Corrigendum: Precursor and mature NGF live tracking: one versus many at a time in the axons

Scientific Reports 6: Article number: 20272; published online: 01 February 2016; updated: 30 March 2016 This Article contains errors in Affiliation 3. The correct affiliation is listed below: Center for Nanotechnology Innovation @NEST, Istituto Italiano di Tecnologia, Piazza San Silvestro 12-56127 Pisa, Italy.

e reciprocal levels of Nerve Growth Factor (NGF) and of its unprocessed precursor (proNGF) play a crucial role in regulating the survival/death balance of several neuronal populations in physiopathological conditions [1][2][3][4] . proNGF, which is the most abundant form of NGF in the brain 5 , can either act as an intracellular or secreted precursor for mature NGF or remain unprocessed, activating survival/di erentiation or apoptosis pathways, respectively. is molecular switch relies on the di erent receptor binding pro les, determining di erent biological outcomes. When administered separately or together, NGF and proNGF activate distinct and peculiar gene expression patterns in target cells 6,7 . Disruption of the NGF to proNGF balance has been causally linked to neurodegeneration 8,9 , and the proNGF versus NGF ratio is increased in the cortex of Alzheimer's disease patients 5 .
us, describing the signaling mechanisms that link NGF and proNGF cellular tra cking to their speci c biological function is of crucial importance.
Axonal transport of neurotrophins (NTs) represents a crucial link between receptor mediated signaling and their biological outcome 10 . Despite its importance, we currently lack the molecular de nition and characterization of NTs axonal signalling endosomes 11 . Furthermore, the question of whether and how pro-NTs move retrogradely or anterogradely along axons, in comparison to their mature counterparts, has remained so far largely unexplored. We have addressed these issues by using a biophysical approach of NT labeling and tracking in living neurons. Ideally, a quantitative comparison between the axonal transport of NGF and proNGF requires the two molecules to be uoro-labeled at the same molecular site and with the same stoichiometry. So far, the mature forms of NTs have been chemically coupled to organic uorophores [12][13][14][15][16] or to biotin [17][18][19][20][21] . While this has allowed obtaining valuable information about NGF tra cking, it has not been possible to control the exact number and site of conjugated probes, so that mixed and not fully reproducible labeled protein populations are obtained. Similar approaches are not recommended for the purpose of labeling proNGF, whose pro-domain has features of an intrinsically unfolded protein 22,23 , nor for a detailed side-by-side comparison of the mature and precursor forms.
Here a novel uorolabeling strategy is described, allowing for the production of "homologous" uorescent human NGF and proNGF, based on the insertion of an 11 amino acid tag at the at the C-terminus of the protomer sequence. e inserted tag is a target for a site-speci c enzymatic covalent binding and it is used here to BioSNS Laboratory, Scuola Normale Superiore and Istituto di Neuroscienze - CNR, Pisa, Italy. NEST, Scuola Normale Superiore and Istituto Nanoscienze-CNR, Pisa, Italy. IIT@NEST, Center for Nanotechnology Innovation, Pisa, Italy. EBRI, European Brain Research Institute, Rome, Italy. Correspondence and requests for materials should be addressed to A.C. (email: antonino.cattaneo@sns.it) bind a small organic dye, so that a site-speci c uorophore conjugation with 1:1 (label:NT-monomer) stoichiometry is obtained for both the unprocessed and mature forms of NGF. e technique allows for a high (≈ 80%) uorescent NT production yield and for an optimal puri cation from the unlabeled counterparts. e obtained labelled species retain the same functional features of the native proteins. Fluorescence microscopy experiments were performed on compartmentalized living cultures of rat dorsal root ganglion (DRG) neurons, in which uorescent proNGF and NGF were administered either separately or together. e rst direct evidence that proNGF is retrogradely transported like mature NGF is provided here, although important di erences of the axonal transport of the two molecules have been uncovered. Crucially, the controlled stoichiometry of the labelling reaction allowed quantifying the number of NTs in each vesicle, showing a signi cant di erence between their distributions in the two cases. Moreover, by coadministering both neurotrophins labelled with di erent uorophores, we were able to analyse the cotransport of precursor and mature neurotrophins in neurons.

Results
Fluo--proNGF and Fluo--NGF synthesis and functional validation. We recently explored the possibility of chemically modifying NTs by the insertion of short tags derived from the acyl and peptidyl carrier proteins 24 . Following this approach, a single uorophore per monomer of proNGF or NGF was conjugated by introducing a short amino acidic tag at the C-terminus of the proNGF sequence (Fig. 1A,B), by exploiting a site-speci c enzymatic reaction to covalently link the small organic dye to a serine residue of the tag (Fig. 1C). e C-terminal portion of NTs was chosen as a permissive site to insert the tag with minimal interference with NGF structure and receptor binding 24 .
In order to optimize both expression and labelling yields, as well as to assure the maintenance of NT physiological activity, four di erent tag sequences were compared (spanning from 8 to 12 amino acids): YBBR, A4, A1, and S6 (Fig. 1C) [25][26][27] . e production yields obtained for all the tagged NTs a er their expression, refolding and puri cation from E.Coli inclusion bodies 22,28 (Supplementary Figure 1A) were compared to those obtained for the untagged counterparts ( Fig. 1D): while for all tagged constructs the precursor NT could be successfully puri ed (although with di erent yields: YBBR > A4 > A1 > S6), only two of the tags (YBBR > A4) consistently yielded measurable amounts of puri ed mature protein. Accordingly, NGF-YBBR and NGF-A4 were further characterized. Tag insertion does not interfere with NGF functionality, according to a number of signaling, cell di erentiation and proliferation assays ( Supplementary Figs 1B-D). For labelling purposes, YBBR-and A4-tagged precursor or mature NGF proteins were incubated with Coenzyme A (CoA) labelled biotin substrate and Acp-or Sfp-synthase PPTases (or no enzyme as control) (Fig. 1E). Data demonstrate that a speci c biotin labelling is achieved for NGF-YBBR in the presence of both enzymes, with a higher labelling yield provided by the reaction with Sfp-synthase, while NGF-A4 displays signi cant labelling only in presence of Acp-synthase but with a lower yield compared to the NGF-YBBR/Sfp-synthase reaction. Based on these data, YBBR sequence was identi ed to be the best tag for NGF and proNGF labelling.
Next, YBBR tag was exploited to produce uorescent proNGF and NGF, herea er referred as uo-proNGF and uo-NGF. To this purpose, puri ed proNGF-YBBR or NGF-YBBR were incubated with CoA-Alexa488 substrate in the presence of Sfp-synthase (see Materials and Methods). Upon labelling, uo-proNGF and uo-NGF were puri ed by ion exchange chromatography to remove both the free uorophore and the non-reacted NT, so as to recover exclusively uorescent NTs (Fig. 1F). e comparison of the various peaks integrals in the chromatogram allowed us to estimate a labelling yield of about 80% (see Materials and Methods). In order to validate whether the uorescent NTs retain functional activity, we performed a DNA microarray analysis of gene expression activated by 1 hour treatment of PC12 cells with uo-proNGF or uo-NGF respectively, and compared the expression pro le with those obtained a er incubation of equal amounts of the corresponding wt, untagged counterparts. Indeed, we have previously demonstrated that this assay is able to nely discriminate proNGF from NGF functional activity in relatively short timescales, so as to minimize the impact of proNGF cleavage to NGF 6 . Clustering analysis of gene expression pro les showed that treatments with wt and uorescent proNGF cluster together, while the gene expression pro le obtained with wt NGF is farther away in the tree and is clustered together with uo-NGF (Fig. 1I). us, each uo-NGF and uo-proNGF neurotrophin induces a global signaling very similar to that of the corresponding unmodi ed neurotrophin. Furthermore, uo-NGF is able to promote PC12 cells differentiation, inducing neurite outgrowth to the same extent as wt NGF (Fig. 1G). Moreover, uo-NGF induces a robust activation of the downstream signaling e ectors involved in NGF-signaling pathways, like phosphorylated Erk1/2, PLC-γ and AKT proteins (Fig. 1H). ese data indicate that uorescent NTs are biologically active and induce physiological responses comparable to those of the respective native proteins. fluo--NGF versus fluo--proNGF Axonal Transport. Purified fluorolabeled proNGF and NGF were employed for axonal transport studies in living DRG neurons, cultured in a micro uidic chamber divided in three compartments 29 : a soma compartment (SC), where neurons are plated, a channel compartment (CC), where neurons extend their axons through micrometer channels, and an axon compartment (AC) where neurons spread their axonal tips ( Fig. 2A,E,I,M). Each experiment was performed by administering the NT at 2 nM concentration in the AC or in the SC and measuring the axonal transport of the protein by epi-uorescence microscopy and single particle tracking of uorescent vesicles.
First, uo-NGF was administered to the AC or to the SC and the vesicles tra cking was measured in the CC ( Fig. 2A,E). In both cases, uo-NGF lled vesicles started to move about 20 minutes a er NT administration. Fluo-NGF vesicles were found to move not only when NGF was delivered to the AC, from the axon tip to the cell soma (retrograde movement), but also when delivered to SC, in the opposite direction (anterograde movement). Typical recorded trajectories are represented in Fig. 2B, F. Most of the uo-NGF vesicles move with a stop-and-go dynamics, as previously reported 17 , characterized by variable pausing times (Fig. 2C,G). On the whole, traveling vesicles spend about 55% of their time moving (see also Supplementary Fig. 2 and Supplementary Information).  Table 1). Notably, in the case of AC-administered uo-NGF, also anterogradely moving vesicles have been observed, about 10 minutes a er the rst retrogradely moving vesicle was measured. e speed distribution during the active phase of the overall movement, evaluated by separating the "stop" and "go" parts of the trajectories, is shown as solid lines in Fig. 2D,H. e high number of trajectories acquired with SC-applied uo-NGF allows to conclude that the anterograde transport is slightly, but signi cantly (p < 0.05, Dunns Test), slower than the retrograde transport. On the contrary, the low number of anterograde moving vesicles did not allow this comparison in the case of AC-applied uo-NGF. e retrograde and anterograde average velocities measured by administering uo-NGF to the AC or the SC are the same within uncertainties (see also Supplementary Table 1).
In the case of uo-NGF applied to the AC, the anterograde population of moving vesicles increases during the acquisition, with an average value of 9% ( Supplementary Fig. 3A). is anterograde movement is di erent from the short anterograde displacement seen during vesicle retrograde transport, as frequently observed here and previously reported 17 , and persists for large displacements (up to about 100 µ m, which represents the length of the eld of view) and for long times (more than 60 sec). Furthermore, when uo-NGF is applied to the SC, about 20 minutes a er the appearance of the rst anterograde-moving vesicles, we observe the appearance of a robust transport back to the SC. In this case, the number of vesicles transported back (to the SC), a er they have had the chance of an outward movement, is signi cantly higher than that (from SC to AC) observed for AC-administered NGF (25% of the cases compared to 9%) and, once started, this retrograde ux to the soma persists quite constantly for hours ( Supplementary Fig. 3B).
Next, proNGF vesicular trafficking was examined under the same experimental conditions as above. Fluo-proNGF was administered to the AC or to the SC (Fig. 2I,M). We found that a retrograde flux of uo-proNGF vesicles moving from the AC to the SC can be recorded (Fig. 2J), although this is rst visible a er a longer lag time compared to that of uo-NGF (starting about 35 minutes a er NTs administration). Fluo-proNGF moving vesicles display the same characteristic stop and go movement seen with uo-NGF (Fig. 2K,O). Vesicles exhibit similar distributions of average speeds and speeds during active movement, compared to the same experiment conducted with mature uo-NGF. Strikingly, unlike uo-NGF, when uo-proNGF was administered to the SC we observed no signi cant anterograde transport through the axons (Fig. 2M); however, as a proof of neurotrophin internalization, vesicles tra cking with directed motion inside axons in the SC was observed ( Fig. 2N-P; note that in this case a polarity to the movement couldn't be assigned, hence only positive speed is reported, by convention). For comparison, similar measurements for SC applied uo-NGF are reported in Supplementary  Fig. 4. Nevertheless, when uo-proNGF was given to the AC, a small number of anterogradely moving vesicles (less than 5%), moving just for short displacements, was observed (the maximum anterograde displacement observed is 25 µ m). ese movements can be con dently attributed to little steps backwards while the vesicle moves retrogradely.
As TrkA and p75 are the main receptors for NGF and proNGF, we investigated by immuno uorescence which of the two receptors mediated the intracellular tra cking of NGF and proNGF. e vast majority of uo-NGF particles are associated with TrkA receptors; only a subset of them associates with p75, most prominently at the axonal level. Conversely, uo-proNGF associates with both TrkA and P75 receptors. Notably, we found that when uo-proNGF is present, p75 also colocalizes with TrkA, and that uo-proNGF particles are associated with p75/ TrkA positive puncta ( Supplementary Fig. 5).
Visual inspection of the experimental acquisitions of uorescent NGF and proNGF tra cking (Supplementary Video 1 and 2) suggests that the uorescent intensity of tra cking vesicles are substantially di erent in the two cases, despite the identical labelling site and stoichiometry for the two proteins. Two representative uo-NGF and uo-proNGF labelled vesicles are presented in Fig. 3A. We quantitatively analyze the di erent number of NTs transported by a single vesicle in the two cases, thanks to the controlled 1:1 uorophore:NT-monomer stoichiometry ensured by the labelling strategy. To this end, the mean uorescence intensity from each vesicle was quanti ed and normalized to the covalently conjugated to the uorophore highlighted in red. (D) Puri cation yields (mg of product per litre of bacterial culture) of tagged proNGF-tag (gray) and NGF-tag (blue), compared to wt (pro)NGF. (E) Western blot analysis of the biotinylation reaction of puri ed NGF-YBBR and NGF-A4 using CoA-biotin substrate and AcpS or SfpS PPTases. e same biotinylation reaction is performed in parallel using untagged wt-NGF as negative control. Streptavidin-HRP is used for biotin detection. e anti-NGF blot (NGF and proNGF lines) is the loading control. (F) e HPLC chromatogram of NGF-YBBR, incubated with CoA-Alexa488 substrate in the presence and absence of Sfp-synthase, showing the di erent retention times of uorescent and non-uorescent NTs. Absorbance curves at 280 and 490 nm are reported. (G) Typical DIC images of PC12 di erentiation assay using equimolar amounts of wt-NGF and uo-NGF. Untreated cells are the control. Scale bars represent 20 µ m. (H) Western blot analysis of phosphorylated Akt (pAkt), phosphorylated Erk1/2 (pErk1/2) and phosphorylated PLCγ (pPLCγ ) protein levels in PC12 cells in response to wt NGF and uo-NGF, compared to the same obtained for untreated cells (No NGF); the signal of the total corresponding signalling e ectors is the loading control. (I) Hierarchical clustering tree of samples, corresponding to the di erent experimental points (for each neurotrophin type four individual PC12 cells administrations were performed). e trees show the gene expression similarity between samples. e x-axis indicates the distance between samples. Euclidean distance is the chosen metric, with average linkage clustering, using all normalized Log2 data. reference intensities measured from immobilized single uorophores (see Fig. 3A and Supplementary Methods for details). e measured number of NT dimers, each carrying two uorophore molecules, is then used to produce a 2D histogram in which the number of NTs is presented versus its average speed in Fig. 3B. e position on the Y axis con rms that the average speed of the NGF and proNGF vesicles is the same. e X axis shows that uo-NGF vesicles carry a range of NT molecules, spanning from 2 to 8 (full width at half maximum) with a mode of 4 dimers per vesicle, while uo-proNGF vesicles mostly contain a lower number of NTs, from 1 to 4 proNGF dimers with a mode value of 1 dimer per vesicle (see also Supplementary Fig. 6).
NGF competes with proNGF for axonal transport. NGF and proNGF coexist in-vivo, but a direct observation of simultaneous axonal transport is still lacking. In order to study the simultaneous axonal transport of proNGF vs NGF, NTs were conjugated to spectrally distinct probes by using two di erent CoA-uorophore substrates: CoA-Alexa647 and CoA-Alexa488 respectively (see Materials and Methods). uo-NGF and uo-proNGF (Fig. 4A Top) were co-administered, at equimolar (2 nM) concentration, to the AC. A typical dual-colour acquisition of the trajectories recorded in the CC is represented in Fig. 4A, with green and red colours representing putative NGF and proNGF vesicles, respectively. Notably, a population of vesicles carrying both uorophores, represented in yellow in Fig. 4A,B, was registered. e velocity of dual-labelled vesicles is not signi cantly di erent from that of vesicles carrying the individual NTs (Fig. 4B), and from the velocities registered for the individual administrations of proNGF and NGF ( Supplementary Fig. 7); however, we noticed that the number of anterogradely moving vesicles of NGF is decreased to ~3% of the total in this case ( Supplementary Fig. 8). en, the number of neurotrophin molecules carried by vesicles was determined (Fig. 4C). e distributions of uo-NGF is quite similar to that observed a er a single administration, while, interestingly, the number of proNGF dimers was markedly reduced at 1 in the vast majority of cases, demonstrating a great preference of vesicles to transport mature NGF instead of the precursor form ( Fig. 4C and Supplementary Fig. 6). Similar conclusions can be drawn by measuring the vesicle uxes with uo-NGF and uo-proNGF, given separately or simultaneously at the AC (Fig. 4D). In the case of single administrations, the number of moving vesicles per 1 mm of axon is 110 ± 42 and 72 ± 38 (mean ± SD) for uo-NGF and for uo-proNGF respectively. is indicates that, when administered alone and at the same concentration, proNGF vesicles ux is ~65% of the NGF one. e uo-proNGF ux dramatically drops down to 19 ± 14 vesicles per 1 mm of axon (i.e. ~18% of NGF ux) when the two NT forms are given together, while uo-NGF vesicles ux is similar to that observed a er the individual NT administration. Actually, the real proNGF ux could be even lower than that observed due to the possibility of a small fraction of uo-proNGF being cleaved to its mature counterpart during the acquisition time window (Supplementary Fig. 9). us, upon co-administration we observed for proNGF a reduction both in the ux of vesicles and in the number of neurotrophins per vesicle that might stem from competition with NGF or from signaling events caused by NGF and proNGF simultaneous administration (see Discussion for more details) 7 .

SCIENTIFIC REPORTS
Discussion e axonal transport of NTs is a crucial aspect of their mode of action in the nervous system. In this context, despite the fact that proNGF has a receptor binding pro le distinct from that of NGF 30,31 , very little is known on how this re ects in the transport properties of proNGF versus those of mature NGF. We address this issue through a novel strategy to uorolabel puri ed recombinant precursor and mature NTs with small organic dyes, exploiting a recently reported chemical-tag based NT labeling 24 . e method has several advantages over previously used NT labelling procedures: i) the precise control of stoichiometry (1:1 NT-monomer:probe) and site (C-terminus) of uorophore conjugation; ii) the complete puri cation of the labelled species from unlabelled counterparts as well as from free uorophores (Fig. 1F); iii) the versatility of the used tags, which can be functionalized by virtually any kind of probe, e.g. biotin or uorophores (Fig. 1E,F), that can be carried by coA substrates; iv) the possibility to simultaneously study two "homologue" molecules with orthogonal uorolabels (Fig. 4). e obtained uo-NGF and uo-proNGF probes allowed here the rst comparative imaging and tracking of proNGF and NGF axonal transport in living DRG neurons by single molecule fluorescence microscopy. Results unambiguously show that proNGF is retrogradely transported inside vesicles like its mature counterpart (Figs 2-4). proNGF displays the same stop-and-go movements previously reported for NGF and an average speed distribution similar to that of NGF. ese data suggest that the endocytic pathway and related engaged molecular motors are conserved between proNGF and NGF. Beside these similarities, the axonal transport of these two proteins di ers from each other in several signi cant aspects.
First, proNGF vesicles move exclusively from the axon tip to the cell soma of neurons, while NGF vesicles exhibit movements in both directions (Fig. 2). While a retrograde tra cking has been paradigmatically described

SCIENTIFIC REPORTS
for mature NGF 10 , a bidirectional NGF transport has been previously described in neurite-like processes of PC12 cells, either directly using Cy3.5-NGF 13 or indirectly by studying TrkA tra cking 32,33 . It should be considered, however, that neurites in PC12 do not have comparable biochemical and functional features to DRG ones. In our experiments, anterograde (centrifugal) movement when NGF is administered at the AC accounts for ~10% of the total vesicles analyzed (Fig. 2C), consistently to what reported, for tracked quantum-dot conjugated NGF 17 , in the same neurons. e latter study, however, did not quantitatively analyze the 10% anterograde trajectories and did not study proNGF. Considering NGF axonal transport as purely retrograde was therefore previously suggested 17 . Nevertheless, we believe this smaller population of anterogradely moving vesicles containing NGF should not be neglected because: 1) it is increased (up to ~75%) when NGF is administered to the SC; 2) it is not observed with uo-proNGF applied to the AC nor to the SC; 3) it is decreased to ~3% upon the simultaneous administration of proNGF ( Supplementary Fig. 8). ese data point to anterograde centrifugal movement as a speci c feature of NGF but not of proNGF, with the latter interfering with the transport of the former. While these considerations strengthen the idea that NGF and proNGF are indeed two di erent signalling molecules, the biological significance of the observed anterograde movement remains to be established. It could indeed be that this is a key feature of survival/di erentiation responses; however, it could also be that in our experimental conditions NGF in the SC passively exploits the well-known Trk anterograde transport 34,35 and that p75 does not support proNGF anterograde movement in this cellular model.
Secondly, proNGF and NGF vesicles contain a di erent number of molecules per vesicle (Fig. 3). Data demonstrate that each vesicle mostly hosts 1 or 2 proNGF and between 2 and 8 NGF dimers, when NTs are administered separately, while proNGF number is drastically decreased to one, when NTs are co-administered, indicating that NGF provides a veto or competing signal for proNGF internalization and transport. is nding does not appear to match with the recently proposed idea that the typical functional signaling endosome consists of a single NT dimer bound to a single pair of TrkA receptors 17 . Rather, it is consistent with a model that envisions the existence of a larger ligand-receptor complex per vesicle, in which a higher number of NTs are clustered together. is discrepancy could arise from the largely di erent steric hindrance of the two uorolabels used in the two studies. In fact, Cui et al. used Quantum Dots (QD)-conjugated NGF (each QD putatively coupled to a single NGF dimer) 17 . As QD volume is up to 70 times that of NGF ( Supplementary Fig.  10), the QD-NGF conjugate might impair NGF clustering 14 , thus artifactually preventing the simultaneous internalization of several NGF molecules. is would lead to a decrease in the observed number of QD-NGF internalized and transported per vesicle. Conversely, using much smaller organic dyes to uorolabel NGF, as done here, might allow accommodating a physiologically higher number of clustered NGF molecules that could easily nd room in the same vesicle. In this context, it is worth mentioning that, using 125 I labelled NGF ( 125 I -NGF), Campenot and coworkers 36 argued that the total amount of 125 I -NGF transported to the soma would require two orders of magnitude more vesicles with respect to what measured by Cui et al. 17 , if such vesicles contained only one NGF molecule.
Finally, since in physiological contexts NGF and proNGF coexist, the two forms of the NT were co-administrated to the AC (Fig. 4), highlighting a competitive mechanism between them (proNGF ux and number of molecules per vesicle was strongly inhibited and reduced by NGF co-administration). Taking these data together with immuno uorescence data (Supplementary Figure 5), we shall propose two possible mechanisms for the observed competition, which are interestingly not mutually exclusive. First, there could be simply a competition for the receptors available on the plasma membrane of the axon tip. Immuno uorescence analysis shows that the vast majority of uo-NGF particles are associated with TrkA receptors, whereas only a subset of them associate with P75, and uo-proNGF particles associate with both TrkA and P75 receptors. us, the two neurotrophin forms would compete for TrkA binding. e nature of proNGF -TrkA interaction is still elusive and likely occurs with less a nity than NGF-TrkA interaction 37 ; in any case, free TrkA could be necessary for proNGF internalization, and sequestration of free TrkA by NGF could a ect proNGF transport so that the only possibility is that it enters via p75; the latter has well recognized slower internalization rate compared to TrkA in peripheral neurons 12 , and this could nally result in a reduction of uo-proNGF transport compared to that of uo-NGF in our observation-time window. However, NGF/proNGF reciprocal signalling per se may alter endocytosis and transport. us another possibility is that NGF, rather than being a simple competitor of proNGF for TrkA receptor binding, activates some signalling pathway that prevents proNGF to internalize. It is useful to recall here that di erent proNGF/NGF mixtures exert a signalling pro le that is di erent from that raised separately by the two neurotrophins and that depends on the ratio between the two proteins 7 . In any case, whatever the competition mechanism, the most important nding emerging from these data is that proNGF retrograde signalling, which is likely to be pro-apoptotic, can emerge more when an NGF trophic signalling is absent. One could therefore expect that NGF pro-survival signalling is dominant with respect to proNGF one. Alterations in the NGF to proNGF ratios might have direct consequences in the transport uxes and hence in the availability of the two NTs in a competitive way, adding a new functional consequence to the emerging importance of the homeostatic regulation of proNGF to NGF metabolism 9,38 .
From a general perspective, then, the labeling platform proposed here could also be exploited to achieve the puri cation of vesicles carrying NTs and/or the respective proNTs, leading to a proteomic characterization of NT signalling endosomes. is would lately allow to discover whether the molecular composition and identity of tra c and signalling endosomes are dependent on the transported cargo 39,40 . Furthermore, we envisage the possibility of using our approach for the setup of chemical or functional genomic screening assays for compounds or genes that regulate the velocity of NTs transport.

Materials and Methods
Human proNGF cDNA cloned in pET11 vector 22 was used as template. e cDNA coding sequences of each tag was inserted into proNGF, by SCIENTIFIC REPORTS using an insertional mutagenesis procedure adapted from the site-directed mutagenesis method 24 . wt and tagged proNGF proteins were expressed in E. coli and puri ed using a modi ed protocol 22 from the previously published method 28 .
Labelling of tagged proNGF and NGF. 10 µ g of proNGF-YBBR or NGF-YBBR were incubated for 30 minutes at 37 °C in a thermomixer at 300 rpm with a reaction mix (10 mM MgCl 2 , 10 µ M CoA-alexa488/ CoA-alexa 647 or CoA-biotin and 2 µ M Sfp Synthase (SfpS) (New England Biolabs), in phosphate bu er up to 250 µ l nal volume. e synthesis of CoA-biotin and CoA-uorophore conjugates has been performed as previously described 37 .
Rat Dorsal Root Ganglion Neurons (R-EDRG-515 AMP, Lonza) were maintained in a humidi ed atmosphere at 37 °C, 5% CO 2 in Primary Neuron Basal medium (PNBM, Lonza) supplemented with L-glutamine, NSF-1 (2% nal concentration ) and antibiotics, following the manufacturer's instructions. For DRG neurons survival the media was supplemented by 100 ng/ml of NGF and was replaced every 4-5 days with a pre-warmed fresh one, being careful to always leave the main channels lled. When neurons were plated in micro uidic devices, a NGF gradient (obtained leaving the SC with 50 ng/ml of NGF, and the AC in the presence of 100 ng/ml) was used to induce axons to grow in the CC and to reach the AC. DIV8-15 cultures of DRG neurons were used in all experiments: within this range, we did not detect signi cant changes in the NT vesicles transport. In all the transport experiments, the uorescent NT has been administered in fresh media (not supplemented with NGF) by removing from the administration compartment all the pre-existing media.
In order to quantify proNGF and NGF biotinylation reaction yields, 2 µ l of all NGF/proNGF biotinylation reactions were treated under denaturing conditions (100 °C, 8 minutes in 2X Laemmli Sample Bu er), loaded on two gels (1 µ l for each gel) and electrotransferred to two PVDF membranes respectively. ese were blocked in TBST + 5% w/v BSA, then one of them was blotted with anti-NGF antibody (sc-549, Santa Cruz Biotechnology) (1:2000), while the other one was incubated with HRP-conjugated streptavidin (Zymed ® ) 1:10000 diluted in blocking solution.
To study the signal transduction e ectors, PC12 cells were cultured in P100 Petri dish to reach con uence, starved o.n. in a serum-free medium and then incubated with native NGF, tagged NGF, biot-NGF and uo-NGF (150 ng/ml) at 37 °C. A er 15 min cells were washed in ice-cold PBS and lysed in RIPA bu er supplemented with proteases and phosphatases inhibitors. 50 µ g of each clari ed lysate were loaded on a gel and electrotransferred to PVDF membranes. ese were rst blotted using the antibody anti-Phospho-PLCγ 1, anti-Phospho-p44/42 MAPK and anti-Phospho-Akt. e primary antibody was detected by using an anti-mouse or rabbit secondary antibody HRP-conjugated.
Microscopy. All microscopy measurements have been conducted in a wide eld microscope (Leica DM6000, equipped with a 4-laser TIRF-AM module) at 37 °C, 5% CO 2 . Transmitted light imaging has been performed in di erential interference contrast (DIC) con guration. For epi-uorescence microscopy 488 and 633 solid state lasers have been used to excite Alexa488 and Alexa647 respectively.
Single Particle Tracking and Number analysis. Single Particle Tracking analysis was performed by custom made Matlab scripts following the approach previously described 41 . Colocalizing trajectories were estimated by visual inspections of extracted trajectories in the green and red uorescence channels, a er an initial automatic selection based on the average time and positions of each trajectory in each micro uidic channel. is work is licensed under a Creative Commons Attribution 4.0 International License. e images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/