Syntaphilin-Mediated Docking of Mitochondria at the Growth Cone Is Dispensable for Axon Elongation In Vivo

Abstract Mitochondria are abundantly detected at the growth cone, the dynamic distal tip of developing axons that directs growth and guidance. It is, however, poorly understood how mitochondrial dynamics relate to growth cone behavior in vivo, and which mechanisms are responsible for anchoring mitochondria at the growth cone during axon pathfinding. Here, we show that in retinal axons elongating along the optic tract in zebrafish, mitochondria accumulate in the central area of the growth cone and are occasionally observed in filopodia extending from the growth cone periphery. Mitochondrial behavior at the growth cone in vivo is dynamic, with mitochondrial positioning and anterograde transport strongly correlating with growth cone behavior and axon outgrowth. Using novel zebrafish mutant lines that lack the mitochondrial anchoring proteins Syntaphilin a and b, we further show that Syntaphilins contribute to mitochondrial immobilization at the growth cone. Syntaphilins are, however, not required for proper growth cone morphology and axon growth in vivo, indicating that Syntaphilin-mediated anchoring of mitochondria at the growth cone plays only a minor role in elongating axons.


Introduction
Nervous system formation and function critically rely on mitochondria. The ability of mitochondria to produce ATP via oxidative phosphorylation and to buffer cytosolic cal-cium is especially important in neurons that have a high energy demand and require proper ion homeostasis. Mitochondrial dynamics, including transport, fission, and fusion, contribute to the correct distribution of mitochondria in axons and are therefore essential regulators of mitochondrial functions (Trevisan et al., 2018). Mitochondria are targeted to regions distant from the cell body, such as the axonal growth cone and synaptic terminals, through their active transport along microtubules (Melkov and Abdu, 2018). Overall, the importance of mitochondrial dynamics for proper neuronal development and function is emphasized by the large number of neurologic disorders caused by mutations affecting mitochondrial proteins (Misko et al., 2010;Bertholet et al., 2016).
During development, distal growing axons and growth cones contain higher densities of mitochondria compared with proximal axonal regions (Morris and Hollenbeck, 1993). This asymmetrical distribution is thought to be necessary for axon growth (Smith and Gallo, 2018). For instance, in vitro, increasing mitochondrial density in the distal axon and growth cone by overexpressing the mitochondrial biogenesis regulator peroxisome proliferatoractivated receptor gamma coactivator-1␣ (PGC-1␣) was shown to increase axonal length (Vaarmann et al., 2016). Other studies have reported that factors promoting or inhibiting axon growth regulate mitochondrial localization at the growth cone (Beck et al., 2012;Sainath et al., 2017). For example, adding the growth-promoting factor BDNF to cultured neurons increases distal mitochondrial density, whereas adding the repulsive guidance cue ephrin-A5 or using chondroitin sulfate proteoglycans as a nonpermissive substrate both cause mitochondria to leave the growth cone (Beck et al., 2012;Sainath et al., 2017). While these in vitro studies have highlighted a significant role for mitochondria in axon extension and growth cone morphology, mitochondrial dynamics at the growth cone during axon pathfinding have never been assessed in vivo.
The mechanisms responsible for maintaining mitochondria in distal growing axons are not yet fully understood. Mitochondria are able to attach to both microtubule and actin cytoskeletons (Boldogh and Pon, 2006), and could therefore be immobilized on microtubules in the growth cone central area after being transported along developing axons. Such stalling might be mediated by Syntaphilin (Snph), a mitochondrial membrane protein able to directly tether mitochondria to microtubules, thereby stalling mitochondrial transport (Kang et al., 2008). By immobilizing mitochondria, Snph was shown to reduce synaptic plasticity (Kang et al., 2008) and modulate axon branching (Courchet et al., 2013). Snph was further found to prevent mitochondrial transport toward the distal axon and inhibit axon regeneration after injury (Zhou et al., 2016). However, whether Snph also docks mitochondria at the growth cone during axon elongation has never been tested.
The zebrafish embryo offers a unique accessibility and transparency to monitor mitochondrial distribution and transport along axons in vivo (Pluci Ј nska et al., 2012;Campbell et al., 2014;Drerup et al., 2017;Wehnekamp et al., 2019). Here, we used time-lapse imaging of mitochondria in single retinal axons as they elongate along the optic tract in zebrafish to characterize for the first the time in vivo mitochondrial dynamics at the growth cone during axon pathfinding. We show that mitochondrial distribution at the growth cone correlates with axon growth status, and that mitochondrial transport toward the growth cone correlates significantly with axon elongation. We further demonstrate that Snph contributes to mitochondrial docking at the growth cone during axon pathfinding. However, growth cone morphology and axon elongation are unaffected in snph mutants, indicating that the direct anchoring of mitochondria to growth cone microtubules by Snph only plays a minor role in axon elongation.

Zebrafish husbandry
All animal procedures were performed in accordance with the University of South Carolina's Institutional Animal Care and Use Committee (IACUC). Zebrafish (Danio rerio) wild-type (WT; AB strain) and lakritz (lak) mutant (Kay et al., 2001) embryos were obtained from natural matings, raised at 28.5˚C in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl 2 , and 0.33 mM MgSO 4 ) in the presence of 150 mM 1-phenyl-2-thiourea (PTU; Sigma-Aldrich) to prevent pigment formation, and staged by age and morphology (Kimmel et al., 1995). Zebrafish possess a polygenic sex determination system, and sex-associated chromosomal regions are not fixed for the species (Liew et al., 2012). Our experiments were conducted on embryos before the onset of sexual differentiation, which occurs only at ϳ2.5 months after metamorphosis is complete (Maack and Segner, 2003).

TALEN-mediated mutagenesis and genotyping of Snph mutants
The transcription activator-like effector nucleases (TALENs) to target zebrafish snpha and snphb were designed using TAL Effector Nucleotide Targeter 2.0 (https://tale-nt.cac.cornell.edu/node/add/talen; Cermak et al., 2011). The TALENs for snpha contained the following repeat variable di-residues (RVDs): HD NI HD NN HD HD NI HD NG NN HD NI NN HD NI NN and NI NN HD HD NN NG NN HD NI HD NN NG NI NN NI NG NN HD, which targeted the sequences CACGCCACTGCAGCAG and AGCCGTGCACGTAGATGC, respectively. The TALENs for snphb contained the following RVDs: NN NI NN HD NI NI NG NI HD HD NG HD NI HD HD HD and NN HD HD NG NN NI NG NN HD NI HD NI HD HD NG HD, which targeted GAGCAATACCTCACCC and GCCTGATGCACACCTC, respectively. The TALEN cDNAs were assembled as described previously (Sakuma et al., 2013) and subcloned into pCS2pTAL3DD and pCS2pTALRR (Dahlem et al., 2012). Capped RNAs were synthesized from NotIdigested TALEN expression plasmids using SP6 RNA polymerase (Promega) in the presence of m´G(5´)ppp(5´)G RNA Cap Structure Analog (NEB). One nanoliter of solution containing a pair of TALEN RNAs (0.4 g/l each) was injected into zebrafish embryos at the one-cell stage. Deletion mutations in the target region were detected using heteroduplex mobility analysis (Ota et al., 2013). The following primers were used: AGAATCATGGCATTCGTC-CTC and TGAAGCCTCTCCACATTTTCTT to detect the 14 bp deletion in snpha; and AATGATAACCATGGCATTC-GAC and CTTTAAGCCGTGCTCTCAGGT to detect the 4 bp deletion in snphb. PCR products were separated on 12% or 20% Tris-borate-EDTA acrylamide gels or on a 4% Metaphor gel (Lonza). The snphb mutation was additionally detected using high-resolution melting analysis (Parant et al., 2009) using the following primers: AG-CAATACCTCACTCCACTG and GCCTGATGCACAC-CTCTTTC.
Cloning of snpha and snphb coding sequences mRNA from embryos at 24 h postfertilization (hpf) was isolated using TRIzol (Thermo Fisher Scientific) and the RNeasy Mini Kit (Qiagen). cDNA was prepared from RNA using SuperScript III First-Strand Synthesis System (Thermo Fisher Scientific). The following primers were used to amplify zebrafish snpha and snphb cDNA: snpha forward, TGTCCTTCTGCATCCATGTC; snpha reverse, TCAGATAGGTGTCGCTCTTTC; snphb forward, ATG TCTTCGCCTTCAAATAAAAG; and snphb reverse, TCATATATTCATTCCCCCTGG. Amplicons were subcloned into Invitrogen PCRII-TOPO (Thermo Fisher Scientific), and were sequenced to verify gene identity and confirm sequence orientation for the generation of sense and antisense RNA probes.

DNA plasmid constructs
All expression vectors were constructed using the Tol2kit Gateway cloning system (Kwan et al., 2007). We generated a pME-mitoEGFP entry clone by adding the mitochondrial targeting signal of the zebrafish cox8a gene to the 5´end of EGFP sequence using a BP-compatible forward primer. We generated a pME-Lifeact-TagRFP en-try clone by fusing the Lifeact sequence (Riedl et al., 2008) upstream of TagRFP using a BP-compatible forward primer. Lifeact-TagRFP was used to label F-actin in retinal axons and improve the visualization of growth cone filopodia. We generated a pME-TagBFP-snphb by adding the coding sequences of TagBFP (Evrogen) and a linker peptide (SGLRSRV) to the 5´end of snphb. The p3E-2A-TagRFPCAAX-pA entry clone that encodes a 2A peptide (Provost et al., 2007) and TagRFP targeted to the plasma membrane by the prenylation motif of Ras (Moriyoshi et al., 1996) was a gift from Dr. Kristen Kwan (University of Utah, Salt Lake City, UT). The p5E-isl2b plasmid that drives specific expression in retinal ganglion cells (RGCs) was described previously (Pittman et al., 2008). Final isl2b:mitoEGFP-2A-TagRFPCAAX, isl2b:Lifeact-TagRFP, and isl2b:TagBFP-snphb plasmids were generated using LR reactions with the pDestTol2pA2 backbone (Kwan et al., 2007).

RNA isolation and reverse transcriptase-coupled droplet digital PCR
For RNA isolation from whole embryos, 15 dechorionated WT embryos were lysed and homogenized in 500 l TRIzol at cleavage, pharyngula prim-5 (24 hpf), long-pec (48 hpf), protruding mouth (72 hpf), larval day 4 (96 hpf), and larval day 5 (120 hpf). For RNA isolation from eyes at 72 hpf, 50 eyes were manually dissected from WT or lak mutant embryos. Total RNA was isolated using the Directzol RNA Miniprep Kit (Zymo Research) and eluted into 20 l of nuclease-free water. cDNA was synthesized with 5 g of purified RNA as an input using the SuperScript III System. Digital droplet PCR (ddPCR) was then performed on a QX200 AutoDG Instrument (Bio-Rad) using predesigned TaqMan primers and probes for zebrafish snpha, sphnb, gapdh, and 18s (Table 1). Briefly, ddPCR reactions were assembled using Bio-Rad 2ϫ ddPCR Supermix for Probes (no dUTP) and contained 1 g of cDNA and 250 nM primers/probe. PCR amplification was performed with the following parameters: 39 cycles of 94°C for 30 s and 60°C for 1 min.

In situ hybridization
In vitro transcription of digoxigenin-labeled probes was performed using the RNA Labeling Kit (Roche Diagnostics) according to manufacturer instructions. Embryos were dechorionated at the appropriate developmental stages and fixed in 4% paraformaldehyde in PBS, pH 7.4, for 2 h at room temperature and overnight at 4˚C. Wholemount in situ hybridization (ISH) was performed as described previously (Thisse and Thisse, 2008). Sense probes were used as controls. After staining, embryos were cleared in 80% glycerol. Images were acquired using an Olympus SZX16 stereomicroscope equipped with

Imaging of mitochondria in retinal axons
Both isl2b:mitoEGFP-2A-TagRFPCAAX (30 pg) and isl2b:LifeAct-TagRFP (10 pg) plasmids were injected together with transposase mRNA into one-cell stage WT or snph db embryos. At 30 -32 hpf, embryos were sorted for fluorescence, anesthetized in 0.015% tricaine, and embedded in a lateral view in 1% low-melt agarose in E3 medium ϩ PTU. Their right eye was removed using a pulled glass pipette with a short taper as described in previous studies (Poulain et al., 2010;Poulain and Chien, 2013;Gaynes et al., 2015). Of note, removal of the contralateral eye preserves the underlying neuroepithelium, which ensures that the optic tract environment is not changed by the surgery and that retinal axon elongation and guidance are not affected. Embryos were allowed to recover until 46 hpf, when they were reanesthetized and about one-third of the yolk was removed by squeezing it out through a small hole torn with sharpened tungsten needles. Embryos were allowed to recover, reanesthetized at 50 -52 hpf, and remounted in a lateral view in 1% low-melt agarose in E3 medium ϩ PTU ϩ tricaine in a membrane-bottomed Petri dish for time-lapse imaging on a Leica TCS SP8X laser-scanning confocal microscope equipped with LAS X software, HyD detectors, and a 40ϫ objective (digital zoom 3, pinhole 1.25). Z-series were acquired for up to 3 h with 512 ϫ 512 pixel resolution at 1 min intervals to minimize photobleaching. Z-intervals were 1 m with a z-range of ϳ35-40 m to account for potential movement of the embryo. Chamber temperature was maintained at 28.5°C. Maximal intensity projections for each time point were compiled and aligned using ImageJ software (Schindelin et al., 2012;Schneider et al., 2012; RRID: SCR_002285) and the StackReg plugin (Thévenaz et al., 1998). Kymograph analyses of mitochondrial movement were performed using the ImageJ plugins KymoToolBox (Zala et al., 2013;RRID:SCR_016098) and KymoAnalyzer (Neumann et al., 2017). The number of mitochondria arriving at and leaving the growth cone were counted manually from the time-lapse sequences.
For the visualization of mitochondria in single mature retinal axons at 120 hpf, embryos were injected with isl2b:mitoEGFP-2A-TagRFPCAAX (15 pg) and transposase mRNA at the one-cell stage and were selected for fluorescence in the optic tract at 96 hpf, after which the right eye was removed as described above. Larvae were reanesthetized at 120 hpf and mounted in a lateral view in 1.5% low-melt agarose in E3 medium ϩ PTU ϩ tricaine in a membrane-bottomed Petri dish. Imaging was per-formed using a 40ϫ objective with digital zoom 2.5 and pinhole 1.0. Time-lapse recordings were performed in 35 z-planes (0.80 m z-interval), allowing acquisition at 1 frame every 15 s over 10 min.

Time-lapse image analysis
Image analysis was conducted using ImageJ (Schindelin et al., 2012;Schneider et al., 2012). For 2D analyses, the TagRFP/TagRFPCAAX signals from stack maximal projections were used to manually segment the growth cone total and central areas. The growth cone central area was determined by tracing the perimeter of the growth cone body, not including filopodia. As such, the central area is an arbitrary outline based on growth cone morphology and fluorescent intensity that presumably corresponds to growth cone central domain and transition zone (it does not necessarily correspond to what is commonly defined as the growth cone central domain based on cytoskeletal components). Regions of interest (ROIs) defined as the total and central areas were used to segment the mitoEGFP signals on binary images using the "particle analysis" tool. The following two different thresholds were used to analyze mitochondrial particles: a threshold of 55 was applied to the growth cone central area to segment the main mitochondrial cluster; and a threshold of 20 was applied to the growth cone total area to segment smaller individual mitochondria present in the peripheral area (Fig. 1B). A threshold of 20 was also applied to analyze pioneering mitochondrial clusters using the particle analysis tool. The number, area, and presence (percentage of time of advance) of pioneering mitochondrial clusters were analyzed in every frame of an advancing growth cone central area. The peripheral area, comprising filopodia, was calculated by subtracting the growth cone central area from the growth cone total area. The number of filopodia was counted manually. We drew a straight line from the mitochondrial cluster boundary to the most distal outline of the growth cone central area where two filopodia join together to measure the distance between the main mitochondrial cluster and the growth cone leading edge. We also defined the proximal growth cone as the most proximal boundary of the growth cone central area, where the growth cone transitions into the axon shaft. A growth cone was classified as advancing when the proximal growth cone advanced Ͼ1 m after an elongation of the growth cone central area (i.e. major Ͼ3ϫ minor axis length of the best fitting ellipse of the central area). Elongation rates were quantified by measuring the displacement between the proximal growth cone boundaries before and after advance and expressed in micrometers per minute.
Visualizations of 3D growth cones and mitochondria were prepared using FluoRender (RRID:SCR_014303; Wan et al., 2017). For volumetric analysis of mitochondrial occupancy at the growth cone, z-stacks were processed and analyzed with ImageJ. Images were thresholded to the lowest level that excludes the majority of noise pixels to obtain binary z-projections similar to raw z-projections, after which volumes were calculated using the Voxel Counter plugin. Volume data of mitochondria and the growth cone, expressed in cubic micrometers, were divided to obtain percentages of occupancy. Figures were prepared using Adobe Photoshop and Illustrator, and time-lapse videos were assembled using ImageJ.

Statistics
Data were analyzed and graphs were produced using Prism (GraphPad Software). Data are presented as the mean Ϯ SEM. Statistical tests were applied as indicated and their mitochondria were imaged in a lateral view between 50 and 54 hpf (⌬t ϭ 1 min). A, Anterior; D, dorsal; OT, optic tectum; P, posterior; V, ventral. B, Growth cone total and central areas visualized with TagRFP/TagRFPCAAX were manually segmented and used as ROIs for segmenting mitochondrial signals using the particle analysis tool in ImageJ. Two different thresholds were used to analyze mitochondrial particles: a threshold of 55 was applied to the growth cone central area to segment the main mitochondrial cluster, and a threshold of 20 was applied to the growth cone total area to segment smaller individual mitochondria present in the peripheral area (red arrowheads). Time-lapse recordings were classified according to growth cone behavior. Growth cone leading edge and proximal growth cone are indicated in the merged image; see Materials and Methods for definitions. Lateral view, confocal maximal projections. Scale bar, 5 m. C, Volumes of mitochondria and GC total and central volumes were calculated using 3D analysis. GC total and central volumes visualized with TagRFP/TagRFPCAAX were calculated on binary z-projections and used as volumes of interest for segmenting mitochondrial volumes using the voxel counter plugin in ImageJ (see also Movie 1).

Mitochondrial localization at the growth cone correlates with axon outgrowth
To monitor mitochondrial behavior in developing axons in vivo, we performed high-resolution confocal time-lapse imaging of mitochondria in retinal axons elongating along the optic tract of zebrafish embryos between 50 and 55 hpf (Fig. 1A). We coexpressed EGFP targeted to mitochondria (mitoEGFP) and TagRFP targeted to the plasma membrane (TagRFPCAAX), and to actin filaments (Lifeact-TagRFP) in single retinal axons, and monitored labeled axons elongating along the contralateral optic tract toward the optic tectum ( Fig. 1, Movie 1). As described previously (Bovolenta and Mason, 1987;Holt and Harris, 1993), growth cones of retinal axons in the tract were slender with numerous filopodia protruding and retracting dynamically. We could distinguish two classes of axons depending on their outgrowth status ( Fig. 2A). Advancing axons had very elongated growth cones, whereas pausing, not growing, axons had rounder growth cones with filopodia oriented in all directions, supporting the notion that growth cone shape correlates with growth cone behavior (Bovolenta and Mason, 1987;Mason and Wang, 1997). For the duration of our time-lapse videos (range, 30 -160 min), several growth cones were found to alternate between pausing and advancing behaviors and could therefore be classified into both categories (Movies 2, 3, 4).
As previously observed in vitro (Morris and Hollenbeck, 1993), mitochondria were abundantly detected at the growth cone ( Fig. 2A, Movies 1, 2, 3, 4). The majority of mitochondria clustered in the microtubule-rich growth cone central area, whereas smaller mitochondria regularly appeared in the peripheral area along protrusions or actin filopodia (Movie 3). We analyzed mitochondrial localization and distribution at the growth cone by quantifying mitochondrial occupancy in the growth cone total and central volumes (Fig. 2B,C), peripheral mitochondrial vol-ume (Fig. 2D), mitochondrial occupancy in the growth cone total and central areas (Fig. 2B´,C´), the number of mitochondria in the peripheral area (Fig. 2D´), and the distance between mitochondria and the growth cone leading edge ( Fig. 2E; see also Fig. 1B,C). Mitochondrial occupancy was chosen as a measure to evaluate mitochondrial density in the growth cone since spatial overlap of mitochondria within the growth cone central zone did not allow the visualization and quantification of single mitochondria. Mitochondrial occupancy in the total and central volumes and areas were similar in advancing versus pausing growth cones (Fig. 2B-C´). We did note that when we exclusively compared growth cones that frequently alternated between pausing and advancing behaviors, a significant decrease in mitochondrial occupancy in the central area of advancing growth cones was observed (paired t test, p ϭ 0.006, t (6) ϭ 4.2). Furthermore, the distance of the main mitochondrial cluster to the leading edge was significantly different between pausing and advancing growth cones (Fig. 2E). When a growth cone advanced, its central area elongated substantially, but this forward extension was not accompanied with a forward advance of mitochondria (Fig. 3A, Movie 5). Hence, the distance between them and the leading edge increased twofold (Fig. 2E). Mitochondria repositioned near the leading edge when the growth cone regained a rounder shape (Fig. 3A), which often indicated a transition to a pausing state. Interestingly, although the majority of mitochondria lagged behind during growth cone extension, a smaller subset of mitochondria was always detected adjacent to the leading edge (Fig. 3, Movie 5). We analyzed the number and area of these "pioneering mitochondrial clusters," as well as the amount of time they were present when a growth cone advanced. We found that pioneering mitochondrial clusters had on average a total area of 1.1 Ϯ 0.4 m 2 , and that 2.6 Ϯ 0.3 mitochondrial clusters positioned near the leading edge in 43 Ϯ 9.5% of the time of advance. Importantly, pioneering mitochondria were consistently observed in every growth cone that elongated.
Altogether, our data show that mitochondria localize in the growth cone central area in vivo and are also present along filopodia extending from the growth cone periphery. The position of mitochondria in the growth cone central area further changes with the growth status of the axon.

Mitochondrial transport is coordinated with growth cone behavior and axon growth in vivo
Next, we asked whether transport of mitochondria in the axon shaft proximal to the growth cone also relates to growth status. We performed kymograph analysis to measure net mitochondrial transport as well as the percentage of time mitochondria spent in motion (Fig. 4A-C). In agreement with a higher mobility of mitochondria observed in developing versus mature axons (Lewis et al., 2016), Ͻ5% of all mitochondria were stationary (Fig. 4C). The majority of mobile mitochondria moved anterogradely, with a significantly higher percentage of anterograde versus retrograde trafficking in both pausing and advancing growth cones (Fig. 4B). An analysis of mito-Movie 1. Representative 3D visualization of mitochondria in a pausing growth cone. Video corresponding to Figure 1C showing mitochondria (green) in a distal retinal axon and growth cone (red) pausing along the optic tract.  chondrial transport over shorter timescales showed a similar proportion of mitochondria moving anterogradely (71 Ϯ 5.4% in long movies, 78 Ϯ 7.3% over shorter timescales), retrogradely (26 Ϯ 4.5% in long movies, 20 Ϯ 6.7% over shorter timescales), or remaining stationary (2.3 Ϯ 1.6% in long movies, 2.3 Ϯ 2.3% over shorter timescales), indicating that the length of time-lapse movies had no effect on mitochondrial transport parameters (two-way ANOVA, effect of time, p ϭ 0.9994, F (1.66) Ͻ 0.0001). Mitochondria spent more time moving when a growth cone advanced (Fig. 4C), and we observed a trend toward more mitochondria arriving versus leaving in both pausing and advancing growth cones (Fig. 4D). We finally found a strong correlation between the number of mitochondria arriving at the growth cone and the distance the axon elongated (Fig. 4E). Altogether, these results illustrate that mitochondrial transport toward the growth cone correlates with growth cone advance, suggesting that mitochondrial motility and axon elongation may be functionally linked or coregulated in vivo.

Zebrafish Syntaphilins are expressed in RGCs and localize to the growth cone
How mitochondria are maintained in the growth cone remains poorly understood. One possible mechanism involves the attachment, or docking, of mitochondria to microtubules by Snph. Snph was identified as an outer mitochondrial membrane protein that is able to dock mitochondria directly to microtubules via its microtubulebinding domain (MTB; Kang et al., 2008; see Fig. 6A). Snph was shown to be important in axons for the regulation of branching (Courchet et al., 2013) and synaptic plasticity (Kang et al., 2008). We decided to test whether Snph also plays a role in docking mitochondria at the growth cone during axon elongation in vivo. Due to the whole-genome duplication that occurred in the teleost lineage (Glasauer and Neuhauss, 2014), two snph genes, snpha and snphb, are present in zebrafish. Both Snpha and Snphb share a high sequence conservation with human SNPH in their MTB and mitochondrial transmembrane domains (TMs), strongly suggesting a conserved mitochondrial anchoring function (see Fig. 6A, Extended Data Fig. 6-1). We first analyzed the expression of snpha continued mitochondrial cluster and the growth cone leading edge. When growth cones elongate, the distance to the growth cone leading edge increases significantly. Data from 12 independent experiments (pausing, n ϭ 11; advancing, n ϭ 8) are shown as mean Ϯ SEM. Statistical analysis (B-E): unpaired t test, ‫‪p‬ءءء‬ Ͻ 0.001.

Movie 2.
Representative time-lapse recording of mitochondrial dynamics in an advancing growth cone. Time-lapse video corresponding to Figure 2A showing mitochondria (green, white) in a distal retinal axon and growth cone (red) elongating along the optic tract. Images were acquired at 1 min intervals for 48 min. Each frame is a confocal image stack maximal projection, lateral view, and anterior is on the left. Time stamp format: hours: minutes (hr:min). Scale bar, 5 m. [View online] Movie 3. A subset of mitochondria localizes to the leading edge of the growth cone during elongation. Representative time-lapse video corresponding to Figure 3A showing mitochondria (green, white) in a distal retinal axon and growth cone (red) elongating along the optic tract. Images were acquired at 1 min intervals for 36 min. Note that the growth cone is elongating from 00:10 to 00:25. During this elongation, most mitochondria lag behind, but a subset of smaller mitochondria localizes adjacent to the leading edge. Each frame is a confocal image stack maximal projection, lateral view, and anterior is on the left. Time stamp format: hours:minutes (hr:min). Scale bar, 3 m. [View online] Movie 4. Representative time-lapse recording of mitochondrial dynamics in a pausing growth cone. Time-lapse video corresponding to Figure 2A showing mitochondria (green, white) in a distal retinal axon and growth cone (red) pausing along the optic tract. Images were acquired at 1 min intervals for 48 min. Each frame is a confocal image stack maximal projection, lateral view, and anterior is on the left. Time stamp format: hours:minutes (hr:min). Scale bar, 5 m. [View online] and snphb during zebrafish development. Quantification of snpha and snphb mRNA levels using reverse transcriptase (RT)-coupled ddPCR showed that both snpha and snphb are maternally expressed and that mRNA levels for both snphs increase over later developmental periods, albeit at much lower levels for snpha (Fig. 5A). ISH further revealed a high expression of snphb in the nervous system at 48, 72, and 120 hpf (Fig. 5B), whereas snpha expression was not detectable using this approach (data not shown). Importantly, snphb was detected in the RGC layer at the time of retinal axon elongation (Fig. 5B), suggesting a possible role in anchoring mitochondria in retinal growth cones. To further test whether snphb is expressed by RGCs, we performed ISH on WT and RGCdeficient lak mutant embryos (Kay et al., 2001) at 72 hpf (Fig. 5C). Snphb expression was strongly reduced in the retina of lak compared with WT, while it appeared unchanged in the brain (Fig. 5C). This decreased retinal expression was further confirmed by RT-ddPCR performed on dissected eyes (Fig. 5D). Snphb mRNA levels were highly detected in WT but significantly reduced in lak, indicating that snphb is strongly expressed by RGCs. Interestingly, snpha transcripts were also detected in WT, albeit at much lower levels. Like snphb, snpha eye expression was significantly decreased in lak (Fig. 5D), indicating that both Snphs are present in RGCs at the time of axon elongation.
As snphb is highly expressed in RGCs, we finally tested whether Snphb could localize to the growth cones of retinal axons elongating along the tract. As we could not identify an antibody directed against mammalian Snph that would specifically recognize zebrafish Snphs (data not shown), we decided to coexpress TagBFP-tagged Snphb, mitoEGFP, TagRFPCAAX, and Lifeact-TagRFP in single retinal axons, and to monitor labeled growth cones advancing along the contralateral optic tract (Fig. 5E)  Movie 5. Representative time-lapse recording of mitochondrial dynamics in a growth cone that combines pausing and advancing. Time-lapse video corresponding to Figure 2A showing mitochondria (green, white) in a distal retinal axon and growth cone (red) that alternates between pausing and elongating behaviors. The growth cone is pausing during the majority of the recording, but advances intermittently (e.g., from 00:00 to 00:18, 01:39 to 01:55, and 02:24 to 02:39). Images were acquired at 1 min intervals for 160 min. Each frame is a confocal image stack maximal projection, lateral view, and anterior is on the left.

Syntaphilins contribute to mitochondrial immobilization at the growth cone
To test the function of zebrafish Snphs in retinal axons, we generated snpha and snphb mutants using TALENmediated genome editing (Fig. 6A,B). We targeted a region upstream of the MTB to ensure full loss of function and obtained two alleles consisting of a 14 and 4 bp deletion for snpha and snphb, respectively, which were predicted to cause premature termination of translation (Fig. 6B). As both snpha and sphnb are expressed in RGCs (Fig. 5D) and to avoid potential compensation mechanisms, we generated maternal zygotic double mutants (db) and confirmed by RT-PCR and cDNA sequencing that both mutant mRNAs were expressed and contained the corresponding mutations (Fig. 6B). No alternative transcripts were detected in db (data not shown). db were viable and fertile, and db embryos did not show any obvious morphologic abnormalities, as was reported for Snph knock-out (KO) mice (Kang et al., 2008).
To test whether zebrafish Snphs also contribute to mitochondrial anchoring in axons, we quantified mitochondrial transport in mature retinal axons of WT and db embryos at 120 hpf (Fig. 6C), when most mitochondria are known to become immobile (Lewis et al., 2016;Smit-Rigter et al., 2016). As expected, mitochondrial mobility was strongly increased in axons of db (Fig. 6D,E), demonstrating that the mitochondrial docking function of Snph is conserved in zebrafish.
We next performed time-lapse imaging of mitochondria in retinal axons elongating along the optic tract of db embryos to test whether Snphs also play a role in mitochondrial tethering at the growth cone of growing axons (Movies 6, 7, 8). We compared db with WT embryos that have a similar genetic background (datasets analyzed in Figs. 2-4). Interestingly, mitochondrial distribution at the growth cone was modified in db compared with WT ( Fig.  7A-D´). While a same mitochondrial volume and number were observed in the growth cone peripheral volume and  Figure 4. Anterograde mitochondrial transport correlates with axonal outgrowth. A, Representative kymograph of mitochondria (mitoEGFP, white) in a distal axon whose growth cone alternates between advancing and pausing. The first and last frames of the time-lapse recording (Movie 4) are shown with axon and growth cone delineated in orange. Confocal maximal projections. Scale bar, 5 m. The growth cone leading edge and proximal growth cone (green line) are indicated on the kymograph and kymograph analysis panels. B, Quantification of net transport, analyzed by counting the number of mitochondria moving anterogradely or retrogradely. Data from 12 independent experiments (pausing: n ϭ 11, advancing: n ϭ 8) are shown as the mean Ϯ SEM. Statistical analysis: two-way ANOVA with post hoc Bonferroni test, ‫‪p‬ءء‬ Ͻ 0.01, ‫‪p‬ءءء‬ Ͻ 0.001. C, Quantification of the percentage of time mitochondria spent in a mobile or stationary state. A significant increase in time spent in motion (grey asterisk, p ϭ 0.013) and a decrease in time spent in a stationary state (red asterisk, p ϭ 0.008) are observed proximally to growth cones that are advancing. Data from 12 independent experiments (pausing, n ϭ 11; advancing, n ϭ 8) are shown as the mean Ϯ SEM. Statistical analysis: unpaired t test. D, Quantification of mitochondrial flux, showing a trend toward more mitochondria arriving versus leaving the growth cone in both pausing and advancing growth cones. Data from 12 independent experiments (pausing, n ϭ 11; advancing, n ϭ 8) are shown as the mean Ϯ SEM. Statistical analysis: unpaired t test. E, Linear regression analysis between the number of mitochondria arriving at a growth cone and growth cone advance (n ϭ 12). area in WT and db (Fig. 7C,C´), mitochondria occupied a smaller proportion of the growth cone total and central volumes and areas in db (Fig. 7A,B´, Movies 6, 7, 8). Moreover, mitochondria were located further from the growth cone leading edge in db, which was most pronounced in pausing growth cones (Fig. 7D,D´). Interestingly, the decreased mitochondrial occupancy of the growth cone in db was accompanied with changes in mitochondrial flux. While a similar number of mitochondria arrived at the growth cone, more mitochondria left the growth cone in db versus WT (Fig. 7E), suggesting that Snphs dock mitochondria once they have reached the growth cone. The increased departure of mitochondria from the growth cone of db was associated with changes in net mitochondrial transport proximally to the growth cone. More mitochondria moved retrogradely in snph db, which was accompanied by a decreased percentage of mitochondria moving in the anterograde direction (Fig. 7F). No differences were found between WT and db in the amount of time that mitochondria moved in both directions (Fig. 7G).

Number of arriving mitochondria Axon growth (µm)
Finally, in addition to the decreased mitochondrial occupancy of the growth cone in db, we observed some changes in pioneering mitochondrial clusters. As in WT, we detected pioneering clusters in every db growth cone that advanced ( Fig. 8A-C). The number of clusters (Fig. 8E) and the amount of time they were present at the growth cone (Fig. 8F) did not vary; however, the total area occupied by pioneering mitochondrial clusters was significantly decreased in db compared with WT (Fig. 8D). Altogether, these observations demonstrate for the first time that Snphs contribute to mitochondrial clustering at the growth cone in vivo. Snphs play a role in anchoring mitochondria in the growth cone central domain, including both the main cluster and the pioneering mitochondria that dynamically appear along the leading edge of advancing growth cones. loss of Snphs would affect axon elongation. Analysis of the total and central growth cone areas (Fig. 7H) and the number of filopodia (Fig. 7I) did not reveal any differences in growth cone morphology between WT and snph db. Moreover, retinal axon growth rate was similar between WT and db mutants (Fig. 7J), indicating that Snphmediated mitochondrial docking at the growth cone is dispensable for axon elongation. Thus, we identified Snph as a mediator of mitochondrial docking at the growth cone in elongating axons in vivo, but this function plays only a minor role, if any, in axon growth.

Discussion
By performing confocal live imaging in the zebrafish embryo, we provide the first detailed characterization of mitochondrial dynamics during growth cone behavior in elongating axons in vivo. We show that mitochondrial distribution at and trafficking toward the growth cone are coordinated with axon outgrowth, which is in agreement with previous in vitro observations (Morris and Hollenbeck, 1993;Sainath et al., 2017) and highlights that mitochondria might play a role in axon outgrowth and pathfinding. We further provide evidence that Snph contributes to mitochondrial docking at the growth cone. However, growth cone morphology and axon elongation are unaffected in snph db mutants, indicating that the direct anchoring of mitochondria to growth cone microtubules by Snph plays only a minor role in axon elongation.
An intriguing finding of our study is the localization of mitochondria along growth cone filopodia that is independent of Snph. Since filopodia dynamics are highly dependent on actin filament polymerization and turnover, this observation suggests that mitochondria might associate with the growth cone actin cytoskeleton. Mitochondria have indeed been shown to be able to attach to actin via the myosin 19 molecular motor (Shneyer et al., 2016). Interaction with myosin 19 positioned mitochondria into actin-rich filopodia of U-2 OS osteosarcoma cells follow-ing stress (Shneyer et al., 2017). A localization of mitochondria to leading edge lamellipodia, filopodia, and invadopodia of cancer cells has further been shown to play an important role in cancer cell migration and metastatic potential (Cunniff et al., 2016;Smith and Gallo, 2018). Whether the interaction among mitochondria, myosin 19, and actin is relevant in the growth cone remains unknown. Interestingly, mitochondria appeared along growth cone filopodia in a transient manner, suggesting that they might also be transported along the dynamic microtubules that explore the growth cone periphery and play an essential role in growth cone adhesion and turning (Buck and Zheng, 2002;Suter et al., 2004). Our observation of pioneering mitochondrial clusters at the leading edge of the growth cone during elongation further supports the hypothesis that mitochondria might frequently attach to the plus-end of microtubules. An intriguing consequence of mitochondrial targeting to microtubule plus ends in filopodia would be the possibility to predict the net direction of axon growth based on mitochondrial appearance. Unfortunately, our imaging conditions did not have the resolution required for correlating mitochondrial peripheral distribution with filopodia protrusion, retraction, or stabilization. It will be important in the future to optimize the imaging approaches to quantify fine aspects of mitochondrial behavior in vivo and test the possible interaction between mitochondria and the cytoskeleton in the growth cone. Interestingly, the presence of pioneering mitochondrial clusters close to the leading edge is at least partially Snph dependent, suggesting that the targeting of mitochondria to peripheral filopodia and to the leading edge rely on different mechanisms. While mitochondria might be transported on actin filaments or dynamic microtubules in the periphery, they may accumulate on the stable microtubules that "end" in the central area of the growth cone and be captured there by Snph and other factors.
The interdependence between mitochondrial transport and axon outgrowth that we observed is consistent with in vitro work demonstrating a net anterograde mitochondrial trafficking in growing axons that is abolished when axons encounter a physical barrier (Morris and Hollenbeck, 1993). We now establish such a correlation for the first time in vivo by comparing various parameters of mitochondrial transport in pausing versus advancing retinal axons. Together with the observation of a growth statusdependent mitochondrial distribution at the growth cone, our data indicate that mitochondrial dynamics in the distal axon and growth cone are coordinated with growth cone behavior and axon outgrowth in vivo. Vaarmann et al. (2016) previously showed that axon length could be increased by overexpressing PGC-1␣, a central inducer of mitochondrial biogenesis, supporting the notion that mitochondria provide energy for axon growth. In line with this, pharmacological and genetic disturbance of the mitochondrial fission-fusion balance in cultured retinal neurons affected neurite length and caused neurite guidance errors (Steketee et al., 2012). On the other hand, other studies hinted at the ability of axons to grow with dysfunctional mitochondria (Yoon et al., 2012;Campbell et al., 2014). While depletion of the intermediate filament Movie 8. Representative time-lapse recording of mitochondrial dynamics in a growth cone of a snph db mutant embryo. Timelapse video showing mitochondria (green, white) in a distal retinal axon and the growth cone (red) elongating along the optic tract of a snph db embryo. Images were acquired at 1 min intervals for 120 min. Note that the growth cone is pausing during the first minutes (00:00 to 00:47), after which it advances. Each frame is a confocal image stack maximal projection, lateral view, and anterior is on the left.  protein Lamin B2 caused defects in mitochondrial membrane potential, morphology, and transport, it did not affect retinal axon initial growth and guidance along the optic tract in vivo (Yoon et al., 2012). In a similar manner, lack of Kif5Aa, which transports mitochondria anterogradely into axons, causes a lack of mitochondria in distal peripheral axons that leads to degeneration. Yet, axons grow normally in kif5Aa zebrafish mutant embryos (Campbell et al., 2014). As it remains unclear which cellular processes are specifically dependent on mitochondrial respiration (Smith and Gallo, 2018), it might be conceivable that ATP production through glycolysis can, at least partially, sustain axon growth. Previous studies have even suggested that embryonic neurons predominantly rely on continued volume in the peripheral growth cone (C), and of the number of mitochondria in the growth cone peripheral area (C´) in WT and snph db. D, D´, Quantification of the distance between the largest mitochondrial cluster and the growth cone leading edge in WT and snph db. We quantified the distance from the leading edge separately in pausing and advancing growth cones as it varies depending on growth cone status (Fig. 2E). Mitochondria are located further from the leading edge in retinal growth cones of db compared with WT. E-G, Reduced mitochondrial occupancy in the growth cone of snph db is accompanied by an increased removal of mitochondria from the growth cone. More mitochondria leave the growth cone per minute in db, while no difference is detected for arriving mitochondria (E). Quantification of net mitochondrial transport proximally to the growth cone (F) shows more mitochondria moving retrogradely in snph db, which is accompanied by a decreased percentage of mitochondria moving in the anterograde direction. No differences in percentage time in anterograde or retrograde motion were found (G). H, I, Quantification of growth cone morphology shows no differences in growth cone total and central areas (H) and number of filopodia (I) between db and WT embryos. J, Axon elongation is not statistically different between db and WT embryos. Data from 12 independent experiments per genotype are shown as the mean Ϯ SEM. Statistical analysis: unpaired t test, ‫‪p‬ءء‬ Ͻ 0.01, ‫‪p‬ءءء‬ Ͻ 0.001. glycolytic ATP (Surin et al., 2012). As growth cone turning is regulated by local calcium levels (Gasperini et al., 2017), the ability of growth cone mitochondria to buffer calcium might be more important than its capacity to generate energy. Future work analyzing mitochondrial calcium dynamics at the growth cone in vivo might thus be highly interesting.
We identified Snph as a mediator of mitochondrial docking at the growth cone in elongating axons in vivo. This function is consistent with its previously established role in axonal mitochondrial stalling (Kang et al., 2008;Chen and Sheng, 2013). Yet, our data suggest that Snph does not play an important role in stalling mitochondria along the axon at this early developmental stage. The lack of difference in the number of mitochondria arriving at the growth cone in WT versus snph db axons suggests that Snph does not inhibit mitochondrial anterograde transport toward the growth cone as it does in the context of axon regeneration (Zhou et al., 2016). An increased halting of mitochondria along the axon over time is instead consistent with an increased expression of Snph in the brain during development, both in zebrafish in our study and in the mouse (Das et al., 2003;Kang et al., 2008). Alternatively, Snph localization and functions might be differently regulated in the axon shaft versus growth cone during development.
While the loss of Snphs reduced the number of mitochondria at the growth cone, it did not cause a complete depletion of mitochondria and did not affect axon growth. The remaining mitochondria at the growth cone might thus compensate for a lower density by increasing their activity. In this regard, it is noteworthy that Snph KO mice sciatic nerves show enhanced regrowth capacity after injury (Zhou et al., 2016). Cultured cortical neurons from these Snph KO mice displayed increased ATP content on axotomy (Zhou et al., 2016), but this difference in energy production was not observed in baseline conditions. Alternatively, the number of mitochondria remaining at the growth cone in the absence of Snph might be enough for normal axon development, especially considering that glycolytic ATP might have an important role in sustaining axon growth (Surin et al., 2012). These remaining mitochondria are likely stabilized at the growth cone central domain by other anchoring factors, of which the outer mitochondrial membrane GTPases Miro1 and 2 are especially good candidates. Miro proteins are indeed involved in mitochondrial trafficking along microtubules (Devine et al., 2016) and can arrest mitochondrial movement in a calcium-dependent manner (Saotome et al., 2008), which is particularly relevant at the growth cone where calcium signaling regulates pausing, extension, and turning (Sutherland et al., 2014;Gasperini et al., 2017). In that context, an elegant study recently demonstrated that growth inhibiting substrates such as MAG or chondroitin sulfate proteoglycans decrease mitochondrial axonal transport by promoting the deacetylation of Miro1 by HDAC6 in a calcium-dependent manner (Kalinski et al., 2019). Finally, Miros have also been discovered to play a role in mitochondrial trafficking and positioning along actin filaments (López-Doménech et al., 2018), which might regulate mitochondrial positioning along filopodia in the growth cone peripheral area. Future studies examining the role of Miros in mitochondrial docking at the growth cone in vivo will thus be of great interest to decipher how mitochondrial positioning is regulated during axonal development.