Caveolin 1 is required for axonal outgrowth of motor neurons and affects Xenopus neuromuscular development

Caveolins are essential structural proteins driving the formation of caveolae, specialized invaginations of the plasma membrane. Loss of Caveolin-1 (Cav1) function in mice causes distinct neurological phenotypes leading to impaired motor control, however, the underlying developmental mechanisms are largely unknown. In this study we find that loss-of-function of Xenopus Cav1 results in a striking swimming defect characterized by paralysis of the morphants. High-resolution imaging of muscle cells revealed aberrant sarcomeric structures with disorganized actin fibers. As cav1 is expressed in motor neurons, but not in muscle cells, the muscular abnormalities are likely a consequence of neuronal defects. Indeed, targeting cav1 Morpholino oligonucleotides to neural tissue, but not muscle tissue, disrupts axonal outgrowth of motor neurons and causes swimming defects. Furthermore, inhibition of voltage-gated sodium channels mimicked the Cav1 loss-of-function phenotype. In addition, analyzing axonal morphology we detect that Cav1 loss-of-function causes excessive filopodia and lamellipodia formation. Using rescue experiments, we show that the Cav1 Y14 phosphorylation site is essential and identify a role of RhoA, Rac1, and Cdc42 signaling in this process. Taken together, these results suggest a previously unrecognized function of Cav1 in muscle development by supporting axonal outgrowth of motor neurons.

Cav1L loss-of-function results in muscular defects. Embryonic motility can be affected by defects in muscular development as well as neuromuscular development. To analyze if the musculature of Cav1L-morphant tadpoles is affected, muscular actin was visualized using Phalloidin (Fig. 1F,G). Embryos injected with the Co MO showed a highly organized and characteristic striped pattern (Fig. 1F,F'). However, loss of Cav1L function led to a disorganized actin cytoskeleton and defects in muscular integrity (Fig. 1G,G'). These data were confirmed by ultra-structure analysis of the muscle fibers revealing the characteristic striped pattern of the sarcomeric bundles in controls (Fig. 1H,H'), while the sarcomeric ultrastructure was highly disrupted in Cav1Lmorphant embryos displaying disorganized actin and myosin filaments (Fig. 1I,I'). Interestingly, somitogenesis per se seems not to be affected, as expression of myoD was not compromised in Cav1L morphants (supplementary Fig. S2). Taken together, knockdown of Cav1L results in severe muscular defects, which are likely the cause of the aberrant swimming behavior.
Cav1L is expressed in neural tissue and loss-of-function indicates a function in neural but not muscle tissue. In order to analyze how Cav1L affects muscle development and ultimately swimming behavior, we characterized the expression profile of cav1L during Xenopus laevis embryogenesis. RT-PCR showed that cav1L is maternally expressed at low levels in oocytes and early cleavage stages, while its expression increased at neurula and tadpole stages (supplementary Fig. S3A). Using in situ hybridization we detected cav1L expression at neurula stage 12.5-20 in the notochord and the neural plate ( Fig. 2A  Complementing the in situ data, Cav1 protein expression was analyzed by immunofluorescence using an antibody detecting both isoforms α and β (Fig. 3). Cav1 protein was detected in the notochord (Fig. 3A,C,F,G), the lung (Fig. 3C,D), the heart and the vasculature of the branchial arches and the tail (Fig. 3E,F) and the epidermis (Fig. 3I). Interestingly, Cav1 protein is also expressed in the nervous system, including cranial nerves and motor neurons (Fig. 3A,E,G). Transverse sections of stage 37 embryos confirmed that Cav1 is expressed in neuronal cells of the spinal cord, the notochord as well as the epidermis (Fig. 3H,J). Taken together, Cav1 expression (protein or RNA) was not detected in the musculature, neither by in situ hybridization nor by immunofluorescence staining. Thus, Cav1 likely functions in muscle innervation rather than directly in the musculature.
To further dissect if Cav1L function is required in neural versus mesodermal tissue targeted injections were performed. Swimming behavior was analyzed at stage 38 whereby we distinguished either between a circling movement (severe defect) or a slow forward movement (mild defect). Indeed, mesodermal injections did only cause few swimming defects (Fig. 4A, supplementary Fig. S4A). In contrast, targeted Morpholino injection into the neural tissue resulted in a high percentage of swimming defects. These defects could be significantly improved upon co-injection of cav1L RNA (Fig. 4B, supplementary Fig. S4B). This was also confirmed by analyzing the sarcomeric structure of these embryos using phalloidin staining (Fig. 4C-E). In contrast to Co MO-injected embryos (Fig. 4C), Cav1L morphants showed a disrupted actin network on the injected side (Fig. 4D), which was restored upon co-injection of a wild-type cav1L construct lacking the Morpholino binding site (res-cav1L, Fig. 4E). In contrast, co-injection of lacZ RNA did not improve the swimming behavior of the morphant embryos (Fig. 4B). Thus, these data indicate that Cav1 function is required in the neural circuits controlling muscular integrity.
Inhibition of voltage-gated sodium channels mimics the Cav1L loss-of-function phenotype. Cav1L is expressed in the nervous system, but it is also required for muscular function. Since it is well-known that defects in neuromuscular activity can lead to muscular atrophy [38][39][40] , we asked if the muscular phenotype may be caused by defects in muscular innervation. To this end we treated embryos-injected with the Cav1L MO or Co MO-prior to the onset of the swimming stage with Benzocaine, an inhibitor of voltagegated sodium channels (Fig. 4F). This treatment blocks the transmission of an action potential from the motor www.nature.com/scientificreports/ neurons to the muscle cells mimicking a denervated muscle. Embryos were cultured in the anesthetic until stage 38 when they were stained with Phalloidin to analyze muscle morphology. Actin filaments were correctly aligned in untreated controls, but Cav1L-morphant embryos displayed disorganized actin filaments (Fig. 4G,H). Benzocaine-treated embryos-including the uninjected controls-showed the same muscular defects as untreated Cav1L morphants (Fig. 4I,J). Interestingly, the severity of the morphant phenotype was not increased by Ben-  Scientific RepoRtS | (2020) 10:16446 | https://doi.org/10.1038/s41598-020-73429-x www.nature.com/scientificreports/ zocaine-treatment. Thus, these data suggest that the observed muscle phenotype is likely caused by defects in innervation.

Cav1L loss-of-function causes defects in axon outgrowth and neuronal morphology by affecting Rho GTPases. As Cav1L is expressed in motor neurons we next investigated if loss-of-function of
Cav1L also affects Xenopus nerve morphology. To this end, we analyzed motor neuron morphology by immunostaining using an N-CAM antibody on embryos injected unilaterally with 20 ng MO. In control embryos www.nature.com/scientificreports/ axons from motor neurons innervated the musculature in a characteristic chevron-shaped pattern (Fig. 5A). In contrast, motor neurons were severely affected by loss of Cav1L function; although they were still able to form axons, they randomly projected their axons in the periphery and did not follow the boundaries of the somitic muscles ( Fig. 5B-D). Furthermore, injection of the Cav1L Spl-MO, which is also more effective in blocking Cav1L expression compared to the translation blocking Cav1L MO (Fig. 1B), showed both severe pathfinding ( Fig. 5C) and outgrowth defects (supplementary Fig. S5B). In order to analyze if the neuronal phenotype is specific for Cav1L loss-of-function, rescue experiments were performed. While the injection of the Cav1L Spl-MO caused motor neuron outgrowth and pathfinding defects, co-expression of cav1L-HA RNA significantly rescued these defects (Fig. 5D, supplementary Fig. S5C), indicating that the observed neuronal abnormalities are indeed specific to Cav1L loss-of-function.
To examine if Cav1L also affects early neuronal patterning we analyzed if Cav1L loss-of-function affects the expression of the neuronal marker n-tubulin (neuron specific class IIβ tubulin) at early neurula stages (supplementary Fig. S6). Control embryos, as well as the majority of the Cav1L-morphant embryos showed normal n-tubulin expression in all four domains, indicating that loss-of-function of Cav1L has no impact on early neurogenesis and neuronal patterning in Xenopus embryos (supplementary Fig. S6). Furthermore, shh expression in the floorplate (supplementary Fig. 7) and BMP4 expression in the roofplate (data not shown) was also not affected suggesting that dorsal/ventral patterning of the neural tube is not disturbed in Cav1L morphants. Thus, the muscular defects of Cav1L-morphant embryos are likely caused by impaired muscular innervation due to defects in axonal outgrowth and pathfinding of motor neurons.
To further characterize axon outgrowth, neural tubes of stage 20 embryos, injected with 12 ng Morpholino and the lineage tracer GFP, were explanted and analyzed after one day of incubation. Consistent with the in vivo data Cav1L-morphant axons showed defects in outgrowth and morphology (Fig. 6). Again, we noted that the defects caused by the Cav1L Spl-MO-which is also more potent in blocking Cav1L expression compared to the translation blocking Cav1L MO-were more severe and fewer axons extended in vitro (Fig. 6G). This effect was specific and could be rescued by co-expression of cav1L-HA RNA (Fig. 6G). Furthermore, while control neurons within the explants extended long axons with the typical actin-rich growth cone (Fig. 6A,B), a significant larger number of axons from Cav1L-morphants displayed an increase in actin-rich filopodia-and lamellipodia-like structures ( Fig. 6C-F,I-L). The number of axons with filopodia, as well as the number of filopodia per axon was significantly increased (Fig. 6I,J). Likewise, the number of axons with lamellipodia and the lamellipodia per axon was significantly increased in morphants compared to controls (Fig. 6K,L). The defects in axon morphology were specific for Cav1L loss-of-function and could be rescued by co-injection of cav1L-HA RNA (Fig. 6I,K). Thus, Cav1L is not only expressed in motor neurons but also required for axonal outgrowth and morphology in vivo as well as in vitro.
As Rho GTPases are known regulators of lamellipodia and filopodia formation during axonal growth and guidance 41,42 and as Cav1 is known to modulate the activity of RhoA and Rac1 in mouse embryonic fibroblasts 43,44 we analyzed if constitutive active or dominant negative RhoA, Rac1 and Cdc42 are able to rescue the morphant phenotypes if injected into neural tissue. Indeed, dominant negative Rac1 as well as RhoA, but not the constitutive active mutants could partially rescue the swimming behavior as well as the neuronal defects caused by loss of Cav1L function (Fig. 7A-C). Conversely, constitutive active but not dominant negative Cdc42 partially rescued the aberrant swimming behavior and neuronal defects ( Fig. 7A-C). This indicates that Cav1L affects motor neuron morphology by suppressing RhoA and Rac1 activity and supporting Cdc42 activity.
The Cav1 Y14 phosphorylation site is required for the function of Cav1L in the locomotor system of Xenopus tadpoles. It has been shown that Src-dependent tyrosine (Y14) phosphorylation of Cav1 plays an important role in the regulation of RhoA and Rac1/Cdc42 activity [43][44][45] . In order to determine if this modification is required for Cav1L function in the nervous system of Xenopus tadpoles, we performed rescue studies with a non-phosphorylatable Cav1L mutant (CavY14A). Morpholinos in combination with cavY14A RNA were unilaterally injected at the two-cell stage and swimming behavior was analyzed at tadpole stages. The phosphorylation mutant was not able to rescue swimming defects if co-expressed with the Cav1L Spl-MO, while full-length cav1L partially rescued (Fig. 7D). This indicates that the Y14 phosphorylation is required for Cav1L function and possibly the regulation of Rho GTPases in the locomotor system of Xenopus laevis.

Discussion
A well-established neuromuscular communication is important for muscle development and integrity and loss of neural connectivity causes severe muscle atrophy [46][47][48][49][50] . Here we report a novel function of Cav1 in the development of the Xenopus neuromuscular system: Cav1L is expressed in motor neurons and required for muscular innervation. Loss of function causes impaired muscular integrity and paralysis of morphant embryos. As we find that axonal outgrowth and morphology are severely affected in morphants, while the induction of primary neurons per se is not compromised, we suggest that Cav1L function in neural development is likely limited to post-induction stages. Our data suggest a model whereby Cav1L functions in Xenopus motor neuron outgrowth by regulating lamellipodia and filopodia formation of axons (Fig. 7E). In the wild-type situation this is likely mediated by supporting Cdc42 activity and suppressing RhoA and Rac1 activity. In Cav1L morphant axons, lamellipodia and filopodia are not retracted and the growing axons display an increase in protrusions. Further Cav1L function likely requires the tyrosine 14 phosphorylation site, as this phosphomutant (Y14A) was not able to rescue the morphant swimming defects. Thus, these data indicate a novel function of Cav1L in in vivo development of Xenopus motor neurons and reveal an indirect role in muscular function and embryonic mobility.
Coordinated regulation of the small Rho GTPases Rac1, RhoA, and Cdc42 in lamellipodia and filopodia formation of the growth cone is essential for controlled growth and navigation of axons 41  www.nature.com/scientificreports/ inhibited Rac1/Cdc42-mediated axonal growth in human neurons derived from induced pluripotent stem cells 54 . It is currently unclear how Cav1L affects Rho GTPases in Xenopus motor neurons, however, in respect to the studies on isolated neurons, a mechanism affecting the activity of these GTPases seems likely. Evidence for a neuronal function of Cav1 has already been demonstrated by loss of function studies in mice. Cav1 knockout mice display distinct traits associated with progressive neurodegeneration such as deficits in motor coordination, gait abnormalities (shorter stride length), muscle weakness as well as a clasping and spinning phenotype 29 . Additionally, they also show behavioral changes associated with cholinergic dysfunction, characterized by impaired spatial memory, increased anxiety as well as reduced exploratory behavior in a new environment 29,32 . It remains unclear if Xenopus Cav1L morphants would also show neurodegenerative defects, as our Morpholino-mediated approach allowed us only to analyze its role in early neurodevelopment. In vivo as well as overexpression studies of mouse and human hippocampal cell cultures have shown that Cav1 positively regulates neuronal plasticity and neuronal intracellular signaling by recruiting neurotransmitters and neurotrophic factors to synaptic membrane lipid rafts [54][55][56][57] . Membrane lipid rafts are especially important for pro-survival and pro-growth receptor signaling in neuronal cells, since receptors and proteins required for synaptic communication mainly localize in these membrane domains 58 . Further, membrane lipid rafts at the leading edge of the growth cone are important reservoirs for signaling molecules, such as Rho GTPases, integrins and N-cadherins, which are essential for actin dynamics and adhesion [59][60][61] . Cav1 has been shown to modulate the nanoscale lipid organization of specialized membrane lipid raft domains by regulating the expression of metabolic proteins, such as Ppap2A (Lpp1), B3GNT5 and Siat9 (GM3 synthase), which are involved in glycosphingolipid, sphingolipid as well as ganglioside biosynthesis [62][63][64] . Ganglioside expression is tightly regulated during the development of the peripheral and central nervous system and misexpressions of these lipids are associated with progressive neurodegeneration in mice and humans [65][66][67] . Interestingly, mice lacking Siat9 and thus the expression of complex gangliosides, displays similar neurological deficiencies as Cav1 knockout mice 68,69 . Thus, the neurological defects caused by the loss of Cav1 likely result from a disturbed membrane lipid raft composition and the resulting mislocalization of signaling molecules regulating axonal biology.
Another mechanism by which Cav1 could regulate axonal growth is by affecting voltage-gated sodium channels and thereby regulating neuronal activity. In cardiac cells, it has been shown that the voltage-dependent sodium channel Na V 1.5 localizes to caveolae membrane domains 70 . Interestingly, the inhibition of voltage-gated sodium channels, such as Na V 1.9, by either knockout or Tetrodotoxin (TTX) treatment, impairs axonal growth of cultured mouse motor neurons 71 . These neurons display shorter motor axons as well as reduced spontaneous Ca 2+ transients, which are required for axonal growth and the establishment of synaptic connections [71][72][73][74][75] . It has been suggested that the activity of Na V 1.9 and consequently Na V 1.9-dependent spontaneous Ca 2+ transients are regulated by activated TrkB in neuronal cells 71,76,77 . Egawa et al. were able to show that Cav1 recruits receptor tyrosine kinase B (TrkB) to membrane lipid rafts in vivo and improves TrkB signaling in vitro in hippocampal neurons 36,55,57 . Although a direct link of Cav1 to the regulation of neuronal activity has not yet been described, Cav1 might have a function in this process by recruiting TrkB to MRLs in neuronal cells, which in turns modulates the activity of voltage-gated sodium channels located in these membrane domains.
A question that remains is if the swimming defects of Cav1L morphants are also impacted by defects in notochord development. Indeed, Xenopus Cav1L is highly expressed in the notochord, which serves as important structural element supporting locomotion in free-swimming larvae. Caveolae are essential structures protecting the notochord against mechanical stress [78][79][80] and their depletion causes mechanical-induced collapse of notochord cells 78,81 . In zebrafish embryos, it has been shown that defects in notochord development can affect both the axial skeleton as well as muscular innervation 82,83 . Interestingly, loss of cavin1b function in the zebrafish notochord, which abolishes caveolae-formation, affects swimming behavior 81 . However, in contrast to Xenopus Cav1L-morphant embryos, which were completely paralyzed when Cav1L MO was injected into both blastomeres of a two-cell stage embryo, zebrafish depleted of caveolae only swam shorter distances compared to wild type embryos, while overall swimming behavior was not affected 81 . In support of these findings we also noted that loss of function of Xenopus cavin1 does not affect the swimming behavior of Xenopus embryos (data not shown). Thus, the observed muscular defects of Xenopus Cav1L morphants are likely not caused by defects in notochord development. Moreover, single loss of function of either zebrafish cav1 or cav3 was not sufficient to induce mechanical induced lesions in the notochord suggesting that other caveolin proteins likely compensate for the loss of a single caveolin isoform 78,81 . We expect a similar situation in Xenopus, where different caveolin paralogs are expressed in the notochord 84 . Interestingly, we note that targeted injection of Cav1L morpholino to one side of the embryo caused paralysis only on the injected side; however, notochord cells intercalate during development, thus irrespective of the site of injection, we would expect both sides to be affected. Thus, in combination with our in vitro analysis showing axonal outgrowth and morphology defects these data argue for a direct function of Cav1L in neuromuscular development. For the detection of endogenous Cav1L mRNA, sense and antisense RNAs were amplified from xcaveolin-1α pCMV-Sport6 (RZPD, catalogue number IRBMp990B0725D). The caveolin1L-rescue construct (res-cav1L-HA) was cloned from xcaveolin1-HA pCS2+ 85 using the forward primer 5′-ATG CTA GCA TGG AAG AGG GTG TTC Scientific RepoRtS | (2020) 10:16446 | https://doi.org/10.1038/s41598-020-73429-x www.nature.com/scientificreports/ TCT ACA C-3′ and the reverse primer 5′-ATG CTA GCG AAT CGA TGG GAT CCT GCAAA-3′ to generate a truncated construct lacking the Morpholino binding side. The Cav1L Y14 phosphorylation mutant was cloned by site-directed mutagenesis from xcaveolin1-HA pCS2 + using the forward primer 5′-TGA AGA GGG TGT TCT  CGC CAC CAC GCC GGT CATC-3′ and the reverse primer 5′ GAT GAC CGG CGT GGT GGC GAG AAC ACC CTC  TTCA-3' . For overexpression analysis the following plasmids were used: lacZ pCS2+ 86 mGFP 87 , xcaveolin1-HA pCS2+ 85 , ca Cdc42 V12 pCDNA3.1 88 , dn Cdc42 N17 pCS2+ 89 ; the constructs ca RhoA (V12/V14) N17 pCS2 + , dn RhoA N19 pCS2 + , ca Rac1 V12 pCS2 + and dn Rac1 (N17) pCS2 + were cloned from their original pcDNA3.1 vectors 88,90 into the pCS2 + vector. For microinjection RNA was synthesized using the mMessage Machine Kit (Ambion, Life Technologies) according to the manufacturer's instructions.

Constructs.
Xenopus injection and phenotypic analysis. Xenopus laevis embryos were obtained by in vitro fertilization and staged according to Nieuwkoop and Faber 91 . All procedures were performed according to the German animal use and care law (Tierschutzgesetz) and approved by the German state administration Hesse (Regierungspräsidium Giessen). Microinjections were performed either into one blastomeres of two-cell stage embryos with an injection volume of 10 nl or into one blastomere of eight-cell stage embryos using an injection volume of 5 nl, respectively. To specifically target the musculature, embryos were injected in the dorsal blastomere of the vegetal hemispheres; to target the brain and spinal cord, embryos were injected into the dorsal blastomere of the animal hemispheres. For western blot analysis embryos were injected at the one-cell stage. mGFP RNA was always co-injected as lineage-tracer with a concentration of 75-80 pg.
Swimming behavior was analyzed by stimulating the escape response of the embryos in response to a touch stimulus. Swimming behavior was counted as normal, when the embryos were able to move their tail in both directions. Embryos were defined as having a mild defect, when they were still able to swim slowly, however their tail movement was visible impaired. Embryos with a severe swimming defect were completely paralyzed on the injected site and only moved in circles or were unable to swim. Preparation of protein lysates and Western blotting. Protein extracts were prepared by lysis of injected Xenopus embryos with insulin syringes in 10 µl/embryo Lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% NP40 (v/V), 0.1% SDS, EDTA-free 1 × CompleteProteaseInhibitor (Roche). Lysates were centrifuged for 15 min by 16.000g at 4 °C and supernatants were transferred to a fresh Eppendorf tube. Protein lysates were diluted 1:5 with 6 × Laemmli loading buffer (350 mM Tris-HCl pH 6.8, 9.3% Dithiothreitol, 30% (v/v) glycerol, 10% SDS, 0.02% Bromophenol Blue) denatured for 5 min at 95 °C and subsequently loaded to a 10 or 12% SDS-PAGE gel. Following separation, proteins were transferred to a nitrocellulose membrane (Whatman) by wet-or semi-dry blotting (Mini PROTEAN Tetra System; Trans-Blot-turbo Transfer System, BIO-RAD). The membrane was blocked in TBST-blocking buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl and 0.5% (v/v) Tween 20) containing 5% nonfat dried milk for at least 30 min. The following primary antibodies were used: anticaveolin-1 (Abcam, ab2910, 1:300), anti-GAPDH (AM4300, Thermo Fisher Scientific, 1:5000). The antibodies were removed by washing three times for 10 min each in blocking solution and the secondary anti-mouse-HRP antibody (sc-516 102, Santa Cruz Biotechnology, 1:5000) was applied for 1 h at room temperature. Chemiluminescence was detected using SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) and Odyssey Fc Imaging System (LI-COR Bioscience) and analyzed by Image Studio Software (LI-COR).

RT-PCR.
Whole mount immunofluorescence staining and in situ hybridization. Embryos  www.nature.com/scientificreports/ washed in PBS-TD six times for one hour each and then incubated overnight with secondary antibodies: antimouse Alexa 594 (Invitrogen, A-11005, 1:400), anti-mouse Alexa 488 (Invitrogen, A11029, 1:400), anti-rabbit Alexa 488 (Invitrogen, A-21206, 1:400). Embryos were washed six times for one hour each in PBS-TD at room temperature and then re-fixated in Dents overnight. For imaging, embryos were cleared in Benzyl-alcohol/Benzyl-Benzoate (Sigma Aldrich) (BA/BB; 1:2). Therefore, embryos were washed two times in 100% ethanol. For clearing embryos were incubated first in BA/BB for 10 min and then in fresh BA/BB for imaging. Embryos were imaged using the Zeiss Spinning Disc system (Axio Observer Z1 with a 25 × or 40 × water objective) or a fluorescence stereo-microscope (Leica, M165-FC).
Histological analysis of the musculature. For vibratome sectioning Xenopus embryos were sorted according to their fluorescence and fixed in MEMFA overnight. Embryos were then transferred to 25% glutaraldehyde (Roth) and embedded in a 1/10 mixture of gelatin/albumin (4.4% gelatin, 27% bovine serum albumin (BSA), 18% saccharose in 1 × PBS) and 25% glutaraldehyde. Embedded embryos were then trimmed to square blocks and sectioned with a thickness of 40 µm using the Leica Vt1000S vibratom. During sectioning and following immunostaining, sections were kept on coated microscope slides (Roth) in a humid atmosphere.
For electron microcopy embryos were fixed in 6.25% glutaraldehyde in 0.1 M cacodylate buffer (0.2 M Cacodylat, pH 7.2 adjust with 0.2 N HCl) overnight. Embryos were washed three times in Cacodylate buffer for 30 min each. For contrasting, samples were incubated in 1% OsO4 in 0.1 M Cocadylate (pH 7.2) for 60-90 min while gently shaking. Osmium was washed out through several washes and overnight incubation in 0.1 M Cocadylate. After contrasting the embryos were embedded in 2.5% Agar-Agar, and then trimmed to square blocks to allow orientation of the embryos during embedding in Spurr's resin. Prior to embedding in Spurr, embryos were dehydrated in an ascending alcohol-1.4-dioxid series. Dehydration occurred for 30 min each in a EtOH series (50%, 70%) and subsequently in Dioxan for 45 min. Subsequently, the embryos were first infiltrated in a mixture of Spurr's resin and Dioxan (1:1; 2:1) for 90 min each and then in pure Spurr overnight. The Spurr's resin was replaced with fresh Spurr and incubated 5-8 h under constant stirring. Finally, embryos were placed in embedding molds and embedded in Spurr for 16 h at 70 °C.
For electron microscopy 50-80 nm ultra-thin sections were prepared using an Ultracut microtome (Reichert) and contrasted with lead acetate and uranyl acetate on Copper-Rhodium (CU/Rh) grids (75 × 300 stiches). For light microscopy 2 µm semi-thin sections were prepared in a LKB-Pyramitom and contrasted in a methylene blue solution at 70 °C for 2-3 min.
Neural tube explants. Neural tubes were explanted from stage 19-22 embryos as described in 92 . Dissected neural tube explants were cultured in DFA medium (53 mM NaCl, 5 mM Na 2 CO 3 , 4.5 mM Potassium Gluconate, 32 mM Sodium Gluconate, 1 mM CaCl 2 , 1 mM MgSO 4 , 0.5 g/ml BSA, pH 8.3 with 1 M Bicine) on Poly-llysine (150 µg/ml, P-1399 Sigma-Aldrich) and Laminin (10 µg/ml, L2020 Sigma-Aldrich) coated chamber slides (Sarstedt) for 12-24 h at 18 °C. Explants were imaged using the Zeiss Axio Observer Z1 inverted microscope (63 × oil objective). Outgrowth as well as morphology of spinal neurons was counted for each explant individually. Axon length, filopodia as well as lamellipodia area was determined using ImageJ. The area covered by lamellipodia was calculated by subtracting the area of the axon (Aa) from the total area including lamelipodia (Ta). Lamellipodia area was then normalized to the total area (Ta) per axon. Number of filopodia per axon was calculated by normalizing the number of counted filopodia (Fn) by the axon length (AL). The growth cone was excluded in both calculations. Filopodia number: (Fn/ AL)*100 µm; Lamellipodia area: (Ta-Aa)/Ta.

Statistical analysis.
All experiments, if not indicated otherwise, were conducted at least three times. The total number of analyzed embryos (n) is indicated for each experiment. Normality of datasets was tested using D' Agostine & Pearson test, Shapiro-Wilk test and Kolmogorov-Smirnov test. Significance was calculated by using either a two-tailed unpaired Student's t-test and ordinary one-way ANOVA (Dunnett's multiple comparisons) or Mann-Whitney test (Box plots) (*p-value ≤ 0.05; **p-value ≤ 0.01; ***p-value ≤ 0.001) using Microsoft excel (2013) or GraphPad Prism8. Standard errors of the mean (s.e.m) are shown for each graph, except for supplementary figure SFig.6 where standard deviation (s.d.) is shown. Box plots are presented as Tukey box plots.