is conserved in a different system and cell type, and explored the possibility that the Sema3A receptor, Neuropilin-1 (Nrp1), also interacts with ribosomes in this system. To do this, we performed immunoprecipitation (IP) of endogenous DCC and Nrp1 from Xenopus laevis embryonic brains and eyes followed by mass-spectrometry (LC-MS/MS) analysis of eluted samples. Each IP was performed in triplicate and after raw data processing using MaxQuant software, we determined statistically significant interactors of DCC and Nrp1 compared to an IgG control pulldown using label-free (LFQ) intensities and Perseus software analysis (Figure 1A). Gene-ontology (GO) enrichment analysis revealed that ‘structural constituent of ribosomes’ appeared as the most prominently enriched category in both DCC and Nrp1 pulldowns, indicating that both receptors can interact with ribosomal proteins (Figure 1B). Specifically, 75 out of 79 ribosomal proteins (94.9%) were detected in the DCC and Nrp1 pulldowns. Of these, 51 and 33 RPs were identified as statistically enriched interactors for Nrp1 and DCC, respectively, compared to IgG control pulldowns. There was no bias towards small or large ribosomal subunit proteins (Figure 1A, red dots). The GO analysis also revealed the presence of other groups shared between the receptors, such as ‘vesicle-mediated transport’ (Figure 1B). Interestingly, some categories of proteins were enriched for only one of the receptors, for example the ‘phosphoprotein phosphatase activity’ GO term was significantly enriched only in the DCC pulldown and the ‘barbed-end actin filament capping’ GO term was enriched only in the Nrp1 pulldown (Figure 1B). To confirm the interaction between receptors and ribosomal proteins, we performed Western blot (WB) analysis after IP and validated that both DCC and Nrp1 interact with small (40S) and large (60S) ribosomal subunit proteins (Figure 1C–D). These interactions appear to be conserved, as endogenous IP from the human neuronal cell line SH-SY5Y, which expresses both DCC and Nrp1, also shows ribosomal protein co-precipitation after pulldown of the endogenous receptor (Figure 1—figure supplement 1A–B).

Figure 1 with 1 supplement see all
Multiple guidance cue receptors interact with ribosomes.

(A) Volcano plots showing statistically enriched proteins in DCC-IP and Nrp1-IP samples identified by permutation-based FDR-corrected t-test based on three biological replicates. The LFQ intensity of the DCC or Nrp1 pulldowns over IgG pulldowns are plotted against the -log10 p-value. FDR < 0.05; S0 = 2. (B) Gene enrichment analysis of statistically enriched proteins in the DCC and Nrp1 pulldown samples. The values in each circle denotes protein count. (C–F) Western blot validation of RP co-immunoprecipitation with DCC, Nrp1 and Robo2 but not with EphB2. Each Western blot was repeated 2 to 4 times, representative images are shown. (G–J) Relative 18S and 28S ribosomal RNA abundance after control (IgG) pulldown or receptors pulldowns shows enrichment of rRNA in DCC, Nrp1, and Robo2 but not EphB2 pulldowns (unpaired two-tailed t-test; three biological replicates). Bars indicate means, error bars indicate standard deviation; *p<0.05.

In addition to DCC and Nrp1, Roundabout 2 (Robo2) triggers local protein synthesis after binding to the guidance cue Slit2 (Piper et al., 2006). Therefore, we asked whether Robo2 also interacts with ribosomal proteins. WB after IP from Xenopus embryonic brains and eyes or SH-SY5Y cells showed that Robo2 also interacts with ribosomal proteins of both subunits (Figure 1E, Figure 1—figure supplement 1C). We then looked at EphB2, as growth cone collapse mediated by EphrinB, the ligand for this receptor, is not mediated by local protein synthesis (Mann et al., 2003). In this case, we could not detect co-IP of ribosomal proteins with EphB2 in Xenopus embryonic brains and eyes, indicating that not all guidance receptors interact with ribosomal proteins (Figure 1F), and suggesting that only receptors that require local protein synthesis for their action on growth cones are coupled to ribosomes.

To confirm that receptors bind to ribosomes or ribosomal subunits and not free ribosomal proteins, we isolated RNA after IP and performed quantitative-RT-PCR (qPCR) for 18S (40S small ribosomal subunit) and 28S (60S large ribosomal subunit) ribosomal RNA (rRNA), which should be present only in intact ribosomal subunits in the cytoplasm. Consistent with the WB results, DCC, Nrp1 and Robo2, but not EphB2, exhibit a significant enrichment of both 18S rRNA and 28S rRNA compared to an IgG control pulldown in Xenopus brains (Figure 1G–J), and in SH-SY5Y cells in the case of DCC and Nrp1 (Figure 1—figure supplement 1D–E). Collectively, these findings reveal that multiple receptors known to trigger local protein synthesis can associate with ribosomal subunits.

Guidance cue receptors associate with ribosomes in a mRNA-dependent manner

We next examined the co-sedimentation profiles of DCC and Nrp1 in Xenopus embryonic brains and eyes after sucrose gradient purification of ribosomes in order to see if the receptors were mostly associated with ribosomal subunits, monosomes or polysomes. Consistent with previous findings (Tcherkezian et al., 2010), DCC was prominent in 40S, 60S and 80S fractions but not in polysomal fractions (Figure 2—figure supplement 1A). Nrp1, however, was found in 40S, 60S and 80S fractions, as well as in polysomal fractions (Figure 2—figure supplement 1A), suggesting a possibly different association mechanism or a different translational status of the receptor-bound ribosomes. Both DCC and Nrp1 were also present in ribosome-free fractions indicating that not all receptor molecules are associated with ribosomes (Figure 2—figure supplement 1A,C). EDTA treatment, which dissociates the monosomes/polysomes into separate ribosomal subunits (Simsek et al., 2017), shifted both DCC and Nrp1 to lighter fractions, supporting a valid association with ribosomes (Figure 2—figure supplement 1B,C).

We used qPCR to investigate this association further. When IP samples were treated with EDTA before elution, the enrichment of 18S and 28S rRNA after receptor pulldown was significantly decreased for both DCC and Nrp1 (Figure 2A). A possible explanation for this decrease is that DCC and Nrp1 interact mainly with 80S ribosomes (Tcherkezian et al., 2010). Another possibility is that the binding of ribosomes to receptors is mRNA-dependent. To test the latter hypothesis, we treated the receptor pulldown samples with RNase A/T1, which digests mRNAs and releases any factors bound to ribosomes via mRNA (Simsek et al., 2017). The concentration of RNase A/T1 used here largely preserves the integrity of ribosomes, as evidenced by the co-sedimentation profiles that show successful conversion of polysomes into monosomes, increasing the monosomal (80S) peak (Figure 2—figure supplement 1D), though we cannot exclude that it may still partially cleave rRNA. The significant decrease in the co-precipitation of 18S and 28S rRNA with receptors in these conditions suggests that mRNA is important for the association of 80S ribosomes with receptors (Figure 2A). Consistent with these results, Western blot analysis of IP samples treated with RNase A/T1 or EDTA after pulldown confirms the decrease in ribosomal proteins for both DCC and Nrp1 (Figure 2B,C), while the amounts of DCC and Nrp1 that precipitated were unaffected by the treatment conditions (Figure 2B–C). Together, these results suggest that the interaction of receptors with ribosomes is likely mediated through mRNA.

Figure 2 with 1 supplement see all
Receptor-ribosome coupling is mRNA dependent and DCC and Nrp1 bind to specific RBPs and mRNAs.

(A) Relative 18S and 28S ribosomal RNA abundance after control (IgG) pulldown or receptors pulldowns with or without EDTA or RNase A/T1 treatments (two-way ANOVA with Bonferroni’s multiple comparisons test; three biological replicates; Bars indicate mean, error bars indicate standard deviation; ***p<0.0001). (B) Western blot analysis and quantification of ribosomal proteins after DCC and (C) Nrp1 pulldowns. (two-way ANOVA with Bonferroni’s multiple comparisons test; three biological replicates; Bars indicate mean, error bars indicate standard deviation; **p<0.01; ***p<0.0001). (D) Hierarchically-clustered heatmap of detected RBPs after DCC and Nrp1 pulldown. LFQ intensities are plotted for each IP-MS replicate. (E) Mander’s overlap coefficients analysed using dual immunohistochemistry of DCC and Staufen1 or hnRNPA2B1 in axonal growth cones (unpaired two-tailed t-test; three biological replicates; individual data points are shown, error bars indicate SEM; p=0.03913). (F) Mander’s overlap coefficients analysed using dual immunohistochemistry Nrp1 and Staufen1 or hnRNPA2B1 in axonal growth cones (unpaired two-tailed t-test; three biological replicates; individual data points are shown, error bars indicate SEM; p=0.00161). (G) Volcano plot showing differential expression analysis for DCC and Nrp1 pulldowns. (H) Enrichment analysis plot of known RBP targets of Staufen1 and hnRNPA2B1 detected in RNA-sequencing data after DCC and Nrp1 pulldown (individual data points are shown, error bars indicate standard deviation, Mann-Whitney test, Wilcoxon rank sum test DCC versus Nrp1; p=0.001511).

DCC and Nrp1 bind to specific RNA-binding proteins

The mRNA-dependency of the receptor-ribosome interaction could be explained by mRNAs directly mediating the binding of receptors to ribosomes. Another possibility is that RNA binding proteins are key intermediaries in this binding and that mRNAs have a secondary role. Our MS analysis revealed that several RBPs are significantly enriched after DCC or Nrp1 pulldown (Figure 2D). Of 22 RBPs pulled down with DCC and 37 RBPs pulled down with Nrp1, only 11 are shared between the two receptors (Figure 2D). Several RBPs are significantly enriched in only one of the two receptor IPs. For example, Staufen1 is significantly enriched after Nrp1 IP, but not DCC IP, whereas hnRNPA2B1 is only detected after DCC IP (Figure 2D). This preferential RBP-receptor binding in axonal growth cones was also seen using dual immunocytochemistry with antibodies against DCC and Nrp1 and the RBPs Staufen1 and hnRNPA2B1 (Figure 2E–F). DCC co-localized with hnRNPA2B1 to a higher degree than with Staufen1 (Figure 2E). Conversely, Nrp1 showed a higher degree of co-localization with Staufen1 compared to hnRNPA2B1 (Figure 2F). RNAse A/T1 treatment was then used to test whether mRNA affects these associations. Western blot quantification after pulldown showed that the interaction of Staufen1 with Nrp1 was partly decreased by RNAse A/T1 treatment, suggesting that mRNA may stabilize the interaction between receptors and RBPs (Figure 2—figure supplement 1E). Together with our evidence implicating mRNA in the association of receptors with ribosomes, these results are consistent with a model in which receptors associate with specific RBPs, which bind specific mRNAs, and these mRNAs, in turn, recruit ribosomes.

DCC and Nrp1 bind to specific subsets of mRNAs

Next, we examined if and which mRNAs can associate with DCC and Nrp1 by performing RNA-sequencing (RNA-seq) on RNAs isolated after DCC and Nrp1 IP. We used a human neuronal cell line, SH-SY5Y, for these experiments in order to rule out that any detected difference in the mRNAs is due to the expression of DCC and Nrp1 in different cell types. Co-precipitation of RNA was observed in DCC and Nrp1 pulldowns but not in IgG control pulldowns (Figure 2—figure supplement 1F). A distance matrix analysis revealed that the experimental replicates clustered together for each receptor and we observed a distinct signature of detected mRNAs between DCC, Nrp1 or whole lysate input samples (Figure 2—figure supplement 1G). Differential expression analysis revealed that DCC and Nrp1 each differentially bind to specific subsets of mRNAs, with 541 mRNAs differentially binding between DCC and Nrp1 (158 mRNAs for DCC versus 383 mRNAs for Nrp1) (Figure 2G). Of the highly abundant detected mRNAs (FPKM >100 and FPKM >1000), ~41% and ~70% respectively were differential between DCC and Nrp1, whilst with the low abundant detected mRNAs (FPKM 1–10), only ~5% were differential between DCC and Nrp1. GO enrichment analysis of both all and only high abundance (FPKM >100) differentially expressed mRNAs showed the receptor-specific enrichment of mRNAs involved in different processes (Figure 2—figure supplement 1H,I and Figure 2—source data 2). For the high abundance mRNAs, GO terms that were associated with the mRNAs pulled down with DCC included ‘cell-cell adhesion’ and ‘protein targeting, while ‘translation’ and ‘small GTPase mediated signal transduction’ were associated with Nrp1.

Although these results rely on mRNA populations expressed in SH-SY5Y cells, which may differ from mRNAs binding to these receptors in Xenopus RGC axons, we compared mRNAs that preferentially bind to DCC or Nrp1 (Figure 2G) with known mRNA targets of several RBPs (Staufen1, hnRNPA2B1, Elavl1 and Fxr1), which were identified by previous CLIP studies in other systems (Lebedeva et al., 2011; Martinez et al., 2016; Sugimoto et al., 2015; Ascano et al., 2012). In particular, we focused on Staufen1 and hnRNPA2/B1 because our proteomic analysis revealed that Staufen1 is enriched after Nrp1 pulldown compared to DCC pulldown and hnRNPA2B1 was only detected after DCC pulldown (Figure 2D). The analysis revealed significant enrichment of known targets of Staufen1 and hnRNPA2B1 in Nrp1 versus DCC pulldown, respectively (Mann-Whitney U test, Wilcoxon rank sum test; p=0.001511) (Figure 2H). Overall, the known targets of the 4 RBPs tested (Staufen1, hnRNPA2B1, Elavl1 and Fxr1) can account for 41.1% of the significantly enriched DCC-precipitated RNAs and for 43.1% of the significantly enriched Nrp1-precipitated mRNAs. Collectively, the results support a model where receptor-specific RBPs mediate the differential association of mRNAs to receptors.

Receptor-ribosome coupling occurs in RGC axonal growth cones

As our IP experiments were performed in whole brain lysates (Figure 1), we next searched for evidence that these interactions occur in retinal growth cones. To begin to address this, we cultured eye primordia from Xenopus embryos and performed immunocytochemistry and expansion microscopy (Chen et al., 2015) on retinal axons using antibodies against the intracellular domain of DCC and a ribosomal protein (Figure 3A). DCC and RPL5/uL18 partially co-localized in retinal growth cones and filopodia (Figure 3A, white arrowheads). Similarly, RPS3A/eS1 co-localized with Nrp1 in retinal growth cones (Figure 3B, white arrowheads). Quantification of co-localization in expanded growth cones indicated a positive association between DCC and RPL5/uL18 (Pearson’s correlation = 0.4316 ± 0.011, n = 73) and Nrp1 and RPS3A/eS1 (Pearson’s correlation = 0.6727 ± 0.014, n = 72) (Figure 3—figure supplement 1A). To show close association of receptors and ribosomes in axonal growth cones, we employed the Proximity Ligation Assay (PLA) (Söderberg et al., 2006), modified for use on retinal axons (Yoon et al., 2012), which reports signal when the spatial coincidence of two proteins of interest is closer than 40 nm by using the respective antibodies. As a negative control, PLA was performed using the anti-DCC antibody and an IgG control antibody. This control generated a very low amount of background PLA signal (Figure 3C, Figure 3—figure supplement 1B), while we detected abundant PLA signal between DCC and RPL5/uL18, in line with previous findings (Konopacki et al., 2016), as well as with RPS4X/eS4 or RPL10A/uL1 (Figure 3C, Figure 3—figure supplement 1B). Similarly, Nrp1 generated abundant PLA signal together with RPS3A/eS1 or RPS23/uS12, with no detectable PLA signal in the negative control (Nrp1-IgG PLA) (Figure 3D). Given that EphB2 IP does not show any interaction with ribosomal proteins in Xenopus brain and eyes (Figure 1F,J), we tested whether this is conserved in retinal growth cones. Consistent with the IP results (Figure 1F,J) and with the EphB2-induced local protein synthesis independent growth cone collapse (Mann et al., 2003), PLA between EphB2 and RPL5/uL18 generated almost no detectable signal compared to DCC-RPL5/uL18 or Nrp1-RPS3A/eS1 in growth cones (Figure 3E). To provide further evidence, we performed electron microscopy on unstimulated axonal growth cones, and we observed a remarkable abundance of ribosomes in growth cones (Figure 3F). Strikingly, ribosomes could be seen aligned in rows underneath the plasma membrane (Figure 3F, Figure 3—figure supplement 1C–E), particularly in the regions in closest contact with the culture substrate. Indeed, we observed rows of ribosomes within 50 nm of the plasma membrane in 20 out of 22 axonal growth cones, and the presence of single ‘isolated’ ribosomes in the other two growth cones (Figure 3F, Figure 3—figure supplement 1C). The average distance between two neighboring ribosomes close to the plasma membrane in growth cones was significantly larger than the distance between ribosomes in the cell soma (58.12 ± 19.68 nm, n = 93 from 10 growth cones versus 23.05 ± 3.07 nm, n = 158 from five soma, p<0.00001) (Figure 3G, Figure 3—figure supplement 1C,E), indicative of and consistent with monosomes binding to the intracellular portions of transmembrane receptors, such as DCC or Nrp-1.

Figure 3 with 1 supplement see all
DCC and Nrp1 are in close proximity to ribosomes in axonal growth cones in a cue-dependent manner.

(A) Expansion imaging shows partial co-localization of DCC and (B) Nrp1 with ribosomal proteins (Scale bars, 5 μm). (C) Representative proximity ligation assay signal in axonal growth cones between DCC and RPL5/uL18, RPS4X/eS4 or IgG control (Scale bars, 5 μm). (D) Representative proximity ligation assay signal in axonal growth cones between Nrp1 and RPS3A/eS1, RPS23/uS12 or IgG control (Scale bars, 5 μm). (E) Representative PLA signal in axonal growth cones between EphB2 and RPL5/uL18 (left) and quantification of PLA signal in axonal growth cones compared to DCC-RPL5/uL18 or Nrp1-RPS23/uS12 (right) (Mann-Whitney test; three biological replicates; bars indicate mean, error bars indicate SEM, ***p<0.0001; Scale bars, 5 μm). (F) EM image of an unstimulated axonal growth cone showing ribosomes aligned in a row (red arrows) under plasma membrane (PM). Inset shows the growth cone at lower magnification; the red box indicates the area shown in higher magnification. The section glances through the extreme surface of growth cone, where it attaches to the culture dish, giving rise to areas that lack subcellular structure. (G) Distribution frequency of the inter-ribosome distance in nm of ribosomes in axonal growth cones (n = 20) or in RGC soma (n = 5). All distances larger than 100 nm were pooled together. (H, I, J, K) Quantification of PLA signal in cue-stimulated axonal growth cones relative to control (unpaired two-tailed t-test; bars indicate mean, error bars indicate SEM; ***p<0.0001; *p=0.0423; for n.s. in J p=0.3522; for n.s. in K, p=0.885). (L) Relative PLA quantification of DCC and RPL5/uL18 compared to control after Dynasore pre-treatment (50 μM for 20 min), Netrin-1, or Netrin-1 + Dynasore (one-way ANOVA with Bonferroni’s multiple comparisons test; bars indicate mean, error bars indicate SEM; p=0.001027 for Control vs. Netrin-1, p=0.000402 for Netrin-1 vs Netrin-1 + Dynasore, p=0.590377 for Control vs. Dynasore, p=0.384848 for Control vs Netrin + Dynasore). For all PLA experiments, numbers in bars indicate total number of growth cones quantified from at least three independent experiments.

Dissociation of ribosomes from receptors is triggered by extrinsic cues and requires endocytosis

Tcherkezian et al. (2010) showed that ribosomes uncoupled from the DCC receptor in response to extracellularly applied Netrin-1, stimulating local translation, suggesting a mechanism for the precise spatiotemporal control of the proteome in subcellular compartments. Previous work has also shown that stimulation with the guidance cues Netrin-1 and Sema3A that bind DCC and Nrp1, respectively, triggers the remodelling of the axonal proteome within 5 min (Cagnetta et al., 2018). Therefore, we first asked whether the association between receptors and ribosomal proteins is cue-sensitive. Remarkably, the PLA signal between DCC and the ribosomal proteins RPL5/uL18 and RPS4X/eS4 decreased significantly in retinal axon growth cones after 2 min of Netrin-1 stimulation (Figure 3H), suggesting a rapid dissociation of ribosomes from the receptor. It should be noted that, whereas DCC protein level does not change in response to 5 min Netrin-1 stimulation, both RPL5/uL18 and RPS4X/eS4 are up-regulated in response to 5 min Netrin-1 stimulation (Cagnetta et al., 2018), indicating that the decrease in the PLA signal in response to Netrin-1 may be underestimated. In contrast to the DCC-RP PLA signal, the PLA signal between DCC and the RBP hnRNPA2B1 did not decrease after 2 min of Netrin-1 stimulation, indicating that the receptor-RBP interaction is not affected by cue stimulation (Figure 3—figure supplement 1F).

Extracellular Sema3A at a concentration of 150 ng/ml, which is known to affect local axonal translation (Manns et al., 2012; Nédelec et al., 2012), also triggers a significant decrease in the Nrp1-RPS3A/eS1 and RPS23/uS12 PLA signal within 2 min (Figure 3I). Interestingly, when Sema3A is presented extracellularly at a higher concentration (700 ng/ml), it induces growth cone collapse that is independent of protein synthesis (Nédelec et al., 2012; Manns et al., 2012). Puromycylation of newly synthesized proteins in axon-only cultures and subsequent visualization and quantification of immunofluorescence using an anti-puromycin antibody (Schmidt et al., 2009) in the presence of 700 ng/ml Sema3A shows no increase in global translation in growth cones (Figure 3—figure supplement 1G). In line with this finding, stimulation with 700 ng/ml Sema3A does not cause a rapid decrease in the Nrp1-RPS3A/eS1 PLA signal (Figure 3—figure supplement 1H). This suggested that the dissociation of ribosomes from Nrp1 in response to Sema3A is intimately linked to rapid and local protein synthesis. Importantly, the detected decrease in PLA signal is not due to changes in Nrp1, RPS3A/eS1 and RPS23/uS12 protein levels as these do not change in response to 5 min Sema3A stimulation (Cagnetta et al., 2018).

Next, we tested the specificity of the cue-induced dissociation of RPs from receptors by quantifying the PLA signal between DCC and RPL5/uL18 after Sema3A stimulation and the PLA signal between Nrp1 and RPS23/uS12 after Netrin-1 stimulation. In neither case did we observe a decrease in PLA signal, confirming the ligand-receptor specificity of the cue-induced RP dissociation (Figure 3J–K).

The receptor-RP dissociation in response to an extrinsic cue suggests that this may occur on the plasma membrane but it is also possible that the dissociation happens intracellularly. Indeed, DCC and Nrp1 receptors are known to be rapidly endocytosed after cue stimulation (1–2 min) in growth cones (Piper et al., 2005) and we have recently identified the presence of ribosomal proteins on axonal endosomes which serve as platforms for local translation (Cioni et al., 2019), raising the possibility that the observed dissociation between receptors and ribosomes may also take place on endosomes. Therefore, we asked whether endocytosis plays a role in the cue-induced dissociation of ribosomes from receptors. Indeed, we found that treatment with the inhibitor of endocytosis Dynasore, a small GTPase inhibitor targeting dynamin (Macia et al., 2006), completely blocked the Netrin-1-induced decrease in PLA signal between DCC and RPL5/uL18, indicating that endocytosis is required for the receptor-ribosome dissociation (Figure 3L).

Together, these findings suggest that the rapid cue-specific dissociation of ribosomes in response to extracellular guidance cues is shared among different receptors, is tightly linked to cue-induced local translation-dependent responses, and requires endocytosis.

Integration of multiple cues can affect the cue-induced selective translation of receptor-specific mRNAs

During axon pathfinding and branching, axons encounter multiple cues, such as EphrinB2 and Netrin-1,and can integrate these cues by forming a complex between their respective receptors in a ligand-dependent manner (Morales and Kania, 2017; Dudanova and Klein, 2013; Poliak et al., 2015). The cue EphrinA1 has been reported to decrease local translation in hippocampal axons (Nie et al., 2010) and the rapid local translation of the Translationally controlled tumor protein (Tctp), which is up-regulated by Netrin-1 (Gouveia Roque and Holt, 2018). Therefore, we asked whether co-stimulation with EphrinA1 and Netrin-1 alters the dissociation of ribosomes from DCC. To address this question, we co-stimulated retinal axons with Netrin-1 and EphrinA1 and examined receptor-ribosome coupling using the PLA approach. Whereas Netrin-1 induces a decrease in the DCC-RPL5/uL18 PLA signal within 2 min, both Ephrin-A1 stimulation alone and co-stimulation with Netrin-1 and EphrinA1 do not decrease the DCC-RPL5/uL18 PLA signal, indicating that the Netrin-1-induced dissociation of ribosomes from DCC is blocked by co-stimulation with EphrinA1 (Figure 4A). By contrast, co-stimulation with EphrinA1 and Sema3A does not block the Sema3A-induced decrease in the Nrp1-RPS23/uS12 PLA signal (Figure 4—figure supplement 1A). These results reveal that integration of guidance cues can alter the receptor-ribosome dissociation, possibly by structural changes of the interacting receptors (Morales and Kania, 2017; Dudanova and Klein, 2013; Poliak et al., 2015).

Figure 4 with 1 supplement see all
EphrinA1 co-stimulation blocks Netrin-1 induced receptor-ribosome dissociation and selective translation.

(A) Relative PLA quantification of DCC and RPL5/uL18 compared to control after Netrin-1, EphrinA1, or co-stimulation (one-way ANOVA with Bonferroni’s multiple comparisons test; bars indicate mean, error bars indicate SEM; **p<0.01). (B, C) Puromycin QIF relative to control after Netrin-1, EphrinA1 or co-stimulation (one-way ANOVA with Bonferroni’s multiple comparisons test; bars indicate mean, error bars indicate SEM; ***p<0.0001). (D) Relative mRNA quantification after DCC IP of hnrnph1 and ctnnb1 mRNA (unpaired t-test with Welch’s corrections on dCT values; three biological replicates; bars indicate mean, error bars indicate SEM; *p=0.02 for hnrnph1; **p=0.0018 for ctnnb1). (E, F) β-Catenin QIF relative to control after Netrin-1, EphrinA1, Sema3A or Netrin-1 and EphrinA1 co-stimulation (one-way ANOVA with Bonferroni’s multiple comparisons test; bars indicate mean, error bars indicate SEM; ***p<0.0001). (G, H) hnRNPH1 QIF relative to control after Netrin-1, EphrinA1, Sema3A or Netrin-1 and EphrinA1 co-stimulation (one-way ANOVA with Bonferroni’s multiple comparisons test; bars indicate mean, error bars indicate SEM; ***p<0.0001; *p=0.0164). Scale bars, 5 μm. For all QIF experiments, numbers in bars indicate amount of growth cones quantified collected from at least three independent experiments.

Our data showing that EphrinA1 blocks the Netrin-1-induced ribosome dissociation from DCC, suggest that EphrinA1 may inhibit the axonal translation induced by Netrin-1. To test this hypothesis, we examined the effect of cue integration of Netrin-1 and EphrinA1 on both global and selective local translation in growth cones. In the culture conditions used in this study (Höpker et al., 1999), both Netrin-1 and EphrinA1 decrease global local translation in axons as measured by the puromycylation assay in axon-only cultures (Figure 4B–C). Consistent with this result, both cues decrease pERK1/2 levels (Figure 4—figure supplement 1B), an upstream activator of the TOR signalling pathway, which is known to regulate axonal protein synthesis (Campbell and Holt, 2003).

Despite the decrease in global axonal translation, previous work has revealed that Netrin-1 can induce the rapid selective translation of specific mRNAs (Cagnetta et al., 2018; Shigeoka et al., 2019). The IP-RNA-seq data in human SH-SY5Y cells had revealed that DCC associates with mRNAs encoding β-catenin (ctnnb1) and hnRNPH1 (hnrnph1) significantly more than with Nrp1. Interestingly, ctnnb1 and hnrnph1 mRNAs have been detected in Xenopus retinal axons (Shigeoka et al., 2019) and are selectively synthesised in response to 5 min Netrin-1 stimulation, but not Sema3A (Cagnetta et al., 2018), indicating that receptor-specific mRNAs can underlie the cue-induced selective translation. To further test this, we examined whether these mRNAs associate with DCC also in Xenopus brain and eyes by carrying out IP followed by qPCR. The results showed significant enrichment of ctnnb1 and hnrnph1 mRNAs in DCC pulldown compared to an IgG pulldown, thus confirming their association with DCC (Figure 4D). Finally, quantification of immunofluorescence confirmed that both β-catenin and hnRNPH1 protein levels increase in response to 5 min Netrin-1 stimulation, but not Sema3A (Figure 4E–H), in line with previous axonal translation findings (Cagnetta et al., 2018).

Similar to β-catenin and hnRNPH1, RPS14/uS11 mRNA is present in Xenopus retinal axons (Shigeoka et al., 2019) and is up-regulated in response to 5 min Netrin-1 stimulation, but not Sema3A (Cagnetta et al., 2018), as confirmed by quantification of immunofluorescence (Figure 4—figure supplement 1E). However, rps14 mRNA was not detected to be associated with DCC in SH-SY5Y cells. Therefore, we asked whether this is due to interspecies differences (human (SH-SY5Y) versus Xenopus), or whether rps14 is selectively translated via a DCC interactome-independent mechanism. To address this question, we carried out IP followed by qPCR in Xenopus brain and eyes, which confirmed rps14 association to DCC (Figure 4—figure supplement 1C). Our findings that Netrin-1, but not Sema3A, induces the translation of mRNAs bound to DCC point towards a model where receptor-specific mRNA interactomes act as a hub for rapid cue-specific selective translation.

Finally, we examined the effect of EphrinA1 co-stimulation on the Netrin-1-induced selective translation up-regulation of β-catenin, hnRNPH1 and RPS14/uS11. Quantification of immunofluorescence showed that EphrinA1 stimulation alone does not affect β-catenin and RPS14/uS11 protein levels (Figure 4E–H; Figure 4—figure supplement 1D) and decreases hnRNPH1 protein level in axonal growth cones (Figure 4G–H). Co-stimulation with Netrin-1 and EphrinA1 blocks the Netrin-1-induced increase of all three proteins (Figure 4E–H; Figure 4—figure supplement 1D). Together, the results show that integration of the EphrinA1 and Netrin-1 signals inhibits the Netrin-1-induced selective translation, possibly by inhibiting DCC-ribosome dissociation (Figure 4A).

Discussion

We provide evidence for a receptor-ribosome coupled mechanism by which extrinsic cues cause rapid and selective changes in the local proteome. In support of this model, we show that multiple guidance cue receptors interact with ribosomes, that the interaction between receptors and ribosomes depends on mRNA and rapidly decreases within 2 min of cue stimulation. Moreover, we find that receptors bind to distinct subsets of RBPs and mRNAs, and that cue stimulation induces the selective axonal translation of several receptor-specific mRNAs. Finally, we show that the integration of multiple cues can alter receptor-ribosome dissociation and selective translation.

Based on the candidate receptors tested here, we suggest that whether or not a particular receptor shows receptor-ribosome coupling is related to whether or not the receptors regulate local translation upon ligand binding. Future studies are needed to determine whether receptor-ribosome coupling is restricted to axon guidance receptors and neurons. Interestingly, a previous study has reported the association of a chemokine receptor, CXCR4, with eukaryotic initiation factor 2B (eIF2B), which decreases upon ligand binding in a pre-B cell line (Palmesino et al., 2016). In addition, several adrenergic receptor subtypes have been reported to associate with eIF2B at the plasma membrane (Klein et al., 1997). This raises the intriguing possibility that coupling of translational machinery with receptors extends to other cell types and is a widespread mechanism to rapidly transduce local translation downstream of extracellular signals.

Previous studies have shown that the RBP zipcode binding protein 1 (ZBP1) can be phosphorylated upon cue stimulation, thereby regulating local translation in axons by possibly releasing the bound mRNAs (Hüttelmaier et al., 2005; Sasaki et al., 2010; Lepelletier et al., 2017). DCC and Nrp1 each differentially bind to RBPs and mRNAs, thus providing a way to rapidly achieve cue-induced selective translation. We observed an enrichment of known mRNA targets for RBPs detected specifically in DCC and Nrp1 pulldowns respectively, suggesting a role for RBPs in mediating the differential binding of mRNAs to receptors and their cue-induced selective translation. This hypothesis is supported by the enrichment of the RBP hnRNPA2B1 and ctnnb1 mRNA (encoding β-catenin) specifically in DCC but not Nrp1 pulldown, as hnRNPA2B1 has been reported to control the translation of β-catenin (Stockley et al., 2014), which is selectively translated in response to Netrin-1, but not Sema3A in retinal axons (Cagnetta et al., 2018), in accord with the data reported here.

Our RNA-seq analysis reveals a receptor-specific enrichment of 100–400 mRNAs suggesting that a large number of mRNAs may be regulated by specific receptors and their ligands (Figure 2G). This idea is consistent with our previous proteomics study in Xenopus retinal axons showing that the translation of more than 100 mRNAs is regulated within 5 min in response to Netrin-1 and Sema3A (Cagnetta et al., 2018). It should be noted that, as our RNA-seq data are obtained from the human cell line SH-SY5Y, the number, and exact identity, of receptor-associated mRNAs may be different in axons. This is exemplified by the absence of rps14 mRNA enrichment in SH-SY5Y cells, which was detected in Xenopus brains (Figure 4—figure supplement 1C). In addition, it is possible that not all detected mRNAs interact with DCC and Nrp1 at the plasma membrane as a portion of these mRNAs could also be associated with receptors on endocytic vesicles that are known to contain DCC and Nrp1. Our results point to a model in which different subsets of mRNAs interact via specific RBPs with either DCC or Nrp1, and are released, together with ribosomes, upon specific cue stimulation and thus become available for subsequent translation (Figure 5). To fully understand and validate our model, it will be key to investigate the complex inter-dependency of these interactions.

Model diagram depicting the proposed interactions between receptors, RBPs, mRNAs and ribosomes under basal and cue stimulation conditions.

It should be noted that, in addition to RBPs and mRNAs, several other molecules characterize the receptor-specific interactome. For example, eIF3d, an initiation factor previously shown to regulate specialized translation initiation, is significantly enriched specifically after Nrp1 IP, but not DCC IP, thus raising the interesting possibility that differential binding to initiation factors may contribute to cue-induced selective translation (Lee et al., 2016). Intriguingly, a recent study revealed that an untranslated mRNA can associate with and regulate the signalling of the TrkA receptor in axons via its axon-enriched long 3’UTR (Crerar et al., 2019). It will be interesting to investigate whether any of the DCC and Nrp1 targets identified in our study also play a structural role, for example by regulating the receptor-ribosome association and/or the downstream signalling and local translation.

During axon guidance and branching, axons can encounter a combination of extracellular signals and ample evidence shows that the integration of multiple cues results in different outcomes than those of each single cue (Dudanova and Klein, 2013; Morales and Kania, 2017). Here, we tested the effect of cue integration on receptor-ribosome coupling and found that EphrinA1 blocks the Netrin-1-induced ribosome dissociation from DCC, but not the Sema3A-induced ribosome dissociation from Nrp1. In addition, EphrinA1 blocks the Netrin-1-induced selective increase in translation of several mRNAs. The mechanism by which EphrinA1 affects the coupling of DCC to ribosomes is unknown. One possibility is that, upon co-stimulation of EphrinA1 and Netrin-1, the DCC and Eph receptors may form a complex, thereby altering the receptor structure and association to ribosomes, which could be consistent with a previous study revealing a ligand-dependent interaction between the receptors Unc5 and EphB2 (Poliak et al., 2015).

In conclusion, our findings show that coupling of the translational machinery to guidance cue receptors at the plasma membrane of growth cones is a mechanism to rapidly and selectively control the cue-induced regulation of the local proteome and suggest that this may be a general principle that applies to membrane receptors more broadly.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional
information
Biological
sample
(Xenopus laevis)
Xenopus laevisNASCOCat# LM00715 (male);
RRID:XEP_Xla100;
Cat# LM00535 (female);
RRID:XEP_Xla
Cell line
(Homo-sapiens)
SH-SY5YATCCCat# CRL-2266; RRID:CVCL_0019
Antibodyanti-RPS3A (Rabbit polyclonal)AbcamCat# ab194670;
RRID:AB_2756396
ICC/PLA (1:100)
WB (1:1000)
AntibodyAnti-Neuropilin-1
(Rabbit monoclonal)
AbcamCat# ab81321;
RRID:AB_1640739
ICC/PLA (1:100)
WB (1:2000)
IP (5 µg)
AntibodyAnti-Neuropilin-1
(Mouse monoclonal)
ProteintechCat# 60067–1-Ig;
RRID:AB_2150840
ICC (1:100)
AntibodyAnti-DCC
(mouse monoclonal
BD BiosciencesCat# 554223;
RRID:AB_395314
ICC/PLA (1:100)
WB (1:1000)
IP (5 µg)
AntibodyAnti-RPL5 (rabbit polyclonalProteintechCat# 15430–1-AP;
RRID:AB_2238681
ICC/PLA (1:100)
AntibodyAnti-RPS4X
(Rabbit polyclonal)
ProteintechCat# 14799–1-AP;
RRID:AB_2238567
PLA (1:100)
WB (1:1000)
AntibodyAnti-RPL10A (Rabbit polyclonal)ProteintechCat# 16681–1-AP;
RRID:AB_2181281
PLA (1:100)
WB (1:500)
AntibodyAnti-RPS23 (mouse monoclonal)AbcamCat#: ab57644;
RRID:AB_945314
PLA (1:100)
WB (1:1000)
AntibodyAnti-RPS26
(Rabbit polyclonal
ProteintechCat# 14909–1-AP;
RRID:AB_2180361
WB (1:500)
AntibodyAnti-Robo2
(goat polyclonal)
R and D SystemsCat# AF3147;
RRID:AB_2181857
WB (1:250)
AntibodyAnti-EphB2
(mouse monoclonal)
Santa CruzCat# sc130068;
RRID:AB_2099958
WB (1:100)
IP (5 µg)
AntibodyAnti-EphB2
(mouse monoclonal)
Thermo Fisher ScientificCat# 37–1700;
RRID:AB_2533302
PLA (1:100)
AntibodyAnti-Staufen1
(Rabbit polyclonal)
AbcamCat# ab73478;
RRID:AB_1641030
ICC (1:100)
WB (1:500)
AntibodyAnti-hnRNPA2B1
(Rabbit polyclonal)
AbcamCat# ab31645;
RRID:AB_732978
ICC/PLA (1:100)
AntibodyAnti-RPS14
(Rabbit polyclonal)
AbcamCat# ab174661ICC (1:100)
AntibodyAnti-ß-Catenin
(Rabbit polyclonal)
Sigma-AldrichCat# C2206;
RRID:AB_476831
ICC (1:500)
AntibodyAnti-hnRNPH1AbcamCat# ab154894ICC (1:500)
AntibodyAnti-IgG (Rabbit)AbcamCat# ab37415;
RRID:AB_2631996
PLA (1:100)
IP (5 µg)
AntibodyAnti-IgG1 (Mouse)R and D SystemsCat# MAB002; RRID:AB_357344PLA (1:100)
IP (5 µg)
AntibodyAnti-IgG2b (Mouse)R and D SystemsCat# MAB004; RRID:AB_357346IP (5 µg)
AntibodyAnti-IgG (Goat)R and D SystemsCat# AB-108-C;
RRID:AB_354267
IP (5 µg)
AntibodyAnti-Puromycin-Alexa Fluor 488
conjugate
(mouse monoclonal)
MilliporeCat# MABE343-AF488; RRID:AB_2736875ICC (1:200)
AntibodyAnti-RPL19
(mouse monoclonal)
AbcamCat#ab58328;
RRID:AB_945305
WB (1:1000)
AntibodyAnti-FxRGift from Dr. Edward Khandjan, University of QuebecN/AWB (1:1000)
AntibodyAnti-pERK1/2Cell SignalingCat# 9101;
RRID:AB_331646
ICC (1:250)
AntibodyGoat-anti-rabbit
Alexa Fluor 568
AbcamCat# ab150077;
RRID:AB_2630356
ICC (1:1000)
AntibodyGoat-anti-mouse
Alexa Fluor 568
AbcamCat# ab150117;
RRID:AB_2688012
ICC (1:1000)
AntibodyGoat-anti-mouse-HRPAbcamCat# ab6789;
RRID:AB_955439
WB (1:15000)
AntibodyGoat-anti-rabbit-HRPAbcamCat#: ab97080;
RRID:AB_10679808
WB (1:15000)
Commercial assay or kitRNeasy mini kitQiagenCat# 74104
Commercial assay or kitSuperScript III First-strand
Synthesis kit
Thermo Fisher ScientificCat# 18080051
Commercial assay or kitQuantitect SYBR green PCR kitQiagenCat# 204143
Commercial
assay or kit
KAPA HyperPrep kitRocheCat# KK8503
Commercial
assay or kit
NextSeq 500/550 high output v2
kit (150 cycles)
IlluminaCat# FC-404–2002
Commercial assay or kitDuolink In situ
PLA Detection reagents green
Sigma-AldrichCat# DUO92014
Commercial
assay or kit
Duolink In situ
PLA Detection
reagents red
Sigma-AldrichCat# DUO92008
Commercial
assay or kit
Duolink In situ PLA probe Anti-Rabbit PLUSSigma-AldrichCat# DUO92002
Commercial
assay or kit
Duolink In situ
PLA probe
Anti-Mouse MINUS
Sigma-AldrichCat# DUO92004
Chemical compound,
drug, reagent
CycloheximideSigma AldrichCat# C4859
Chemical
compound,
drug, reagent
RNase AAmbionCat# EN0531
Chemical
compound,
drug, reagent
RNase T1AmbionCat# EN0541
Chemical
compound,
drug, reagent
PuromycinSigma-AldrichCat# P8833
Chemical
compound, drug, reagent
Recombinant mouse Netrin-1R and D systemsCat# 1109-N1
Chemical
compound,
drug, reagent
Recombinant human Sema3AR and D systemsCat# 1250-S3
Chemical compound,
drug, reagent
DynasoreSigma-AldrichCat# D7693
Chemical compound, drug, reagentSUPERase In RNAse inhibitorAmbionCat# AM2696
Software, algorithmVolocityPerkinElmerVersion 6.0.1;
RRID:SCR_002668
Software,
algorithm
GraphPad PrismGraphPadv.5;
RRID:SCR_002798
Software,
algorithm
ROtherv.3.2.2;
RRID:SCR_001905
https://www.r-
project.org
Software,
algorithm
MATLABMathworksv.R2016b; RRID:SCR_001622
Software, algorithmHISAT2Otherv.2.1.0; RRID:SCR_015530https://ccb.jhu.edu/software/hisat2/index.shtml
Software, algorithmCufflinksOtherv.2.2.1;
RRID:SCR014597
http://cole-trapnell-lab.github.io/cufflinks/

Embryos

Request a detailed protocol

Xenopus laevis embryos were fertilized in vitro and raised in 0.1x Modified Barth’s Saline (8.8 mM NaCl, 0.1 mM KCl, 0.24 mM NaHCO3, 0.1 mM HEPES, 82 µM MgSO4, 33 µM Ca(NO3)2, 41 µM CaCl2) at 14–20°C and staged according to the tables of Nieuwkoop and Faber (1994). All animal experiments were approved by the University of Cambridge Ethical Review Committee in compliance with the University of Cambridge Animal Welfare Policy. This research has been regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body (AWERB). All animals used in this study were below stage 45.

Cell line culture

Request a detailed protocol

Human neuroblastoma SH-SY5Y cells (ATCC; Cat# CRL-2266), free of mycoplasma, were cultured in Dulbecco’s minimal essential medium (DMEM) containing antibiotics, L-glutamine and 10% fetal bovine serum (FBS).

Primary Xenopus retinal cultures

Request a detailed protocol

Eye primordia were dissected from Tricaine Methanesulfonate (MS222) (Sigma-Aldrich) anesthetized embryos at stage 35/36 (or stage 32 for EM) and cultured on 10 µg/ml poly-L-lysine- (Sigma-Aldrich) and 10 µg/ml laminin- (Sigma-Aldrich) coated dishes in 60% L-15 medium (Gibco) at 20°C for 24 hr before performing immunohistochemistry or proximity ligation assay, or for 48 hr before the puromycilation assay. Where indicated in the figures and figure legends, cultures were treated with Netrin-1 (600 ng/ml, R and D systems, 1109-N1), Sema3A (150 or 700 ng/ml, R and D systems, 1250-S3), or Dynasore (50 µM, Sigma-Aldrich, D7693).

Immunoprecipitation

Request a detailed protocol

SH-SY5Y cells or Xenopus brains and eyes dissected from stage 40/41 embryos were lysed in lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10 mM MgCl2 and 10% glycerol supplemented with 100 µg/ml cycloheximide (Sigma-Aldrich), EDTA-free protease inhibitors (Roche, 11873580001), phosphatase inhibitors (Thermo Fisher Scientific, A32957) and SuperRNAse In RNAse inhibitor (Ambion, AM2696)). Tissues or cells were lysed for 30 min at 4°C and centrifuged for 5 min at 800 g at 4°C to remove unlysed cells and nuclei and then 15 min at 16000 g at 4°C. The resulting supernatant was incubated with magnetic Dynabeads pre-coupled with antibodies using the Dynabeads antibody coupling kit (Thermo Fisher Scientific, 14311D) for 1.5 hr at 4°C on a rotor. The following antibodies were used: mouse-anti-DCC (BD Biosciences, 554223); rabbit-anti-Nrp1 (Abcam, ab81321); goat-anti-Robo2 (R and D systems, AF3147); mouse-anti-EphB2 (Santa Cruz, sc130068) or an isotype control: rabbit IgG (Abcam, ab37415); mouse IgG1 (R and D systems, MAB002); mouse IgG2b (R and D systems, MAB004); goat IgG (R and D systems, AB-108-C). Beads were then washed three times in lysis buffer and processed for protein or RNA isolation. For EDTA and RNase A/T1 treatment pulldowns, immunoprecipitated samples (samples after incubation of supernatant with antibody-coupled beads) were equally divided into three tubes (tube 1: normal washes as above, tube 2: EDTA treatment washes, tube 3: RNase A/T1 treatment washes). For EDTA treatment, immunoprecipitated samples were washed with EDTA wash buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 25 mM EDTA and 10% glycerol supplemented with EDTA-free protease inhibitors (Roche, 11873580001), phosphatase inhibitors (Thermo Fisher Scientific, A32957) for three times before elution. For RNase A/T1 treatment, immunoprecipitated samples were washed three times for 3 min at RT with RNase A/T1 wash buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10 mM MgCl2% and 10% glycerol supplemented with 100 µg/ml cycloheximide (Sigma-Aldrich), EDTA-free protease inhibitors (Roche, 11873580001), phosphatase inhibitors (Thermo Fisher Scientific, A32957), 10 µg/µl RNase A (Ambion, EN0531) and 250U RNase T1 (Ambion, EN0541). After normal, EDTA, or RNase A/T1 washes, samples were processed for protein or RNA isolation.

For protein isolation, 1x NuPAGE LDS sample buffer (Thermo Fisher Scientific, NP0008) was added to the beads, incubated for 5 min at 95°C and the final protein eluate was collected after magnetic separation of the beads. For RNA isolation, RLT buffer was added to the beads, vortexed for 2 min and then separated from the beads on a magnetic stand.

Polysome fractionation

Request a detailed protocol

For density gradient fractionation, lysate was layered on a sucrose gradient (10–50%) in PLB buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10 mM MgCl2, 100 µg/ml cycloheximide (Sigma-Aldrich), 0.5 mM DTT) and ultracentrifugation was performed using a Beckman SW-40Ti rotor and Beckman Optima L-100 XP ultracentrifuge, with a speed of 35,000 rpm at 4°C for 160 min. Fractionations and UV absorbance profiling were carried out using Density Gradient Fractionation System (Teledyne ISCO). Proteins were precipitated from each fraction using methanol-chloroform precipitation and pellets were resuspended in 1x NuPAGE LDS sample buffer and used for Western blotting as described below.

Western blot

Request a detailed protocol

Proteins were resolved by SDS-PAGE on NuPage 4–12% Bis-Tris gels (Invitrogen, NP0321) and transferred to nitrocellulose membrane (Bio-Rad). The blots were blocked in 5% milk in TBST-T for 60 min at RT and then incubated with primary antibodies in 5% milk in TBS-T overnight at 4°C. After washing three times with TBS-T the blots were incubated with HRP-conjugated secondary antibodies (goat-anti-mouse HRP (Abcam, ab6789); goat-anti-rabbit HRP (Abcam, ab6721) for 1 hr at RT, washed again for three times in TBS-T, followed by ECL-based detection (Pierce ECL plus, Thermo Scientific, 32123). The following primary antibodies were used for Western blot analysis: mouse-anti-DCC (BD Biosciences, 554223), rabbit-anti-neuropilin-1 (Abcam, ab81321), goat-anti-Robo2 (R and D systems, AF3147), mouse-anti-EphB2 (Santa Cruz, sc130068), mouse anti-Rpl19/eL19 (Abcam, ab58328), mouse anti-RPS23/uS12 (Abcam, ab57644), rabbit anti-RPS4X/eS4 (Proteintech, 14799–1-AP), rabbit-anti RPL10A/uL1 (Proteintech, 16681–1-AP), rabbit-anti Rps26 (Proteintech, 14909–1-AP), mouse-anti-Rps3A (Abcam, ab194670), mouse-anti-FxR (gift from dr. Khandjian), rabbit-anti-Staufen1 (Abcam, ab73478).

Quantitative RT-PCR

Request a detailed protocol

RNA was isolated from eluted samples using the RNeasy Mini kit (Qiagen, 74104) and reverse transcribed into cDNA using random hexamers and the SuperScript III First-Strand Synthesis System (Thermo Fisher Scientific, 18080051). The cDNA was used to prepare triplicate reactions for qRT-PCR according to manufacturer’s instructions (QuantiTect SYBR Green PCR kit, Qiagen, 204143), plates were centrifuged shortly and run on a LightCycler 480 (Roche) using the following PCR conditions: denaturation for 15 s at 94°C; annealing for 30 s at 60°C; extension for 30 s at 72°C. The levels for each condition were corrected with their own input. The following primers were used for qPCR:

  • Xenopus 18S rRNA, 5’-GTAACCCGCTGAACCCCGTT-3’ and 5’-CCATCCAATCGGTAGTAGCG-3’;

  • Xenopus 28S rRNA, 5’-CTGTCAAACCGTAACGCAGG-3’ and 5’-CTGACTTAGAGGCGTTCAGTCA-3’.

  • human 18S rRNA, 5’-GTAACCCGTTGAACCCCATT-3’ and 5’-CCATCCAATCGGTAGTAGCG-3’;

  • human 28S rRNA, 5’-AACGGCGGGAGTAACTATGA-3’ and 5’-TAGGGACAGTGGGAATCTCG-3’.

  • Xenopus ctnnb1 mRNA, 5’-GACCACAAGTCGGGTGCTTA-3’ and 5’- CCAGACGTTGGCTTGAGTCT-3’;

  • Xenopus hnrnph1 mRNA, 5’- GGTTGGAAAATCGTGCCAAATG-3’ and 5’- GCCTTTTCAGCTATTTCCTGTGAAG-3’;

  • Xenopus rps14 mRNA, 5’- GTGACTGACCTGTCTGGCAA-3’ and 5’- GCAACATCTTGTGCAGCCAA-3’.

Proximity ligation assay

Request a detailed protocol

These experiments were carried out according to the manufacturer’s protocol (Sigma-Aldrich, Duolink Biosciences) using Duolink In Situ Detection reagents (Sigma-Aldrich, DUO90214 or DUO92008). After 24 hr, cultures were fixed in 2% formaldehyde/7.5% sucrose in PBS for 20 min at RT, washed three times in PBS with 0.001% Triton-X-100, permeabilized for 5 min in 0.1% Triton-X-100 in PBS, washed three times in PBS with 0.001% Triton-X-100, blocked with 5% heat-inactivated goat serum in PBS for 45 min at RT and subsequently incubated with primary antibodies overnight at 4°C. Primary antibodies were diluted at 1:100 for mouse anti-DCC (BD Biosciences, 554223), 1:100 mouse-anti-EphB2 (Thermo Fisher Scientific, 37–1700) 1:100 for rabbit anti-RPL5/uL18 (Proteintech, 15430–1-AP), 1:100 rabbit anti-RPS4X/eS4 (Proteintech, 14799–1-AP), 1:100 rabbit-anti RPL10A/uL1 (Proteintech, 16681–1-AP), 1:100 for rabbit anti-neuropilin-1 (Abcam, ab81321), 1:100 mouse anti-RPS3A/eS1 (Abcam, ab194670),1:100 mouse-anti-RPS23/uS12 (Abcam, ab57644), rabbit-anti-hnRNPA2B1 (Abcam, ab31645), rabbit-IgG isotype control (Abcam, ab37415), mouse IgG1 isotype control (MAB002, R and D Systems). After primary antibody incubation, dishes were washed twice for 5 min with 0.002% Triton X-100 in PBS and incubated with anti-rabbit-PLUS (Sigma-Aldrich, DUO92002) and anti-mouse-MINUS (Sigma-Aldrich, DUO92004) PLA probes for 1 hr at 37°C, with ligase for 30 min at 37°C and with the polymerase mix with red fluorescence for 100–140 min at 37°C. The samples were subsequently mounted with the mounting medium (DUO82040, Duolink) and imaged using a Nikon Eclipse TE2000-U inverted microscope equipped with an EMCCD camera. The number of discrete fluorescent puncta from randomly selected isolated growth cones were counted using Volocity software (Perkin Elmer).

Immunocytochemistry

Request a detailed protocol

After 24 hr, Xenopus retinal cultures were fixed in 2% formaldehyde/7,5% sucrose in PBS for 20 min at RT. For the puromycilation assay, 48 hr old cultures were used, eyes were manually removed and axons were treated with 10 µg/ml puromycin (Sigma-Aldrich, P8833) for 10 min at RT before fixation. The fixed cultures were then washed three times in PBS with 0.001% Triton-X-100, permeabilized for 5 min at RT in 0.1% Triton-X-100 in PBS, washed again for three time in PBS with 0.001% Triton-X-100 and blocked with 5% heat-inactivated goat serum in PBS for 45 min at RT. Primary antibodies were incubated overnight at 4°C, followed by Alexa Fluor-conjugated secondary antibodies for 60 min at RT in the dark. Cultures were mounted in FluorSave (Calbiochem, 345789). Primary antibodies were used at the following dilutions: 1:100 for mouse anti-DCC (BD Biosciences, 554223), 1:100 for rabbit anti-neuropilin-1 (Abcam, ab81321), 1:100 for mouse-anti-neuropilin-1 (Proteintech, 60067–1-Ig), 1:100 for rabbit anti-RPL5/uL18 (Proteintech, 15430–1-AP), 1:100 mouse anti-RPS3A/eS1 (Abcam, ab194670), 1:200 mouse-anti-puromycin-AlexaFluor-488 (Millipore, MABE343-AF488), rabbit-anti-Staufen1 (Abcam, ab73478), rabbit-anti-hnRNPA2B1 (Abcam, ab31645), 1:500 rabbit-anti-β-Catenin (Sigma-Aldrich, C2206), 1:500 rabbit-anti-hnRNPH1 (Abcam, ab154894), rabbit-anti-RPS14/uS11 (Abcam, ab174661), 1:250 rabbit-anti-pERK1/2 (Cell Signaling, 9101). Secondary antibodies were diluted at: 1:1000 goat anti-rabbit Alexa Fluor 568 (Abcam, ab150077), 1:1000 goat anti-mouse Alexa Fluor 568 (Abcam, ab150117).

Expansion microscopy

Request a detailed protocol

For expansion microscopy, RGCs explant cultures were immunostained with primary and secondary antibodies as described above, followed by applying the expansion protocol for cultured cells (Chen et al., 2015). Briefly, cultures were incubated in 0.25% glutaraldehyde in PBS for 20 min at RT and then washed with PBS three times, before adding monomer solution (2M NaCl, 8.625% (w/w) sodium acrylate, 2.5% (w/w) acrylamide, 0.1% (w/w) N,N‘-methylenebisacrylamide in PBS) for 2 min at RT. Subsequently, monomer solution was mixed with 0.2% ammonium persulfate (APS) and 0.2% Tetramethylethylendiamin (TEMED) and added to the samples. Gelation of the polymer occurred at 37°C for 30 min, followed by digestion of the samples with digestion buffer (40 mM Tris (pH 8), 1 mM EDTA, 0.5% Triton-X-100, 0.8M guanidine NaCl, 8 U/ml Proteinase K in water) and incubated at 37°C for 1 hr. To expand the samples, digestion buffer was removed and gels were placed in water for several hours during which water was replaced every 30 min. Once gels detached from the glass dish, they were transferred to a bigger dish to allow expansion. For imaging, expanded gels were cut in pieces and transferred to poly-L-lysine coated glass bottom dishes. Imaging was performed using a 60x/1.3 NA silicone oil objective lens on a Perkin Elmer Spinning Disk UltraVIEW ERS, Olympus IX81 inverted microscope and the Volocity software. Images were processed by using Fiji (NIH) and co-localisation analysis was carried out by using a purpose-written Matlab (The MathWorks) code. For co-localisation analysis, images were multiplied with a mask of a focused area of interest and the average background fluorescence was subtracted, before Pearson’s correlation coefficients were computed.

Quantification of immunofluorescence

Request a detailed protocol

For the quantification of fluorescence intensity, isolated growth cones were randomly selected with phase optics. For each experiment, the images were captured on the same day using the same gain and exposure settings and pixel saturation was avoided. Using Volocity software (Perkin Elmer), a region of interest (ROI) was defined by tracing the outline of each single growth cone using the phase image and the mean pixel intensity per unit area was measured in the fluorescent channel. The background fluorescence was measured in a ROI close to the growth cone that was free of debree or other axons and this was substracted from the mean fluorescence value of the growth cone. For the co-localization analysis of RBPs with receptors (Figure 2E–F), masks of the region of interest of each imaged growth cone were automatically generated using a code written in the wolfram language in Mathematica (https://wolfram.com/mathematica). For this code, training data was generated first by using hand traced outlines of 30 growth cones in two channel fluorescence images using ImageJ (http://imagej.net) to generate 30 corresponding binary growth cone maps. We chose the U-Net architecture (Ronneberger et al., 2015) to learn the growth cone segmentation similar as done in Jakobs et al. (2019). For training, we split the dataset into 25 training images and five validation images and down sampled every image so that the short dimension was 600 pixels long. During training input images were heavily augmented to prevent overfitting by (i) random cropping to 256 × 256 pixel sizes, (ii) random rotations, (iii) random reflections, (iv) random background gradients, (v) random noise, (vi) random nonlinear distortions. U-Net was with batch size eight and cross entropy loss until the validation loss did not decrease any further for 10 consecutive epochs on a nVidia 1080 Ti. The best performing network (using intersection over union benchmarking) was subsequently chosen to generate growth cone masks for our data. Masks were generated by first applying the best U-Net to the downsampled image followed by upsampling. The resulting output images were binarized by a morphological binarization algorithm with foreground threshold 0.3 that treats any pixel that is connected to the foreground and has a value larger than 0.2 also as part of the foreground.

Mass-spectrometry

Request a detailed protocol

1D gel bands were transferred into a 96-well PCR plate. The bands were cut into 1 mm2 pieces, destained, reduced (DTT) and alkylated (iodoacetamide) and subjected to enzymatic digestion with chymotrypsin overnight at 37°C. After digestion, the supernatant was pipetted into a sample vial and loaded onto an autosampler for automated LC-MS/MS analysis.

All LC-MS/MS experiments were performed using a Dionex Ultimate 3000 RSLC nanoUPLC (Thermo Fisher Scientific Inc, Waltham, MA, USA) system and a Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific Inc, Waltham, MA, USA). Separation of peptides was performed by reverse-phase chromatography at a flow rate of 300 nL/min and a Thermo Scientific reverse-phase nano Easy-spray column (Thermo Scientific PepMap C18, 2 μm particle size, 100A pore size, 75 μm i.d. x 50 cm length). Peptides were loaded onto a pre-column (Thermo Scientific PepMap 100 C18, 5 μm particle size, 100A pore size, 300 μm i.d. x 5 mm length) from the Ultimate 3000 autosampler with 0.1% formic acid for 3 min at a flow rate of 10 μL/min. After this period, the column valve was switched to allow elution of peptides from the pre-column onto the analytical column. Solvent A was water + 0.1% formic acid and solvent B was 80% acetonitrile, 20% water + 0.1% formic acid. The linear gradient employed was 2–40% B in 30 min.

The LC eluant was sprayed into the mass spectrometer by means of an Easy-Spray source (Thermo Fisher Scientific Inc). All m/z values of eluting ions were measured in an Orbitrap mass analyzer, set at a resolution of 70000 and was scanned between m/z 380–1500. Data-dependent scans (Top 20) were employed to automatically isolate and generate fragment ions by higher energy collisional dissociation (HCD, NCE:25%) in the HCD collision cell and measurement of the resulting fragment ions was performed in the Orbitrap analyser, set at a resolution of 17500. Singly charged ions and ions with unassigned charge states were excluded from being selected for MS/MS and a dynamic exclusion window of 20 s was employed.

Raw data were processed using Maxquant (version 1.6.1.0) (Cox and Mann, 2008) with default settings. MS/MS spectra were searched against the X. laevis protein sequences from Xenbase (xlaevisProtein.fasta). Enzyme specificity was set to trypsin/P, allowing a maximum of two missed cleavages. The minimal peptide length allowed was set to seven amino acids. Global false discovery rates for peptide and protein identification were set to 1%. The match-between runs option was enabled.

Label-free quantification (LFQ) analysis of proteomics data

Request a detailed protocol

To identify significant interactors, t-test-based statistics were applied on label-free quantification (LFQ) intensity values were performed using Perseus software. Briefly, LFQ intensity values were logarithmized (log2) and missing values were imputed based on the normal distribution (width = 0.3, shift = 1.8). Significant interactors of DCC or Nrp1 pulldowns compared to IgG pulldowns were determined using a two-tailed t-test with correction for multiple testing using a permutation-based false discovery rate (FDR) method.

RNA-sequencing

Request a detailed protocol

RNA was isolated from immunoprecipitated samples from SH-SY5Y cells as described above using RLT buffer (Qiagen) containing β-mercaptoethanol and the RNeasy Mini kit (Qiagen) followed by in-column DNase I treatment to remove genomic DNA contamination. RNA quality was analysed using Agilent RNA 6000 Pico kit and reagents (Agilent, 5067–1514,1535,1513) on a Agilent 2100 Bioanalyzer (Agilent). cDNA was then amplified using a method developed for single cell transcriptomics (Tang et al., 2009) with minor modifications (Shigeoka et al., 2016). The cDNA library preparation was performed using a KAPA Hyperprep kit (Roche) and cDNA libraries were subjected to a RNA-sequencing run on a Next-seq 500 instrument (Illumina) using the 150 cycles high output kit (Illumina).

Bioinformatic analysis of RNA-sequencing data

Request a detailed protocol

The sequence reads were mapped using HISAT 2 version 2.1.0, and FPKM values were estimated using Cufflinks version 2.2.1. Read counts for each gene were determined using HTSeq version 0.11.0. Differential expression analysis was performed using edgeR in R version 3.5.0 (FDR < 0.05). The GO enrichment analysis was performed using topGO version 2.32.0. The mRNA targets of RBPs were obtained from previously published studies as listed in the main text. To analyse the enrichment of Staufen1 and hnRNPA2B1 targets, all RBP targets that showed a significant difference between DCC and Nrp1 pulldowns were first selected and the log2 fold change values between DCC and Nrp1 were used for a Mann-Whitney U test (Wilcoxon rank sum test).

Electron microscopy of axonal growth cones

Request a detailed protocol

Cultured neurons were fixed at 37°C for 45 min in 2.5% glutaraldehyde, sodium cacodylate buffer 0.1M pH7.4 containing 2 mM CaCl2 and 2 mM MgCl2. Samples were post-fixed for 15 min at RT in 1% osmium and embedded in epoxy resin. Ultrathin sections were imaged with a ZEISS EM 912 microscope. Ribosomes were identified based on size and shape. To quantify the inter-ribosome distance, the center-to-center distance was measured using ImageJ. For axonal growth cones, ribosomes were selected that were located within 50 nm of the plasma membrane and the distance to its closest neighbor was quantified.

Statistical analysis

Request a detailed protocol

All experiments were performed in at least three independent biological replicates unless explicitly stated otherwise. The order of data collection was randomized, and no data were excluded from analysis. Statistical analysis was performed using GraphPad Prism, R or MATLAB. Statistical tests used are described in the figure legends.

Data availability

Request a detailed protocol

RNA-sequencing data associated with this manuscript has been deposited on the GEO database (identifier GSE135338). All proteomics data associated with this manuscript has been uploaded to the PRIDE online repository (identifier: PXD015650).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
    Normal Table of Xenopus Laevis (Daudin): a Systematical and Chronological Survey of the Development From the Fertilized Egg Till the End of Metamorphosis
    1. PD Nieuwkoop
    2. J Faber
    (1994)
    Garland Pub.
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
    Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015
    1. O Ronneberger
    2. P Fischer
    3. T Brox
    (2015)
    U-Net: Convolutional Networks for Biomedical Image Segmentation, Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015, Springer.
  44. 44
  45. 45
  46. 46
  47. 47
    On-site ribosome remodeling by locally synthesized ribosomal proteins in axons
    1. T Shigeoka
    2. M Koppers
    3. H-w Wong
    4. JQ Lin
    5. R Cagnetta
    6. A Dwivedy
    7. J de Freitas Nascimento
    8. F van Tartwijk
    9. F Ströhl
    10. J-M Cioni
    11. J Schaeffer
    12. M Carrington
    13. CF Kaminski
    14. H Jung
    15. WA Harris
    16. CE Holt
    (2019)
    Cell Reports In press.
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59

Decision letter

  1. Carol A Mason
    Reviewing Editor; Columbia University, United States
  2. Catherine Dulac
    Senior Editor; Harvard University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Your paper is a welcome follow-up to John Flanagan's previous study that showed that ribosome and mRNA-protein complexes assemble at the cytoplasmic tail of DCC, to provide translational specificity upon ligand binding. Your demonstration that unique RBPs and mRNAs associate with Neuropilin1, Robo2 and DCC but not EphB2 expands upon this process. The receptor-specific interactomes you describe provide an excellent model for the control of cue-specific local translation and will be appreciated by those who work on molecular mechanisms of the receptors and ligands formerly thought to function strictly in axonal growth and guidance but are now implicated in synaptogenesis, vasculogenesis and tumor regulation.

Decision letter after peer review:

[Editors’ note: this article was originally rejected after discussions between the reviewers, but the authors were invited to resubmit after an appeal against the decision.]

Thank you for submitting your work entitled "Receptor-specific interactome as a hub for rapid cue-induced selective translation in axons" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation amongst the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work is not acceptable for publication in eLife at this time.

All three reviewers had high praise for this study done in collaboration with John Flanagan's group, that ribosomes can couple with surface receptors in subcellular compartments, and that specific RBPs recruit distinct subsets of mRNAs to these receptors, to provide a potential mechanism of how specificity of translation is achieved following an extracellular cue. That the Semaphorin receptor Neuropilin-1 in addition to DCC, but not EphB2, is associated with ribosomal proteins and RNAs, as well as with RNA-binding proteins and mRNAs, is welcome news. The work is exciting and we applaud your execution of exceedingly difficult analyses.

However, as you can read in the appended reviews, the reviewers were critical of a number of aspects and call for amendments, several of them non-overlapping, that should be addressed before publication. The key categories of requested amendments include:

- You report a vast number of mRNAs associated with DCC and Nrp1. In the consultation session among reviewers, and from reviewer 2's critique, they wonder if there is a distinction between the dominant mRNAs and others at low levels, questioning the relevance of the latter. Reporting the raw data/data files for the proteomics and RNA-seq experiments would be essential in your revision, for the reader to have access to the full list.

- The authors assume that the binding of guidance cues to their respective receptors leads to dissociation of the ribosomes and subsequent local translation. They ask for evidence that this is the case and that the transcripts are translated into proteins.

In line with this critique, reviewer 3 states that assuming that the RNAseA/T1 treatment effectively shows that mRNA/protein interactions are needed for the ribosome subunits to interact with DCC and Nrp1, they wondered whether you could determine whether RNA is needed for the RBP associations with these receptors, and similarly whether ribosome subunits association is needed for mRNA interaction with DCC and Nrp1.

Reviewer 2 queries whether the reported association takes place on vesicles (either during the transport from the Golgi or upon endocytosis) as per the Holt lab's study 'Late Endosomes Act as mRNA Translation Platforms' (Cioni et al., 2019).

- While reviewer 1 commends you on the striking and convincing EM images of ribosomes within growth cones, one reviewer in the consultation session wondered whether your point on ER and Golgi association could be sorted out by FISH and that including such an analysis for at least a few key mRNAs would be helpful.

- Finally, the reviewers all had comments on rigor/quantification for replication's sake: several figures lack unbiased analyses and biological replication, and there are numerous single, non-repeated experiments.

Reviewer #1:

In this interesting study the authors provide evidences that ribosomes can couple with surface receptors in subcellular compartments, allowing a tight spatiotemporal control of local translation in response to extracellular stimuli. The authors report that this is indeed not a limited phenomenon and show that several receptors whose activation leads to local translation are coupled with ribosomes. Moreover, specific RBPs recruit distinct subsets of mRNAs to these receptors, thereby providing a potential mechanism explaining how specificity of translation is achieved following an extracellular cue. The study is novel and interesting, and most experiments are well controlled. However, there are some issues that need to be addressed before this would be suitable for publication.

- One of the main findings of this study is that specific populations of mRNAs are targeted to different receptors through interactions with different RBPs. However, it is unclear whether these interactions are maintained in the absence of stimulation or during stimulation paradigms that do not lead to ribosomal dissociation, such as EphrinA1+Netrin1 stimulation of the DCC receptor.

- The authors suggest that the binding of guidance cues to their respective receptors leads to dissociation of the ribosomes and subsequent local translation. They should provide evidence that this is the case and that the transcripts are translate into proteins.

- The controls for the co-sedimentation polysome experiments in Figure 2—figure supplement 1 are weak. EDTA slightly reduces the levels of receptors identified in the heavier fractions, but this may be due to different exposure, as there seems to be an overall decrease in total levels of protein.

- For Figures 2B and 2C the authors should show the input levels for Rps3A and Rps26 to demonstrate that the treatment leading to ribosomal dissociation doesn't induce an overall reduction of these components, especially given that the binding with the receptor is quite limited.

Reviewer #2:

In this manuscript, the authors follow up on a 2010 paper by the Flanagan group (Tcherkezian et al.) that reported the association of ribosomal subunits with the intracellular end of the Netrin-1 receptor DCC. Here, Koppers et al. describe that not only DCC but also the Semaphorin receptor Neuropilin-1, but not EphB2 are associated with ribosomal proteins and RNAs, as well as with RNA-binding proteins and mRNAs. The ribosomes and RNAs dissociate from the receptors upon stimulation with Netrin-1 or Sem3A, respectively, and in the case of DCC, this effect is blocked by co-stimulation with EphrinA1. Together, the authors propose that receptor-specific interactomes provide a model for the control of cue-specific local translation.

The findings reported in this manuscript would be of interest to neurobiologists and would represent a significant advance in the understanding of the control of local protein synthesis in developing axons. However, several key aspects of the manuscript are currently underdeveloped, there are unaddressed conceptual questions, and several figures are problematic due to lacking unbiased analyses and biological replications.

Conceptual questions:

The authors find an astonishingly great number of mRNAs associated with DCC and Nrp1. Just the differentially associated mRNAs are 158 and 383, respectively. The total number is not even reported. How would these numbers be possibly compatible with the proposed model of 'selective' control of local translation? Does every receptor associate with several hundred mRNAs and all of them get translated upon ligand binding? If conversely every receptor is associated with just a couple of mRNA species, how can this model ensure reliable translational responses to ligand binding?

The authors focus entirely on DCC and Nrp1 on the cell surface. Would it not be possible that the reported association takes place on vesicles (either during the transport from the Golgi or upon endocytosis)? After all, the same group just published that 'Late Endosomes Act as mRNA Translation Platforms' (Cioni et al., 2019). The finding of vesicle-mediated transport as a category for interactors of DCC and Nrp1 seems to support this alternative idea.

Translational targets for Netrin-1 and Sema3A have been described by the authors and others but are being ignored in this manuscript. Are the mRNAs coding for β-actin, Par-3, Tctp associated with DCC, or is RhoA mRNA associated with Nrp-1? If they are not part of the interactomes, what does this mean for the proposed model?

Other major points in order of occurrence:

Figure 1G-J: The presented western blots are single experiments without replication and are not quantified. The presentation of single, non-repeated experiments is a recurrent problem in this manuscript.

Figure 1—figure supplement 1F, G: RPL39 is mentioned in the main text, but does not appear in the figure. The variances in this figure are quite large, and as the authors state, the conclusion is preliminary. The authors should either perform more experiments to elevate the conclusion from preliminary or remove this entire figure/chapter.

Figures 2—figure supplement 1A, B: Single, unrepeated experiments without a proper quantification. This is especially troublesome, as the purported shift for DCC upon EDTA treatment is not obvious at all.

Figure 2A: The RNAseA/T1 treatment is an elegant approach but to follow the authors interpretation it would be critical to experimentally prove that the rRNAs stay intact.

Figures 2D, E: The author present the same plots as in Figure 1A without mentioning this fact. Also, there is no congruency between the main text, the figure legend and the actual figure: where are the 10 shared RBPs?

Figure 2—figure supplement 1C, D: Single unrepeated, non-quantified experiments.

Figures 3A, B: Single growth cones are presented without replication, quantification or any unbiased co-localization analysis.

Figure 3E: Why did the authors normalize the data instead of presenting the actual counts per sq. mm?

Figure 3F: The authors are correct by stating that association of DCC with RPL5 and RPS4X is significantly decreased, however, they ignore the vastly different effect sizes. RPL5 seems reduced by ~80% while RPS4X is reduced by less than 20%. What is the explanation for this and shouldn't the reduction of the ribosomal subunits be stoichiometrically matched to form functioning ribosomes? The authors argue that local production of RPLs might obscure their results; they should perform the experiments in the presence of CHX to excluded this potentially confounding possibility.

Figure 4A: It is suprising to see that Netrin-1 does not induce local translation as measured by puromycylation. How do the author reconcile this finding with previous reports that Netrin-1 induces local translation?

Figure 5A: A single image without biological repeats or (unbiased) quantification.

Reviewer #3:

This manuscript follows up on previous work from the Flanagan lab showing that ribosome and mRNA-protein complexes assemble at the cytoplasmic tail of DCC, to provide translational specificity upon ligand binding. Koppers et al. show that unique RBPs and mRNAs associate with Neuropilin 1, Robo2 and DCC but not EphB2. This brings an appealing mechanism to modulate specificity of translational regulation in response to extracellular stimuli, and the data are strengthened by lack of RNA and ribosome association EphB2 fits since its output is reportedly protein synthesis independent. Overall, the work is very well done, the data are provocative, and I think this represents an advance appropriate for eLife. However, there are some weaknesses that detract from my enthusiasm.

1) The differential association of specific RPs with DCC vs. Neuropilin-1 is intriguing and distinct ribosome populations have been suggested in a number of systems but unequivocally proven. Some of the techniques used here and methodologies rely on those in the Shigeoka et al., 2018 reference from the authors. Particularly, the experiments with RNAseA/T1 in Figure 2. The reference there is incomplete and I do not see it has been published except for bioRxiv (I am not certain on eLife's policy for referencing unreviewed work like this).

2) Assuming the RNAseA/T1 treatment effectively shows that mRNA/protein interactions are needed for the ribosome subunits to interact with DCC and Nrp1, it is surprising that no efforts are made to determine whether RNA is needed for the RBP associations with these receptors. Similarly whether ribosome subunits association is needed for mRNA interaction with DCC and Nrp1.

3) The MS data for different RPs in Figure 1—figure supplement 1 is surprising and does support the authors' suggestion for different ribosome populations. It is surprising that the authors did not validate those showing greatest differentials (e.g., L4, L35, S6 and S9). Also, it seems like there are more 60S than 40S RPs showing differential association, but the authors did not comment on this point.

4) For Figure 2—figure supplement 1A-B, the majority of Nrp1 and DCC signals do not fractionate with subunits. Granted, these are very hard experiments and the EDTA treatments do shifts signals for Nrp1 and DCC upwards in the gradients. I think this would be strengthened EDTA treatments and some quantitation for the colocalization data shown in Figure 3 and Figure 3—figure supplement 1.

5) It is surprising that no colocalization for mRNAs with the receptors is provided beyond the co-IP, and such RIP experiments are known to have artefacts from in solution RNA protein-interactions after lysis.

6) For Figure 5A, the EM is compelling and shows a surprisingly high number of electron dense ribosome like structures. The authors do not note whether this is a stimulated or naïve culture. Comparison of the two would be informative and strengthen the authors' conclusions (though I do recognize the difficulty in this request).

7) Robo3.3 and EphA2 have been shown to be translated in axons. This is probably worth mentioning in the Discussion since the authors show that EphB2 does not use this mechanism for sequestering RNPs and ribosome-bound mRNAs.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "Receptor-specific interactome as a hub for rapid cue-induced selective translation in axons" for further consideration by eLife. Your revised article has been evaluated by Catherine Dulac as the Senior Editor, and a Reviewing Editor.

The manuscript has been improved and should be a welcome addition to the work of your laboratory and John Flanagan's group on local translation, in particular, that ribosomes can couple with specific RBPs that in turn recruit distinct subsets of mRNAs, to provide a potential mechanism of how specificity of translation is achieved following an extracellular cue.

Two of the reviewers found your manuscript satisfactorily amended, but one of the reviewers cited a number of points that should be addressed before publication, all of them addressed by textual amendments:

1) There are a number of places where your results are somewhat overstated and should be toned down:

a) Abstract: 'Our findings.… provide a general model for the rapid, localized and selective control of cue-induced translation.' Compare this to the much more accurate and measured summary: '… this study provides evidence…and suggests…'. Your results are interesting and represent a significant advance, but it may be premature to consider that they comprise a general model.

b) The findings from Figure 2B, C are summarized: '…these results suggest that the interaction.… is likely mediated through mRNA'. One page later, this sentence has morphed into the definitive statement: '…our finding that mRNA mediates the association of receptors with specific ribosomes…'

c) '…reveal that multiple receptors.… can associate with ribosomes.' This statement is imprecise; the results up to this point show association with ribosomal subunits, not ribosomes.

Please carefully edit the Results section and relegate the interpretation especially statements regarding the general impact/significance, to the Discussion.

2) A major concern in the first review was that some experiments were not repeated and quantified. Most of these issues have been fixed, but in Figure 1C-E, Figure 1—figure supplement 1A-C, results are still presented without quantification. The sentence in the figure legend that these experiments have been repeated three times but the actual results should be presented.

3) Figure 2—figure supplement 1C: the n for these experiments are 2 or 1. Showing bar graphs with SDs is not very meaningful for such low n numbers. The individual data points should be shown instead.

4) Subsection “DCC and Nrp1 bind to specific subsets of mRNAs”, first paragraph: The finding that high abundance mRNAs are much more differentially associated than low abundance mRNAs seems to indicate that these mRNAs are the mRNAs that conform to the authors proposed model. It is a missed opportunity to not talk more about these mRNAs: provide numbers, identity, GO analysis.

5) In response to conceptual question 2 (reviewer 2, first review), you provide the interesting finding that endocytosis is required for the dissociation between DCC and RP. However, the question was in relation to the mRNA-seq findings: it is my understanding that the RNA-seq experiments cannot distinguish between mRNAs associated with the receptors at the membrane vs. receptors at vesicles. For example, if mRNAs/RNPs are associated receptors on vesicles that transport the receptors into axons, these mRNAs would show up in the RNA-seq, even if they no longer associate with receptors after insertion into the plasma membrane. Please discuss this point to make clear that the RNA-seq results might include mRNAs that are not associated with receptors in the membrane.

6) The response to the comment regarding Figure 3F is unsatisfactory: You seem to indicate that the data as presented are unreliable due to problems with the reagents. Expand on this point.

https://doi.org/10.7554/eLife.48718.sa1

Author response

[Editors’ note: the author responses to the first round of peer review follow.]

[…] As you can read in the appended reviews, the reviewers were critical of a number of aspects and call for amendments, several of them non-overlapping, that should be addressed before publication. The key categories of requested amendments include:

- You report a vast number of mRNAs associated with DCC and Nrp1. In the Consultation session among reviewers, and from reviewer 2's critique, they wonder if there is a distinction between the dominant mRNAs and others at low levels, questioning the relevance of the latter. Reporting the raw data / data files for the proteomics and RNA-seq experiments would be essential in your revision, for the reader to have access to the full list.

We have deposited the (raw) data files from our RNA-sequencing experiments on the GEO database (GSE135338). The proteomics raw data files have been deposited on PRIDE (PXD015650). This information has been added to the Data availability section.

The reviewers raise a good point about dominant mRNAs. Therefore, we have performed further bioinformatic analysis to look at any differences between dominant versus low abundant mRNAs. This showed that the majority (67.9%) of high abundant mRNAs (FPKM > 1000) are differential mRNAs between DCC and Nrp1 while less than 5% of low abundant mRNAs (FPKM 1-10) are differential. This is interesting as it indicates that different receptors tend to associate differentially with more dominant mRNAs and suggests that these mRNAs are most relevant for cue-induced selective translation. We have included this in the subsection “DCC and Nrp1 bind to specific subsets of mRNAs”.

- The authors assume that the binding of guidance cues to their respective receptors leads to dissociation of the ribosomes and subsequent local translation. They ask for evidence that this is the case and that the transcripts are translated into proteins.

We find this comment puzzling because we did provide direct evidence for both the cue-induced receptor-ribosome dissociation and subsequent translation in axons in the original submitted manuscript.

Original Figure 3 shows that cue stimulation leads to receptor-ribosome dissociation for both DCC and Nrp1 in axonal growth cones, using proximity-based assays (Figure 3H-I). An important further question, one which the reviewers did not raise but is related to this issue, is to ask whether the receptor-ribosome dissociation is cue-specific. We have now conducted a new set of experiments to investigate this and we find that the dissociation is, indeed, highly cue-specific: Netrin-1 causes dissociation of ribosomal proteins (RPs) from DCC, but not Nrp1, whereas Sema3A causes dissociation of RPs from Nrp1, but not DCC. We have added these new data in Figure 3J-K and related text in the third paragraph of the subsection “Dissociation of ribosomes from receptors is triggered by extrinsic cues and requires endocytosis”.

Regarding the second point, whether receptor-associated mRNAs are locally translated into proteins on stimulation, we provided direct evidence in the original manuscript that this, indeed, is the case.

Figure 4E-H shows the validation of the protein changes in response to Netrin-1 but not Sema3A, for two corresponding mRNAs (ctnnb1 and hnrnph1) that are identified as significantly enriched after DCC versus Nrp1 pulldown in our RNA-seq dataset, consistent with our previously reported cue-induced nascent proteome data (Cagnetta et al., 2018). In addition, Figure 4—figure supplement 1C-E shows the association of another mRNA – rps14 – with DCC in Xenopus brains and the cue-specific protein increase of its corresponding protein in response to Netrin-1 but not Sema3A, again consistent with our previous proteomic data (Cagnetta et al., 2018). We have now made the text clearer (subsection “Integration of multiple cues can affect the cue-induced selective translation of 332 receptor-specific mRNAs”, third paragraph) stating that β-catenin, hnRNP-H1 and RPS14 have been detected in our recent axonal nascent proteome study to be selectively translated in response to Netrin-1 but not Sema3A (Cagnetta et al., 2018), and we also specified that their corresponding mRNAs have been detected in retinal axons by RNA-seq (Shigeoka et al., 2018). We hope that this clarifies any misunderstanding.

In line with this critique, reviewer 3 states that assuming that the RNAseA/T1 treatment effectively shows that mRNA/protein interactions are needed for the ribosome subunits to interact with DCC and Nrp1, he/she wondered whether you could determine whether RNA is needed for the RBP associations with these receptors, and similarly whether ribosome subunits association is needed for mRNA interaction with DCC and Nrp1.

To address these questions, we have now quantified the effect of RNAseA/T1 treatment on the binding between Nrp1 and the RNA-binding protein (RBP), Staufen1. The analysis shows a small, but significant decrease in Staufen1 binding to Nrp1 (Figure 2—figure supplement 1E). The decrease is smaller than that seen for ribosomal proteins but, nonetheless, suggests that mRNA can contribute to stabilizing the interaction between receptors and RBPs.

Additionally, to show our treatments work effectively, we have now added data from sucrose density gradients showing that our RNAseA/T1 treatment effectively cleaves unprotected mRNA but leaves rRNA intact (Figure 2—figure supplement 1D and subsection “Guidance cue receptors associate with ribosomes in a mRNA-dependent manner”, last paragraph).

The question about whether ribosome subunit association is needed for the association of mRNA with receptors is difficult to answer experimentally because of the lack of a clean way to dissociate ribosome subunits. We have tried to address this question by performing qPCR quantification of β-actin mRNA after DCC pulldown with and without EDTA treatment (which decreases the association of ribosome subunits with DCC). EDTA treatment resulted in a loss of β-actin mRNA binding to DCC suggesting that ribosome binding is needed for the mRNA interaction with DCC. Because the EDTA treatment may have affected the mRNA-DCC interaction directly or the DCC-RBP interaction it seems hard to draw a definitive conclusion from this experiment so we have not included it in the current manuscript, although we could if the reviewer and editor feel this is important.

Reviewer 2 queries whether the reported association takes place on vesicles (either during the transport from the Golgi or upon endocytosis) as per the Holt lab's study 'Late Endosomes Act as mRNA Translation Platforms' (Cioni et al., 2019).

This is an excellent question. We argue that the enrichment of the functional group “vesicle-mediated transport” associated to the receptors (Figure 1B), highlighted by the reviewer, may likely serve to mediate the endocytosis of the receptor upon cue stimulation (e.g. Nrp1 endocytosis upon Sema3A stimulation; (Zylbersztejn et al., 2012). Netrin-1 and Sema3A are indeed well known to cause rapid (1-2 min) endocytosis of their respective receptors, DCC and Nrp1, and inhibition of such endocytosis blocks local protein synthesis in response to ligand (Konopacki et al., 2016; Piper et al., 2005). Therefore, it is possible that the dissociation of the ribosome from the receptor also depends on endocytosis. For this reason, we have now performed new experiments to assess whether endocytosis is required for cue-induced receptor-ribosome dissociation by using Dynasore, an endocytosis inhibitor. The results show that inhibition of endocytosis completely blocks the cue-induced DCC-RP dissociation (now Figure 3L).

Furthermore, our focus in this manuscript concerns the role of receptor-ribosome interactions in responses to extracellularly applied ligand in vitro, leading to a protein synthesis response. Guidance molecules are also present extracellularly in vivo along the retinotectal pathway, therefore, we reason that the interaction between the ligand and the receptor is likely to occur at the cell surface.

In addition to this, we have performed new analyses on EM images obtained from axonal growth cones which show that 20 out of 22 axonal growth cones contain rows of ribosomes aligned just under the plasma membrane (within 50nm), again supportive of our model (subsection “Receptor-ribosome coupling occurs in RGC axonal growth cones”). We have added these new data and analyses in Figure 3F, 3G and Figure 3—figure supplement 1C-E.

- While reviewer 1 commends you on the striking and convincing EM images of ribosomes within growth cones, one reviewer in the Consultation session wondered whether your point on ER and Golgi association could be sorted out by FISH and that including such an analysis for at least a few key mRNAs would be helpful.

We are pleased that reviewer 1 liked our EM image of ribosomes in growth cones. To address the concern raised here, we have now added additional EM images as well as new quantitative analysis on these EM images (Figure 3F-G and Figure 3—figure supplement 1C-E) which are discussed in more detail below.

We are not sure exactly what is being referred to here regarding “…your point on ER and Golgi association…” since we do not mention ER or Golgi in our manuscript. The possible involvement of ER and Golgi in local translation in axons is currently controversial as ultrastructural evidence of canonical rough (ribosome-containing) ER or Golgi in axons is unclear (Gonzalez et al., 2018). Furthermore, it is important to emphasize that what we are focusing on here is evidence that different extracellular cues regulate the interaction of their specific receptors with ribosomes, and the subsequent translation of different sets of associated mRNAs. Therefore, while it is an interesting possibility that ER and Golgi components in axons are, indeed, involved in local translation, we believe that investigating this properly would constitute a further full study on its own and argue that this is beyond the scope of the current study. In the meantime, we prefer to refrain from speculations on this topic.

- Finally, the reviewers all had comments on rigor/quantification for replication's sake: several figures lack unbiased analyses and biological replication, and there are numerous single, non-repeated experiments.

We sincerely apologize for the omission in the presentation of the original manuscript. All of the experiments were repeated several times with biological replicates. As discussed in more detail below, where suitable we have also added unbiased analyses. We have now clarified this in the figure legends and Materials and methods section.

Reviewer #1:

[…]

- One of the main findings of this study is that specific populations of mRNAs are targeted to different receptors through interactions with different RBPs. However, it is unclear whether these interactions are maintained in the absence of stimulation or during stimulation paradigms that do not lead to ribosomal dissociation, such as EphrinA1+Netrin1 stimulation of the DCC receptor.

The reviewer raises an interesting point, the initial receptor-IP-MS was done on whole brains/eyes where the activation state of the receptors is unknown. For the RNA-seq experiments, we used a human neuronal cell line, showing that Nrp1 and DCC associate with different mRNAs in the absence of stimulation (Figure 2G, Figure 2—figure supplement 1F-I). Unfortunately, it is not possible to perform receptor-IPs, followed by RNA-seq analysis, on axons in response to different ligands because of the limited amount of axon-only material.

While the use of different systems was mentioned in the original manuscript, we have now clarified these points further in the subsections “DCC and Nrp1 bind to specific subsets of mRNAs” and “Integration of multiple cues can affect the cue-induced selective translation of receptor-specific mRNAs”.

- The authors suggest that the binding of guidance cues to their respective receptors leads to dissociation of the ribosomes and subsequent local translation. They should provide evidence that this is the case and that the transcripts are translate into proteins.

We were puzzled by this comment as we provided direct evidence that cue stimulation leads to receptor-ribosome dissociation in axonal growth cones (Figure 3H-I). The same conclusion was previously reported for DCC (Tcherkezian et al., 2010). We have now also added new control experiments showing that stimulation with the receptor-specific guidance cue is needed for receptor-ribosome dissociation, highlighting the specificity of these events. We have added these new data in Figure 3J-K and in the third paragraph of the subsection “Dissociation of ribosomes from receptors is triggered by extrinsic cues and requires endocytosis”.

In Figure 4E-H and Figure 4—figure supplement 1D-E we show that hnRNPH1, β-catenin (whose mRNAs were identified as significantly enriched after DCC versus Nrp1 pulldown in our receptor IP-RNA-seq analysis) and RPS14 increase upon Netrin-1, but not Sema3A, stimulation. Importantly, this is consistent with our previous proteomic-based study which showed that hnRNPH1, β-catenin and RPS14 are axonally translated in response to Netrin-1 but not Sema3A stimulation (Cagnetta et al., 2018), and we previously found that their transcripts are present in RGC axons (Shigeoka et al., 2018). Therefore, we do provide evidence that guidance cue stimulation leads to receptor-ribosome dissociation and that receptor-bound mRNA transcripts are translated into proteins after cue stimulation in axons. We have now made this clearer in the subsection “Integration of multiple cues can affect the cue-induced selective translation of receptor-specific mRNAs”.

- The controls for the co-sedimentation polysome experiments in Figure 2—figure supplement 1 are weak. EDTA slightly reduces the levels of receptors identified in the heavier fractions, but this may be due to different exposure, as there seems to be an overall decrease in total levels of protein.

We performed these experiments to provide further evidence in support of the association of receptors with ribosomes. This is in addition to evidence that we provide by: i) IP followed by Western blot and mass spectrometry (Figure 1A-F), ii) IP followed by qPCR (Figure 1G-J), iii) expansion microscopy imaging (Figure 3A-B), iv) proximity ligation assay (PLA) (Figure 3C-D). It is noteworthy that these polysome EDTA experiments were also performed for DCC in Tcherkezian et al., 2010, and our results are consistent with their findings. Furthermore, the results shown in Figure 2A-C strongly support that EDTA treatment results in loss of association of receptors to ribosomes, consistent with these polysome experiments.

Care was taken to use the same amount of starting material for these experiments to rule out differences in total protein levels. Control and EDTA treated samples were compared directly, twice for DCC and once for Nrp1. We then quantified the relative amounts of DCC (n = 2 for control and EDTA) and Nrp1 (n = 2 for control, n = 1 for EDTA) in ribosome-free and ribosome-containing fractions. This relative quantification corrects for any possible decrease in overall protein levels. These quantifications show a clear trend towards a shift to lighter fractions (i.e. ribosome-free fractions) for both DCC and Nrp1 after EDTA treatment (Figure 2—figure supplement 1C and subsection “Guidance cue receptors associate with ribosomes in a mRNA-dependent manner”). We hope that the reviewer appreciates that these data further support the association of DCC and Nrp1 with ribosomes. Lastly, to improve the presentation of the figure we have now aligned the UV profiles with the corresponding Western blot (Figure 2—figure supplement 1A-B).

- For Figures 2B and 2C the authors should show the input levels for Rps3A and Rps26 to demonstrate that the treatment leading to ribosomal dissociation doesn't induce an overall reduction of these components, especially given that the binding with the receptor is quite limited.

In these experiments, all the conditions come from the same starting material and thus have the same input levels. The samples were treated with EDTA and RNAseA/T1 only after pulldown so the results cannot be explained by an overall reduction in components. We followed this procedure based on Simsek et al., 2017, where they examine the RNA dependency of ribosomal protein interactors using this method. The fact that the pulled-down proteins (i.e. DCC and Nrp1) do not change is a good control with this method and quantification was corrected for the amount of pulled-down protein. We apologize for not explaining this clearly in the original manuscript and we have now clarified this in subsection “Guidance cue receptors associate with ribosomes in a mRNA-dependent manner” and the Materials and methods section.

Reviewer #2:

[…]

Conceptual questions:

The authors find an astonishingly great number of mRNAs associated with DCC and Nrp1. Just the differentially associated mRNAs are 158 and 383, respectively. The total number is not even reported. How would these numbers be possibly compatible with the proposed model of 'selective' control of local translation? Does every receptor associate with several hundred mRNAs and all of them get translated upon ligand binding? If conversely every receptor is associated with just a couple of mRNA species, how can this model ensure reliable translational responses to ligand binding?

We have now deposited the (raw) data files from our RNA-sequencing experiments on the GEO database (GSE135338).

We detected several thousand mRNAs in the human cell line SH-SY5Y, with 100-400 that are differential between DCC and Nrp1. Cagnetta et al., 2018, showed that the translation of more than 100 mRNAs is regulated within 5 min in Xenopus retinal axons in response to Netrin and Sema3A, therefore, the numbers that we find in this manuscript do not seem very surprising, although differences in specific mRNAs or number of mRNAs between the two different systems may exist. This is exemplified by the absence of rps14 mRNA enrichment in SH-SY5Y cells, which was detected in Xenopus brain (Figure 4—figure supplement 1C).

More broadly, the conceptual question of how specificity of translation is achieved when there are so many mRNAs associated with each receptor is an interesting one. This question applies to many studies in the field, which have detected a large number of mRNAs in the axon. The answer may, in part, be that different mRNAs exist in common complexes (Buxbaum et al., 2015), but it clearly does not appear to be a simple one-receptor one-message pathway. It seems more probable that every receptor molecule exposed to the ligand has a certain probability of triggering the translation of a few associated mRNAs, and this could explain a consistent response of hundreds of mRNAs being translated when a particular ligand is added. In addition, the receptor-associated mRNAs were obtained from a cell line, as stated clearly in the original manuscript, so the number of receptor-associated mRNAs in axons may be different. Finally, it is interesting to speculate that a further fraction of the mRNAs detected may play a structural role, as recently reported in (Crerar et al., 2019). We have now addressed this conceptual point in the fourth paragraph of the Discussion.

The authors focus entirely on DCC and Nrp1 on the cell surface. Would it not be possible that the reported association takes place on vesicles (either during the transport from the Golgi or upon endocytosis)? After all, the same group just published that 'Late Endosomes Act as mRNA Translation Platforms' (Cioni et al., 2019). The finding of vesicle-mediated transport as a category for interactors of DCC and Nrp1 seems to support this alternative idea.

We thank the reviewer for raising this point. We argue that the enrichment of the functional group “vesicle-mediated transport” associated to the receptors (Figure 1B), highlighted by the reviewer, may likely serve to mediating the endocytosis of the receptor upon cue stimulation (e.g. Nrp1 endocytosis upon Sema3A stimulation; Zylbersztejn et al., 2012). Netrin-1 and Sema3A are indeed well known to cause rapid (1-2 min) endocytosis of their respective receptors, DCC and Nrp1, and inhibition of such endocytosis blocks local protein synthesis in response to ligand (Konopacki et al., 2016; Piper et al., 2005). Therefore, it is possible that the dissociation of the ribosome from the receptor also depends on endocytosis. For this reason, we have now performed new experiments to assess whether endocytosis is required for cue-induced receptor-ribosome dissociation using Dynasore, an endocytosis inhibitor. The results show that inhibition of endocytosis blocks the cue-induced DCC-RP dissociation (now Figure 3L), supporting our model that the receptor-ribosome complex is present at the cell surface.

Furthermore, we show that the dissociation of ribosomes from receptors happens in response to extracellularly applied ligand in vitro, leading to a protein synthesis response. This strongly supports the view that the interaction between the ligand and the receptor occurs at the cell surface in these in vitro experiments. Guidance molecules are also presented extracellularly in vivo along the retinotectal pathway. Therefore, it seems most likely that the initial interaction between ligand and receptor also occurs at the cell surface in vivo, and it is this ligand receptor interaction at the cell surface that triggers the dissociation of the receptor and the ribosome intracellularly. The subsequent intimate association between specific mRNAs and the ribosome may indeed require endocytosis and an endosomal platform.

In addition to this, we have performed new analyses on EM images obtained from axonal growth cones which show that 20 out of 22 axonal growth cones contain rows of ribosomes aligned just under the plasma membrane (within 50nm), again supportive of our model. We have added these new results in Figure 3F, 3G and Figure 3—figure supplement 1C-D and in the subsection “Receptor-ribosome coupling occurs in RGC axonal growth cones”.

Translational targets for Netrin-1 and Sema3A have been described by the authors and others but are being ignored in this manuscript. Are the mRNAs coding for β-actin, Par-3, Tctp associated with DCC, or is RhoA mRNA associated with Nrp-1? If they are not part of the interactomes, what does this mean for the proposed model?

We probed for the mRNAs listed by the reviewer:

- β-actin is enriched in DCC compared to Nrp1;

- TCTP is higher in DCC, although not significantly and it has a very low FPKM;

- Par-3 is not detected;

- RhoA is detected at very low FPKM and it is not significantly different.

As specified in the original manuscript, we used the human cell line SH-SY5Y rather than whole Xenopus brains to ensure that any detected differences in mRNA binding were not due to the expression of DCC and Nrp1 in different cell types. We were not able to use retinal axons because this would not generate sufficient material to perform pulldowns. Therefore, the receptor interactomes in this cell line are likely to partially differ from the receptor interactomes of the retinal axons where the particular proteins named by the reviewer are known to have key roles in axon guidance. This is exemplified by the absence of rps14 mRNA enrichment in SH-SY5Y cells, which was instead detected in the Xenopus brain (Figure 4—figure supplement 1C). We validated, three mRNAs – ctnnb1, hnrnph1 and rps14 – to interact with DCC in the Xenopus brain (Figure 4D, Figure 4—figure supplement 1E), and we validated that their corresponding proteins are selectively increased in axons in response Netrin-1 but not Sema3A, consistent with our axonal proteome study (Cagnetta et al., 2018) and with the detection of their mRNAs in retinal axons (Shigeoka et al., 2018). These results point to a model where receptor-specific interactomes act as a hub for cue-induced selective translation. As mentioned in the Introduction of the original manuscript, there are several other non-mutually exclusive mechanisms that contribute to the cue-induced selective translation in axons, including microRNA regulation (Bellon et al., 2017), mRNA modification (Yu et al., 2018), modulation of the phosphorylation of eukaryotic initiation factors (Cagnetta et al., 2019), and RBP phosphorylation (Huttelmaier et al., 2005; Lepelletier et al., 2017; Sasaki et al., 2010).

Other major points in order of occurrence:

Figure 1G-J: The presented western blots are single experiments without replication and are not quantified. The presentation of single, non-repeated experiments is a recurrent problem in this manuscript.

We apologize for causing this misunderstanding. These Western blots have been repeated at least 3 times, some of them many more times. We have now made this clear in the revised manuscript in the figure legends and Material and methods section.

Figure 1—figure supplement 1F, G: RPL39 is mentioned in the main text, but does not appear in the figure. The variances in this figure are quite large, and as the authors state, the conclusion is preliminary. The authors should either perform more experiments to elevate the conclusion from preliminary or remove this entire figure/chapter.

We only performed relative quantifications for RPs that were present in all three replicates of at least one of the pulldowns. RPL39 was not detected in the Nrp1 pulldowns and present in only 2 out of 3 replicates of the DCC pulldown (as mentioned in the original manuscript) and did therefore not appear in the figure.

We first tried to validate the differential association of two RPs (RPL39 and RPL4) by IP-Western blot but the antibodies were not suitable for Western blot on Xenopus laevis samples as they generated non-specific bands. We then tried to strengthen this point by testing whether knocking down one of the differentially enriched ribosomal proteins affects the cue-specific selective translation. Specifically, we have performed knockdown of RPL4 and RPL39 with morpholinos. Unfortunately, these results show that while RPL39 morpholino does decrease RPL39 levels (23% decrease), it also decreases the levels of two other ribosomal proteins we tested (RPS7 – 28% decrease and RPL19 – 27% decrease) in axonal growth cones. This result makes it impossible to use this approach for testing the effect of differential ribosomal proteins on cue-induced selective translation. In addition, the RPL4 morpholino that we tested did not result in a knockdown of RPL4 in axonal growth cones. Since we are unable to strengthen our conclusions, we have, as the reviewer suggests, now removed this data from the manuscript. It is important to note that these results are not necessary for the main conclusions of our study.

Figures 2—figure supplement 1A, B: Single, unrepeated experiments without a proper quantification. This is especially troublesome, as the purported shift for DCC upon EDTA treatment is not obvious at all.

We apologize again for not including this information about replication in the original manuscript. We have replicated and quantified these fractionations for DCC. We also quantified the experiments we have (Figure 2—figure supplement 1C). Our results are consistent with those from Tcherkezian et al., 2010, where they also performed polysome fractionation with and without EDTA for the DCC receptors. Furthermore, the results shown in Figure 2A-C strongly support that EDTA treatment results in loss of association of receptors to ribosomes, consistent with these polysome experiments. Importantly, we performed these experiments to provide further evidence of the association of receptors with ribosomes detected using several different approaches: i) IP followed by Western blot and mass spectrometry (Figure 1A-F), ii) IP followed by qPCR (Figure 1G-J), iii) expansion microscopy imaging (Figure 3A-B), iv) Proximity Ligation Assay (Figure 3C-D).

We added the new data in Figure 2—figure supplement 1C and related text in the subsection “Guidance cue receptors associate with ribosomes in a mRNA-dependent manner”.

Figure 2A: The RNAseA/T1 treatment is an elegant approach but to follow the authors interpretation it would be critical to experimentally prove that the rRNAs stay intact.

We agree that it is critical to show that rRNA is not affected by our RNAseA/T1 treatment. We have now provided new data from sucrose density gradients with and without RNAseA/T1 treatment. The UV absorbance profiles clearly show that unprotected mRNAs are cleaved, resulting in a loss of polysomes. The monosomal peak clearly increases indicating no or minimal loss of ribosomes/ribosomal RNA during our treatment. Prolonging the treatment to 30 min (instead of the 15 minutes that we used in our experiments) shows the same result. We have now added these new data in Figure 2—figure supplement 1D and related text in the subsection “Guidance cue receptors associate with ribosomes in a mRNA-dependent manner”.

Figures 2D, E: The author present the same plots as in Figure 1A without mentioning this fact. Also, there is no congruency between the main text, the figure legend and the actual figure: where are the 10 shared RBPs?

We thank the reviewer for this feedback. We had highlighted the ribosomal proteins (RPs) in Figure 1 and the RNA binding proteins (RBPs) in Figure 2. We have now replaced these panels with a new heatmap analysis of all the RBPs pulled-down after DCC and Nrp1 pulldown (now Figure 2D). This new heatmap provides a clearer overview of which RBPs are pulled-down and which of these are specific or common between the two receptors.

Figure 2—figure supplement 1C, D: Single unrepeated, non-quantified experiments.

These IP-Western blot experiments were repeated two times and the mass spectrometry experiments, which were also repeated 3 times, showed consistent results for this RBP (Figure 2D). Furthermore, we have now replaced this figure with new, more informative data performed in axonal growth cones. In this new experiment, we have quantified the co-localization between DCC, Nrp1 and the RBPs Staufen1 and hnRNPA2B1. The results show that DCC co-localizes to a significantly higher degree with hnRNPA2B1 compared to Staufen1 and, conversely, that Nrp1 co-localizes to a significantly higher degree with Staufen1 compared to hnRNPA2B1. These results support the preferential or selective binding of specific receptors to specific RBPs. We have now added this data in Figure 2E-F and related text in the subsection “DCC and Nrp1 bind to specific RNA-binding proteins”.

Figures 3A, B: Single growth cones are presented without replication, quantification or any unbiased co-localization analysis.

We performed these experiments on 4 independent biological replicates for DCC (total growth cones = 73) and 4 independent biological replicates for Nrp1 (total growth cones = 72). This information has now been added to the figure legend of Figure 3—figure supplement 1. We have quantified the co-localization in unstimulated conditions. We have performed unbiased co-localization analysis and found a Pearson’s correlation of 0.432 and Mander’s overlap coefficient of 0.434 for DCC and RPL5 and a Pearson’s correlation of 0.673 and Mander’s overlap coefficient of 0.675 for Nrp1 and RPS3a, confirming the partial co-localization claimed in the original manuscript. We have now added the Pearson’s correlation values to Figure 3—figure supplement 1A and related text in the subsection “Receptor-ribosome coupling occurs in RGC axonal growth cones” and have added the number of biological replicates to the corresponding figure legend.

Figure 3E: Why did the authors normalize the data instead of presenting the actual counts per sq. mm?

To ensure the highest quality of analysis, we always normalize each experimental replicate to the control condition and finally pool the normalized replicates together.

After having performed many PLA experiments, we have noticed that the absolute counts of PLA signal sometimes differ between experiments and between the use of different kits. To account for these differences, we therefore normalized our PLA data.

In these specific experiments, the average absolute count per growth cone for EphB2-RPL5 was equal to 0.54, with 75% of growth cones showing no PLA signal. By contrast, the average absolute count per growth cone for DCC-RPL5 was equal to 7.31 counts per growth cones, with less than 5% of growth cones without PLA signal. The average absolute count per growth cones for Nrp1-RPS23 was equal to 2.14, with less than 25% of growth cones without PLA signal.

Figure 3F: The authors are correct by stating that association of DCC with RPL5 and RPS4X is significantly decreased, however, they ignore that vastly different effect sizes. RPL5 seems reduced by ~80% while RPS4X is reduced by less than 20%. What is the explanation for this and shouldn't the reduction of the ribosomal subunits be stoichiometrically matched to form functioning ribosomes? The authors argue that local production of RPLs might obscure their results; they should perform the experiments in the presence of CHX to excluded this potentially confounding possibility.

The reviewer is correct, we know that both RPL5 and RPS4X are locally translated in response to Netrin-1 (Cagnetta et al., 2018; Shigeoka et al., 2018), suggesting that the significant decrease in PLA signal detected is even underestimated. Therefore, we think that repeating these experiments with CHX would not provide significant new information. In addition, it should be noted that in other experiments (Figure 4A and new Figure 3L), the Netrin-1-induced decrease in PLA signal of DCC-RPL5 is more in line with the Netrin-1-induced decrease in DCC-RPS4X PLA signal and with the Sema3A-induced decrease in the Nrp1-RPS3A and Nrp1-RPS23 PLA signals. It is possible that some variability may be due to the company selling the PLA kits changing hands during the course of this study. What is key is that in all these experiments there is a significant decrease in the PLA signal for all the RPs tested. Additionally, we have performed new strengthening control experiments showing that Sema3A does not affect the DCC-RPL5 PLA signal and Netrin-1 does not affect the Nrp1-RPS23 PLA signal (Figure 3J-K), thus corroborating the cue-specificity of the decrease in the receptor-RP PLA signal.

These new data have been added in Figure 3J-K and related text in the third paragraph of the subsection “Dissociation of ribosomes from receptors is triggered by extrinsic cues and requires endocytosis”.

Figure 4A: It is surprising to see that Netrin-1 does not induce local translation as measured by puromycylation. How do the author reconcile this finding with previous reports that Netrin-1 induces local translation?

Down-regulation of global translation does not exclude the selective translation of a subset of mRNAs. This is exemplified by the stress response, where, despite a strong decrease in global translation, ~5% of the genome is selectively up-regulated. In our culture conditions, where profuse axon growth is obtained on high laminin substrate, Netrin-1 acts as a repulsive cue (Hopker et al., 1999) and, although repulsive Netrin-1 down-regulates axonal translation (Figure 4B-C), the local synthesis of a subset of proteins is simultaneously up-regulated (Cagnetta et al., 2018). We have now clarified this point in the subsection “Integration of multiple cues can affect the cue-induced selective translation of receptor-specific mRNAs”.

Figure 5A: A single image without biological repeats or (unbiased) quantification.

This is a representative EM image of many that we obtained from 22 different growth cones. We have now included the n numbers and additional EM images, as well as new quantitative analyses in Figure 3F-G and Figure 3—figure supplement 1C-E and in the subsection “Receptor-ribosome coupling occurs in RGC axonal growth cones”. These analyses shows that 20 out of 22 axonal growth cones contain rows of ribosomes aligned just under the plasma membrane (within 50 nm) and the remaining two growth cones contain single ‘isolated’ ribosomes under the plasma membrane. In addition, the inter-ribosome distance measurements revealed, interestingly, that the ribosomes are spaced significantly further apart in growth cones than in cell somas. This is consistent with single ribosomes, monosomes, binding to the intracellular portions of transmembrane receptors, such as DCC or Nrp-1. We have added these new data and EM images in Figure 3F, 3G and Figure 3—figure supplement 1C-E and related text in the aforementioned subsection.

We included this EM image because it demonstrates the striking spatial arrangement of ribosomes lined-up under the plasma membrane in growth cones. We have now moved it to earlier in the manuscript to Figure 3, to clarify that the EM data provide a “snap-shot” view further confirming that ribosomes lie in close proximity to the plasma membrane.

Reviewer #3:

[…]

1) The differential association of specific RPs with DCC vs. Neuropilin-1 is intriguing and distinct ribosome populations have been suggested in a number of systems but unequivocally proven. Some of the techniques used here and methodologies rely on those in the Shigeoka et al., 2018 reference from the authors. Particularly, the experiments with RNAseA/T1 in Figure 2. The reference there is incomplete and I do not see it has been published except for bioRxiv (I am not certain on eLife's policy for referencing unreviewed work like this).

We agree that it is crucial to show that our experimental treatments are working effectively. To provide independent evidence that this is the case, we have performed polysome fractionations and the resulting profiles show that RNAseA/T1 treatment decreases polysomes, whereas monosomes and subunits stay intact, as expected (now Figure 2—figure supplement 1D). We have additionally cited another article that uses similar conditions (Simsek et al., 2017).

These new data are added in Figure 2—figure supplement 1D and related text in the subsection “Guidance cue receptors associate with ribosomes in a mRNA-dependent manner”.

2) Assuming the RNAseA/T1 treatment effectively shows that mRNA/protein interactions are needed for the ribosome subunits to interact with DCC and Nrp1, it is surprising that no efforts are made to determine whether RNA is needed for the RBP associations with these receptors. Similarly whether ribosome subunits association is needed for mRNA interaction with DCC and Nrp1.

These are interesting questions. We have added new data using Western blot analysis in which we quantified the effect of RNAseA/T1 treatment on the interaction of the RBP, Staufen1, with Nrp1 (new Figure 2—figure supplement 1E). This shows a small but significant decrease in Staufen1 binding to Nrp1. This result suggests that mRNA may partly stabilize the interaction between receptors and RBPs. In addition, we have performed new experiments that show cue stimulation (with Netrin-1) did not result in a decrease in DCC-hnRNPA2B1 PLA signal, indicating that RBPs remain attached to receptors after cue stimulation (new Figure 3—figure supplement 1F).

The second question has been more difficult to answer. We performed new qPCR experiments after DCC pulldown in the presence/absence of EDTA and the results indicated that ribosome subunit association does affect the interaction of β-actin mRNA with DCC. However, because EDTA treatment may have other effects in this situation, we are not sufficiently confident to include the data in the manuscript.

We have now added the new data in Figure 2—figure supplement 1E and related text in the subsection “DCC and Nrp1 bind to specific RNA-binding proteins” and Figure 3—figure supplement 1F and related text in the subsection “Dissociation of ribosomes from receptors is triggered by extrinsic cues and requires endocytosis”.

3) The MS data for different RPs in Figure 1—figure supplement 1 is surprising and does support the authors' suggestion for different ribosome populations. It is surprising that the authors did not validate those showing greatest differentials (e.g., L4, L35, S6 and S9). Also, it seems like there are more 60S than 40S RPs showing differential association, but the authors did not comment on this point.

This concern was also raised by the other reviewers, and as we said in response to them, we attempted to strengthen the findings by IP-Western blot quantification, based on the antibodies tested in Xenopus laevis available in our lab. However, the antibodies against RPL39 and RPL4 generated non-specific bands and were not suitable for Western blot on Xenopus laevis samples. We therefore attempted a functional validation by performing morpholino based KD experiments against two differential RPs (RPL39 and RPL4) in axonal growth cones in order to test their possible role in selective translation. These results show that RPL39 KD does decrease RPL39 levels (23% decrease) but also decreases the levels of two other ribosomal proteins (RPS7 – 28% decrease and RPL19 – 27% decrease) in axonal growth cones. This makes it impossible to use this approach for testing the role of a specific RP on cue-induced selective translation. Morpholinos to RPL4 did not knock down the protein in axonal growth cones. Therefore, as it has not been possible for us to independently verify, either by Western blot or function, the differential association of specific RPs with specific receptors, we have eliminated this point from the manuscript as the other reviewers suggested.

We thank the reviewer for pointing our attention to the difference in 60S and 40S RPs that showed differential association. Based on the current literature on heterogeneous/specialized ribosomes, there does not seem to be a bias towards 60S RPs in heterogenous ribosomes. We can only speculate as to why we detect more 60S RPs to be differential but, since we could not validate our findings, we have removed these data and the accompanying discussion.

4) For Figure 2—figure supplement 1A-B, the majority of Nrp1 and DCC signals do not fractionate with subunits. Granted, these are very hard experiments and the EDTA treatments do shifts signals for Nrp1 and DCC upwards in the gradients. I think this would be strengthened EDTA treatments and some quantitation for the colocalization data shown in Figure 3 and Figure 3—figure supplement 1.

We appreciate the acknowledgement that these are not easy experiments to perform. We presented this data to provide further support for the association of receptors with ribosomes, in addition to the other experimental approaches that we used: i) IP followed by Western blot and mass spectrometry (Figure 1A-F), ii) IP followed by qPCR (Figure 1G-J), iii) expansion microscopy imaging (Figure 3A-B), iv) proximity ligation assay (PLA) (Figure 3C-D). Control and EDTA treated samples were compared directly, twice for DCC and once for Nrp1. We have quantified the relative amounts of DCC (n = 2 for control and EDTA) and Nrp1 (n = 2 for control, n = 1 for EDTA) in ribosome-free and ribosome-containing fractions. This relative quantification corrects for any possible decreases in overall protein levels, showing a clear shift to lighter fractions (i.e. ribosome-free fractions) for both DCC and Nrp1 after EDTA treatment (new Figure 2—figure supplement 1C). It is noteworthy that our results are consistent with those from Tcherkezian et al., 2010, which performed similar experiments for DCC. Furthermore, the results shown in Figure 2A-C strongly support that EDTA treatment results in loss of association of receptors to ribosomes, consistent with the polysome experiments.

We have obtained many expansion microscopy images from four biological replicates and we have added the quantification of co-localization (new Figure 3—figure supplement 1A).

The new data are added in Figure 2—figure supplement 1C and Figure 3—figure supplement 1A and related text in the subsections “Guidance cue receptors associate with ribosomes in a mRNA-dependent manner” and “Receptor-ribosome coupling occurs in RGC axonal growth cones”.

5) It is surprising that no colocalization for mRNAs with the receptors is provided beyond the co-IP, and such RIP experiments are known to have artefacts from in solution RNA protein-interactions after lysis.

In this study we have reported that replicates of IP for DCC and Nrp1, followed by RNA-seq, show distinct but consistent signatures of mRNAs. It is technically very challenging to test the co-localization of proteins with specific mRNAs as mRNAs may be masked by the presence of binding proteins (Buxbaum et al., 2014). Furthermore, for any specific mRNA we have tested previously with in situ hybridization, or live, there are rarely more than 8-15 fluorescent mRNA puncta per growth cone. By contrast, DCC and Nrp1 receptor immunostaining gives an abundant signal throughout, with hundreds of fluorescent puncta per growth cone. Therefore, we think that co-localization of receptors with specific mRNA probes (FISH) would not be a convincing approach.

Since mRNAs bind to RBPs and these are likely essential for the mRNA-receptor association, we reasoned that a more feasible and equally informative experiment to address this question would be to test the co-localisation of RBPs with receptors. To this end, we performed dual immunocytochemistry and analysed the co-localization of DCC and Nrp1 with hnRNPA2B1 and Staufen1. The new data show that DCC co-localizes to a higher degree with hnRNPA2B1 compared to Staufen1 and that Nrp1 co-localizes more with Staufen1 compared to hnRNPA2B1. This is in line with, and validates, the differential binding results of our co-IP data.

We have added these data in Figure 2E-F and related text in the subsection “DCC and Nrp1 bind to specific RNA-binding proteins”.

6) For Figure 5A, the EM is compelling an shows a surprisingly high number of electron dense ribosome like structures. The authors do not note whether this is a stimulated or naïve culture. Comparison of the two would be informative and strengthen the authors' conclusions (though I do recognize the difficulty in this request).

We apologise for the lack of clarity. The EM experiments were carried out on unstimulated growth cones. We have now specified this in the subsection “Receptor-ribosome coupling occurs in RGC axonal growth cones”, and the figure legend of Figure 3.

Although it would be nice to show stimulated growth cones by EM too, and compare them quantitatively, we think that an ultrastructural analysis of this sort would take many months to complete and would be a major study in itself. Here we aimed to provide further visual evidence that ribosomes can be found in close proximity to the plasma membrane in axonal growth cones. This finding has been extensively corroborated by the several receptor-ribosome proximity-based (PLA) experiments, quantified and in triplicate, together with negative controls (Figure 3C-E, H-I and Figure 3—figure supplement 1B, H), in stimulated versus naïve conditions. Furthermore, we have now added new strengthening control experiments showing that Sema3A does not affect the DCC-RPL5 PLA signal and Netrin-1 does not affect the Nrp1-RPS23 PLA signal, confirming the cue-specificity of the decrease in the receptor-RP PLA signal (Figure 3J-K and related text in the third paragraph of the subsection “Dissociation of ribosomes from receptors is triggered by extrinsic cues and requires endocytosis”).

To strengthen our EM findings further, we have now added additional images, as well as new analysis of these EM images in Figure 3F-G and Figure 3—figure supplement 1C-E. These analyses show that 20 out of 22 axonal growth cones contain rows of ribosomes aligned just under the plasma membrane (within 50 nm) and that the remaining two growth cones contain single ‘isolated’ ribosomes under the plasma membrane. Interestingly, we observed that the inter-ribosome distance in these growth cone ribosomes is significantly larger than ribosomes in the cell soma. This is consistent with single ribosomes, monosomes, binding to the intracellular portions of transmembrane receptors, such as DCC or Nrp-1. We have added these new data and analyses in Figure 3F, 3G and Figure 3—figure supplement 1C-E and related text in the subsection “Receptor-ribosome coupling occurs in RGC axonal growth cones”.

7) Robo3.3 and EphA2 have been shown to be translated in axons. This is probably worth mentioning in the Discussion since the authors show that EphB2 does not use this mechanism for sequestering RNPs and ribosome-bound mRNAs.

This point is unclear as we think that the lack of association of a receptor with ribosomal proteins does not exclude that the receptor itself may be potentially locally translated.

[Editors’ note: the author responses to the re-review follow.]

[…] Two of the reviewers found your manuscript satisfactorily amended, but one of the reviewers cited a number of points that should be addressed before publication, all of them addressed by textual amendments:

1) There are a number of places where your results are somewhat overstated and should be toned down:

a) Abstract: 'Our findings.… provide a general model for the rapid, localized and selective control of cue-induced translation.' Compare this to the much more accurate and measured summary: '… this study provides evidence…and suggests…'. Your results are interesting and represent a significant advance, but it may be premature to consider that they comprise a general model.

We have now changed the Abstract, stating:

“Our findings reveal receptor-specific interactomes and suggest a generalizable model

for cue-selective control of the local proteome”.

b) The findings from Figure 2B, C in are summarized: '…these results suggest that the interaction.… is likely mediated through mRNA'. One page later, this sentence has morphed into the definitive statement: '…our finding that mRNA mediates the association of receptors with specific ribosomes…'

We understand the reason for this comment, but the results of the paper, as it has been revised with new data, are stronger with respect to this point. We have therefore changed the statement into the following sentence:

“Together with our evidence implicating mRNA in the association of receptors with ribosomes, these results are consistent with a model in which receptors associate with specific RBPs, which bind specific mRNAs, and these mRNAs, in turn, recruit ribosomes.”

c) '…reveal that multiple receptors.… can associate with ribosomes.' This statement is imprecise; the results up to this point show association with ribosomal subunits, not ribosomes.

We agree with the reviewer that it is more accurate to describe it is an association with ribosomal subunits and have now adapted this in the revised manuscript: “…reveal that multiple receptors.… can associate with ribosomal subunits.”

Please carefully edit the Results section and relegate the interpretation especially statements regarding the general impact/significance, to the Discussion.

We have done this.

2) A major concern in the first review was that some experiments were not repeated and quantified. Most of these issues have been fixed, but in Figure 1C-E, Figure 1 —figure supplement 1A-C, results are still presented without quantification. The sentence in the figure legend that these experiments have been repeated three times but the actual results should be presented.

We have now quantified the protein bands in the IP samples relative to the input by densitometry analysis using ImageJ software on the co-IP-Western blot results presented in Figures 1C-F and Figure 1—figure supplement 1A-C (see Author response image 1). These co-IP-Western blot experiments were done to validate whether or not ribosomal proteins bind to the different receptors in a simple ‘yes’ or ‘no’ assay. In general, co-IP-Western blots are used to test whether one protein binds another. Binding is confirmed by the existence of a band of the appropriate molecular size and antigenicity. The main result here, which is clear in the blots presented, and also in the graphs shown below, is that DCC, Nrp1 and Robo2, but not EphB2, bind to ribosomal proteins. These IP results support the mass spectrometry data (Figure 1A, B) and IP-qPCR data (Figure 1G-J and Figure 1—figure supplement 1D, E). These two other techniques provide useful quantitative information about protein levels and interaction with ribosomal subunits (i.e. ribosomal RNA). However, the quantifications below on the co-IP-Western blots do not provide useful information about quantities of protein as replicates were ran on different gels and many different factors influence the density of the bands (e.g. transfer efficiency, antibody incubation time, antibody lot numbers, development time). As quantification of these results is not informative other than providing some assurance that the bands are real, we would prefer not to include these graphs in the revised manuscript, but provide them here, in response to the comment above, to show that the results are robust.

Author response image 1

3) Figure 2—figure supplement 1C: the n for these experiments are 2 or 1. Showing bar graphs with SDs is not very meaningful for such low n numbers. The individual data points should be shown instead.

We agree that presenting the individual data points is more informative. We have therefore replaced the bar graphs in Figure 2—figure supplement 1C with a graph showing the individual data points.

4) Subsection “DCC and Nrp1 bind to specific subsets of mRNAs”, first paragraph: The finding that high abundance mRNAs are much more differentially associated than low abundance mRNAs seems to indicate that these mRNAs are the mRNAs that conform to the authors proposed model. It is a missed opportunity to not talk more about these mRNAs: provide numbers, identity, GO analysis.

We thank the reviewer for this comment. Indeed, GO analysis of high abundance mRNAs (FPKM >100) reveals specific GO categories associated with Nrp1 and DCC, such as ‘cell-cell adhesion’ and ‘protein targeting’ for DCC and ‘translation’ and ‘small GTPase mediated signal transduction’ for Nrp1. We have now added the results in Figure 2—source data 2. This supplementary table also contains the numbers and identity of all differentially expressed genes (DEG) and high abundant genes with an FKPM >100 are highlighted.

In addition, we have added a discussion of these results to the subsection “DCC and Nrp1 bind to specific subsets of mRNAs”.

5) In response to conceptual question 2 (reviewer 2, first review), you provide the interesting finding that endocytosis is required for the dissociation between DCC and RP. However, the question was in relation to the mRNA-seq findings: it is my understanding that the RNA-seq experiments cannot distinguish between mRNAs associated with the receptors at the membrane vs. receptors at vesicles. For example, if mRNAs/RNPs are associated receptors on vesicles that transport the receptors into axons, these mRNAs would show up in the RNA-seq, even if they no longer associate with receptors after insertion into the plasma membrane. Please discuss this point to make clear that the RNA-seq results might include mRNAs that are not associated with receptors in the membrane.

The reviewer is correct to say that we cannot determine where exactly the association of receptors with mRNAs takes place from the RNA-seq experiments. It is indeed possible that some of the detected mRNAs are associated with receptors while traveling on endocytic vesicles and we have now discussed this possibility in the fourth paragraph of the Discussion.

6) The response to the comment regarding Figure 3F is unsatisfactory: You seem to indicate that the data as presented are unreliable due to problems with the reagents. Expand on this point.

The data presented in Figure 3F (Figure 3F in the initial submission, now Figure 3H) are reliable. They were obtained from three independent biological replicates with a total of more than a hundred growth cones per condition. There were no problems with the reagents or PLA signal in these experiments. However, as we indicated, the sensitivity of the technique varies between trials, especially between trials performed with kits from DuoLink and trials with the kits that were provided by Sigma-Aldrich. The DCC-RPL5 data in Figure 3H was performed with a PLA kit from the company Duolink before they were taken over by Sigma-Aldrich. All other experiments for DCC (Figures 3H – DCC/RPS4X, 3J, 3L and Figure 4A) were performed after this takeover and showed a smaller decrease in PLA signal after Netrin-1 stimulation compared to the first set of experiments. We are unsure as to what exactly caused this difference, but it co-occurred with the change of companies. What is most important in these experiments is that there is a significant decrease in PLA signal after Netrin-1 stimulation for all tested RPs, whereas Sema3A stimulation does not affect DCC-RPL5 PLA (Figure 3J), and that this is true with both the DuoLink and the Sigma-Aldrich kit. Indeed, although the sensitivity between kits may be different, the experiments are all internally consistent and are always done with appropriate controls. So the results are even perhaps more reliable simply because they were reproducible with different kits.

https://doi.org/10.7554/eLife.48718.sa2

Article and author information

Author details

  1. Max Koppers

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7751-1082
  2. Roberta Cagnetta

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Formal analysis, Investigation, Visualization, Methodology, Writing—review and editing
    Contributed equally with
    Toshiaki Shigeoka
    Competing interests
    No competing interests declared
  3. Toshiaki Shigeoka

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Writing—review and editing
    Contributed equally with
    Roberta Cagnetta
    Competing interests
    No competing interests declared
  4. Lucia CS Wunderlich

    Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7200-1713
  5. Pedro Vallejo-Ramirez

    Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Software, Formal analysis
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7879-6761
  6. Julie Qiaojin Lin

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2669-6478
  7. Sixian Zhao

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  8. Maximilian AH Jakobs

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Data curation, Wrote a script to automatically detect axonal growth cones from microscopy images and create masks from these enabling imaging quantification, Wrote the methods part for this
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0879-7937
  9. Asha Dwivedy

    1. Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    2. Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Performed in vitro retinal cultures and processed them for immunocytochemistry or PLA
    Competing interests
    No competing interests declared
  10. Michael S Minett

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Carried out in vitro retinal cultures, performed PLA and acquired imaging data
    Competing interests
    No competing interests declared
  11. Anaïs Bellon

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Present address
    Aix Marseille University, INSERM, Institute de Neurobiologie de la Méditerranée, Marseille, France
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  12. Clemens F Kaminski

    Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Resources, Supervision
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5194-0962
  13. William A Harris

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Conceptualization, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9995-8096
  14. John G Flanagan

    Department of Cell Biology, Harvard Medical School, Boston, United States
    Contribution
    Conceptualization, Writing—review and editing
    Competing interests
    No competing interests declared
  15. Christine E Holt

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Writing—review and editing
    For correspondence
    ceh33@cam.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2829-121X

Funding

Netherlands Organisation for Scientific Research (Rubicon 019.161LW.033)

  • Max Koppers

EPSRC (EP/L015889/1)

  • Clemens F Kaminski

EPSRC (EP/H018301/1)

  • Clemens F Kaminski

Wellcome Trust (3-3249/Z/16/Z)

  • Clemens F Kaminski

Wellcome Trust (089703/Z/09/Z)

  • Clemens F Kaminski

Wellcome Trust (085314/Z/08/Z)

  • Christine E Holt

Wellcome Trust (203249/Z/16/Z)

  • Christine E Holt

European Research Council (Advanced Grant 322817)

  • Christine E Holt

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

Acknowledgements

We thank Nicola Lawrence, Caia Duncan (Juan Mata lab, Unversity of Cambridge), and Katrin Mooslehner for technical assistance and Fabrice Richard (PiCSL-FBI core facility, IBDM, CNRS, Aix-Marseille University; member of the France-BioImaging national research infrastructure (ANR-10-INBS-04)) for assistance with EM experiments. This work was supported by the Netherlands Organization for Scientific Research (NWO Rubicon 019.161LW.033) (MK), UK Engineering and Physical Sciences Research Council, EPSRC Grants (EP/L015889/1 and EP/H018301/1) and Wellcome Trust Grants (3-3249/Z/16/Z and 089703/Z/09/Z) (CFK), Wellcome Trust Grants (085314/Z/08/Z and 203249/Z/16/Z) and European Research Council Advanced Grant (322817) (CEH).

Ethics

Animal experimentation: All animal experiments were approved by the University of Cambridge Ethical Review Committee in compliance with the University of Cambridge Animal Welfare Policy. This research has been regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body (AWERB) and under project license PPL80/2198.

Senior Editor

  1. Catherine Dulac, Harvard University, United States

Reviewing Editor

  1. Carol A Mason, Columbia University, United States

Publication history

  1. Received: May 23, 2019
  2. Accepted: November 19, 2019
  3. Accepted Manuscript published: November 20, 2019 (version 1)
  4. Version of Record published: December 5, 2019 (version 2)

Copyright

© 2019, Koppers et al.

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.

Metrics

  • 1,751
    Page views
  • 280
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)