SCG10 is required for peripheral axon maintenance and regeneration in mice Free

Microtubules not only shape cell morphology, but also serve as tracks for intracellular transport (Conde and Caceres, 2009; Kapitein and Hoogenraad, 2015), and dysregulation of microtubule dynamics in neurons leads to a variety of neurodevelopmental and neurodegenerative diseases (Breuss et al., 2017; Sferra et al., 2020). Through decades of studies, many factors including microtubule regulators and microtubule-associated proteins (MAPs) have been identified to regulate microtubule dynamics and organization in a variety of ways (Bodakuntla et al., 2019; Goodson and Jonasson, 2018). In neurons, the microtubule network is tightly and spatiotemporally regulated by these factors (Kapitein and Hoogenraad, 2015). SCG10 (also known as stathmin-2 or STMN2), a neuron-specific microtubule regulator (Stein et al., 1988), belongs to the stathmin family, which includes other members such as stathmin (STMN1), SCLIP (STMN3) and RB3 (STMN4) (Charbaut et al., 2005). All members contain a conserved stathmin-like domain, which sequesters two tubulin heterodimers to decrease the effective concentration of tubulin, which in turn might increase the microtubule catastrophe rate (Charbaut et al., 2001). In the stathmin-like domain of SCG10, four conserved serines (Ser50, Ser62, Ser73 and Ser97) can be phosphorylated by different kinases, for example, Ser62 and Ser73 can be phosphorylated by mitogen-activated protein kinase and c-Jun NH2-terminal kinase, and Ser50 and Ser97 by cAMP-dependent protein kinase, and the phosphorylation state serves as a switch to control the tubulin binding activity of SCG10 (Antonsson et al., 1998). Expressing wild-type or non-phosphorylatable SCG10 mutants disassembles microtubules, whereas expression of the SCG10 phospho-mimicking mutant has no effect on microtubule stability (Wang et al., 2013b). Furthermore, the SCG10 N-terminus contains 34 more amino acids than stathmin; of these, cysteines 22 and 24 are both palmitoylated, and guide SCG10 to the Golgi apparatus and then for transport with vesicles to growth cones (Di Paolo et al., 1997; Lutjens et al., 2000). Several studies reported that the SCG10-C22A&C24A mutant or N-terminus deletion mutant of SCG10 distributed diffusely, as does stathmin, but after adding the SCG10 N-terminus to stathmin, the resulting chimeric protein collected at the Golgi apparatus, a behavior that is similar to that of SCG10 (Di Paolo et al., 1997; Wang et al., 2013b). Therefore, SCG10 is speculated to sequester tubulin dimers at growth cones and promote microtubule assembly locally (Holmfeldt et al., 2003). In the mouse brain, SCG10 expression level peaks after birth and then decreases to a low level after synapse formation (Himi et al., 1994; Sugiura and Mori, 1995), and in the spinal motor and dorsal root ganglion (DRG) neurons, SCG10 mRNA is temporally upregulated after a nerve crush (Mason et al., 2002). Thus, SCG10 likely promotes neurite growth and extension during both development and axonal regeneration. A previous study reported that overexpressing the SCG10-S62A&S73A mutant, which has a constitutive microtubule-depolymerizing ability, decreased neurite length in cultured mouse cortical neurons (Tararuk et al., 2006). Another study using in utero electroporation of the cortex in embryonic day (E) 15.5 mice found that SCG10 knockdown increased the cortical neuron migration rate, and the defects could be rescued by overexpressing SCG10-S62D&S73D, a phospho-mimicking SCG10 mutant lacking the microtubule-regulating function (Westerlund et al., 2011). Recent studies reported the neuroprotective role of SCG10 in motor neurons, as amyotrophic lateral sclerosis-like phenotypes were observed in Scg10 knockout mice (Guerra San Juan et al., 2022; Krus et al., 2022). However, the functions of SCG10-regulated microtubule dynamics in the peripheral nervous system (PNS) are still ill-defined.

Besides the precise regulation of the microtubule network during neuron development, microtubule organization also needs to be tightly regulated for neuronal maintenance, and disturbed microtubule dynamics lead to axon degeneration (Clark et al., 2016). For example, two traditional chemotherapeutic drugs – paclitaxel, which stabilizes microtubules, and vincristine, which depolymerizes microtubules – cause clinical sensory polyneuropathy and axonal degeneration in vitro (Lavoie Smith et al., 2015; Yang et al., 2009). Stathmin facilitates microtubule disassembly by sequestering tubulin dimers (Howell et al., 1999), and its depletion causes late-onset axonopathy in mice, with large-caliber axon degeneration at 20 months of age (Liedtke et al., 2002). Additionally, in distally injured axons of cultured DRG neurons, SCG10 protein levels positively correlate with the Wallerian degeneration rate, and maintaining SCG10 delays Wallerian degeneration (Shin et al., 2012). Whether SCG10 regulates long-range viability of projection neurons in vivo remains unknown.

In this study, we constructed Scg10 knockout mice to study the function of SCG10 in vivo. Remarkably, age-dependent defective coordination movements were detected in Scg10 knockout mice, with significant sciatic nerve atrophy and extensor hallucis longus (EHL) muscle denervation. Our results also reveal that SCG10 loss inhibits peripheral neuron regeneration in vivo, and the microtubule-facilitating function of SCG10 participates in that process. These results highlight the importance of SCG10-mediated microtubule dynamics for peripheral axon maintenance and regeneration.

To reveal how SCG10 functions in vivo, we generated Scg10 knockout mice using a CRISPR-Cas9 approach and designed a gRNA that targeted exon 3 of Scg10 (Fig. S1A). Subsequent genotyping identified offspring with Scg10 gene editing (Fig. S1B), and this genetic mutant led to a frameshift at the beginning of the stathmin-like domain of SCG10 (Fig. S1C). Western blotting using specific anti-SCG10 antibodies validated SCG10 expression in the brains of wild-type (Scg10^+/+) postnatal (P) day 0 (P0) mice but not in the brains of the knockout (Scg10^−/−) mice. SCG10 levels were also much decreased in heterozygous (Scg10^+/−) mouse brains (Fig. S1D). Additionally, the Scg10 transcript was hardly detected in brains of Scg10^−/− mice by real-time PCR analysis (Fig. S1E). The majority of Scg10 knockout mice (∼85%) died at P0 (Fig. S1F). However, no apparent abnormalities were observed in brain structure (Fig. S1G,H) and the cerebral cortex layer formation [labelled by layer 6 and 5 markers Tbr1 and Ctip2 (or Bcl11b), respectively] of Scg10^−/− mice (Fig. S1I–K). To check whether a compensation effect exists in the stathmin family after SCG10 depletion, protein expression analyses of stathmin family members were conducted. SCG10 depletion showed no significant change on the expression of other stathmin family members in both the DRG and brain (Fig. S1L,M), a result that is consistent with a previous report (Krus et al., 2022).

Unlike the brain, where SCG10 levels were high during embryonic development and downregulated to undetectable levels at P30 (Sugiura and Mori, 1995), we found that SCG10 was maintained at high levels in DRG neurons, a type of sensory neurons in the PNS, even in adult or aged mice (Fig. 1A), so we hypothesized that SCG10 might also function in PNS neuronal maintenance in adult mice. Limb clasping is a functional motor test used mainly to assess disease progression in neurodegeneration (Mao et al., 2016; Miedel et al., 2017). When tested, Scg10^−/− mice showed an obvious hind-limb clasping posture at 6 months of age (Fig. 1B). Furthermore, Scg10^−/− mice displayed apparent neurodegenerative symptoms at 12 months of age, such as toe degeneration (Fig. S2A), rigidity and resting tremor (Movie 1).

To further explore the behavioral characters in 6-month-old Scg10^−/− mice, we conducted a series of behavioral tests. First, results of the rotarod test, which determines deficits in motor coordination, revealed that Scg10^−/− mice showed significantly lower latency to fall (Fig. 1C; Movie 2) and lower rotarod speed adaptabilities (Fig. 1D), compared with those in Scg10^+/+ mice. The pole test is widely used to assess motor or sensorimotor impairments in mice, with lower speeds or poorer accuracies of the escape behavior reflecting motor dysfunctions (Brooks and Dunnett, 2009). The pole test results further revealed that Scg10 knockout mice failed to turn around and then fell from the bar (Fig. 1E,F; Movie 3). However, the open-field test was used to evaluate the spontaneous locomotor activity of 6-month-old mice, and we observed that Scg10^−/− mice traveled significantly more in both the peripheral and central regions of the box (Fig. S2B–D), especially spending more time, traveling more distance in the center region and entering the center more often than the Scg10^+/+ mice did (Fig. S2B,E–G), indicating that Scg10^−/− mice are hyperactive and might be less anxious. In the elevated plus maze test, Scg10^−/− mice spent more time in the open arms and less time in the closed arms of the maze, further verifying their weakened anxious phenotypes (Fig. S2H).

In addition, we evaluated mouse sensory functions via von Frey and hot plate tests to examine mechanical and thermal sensitivities of nociceptive pain. SCG10 depletion caused significant mechanical and thermal sensory defects (Fig. 1G,H). Additionally, both motor and sensory impairments progressed with age (Fig. 1C–H).

Taken together, these results suggest that SCG10 deficiency in vivo could lead to early-onset and age-progressed impairment of motor and sensory functions in the PNS.

To further verify the peripheral dysfunction of Scg10 knockout mice, we looked for ultrastructural changes in the sciatic nerves from 14-month-old mice (Fig. 2A). Myelin is essential for maintenance of axonal integrity and its abnormalities lead to degeneration of the axon (Nave, 2010; Ohno and Ikenaka, 2019). Therefore, we calculated the g-ratios, the ratio of inner axon circumference to outer myelin circumference, which is widely used to assess the integrity of myelination (Chen et al., 2017; Sheykholeslami et al., 2001), and found that the average g-ratios of axons from Scg10 knockout mice were significantly lower compared to those of wild-type mice (Fig. 2B). A scatterplot of g-ratios against axon diameters showed that the slopes of the lines of best fit generated by linear regressions were significantly higher for Scg10^−/− mice compared to that of Scg10^+/+ mice (Fig. 2C,D). A consistent left shift in the frequency distribution of Scg10^−/− mouse axons at given g-ratios (Fig. 2E) indicates that the g-ratios of smaller-diameter axons were reduced in Scg10^−/− mice versus those of Scg10^+/+ mice. The areas of axons and average axon diameter in Scg10^−/− mice were significantly less than those of the controls (Fig. 2F,G). However, because the axonal diameter frequency distributions between the two mouse groups were not remarkably different (Fig. 2H), the lowered g-ratio in Scg10^−/− mice might have been caused by hypermyelination, decreased axon diameter, or both, and these explanations need to be further examined. Additionally, we found significantly fewer microtubules in the small-diameter axons of Scg10^−/− mice than in those of Scg10^+/+ mice (Fig. 2I,J; for axon diameter <2 μm). Because the lowered g-ratio and microtubule number in Scg10^−/− mice were observed mainly in small-diameter axons, we further characterized the SCG10 expression pattern in the DRG neurons. Immunostaining results showed that SCG10 preferentially localized to the neurons expressing calcitonin gene-related peptide (CGRP), a marker for c-fiber nociceptor DRG neurons (Fig. 2K,L). Collectively, these results indicate that SCG10 deletion causes myelin deficits and decreased numbers of microtubules in small-diameter axons.

Neuromuscular junction (NMJ) dysfunction is a frequent consequence of neurodegenerative diseases (Taetzsch et al., 2017; Wang et al., 2017). To determine whether NMJs are impaired in Scg10 knockout mice, we used immunostaining to examine the NMJs of EHL muscles in 14-month-old mice. Because the EHL muscle assists with the foot eversion and inversion, it is a suitable model to determine NMJ status during axon degeneration or regeneration after injury (Bhattacharya et al., 2016; Vannucci et al., 2019). The pre-synapses were immunostained with anti-synapsin-1 antibody, and the post-synaptic endplates were labeled with sulforhodamine 101-α-bungarotoxin (BTX). Colocalization of synapsin-1 and BTX indicates healthy NMJs, whereas visible post-synaptic endplates without pre-synapse overlapping indicates denervated NMJs (Ueta et al., 2020). Our results showed that Scg10 knockout mice had a significantly greater muscle denervation rate than that of Scg10^+/+ mice (Fig. 3A–C), and that the post-synaptic endplate perimeters of Scg10 knockout mice were significantly smaller than those of Scg10^+/+ mice (Fig. 3D), suggesting that SCG10 is necessary for neuronal maintenance.

To investigate the mechanisms that result in peripheral axonopathy after SCG10 depletion, we immunostained axons to examine the contents of different post-translationally modified tubulins using in vitro-cultured DRG neurons, with tubulin polyglutamylation or acetylation representing stable microtubules (Infante et al., 2000; Matsuyama et al., 2002) and tyrosination representing dynamic microtubules (Webster et al., 1987). Corresponding with previous reports (Robson and Burgoyne, 1989), our results showed that microtubules were relatively stable in shafts, with a high degree of acetylation and polyglutamylation (Fig. 4A,C); but in growth cones, where more dynamic microtubules are required to enable axon extension, tyrosinated tubulin predominated (Fig. 4B). In Scg10^−/− mouse neurons, the acetylated tubulin content normalized to total tubulin content (marked by βIII-tubulin) increased significantly compared to that in Scg10^+/+ mouse neurons (Fig. 4D), whereas tyrosinated tubulin content decreased significantly (Fig. 4E). However, we detected no difference in polyglutamylated tubulin levels between wild-type and Scg10 knockout mouse neurons (Fig. 4F). Overall, the changes in tubulin modifications suggest that SCG10 depletion increases microtubule stability and attenuates microtubule dynamics.

Long-term axonal transport is mainly regulated by microtubule stability and post-translational modifications (Nirschl et al., 2017), and proper axonal transport ensures neuronal survival, function and regeneration (Millecamps and Julien, 2013; Sheng, 2017). Therefore, we investigated whether alterations of microtubule post-translational modifications after SCG10 depletion could affect axonal transport. As axonal mitochondria move along microtubules (Nangaku et al., 1994; Zhou et al., 2016) and supply energy for distal axons (Cheng et al., 2022; Vaarmann et al., 2016), we first tracked axonal mitochondria in cultured DRG neurons. To help distinguish between proximal and distal axons, we cultured the DRG neurons in a microfluidic chamber for 2 days, such that the axons had extended into the axonal compartment. Then, we stained mitochondria with MitoTracker, monitored their movement under a spinning-disk microscope, and counted the number of motile mitochondria (Fig. 5A). The results showed that the mitochondrial motility rates, including anterograde and retrograde transport, in Scg10^−/− axons were significantly less than those of Scg10^+/+ axons (Fig. 5B). Additionally, both anterograde and retrograde mitochondrial movement velocities in Scg10^−/− axons were significantly less than those in Scg10^+/+ axons (Fig. 5C).

Next, we examined lysosome transport in axons with LysoTracker (Xie et al., 2015; Zheng et al., 2010). As with the mitochondria results, in Scg10^−/− DRG axons, the total motility of lysosomes, including anterograde and retrograde transport, was impaired (Fig. 5D,E) and transport velocities were reduced (Fig. 5F) significantly compared to those of Scg10^+/+ axons. In summary, SCG10 depletion suppresses microtubule-based axonal transport.

When an axon is injured, the proximal site initiates a new growth cone that needs microtubule reorganization (Bradke et al., 2012; He and Jin, 2016), and after a rat sciatic nerve crush, SCG10 mRNA levels are strongly upregulated in the motor and DRG neurons (Mason et al., 2002). Additionally, the SCG10 protein accumulates at the proximal injury site of cultured DRG neurons. Therefore, SCG10 can be used as a marker for regenerative ability (Shin et al., 2014). We examined SCG10 functions during axonal regeneration in vivo using the sciatic nerve crush model, in which GAP43 was used as regenerating marker for its ample immunostaining after injury (Nangaku et al., 1994; Shin et al., 2014). Our results showed that the extension of regenerative axons from Scg10^−/− mice was significantly less than that of the wild-type ones (Fig. 6A,B).

To investigate whether SCG10 depletion delays synapse reconstruction during regeneration, we examined NMJ reformations using Thy1-YFP16 transgenic mice to indicate regenerating axons, as those mice express YFP specifically in neurons, including motor and sensory neurons (Di Maio et al., 2011). We crossed Scg10^+/− mice with Thy1-YFP16 transgenic mice to get Thy1-YFP:Scg10^−/− mice, which we used to assess synapse reformation based on the colocalization of BTX and YFP in EHL muscles 1 month after sciatic nerve crush (Cho and Cavalli, 2012). Consistently, we found that SCG10 depletion delayed NMJ reformation (Fig. 6C,D).

Moreover, we determined the extent of axon regeneration under injury-mimicking conditions in cultured DRG neurons by using a previously reported culture and replating method (Saijilafu et al., 2013). Before injury, total axon length (marked by βIII-tubulin) showed no significant difference between Scg10^+/+ and Scg10^−/− DRG neurons (Fig. S3A,B). However, 24 h after replating, the total axon length in Scg10 knockout neurons was significantly less than that of the wild-type ones (Fig. S3A,C). Overall, our results indicate that axonal regeneration requires SCG10 both in vivo and in vitro.

We next investigated whether SCG10 promotes axonal regeneration by regulating the dynamic properties of microtubules. Our previous work has reported that a phospho-dead mutant of SCG10, SCG10-4A (serines at positions 50, 62, 73 and 97 all mutated to alanines), facilitates its microtubule-depolymerizing ability, whereas a phospho-mimicking mutant of SCG10, SCG10-4D (all four serines mutated to asparagines) has significantly less microtubule-destabilizing activity (Wang et al., 2013b). Thus, we overexpressed wild-type SCG10, SCG10-4A and SCG10-4D mutants, using adeno-associated virus 2 (AAV2) in cultured Scg10^−/− DRG neurons on the first day in vitro (DIV 1), and replated the neurons at DIV 3. After 24 h, overexpressed SCG10 protein levels were comparable to endogenous SCG10 levels (Fig. 7A) and immunofluorescence analysis revealed that SCG10 was expressed in AAV2-infected cells (including glial cells) with similar localization patterns (Fig. 7B). Analyses of total axon length showed that the expression of SCG10-4A, but not SCG10-4D, could rescue axon length defects as well as wild-type SCG10 (Fig. 7B,C), suggesting that SCG10-mediated microtubule dynamics are critical for axonal regeneration.

Here, we used the CRISPR/Cas9 approach to construct Scg10 knockout mice. Unlike stathmin knockout mice, which live normally (Schubart et al., 1996), most of the Scg10 knockout mice died at or shortly after birth, illustrating the importance of the neuronal-specific SCG10 for survival, although the reason for the death needs to be further studied. SCG10 knockdown, using in utero-electroporated E15.5 mice, has been reported to increase the cortical neuron migration rate (Westerlund et al., 2011); however, we detected no significant layer formation defects in Scg10 knockout mouse brain sections at P0, and the brain structures were generally normal. This discrepancy might be due to compensation by other microtubule dynamics regulators, a possibility that needs further investigation. Nevertheless, some specific roles of SCG10 are necessary for mice viability.

A genetic screen in third-instar Drosophila larvae found that mutating stathmin by inserting transposable elements interfered with axonal transport and destabilized the NMJ, with the presynaptic nerve terminal retracting from the postsynaptic membrane (Graf et al., 2011), suggesting that synapse growth and maintenance requires stathmin. Although animal experiments showed that newborn stathmin-deficient mice initially showed no anatomical defects and developed normally and that, at 14 months of age, they began to develop mild axoplasm irregularities in the sciatic nerve and spinal cord (Liedtke et al., 2002). Unlike Drosophila cells that possess a single stathmin gene, vertebrate cells encode three other related genes, including Scg10, Sclip and Rb3. Scg10 has been studied as a regeneration-associated gene (Mason et al., 2002; Shin et al., 2014), whereas its axon maintenance functions were neglected. Here, we highlight the roles of SCG10 in peripheral neuronal maintenance and show that its expression persists in DRG neurons throughout the life of a mouse. Scg10 knockout mice showed peripheral neuropathy with movement coordination and sensory defects. Mechanistically, we found that SCG10 depletion caused a dramatic decrease in tyrosinated tubulin levels, whereas acetylated tubulin levels increased, indicating attenuated microtubule dynamics. Interestingly, loss of SCG10 has marginal effects on polyglutamated tubulin, even though excessive polyglutamylation has been reported to induce neurodegeneration (Magiera et al., 2018). How SCG10 affects specific tubulin post-translational modifications is an interesting topic for future study. Microtubule post-translational modifications might determine axonal transport. Tyrosinated tubulin promotes dynactin loading on microtubules and facilitates retrograde axonal transport (McKenney et al., 2016). Additionally, acetylated tubulin likely promotes the binding with kinesin-1, thus enhancing the axonal transport of kinesin-1 cargos such as JNK-interacting protein 1 (JIP1 or MAPK8IP1) (Reed et al., 2006). Pharmacologically increasing the levels of acetylated tubulin further rescued axonal transport defects in HSPB1 mutant transgenic mice (d'Ydewalle et al., 2011). Thus, alteration of microtubule modifications in SCG10-depleted neurons, forming over-stabilized microtubules, might be responsible for impaired axonal transport and subsequent peripheral neuropathy. The peripheral nerve myelin sheath is formed by Schwann cells (Salzer, 2015). Besides Schwann-cell-autonomous effects, the axon–Schwann cell interplay also contributes to myelin dimensions, as Neuregulin-1 levels on the axon membrane surface have been reported to regulate myelin sheath thickness (Birchmeier and Nave, 2008; Pereira et al., 2012). We propose that the dysregulation of axonal transport after Scg10 deletion might affect the distribution of myelination regulators along the axon, thus indirectly leading to myelination abnormalities.

Long regarded as a regeneration-promotion factor, the levels of SCG10 and its mRNA increase in neurons after injuries (Mason et al., 2002), making it a good marker for axonal regeneration (Cho et al., 2013; Shin et al., 2014). Our results show that Scg10 knockout mice regenerate DRG neurons poorly, providing direct evidence that SCG10 functions in regeneration in vivo. Interestingly, autophagy elicitation stabilizes microtubules by degrading SCG10 and promotes spinal cord regeneration (He et al., 2016). This discrepancy between our findings and previous results might reflect the differences in microtubule organization between the PNS and the central nervous system (CNS). Lesioned CNS axons usually form retraction bulbs with disorganized microtubule arrays (Erturk et al., 2007), and increasing the microtubule stability by taxol promotes spinal cord neurite regrowth (Hellal et al., 2011). Without SCG10, microtubules are more stable against retraction, like the effects of taxol treatment (Sengottuvel et al., 2011). However, in our study, when PNS axons lacked SCG10, microtubule stability increased and axonal regeneration after injury was inhibited. Furthermore, the axon length defect in Scg10 knockout DRG neurons was partially rescued by a phospho-dead mutant (SCG10-4A), but not the phospho-mimicking mutant (SCG10-4D), indicating the inhibitory roles of over-stabilized microtubules in regeneration. Therefore, microtubule dynamics are strictly, but distinctly, regulated in different types of neurons during axonal regeneration.

SCG10 has been reported to be involved in Alzheimer's disease (Okazaki et al., 1995; Wang et al., 2013b) and Goldberg–Shprintzen syndrome (Alves et al., 2010). Recently, a genomic study of inherited Parkinson’s disease postmortem found that STMN2 was downregulated, and knockdown of SCG10 by injecting AAV2/1 in the mouse substantia nigra caused degeneration of dopaminergic neurons (Wang et al., 2019). Moreover, SCG10 expression in postmortem amyotrophic lateral sclerosis spinal cord also decreased, and motor neurons differentiated from STMN2^−/− human embryonic stem cells showed reduced neurite extension (Klim et al., 2019; Melamed et al., 2019). Recently, similar to this study, the functions of SCG10 in motor neurons were reported through the use of Scg10 knockout mice, and it was found that the improper microtubule dynamics after SCG10 depletion ultimately caused motor system defects (Guerra San Juan et al., 2022; Krus et al., 2022). Here, we provide in vivo evidence that SCG10 is essential for perinatal viability, peripheral axon maintenance and regeneration. These data support the notion that the microtubule dynamics regulated by SCG10 are critical for viability and disease pathogenesis across diverse neurons. Scg10^−/− mouse recapitulated peripheral neurodegeneration features, such as motor deficits and sensory neuropathy, and could be used as a mouse model to select therapeutic drugs for different neurologic diseases.

Scg10 knockout mice were generated using a CRISPR/Cas9 approach on a C57BL/6J background (Wang et al., 2013a), and a guide RNA (5′-GGTGAAGCAGATCAACAAGC-3′) was designed to target exon 3 of the Scg10 gene. Genomic DNA extracted from tail biopsies was used as PCR amplification templates for genotyping, with the following sequencing primers for targeted regions: F-Scg10 genome-43780 (5′-TGCAGATGCTGCACTCCATGACA-3′) and R-Scg10 genome-43977 (5′-CTGAATCTCCTCCAGAGACAGG-3′). We crossed the F0 heterozygous mice back to wild-type C57BL/6J mice twice, and then bred the third-generation heterozygous Scg10 knockout mice for homozygosity. Thy1-YFP transgenic mice [B6.Cg-Tg (Thy1-YFP)16Jrs/J] were a kind gift from the Guangdong Pharmaceutical University Vascular Biology Research Institute. All animal procedures were performed according to the Institutional Animal Care and Use Committee guidelines of Peking University (followed GB/T 35892-2018) (MacArthur Clark and Sun, 2020).

DRG neurons harvested from 1-month-old, same-sex wild-type or Scg10 KO mice were incubated first in 0.1% collagenase type IA (Sigma-Aldrich) at 37°C for 45 min, then in 0.25% trypsin for 20 min, followed by trituration by pipetting. The dissociated DRG neurons were then plated onto poly-L-lysine- (0.25 mg/ml, Sigma-Aldrich) and laminin- (20 μg/ml, Life Technologies) coated coverslips in Dulbecco's modified Eagle medium (DMEM; Thermo Fisher Scientific) containing 10% fetal bovine serum (FBS; Gibco) for 3 h and then transferred to Minimum Essential Medium (Thermo Fisher Scientific) supplemented with 20% glucose, 1.1% pyruvate, 1% GlutaMAX (Thermo Fisher Scientific), 2% horse serum (Gibco), 2% B27 supplement (Thermo Fisher Scientific), penicillin/streptomycin (500 U/ml, Gibco) and 50 ng/ml nerve growth factor (Thermo Fisher Scientific) at 37°C and 5% CO₂. We then transduced the DRG neurons by adding adeno-associated virus (AAV2) to the culture medium.

We obtained SCG10 cDNA by PCR cloning from the E17 mouse brain cDNA library, and sub-cloned it into AAV vectors (gifts from Dr Chenjian Li, Peking University) and used site-directed mutagenesis based on wild-type SCG10 to construct SCG10-S50A, -S62A, -S73A, -S97A, -S50D, -S62D, -S73D and -S97D mutants (Wang et al., 2013b). In preparation for transfection, we first cultured HEK293T cells in DMEM and 10% FBS at 37°C and 5% CO₂, and used polyethylenimine (PEI; Polysciences) to transfect cells with AAV2 vectors containing SCG10, pAAV-RC and pHelper plasmids (gifts from Dr Chenjian Li, Peking University). At 72 h after transfection, the cells were collected in 5 ml cell lysis buffer (150 mM NaCl, 20 mM Tris, pH 8.0), and subjected to a freeze-thaw sequence using liquid nitrogen and a 37°C water bath. After centrifuging the cell lysates at 4,000 g for 30 min, we collected the supernatant and purified the virus via a discontinuous iodixanol gradient centrifuge (SW 40Ti, Beckman Coulter) at 288,350 g for 3 h. The viral fraction, harvested using a 10 ml syringe, was washed and concentrated using a 100K centrifugal filter (Millipore), and the proper volume of virus particles was aliquoted and stored at −80°C.

Tissues or cultured neurons were homogenized in lysis buffer (25 mM HEPES, 150 mM NaCl, 2 mM MgCl₂, 1% Triton X-100, pH 7.4) with protease inhibitor cocktails (Amresco), and cell lysates were centrifuged at 12,000 g for 15 min. The supernatants were added with loading buffer, heated at 100°C for 5 min, and subjected to western blot analyses as previously described (Lv et al., 2015). The primary antibodies we used were anti-SCG10 (Proteintech, 10586-1-AP, 1:1000), anti-STMN1 (Abclonal, A4379, 1: 1000), anti-STMN3 (Proteintech, 11311-1-AP, 1: 2000), anti-STMN4 (Proteintech, 12027-1-AP, 1: 2000), anti-GAPDH (CWBIO, CW0100, 1:5000) and rabbit anti-βIII-tubulin (Cell Signaling Technology, 5568, 1:1000). We used horseradish peroxidase-conjugated goat anti-mouse and anti-rabbit IgGs (Jackson ImmunoResearch Laboratories, 115-035-003 and 111-035-144, 1:5000) as the secondary antibodies.

The brains from P0 mice with random sex were homogenized in TRIzol reagent (Ambion, 15596018), and total RNA was isolated according to the protocol. Reverse transcription was conducted using M-MLV reverse transcriptase (Promega, M1705) with oligo-dT primers. Quantitative real-time PCR was conducted using the SYBR qPCR master mix (Vazyme, Q711-03), and samples were analyzed on a Bio-Rad CFX Connect Real-Time PCR instrument. Gene expression was related to Gapdh using the 2^−ΔΔCT method (Ji et al., 2021). The following primers were used: Scg10 F, 5′-ATGAAGGAGCTGTCTATGC-3′, and Scg10 R, 5′-CTGAGATGGGAGATGGTG-3′. The forward primer was about 90 bp upstream of the gRNA recognition site that belongs to exon 2 of Scg10; whereas the reverse primer was 50 bp downstream of the gRNA recognition site found within exon 3 of Scg10. For Gapdh, the primers used were: Gapdh F, 5′-CTGCACCACCAACTGCTTAG-3′, and Gapdh R, 5′-ACAGTCTTCTGGGTGGCAGT-3′.

To examine mitochondria and lysosome transport, we first cultured DRG neurons in homemade microfluidic devices for 2 days to allow the axons to extend into the axonal compartment. Then, 10 μM of either MitoTracker (Thermo Fisher Scientific, M7512) or LysoTracker (Thermo Fisher Scientific, L7528) were added to the culture medium for 5 min, and were exchanged with fresh medium. DRG neurons were kept in the environment chamber of the spinning-disk confocal microscope (UltraView VoX; PerkinElmer) at 37°C and 5% CO₂, and time-lapse imaging was performed with a 40×/0.95 NA objective lens. The images were processed with the kymograph plugin in ImageJ (National Institutes of Health) and the organelle movement parameters were calculated using macro plugins. Organelles with displacement more than 10 μm were classified as motile.

For immunofluorescence, DRG neurons were fixed and immunostained as described previously (Lv et al., 2015). The following primary antibodies were used: anti-acetylated tubulin (Sigma-Aldrich, T7451, 1:500), anti-tyrosinated tubulin (Millipore, MAB1864-1, 1:500), anti-polyglutamylated tubulin (AdipoGene, GT335, 1:500) and rabbit anti-βIII-tubulin (Cell Signaling Technology, 5568, 1:500). We used Alexa Fluor 568-conjugated goat anti-rabbit-IgG (Life Technologies, A-11036, 1:1000), Alexa Fluor 488-conjugated goat anti-mouse-IgG (Life Technologies, A-11029, 1:1000) and Alexa Fluor 568-conjugated goat anti-rat IgG (Thermo Fisher Scientific, A-11077, 1:1000) as secondary antibodies. For immunohistochemistry, the mice were anesthetized with tribromoethanol (Sigma-Aldrich, T48402, 240 mg/kg) and perfused with 4% paraformaldehyde (PFA), and then their brains or DRGs were collected quickly and post-fixed in fresh 4% PFA overnight. Brains were dehydrated in graded sucrose series, transferred to optimal cutting temperature (OCT) compound (Sakura Finetek), and sectioned coronally with a microtome-cryostat (CM1850, Leica) at a thickness of 20 μm. The tissue sections were post-fixed in 4% PFA for 10 min, permeabilized in 0.2% Triton X-100 in PBS for 15 min, and immunostained using the indicating antibodies: anti-Ctip2 (Abcam, ab18465, 1:500), anti-Tbr1 (Abcam, ab31940, 1:500), anti-SCG10 (Proteintech, 10586-1-AP, 1: 500), anti-CGRP (Abcam, ab36001, 1:500), Alexa Fluor 568-conjugated goat anti-rabbit-IgG (Life Technologies, A-11036, 1:1000), Alexa Fluor 488-conjugated goat anti-rat IgG (Thermo Fisher Scientific, A-11006, 1:1000), Alexa Fluor 568-conjugated donkey anti-goat-IgG (Thermo Fisher Scientific, A-11057, 1:1000) and Alexa Fluor 488-conjugated donkey anti-rabbit-IgG (Thermo Fisher Scientific, A-21206, 1:1000). For Hematoxylin and Eosin staining, the post-fixed brains were dehydrated in series of alcohol and xylene, infiltrated with paraffin, sectioned at a thickness of 5 μm, and stained as reported previously (Fischer et al., 2008a,b,c).

We anesthetized 3- to 4-month-old female mice using tribromoethanol, made an incision to access their right-side sciatic nerves, crushed them with forceps for 20 s and then sutured the muscle and skin. Three days later, we anesthetized the mice with tribromoethanol, perfused with 4% PFA, and removed the sciatic nerves. The nerves were fixed in 4% PFA at room temperature for 1 h, immersed in 30% sucrose overnight, and embedded in OCT compound. Finally, we immunostained longitudinal sections (10 μm) of the sciatic nerves with anti-βIII-tubulin (Santa Cruz Biotechnology, sc-80005, 1:500) and anti-GAP43 (Proteintech, 16971-1-AP, 1:1000) (Shin et al., 2014), used Alexa Fluor 568-conjugated goat anti-mouse-IgG (Life Technologies, A-11031, 1:1000) and Alexa Fluor 488-conjugated goat anti-rabbit-IgG (Life Technologies, A-11034, 1:1000) as the secondary antibodies, and imaged those sections under the spinning-disk microscope with a 10×/0.45 NA objective lens. Axonal regeneration was quantified by the lengths and intensities of GAP43-positive axons.

For the denervation assay, 14-month-old female mice were anesthetized using tribromoethanol, perfused with 4% PFA. EHL muscles were dissected out, fixed in 4% PFA overnight, treated with 1% Triton X-100 for 6 h, and then stained with anti-synapsin-1 antibody (Cell Signaling Technology, 5297, 1:200) in 4% bovine serum albumin overnight. The next day, we stained the muscles with Alexa Fluor 488-conjugated goat anti-rabbit-IgG (Life Technologies, A-11034, 1:1000) and sulforhodamine 101-α-bungarotoxin (BTX) (Biotium, 2 μg/ml), and mounted them with 50% glycerol in PBS. The images were captured under an LSM 710 NLO microscope (Zeiss) with a 10×/0.5 NA objective lens. The perimeters of single postsynaptic endplates were calculated in ImageJ. For the re-innervation assay, 1 month after the sciatic nerve injury, EHL muscles were stained with BTX and mounted for confocal imaging.

Fourteen-month-old female mice were anesthetized using tribromoethanol, perfused with 4% PFA and 1% glutaraldehyde in 0.1 M PBS, and then their sciatic nerves at the mid-thigh were removed and fixed overnight with 2.5% glutaraldehyde in 0.1 M PBS. The next day, the samples were rinsed three times with 0.1 M PBS, incubated in 1% osmium tetroxide for 30 min on ice, washed with 0.1 M PBS, and then stained with uranyl acetate in 50% ethanol for 30 min. We dehydrated the samples with gradient ethanol, embedded them in epon, and used an EM UC7 ultramicrotome (Leica) to create ultrathin (70 nm) sections. To increase electron density, the samples were fast stained with uranyl acetate for 20 min and then lead citrate for 10 min. We scanned the samples in a 120 kV Tecnai G2 Spirit transmission electron microscope (FEI) (Chen et al., 2019) and used ImageJ for image analysis.

For the open-field test, female mice were allowed to adapt to the test environment for 30 min before beginning the experiment, and all experiments were double anonymized (Tian et al., 2017). After each test, we cleaned the instrument and test area with 75% ethanol to remove olfactory cues. The open-field apparatus was enclosed by Plexiglas (40×40×40 cm³) and had an automatic video tracking system. The area within 10 cm of the central point was defined as the central area and other zones (from 10 to 20 cm) were the peripheral areas. To begin, a mouse was gently placed in the center of the apparatus. We then measured and recorded the movements of the mouse (travel distance, time and number of entries into the center) over a 6×5 min (30 min) period. The elevated plus maze consisted of a center platform (5×5 cm²) and four arms (30 cm length×5 cm width) (Shumyatsky et al., 2005) that created a ‘+’ shape, and it stood 50 cm above the floor. Two open arms had 1 cm-high rims and the other two closed arms had 15 cm-high walls. Each mouse was placed in the center platform facing an open arm. We recorded its behavior for 5 min, noting the time spent in the center platform and in the open and closed arms.

We conducted the limb-clasping test using a cellphone to record image data for each female mouse. In short, each mouse was suspended by the tail for 5–10 s, during which the hind and forepaws of the animal were recorded and each mouse was evaluated according to defined limb-clasping scores, which have been previously described (Miedel et al., 2017). For rotarod training, each mouse was placed on the rotating rod (Med Associates) with a rotation rate of 10 rpm for 1 min as previously reported (Lüesse et al., 2001). In the test, the rotation rate accelerated from 5 to 45 rpm over 5 min. The elapsed time and the rotation speed at which each mouse fell to the ground were recorded. For the pole test (Balkaya et al., 2013; Ogawa et al., 1985; Zhao et al., 2001), we placed each mouse, head up, on a homemade pole (60 cm height and 10 mm diameter) and recorded the time to make a 180° turn (T-Turn) and the time to reach the ground (T-Total). If a mouse either failed to turn around or dropped down along the bar, T-Total was recorded as 120 s.

To evaluate the sensitivity of mouse hind limbs, we tested the mechanical and thermal sensitivities of nociceptive pain. For mechanical sensitivity, each female mouse was placed separately in a test arena and left to habituate for 5 min. When a mouse stopped moving around freely in the testing arena, we applied von Frey filaments (Aesthesio) to its hind paw mid-plantar surface. When the mouse responded (e.g. by pulling back its paw), the corresponding applied force was recorded by the system. If the mouse failed to react to the maximum force, the test value was recorded as 2. We used a model 390 apparatus (IITC Life Science) for the thermal sensitivity test (Qiu et al., 2011). Briefly, each mouse was placed on a hot plate, covered by a plastic box (10×10×15 cm³) and left to habituate for 30 min. During the test, we placed each mouse on the 55°C hot plate and recorded the time at which it began licking its paws or jumping. A 30 s cut-off time was assigned to avoid tissue injury. Each mouse was measured three times with 15–20 min resting intervals.

All behavioral tests were anonymized to mouse genotype. For morphometry, we randomly viewed microscopic fields of the sciatic nerves and NMJs, making quantitative assessments anonymized to the genotype. The g-ratios (the ratio of the axonal diameter to the myelin diameter) were measured using ImageJ, and linear regression of the g-ratios was performed using Prism 8.0.1 (GraphPad). We randomly chose images of cultured cells to measure axon lengths and cortical layer thicknesses using ImageJ. The immunofluorescence intensities were measured using Velocity 3D Image Analysis software 6.1.1 (PerkinElmer) and profiled using Prism. Data are presented as mean±standard error of mean (s.e.m.) or standard deviation (s.d.). We used repeated-measures two-way ANOVA with Sidak post hoc test to evaluate the differences between two curves and indicated quantification methods to assess the significance. n.s, not significant; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001.

We thank Dr Chenjian Li (College of Life Sciences, Peking University) for the AAV2 vectors. We also thank National Center for Protein Sciences at Peking University in Beijing, China, for assistance with the behavioral experiments, image acquisition and cell sorting, particularly Chunyan Shan and Hongxia Lv, for technical help. We thank the Core Facilities at the School of Life Sciences, Peking University for help with electron microscopy sample preparation and image analysis, particularly Yingchun Hu, Pengyuan Dong, and Yunchao Xie. Finally, we thank everyone who gave helpful suggestions.