Intrathecal delivery of AAV-NDNF ameliorates disease progression of ALS mice

Abstract

Amyotrophic lateral sclerosis (ALS) is a uniformly lethal neurodegenerative disease characterized by progressive deterioration of motor neurons and neuromuscular denervation. Adeno-associated virus (AAV)-mediated delivery of trophic factors is being considered as a potential disease-modifying therapeutic avenue. Here we show a marked effect of AAV-mediated over-expression of neuron-derived neurotrophic factor (NDNF) on SOD1^G93A ALS model mice. First, we adopt AAV-PHP.eB capsid to enable widespread expression of target proteins in the brain and spinal cord when delivered intrathecally. Then we tested the effects of AAV-NDNF on SOD1^G93A mice at different stages of disease. Interestingly, AAV-NDNF markedly improved motor performance and alleviated weight loss when delivered at early post-symptomatic stage. Injection in the middle post-symptomatic stages still improved the locomotion ability, although it did not alleviate the loss of body weight. Injection in the late stage also extended the life span of SOD1^G93A mice. Furthermore, NDNF expression promoted the survival of spinal motoneurons, reduced abnormal protein aggregation, and preserved the innervated neuromuscular functions. We further analyzed the signaling pathways of NDNF expression and found that it activates cell survival and growth-associated mammalian target of rapamycin signaling pathway and downregulates apoptosis-related pathways. Thus, intrathecally AAV-NDNF delivery has provided a potential strategy for the treatment of ALS.

Graphical abstract

Luo and colleagues demonstrate a beneficial effect of AAV-mediated expression of the less studied neurotrophic factor NDNF on disease progression of ALS model mice. It holds promise as a potential disease-modifying strategy for ALS treatment.

Introduction

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease resulting from selective and progressive deterioration of both upper and lower motor neurons in the brain and spinal cord.¹ Approximately 90% of ALS patients are sporadic and 10% are hereditary, with pathogenic variants occurring mostly in genes encoding superoxide dismutase 1 (SOD1),² TAR DNA-binding protein 43,^3,4 an RNA-binding protein fused in sarcoma,^5,6 a hexanucleotide expansion in the C9ORF72 gene,^7,8 and other genes with minor involvement.⁹ The variants of the first ALS-associated SOD1 gene account for 12%–20% of familial and 1%–2% of sporadic ALS cases.¹⁰ Multiple factors have been shown to be involved in the ALS pathogenesis, including impaired RNA and protein homeostasis, abnormal nucleocytoplasmic and axonal transport, mitochondrial dysfunction related oxidative stress, excitotoxicity, and neuroinflammation.⁹ For example, mutations in SOD1 lead to protein misfolding and aggregation, which is toxic to affected neurons via several mechanisms.^9,11

Some small molecule chemicals including riluzole, edaravone (Radicava), and more recently sodium phenylbutyrate-taurursodiol, which are proposed to act via the modulation of glutamatergic transmission or reducing oxidative stress, have been tested in clinical treatments, but the effects are very mild and limited.^12,13,14,15 Thus, alternative approaches like gene therapy are being considered, such as the applications of antisense oligonucleotides that mediate the degradation of SOD1 messenger RNA to reduce SOD1 protein synthesis,^16,17 adeno-associated virus (AAV)-mediated expression of a microRNA targeting SOD1,^18,19 and genome editing.^20,21,22 Because of non-pathogenicity, low immunogenicity, and different serotypes that have different affinities to various cell types, AAV is widely used in research and the clinic to deliver genetic materials modulating target gene expression, correcting variant sequences, or expressing neurotrophic factors.²³ Over the past more than 30 years, approximately three dozen clinical studies have reported the safety and potential effects of neurotrophic factors in treated patients with ALS or other neurodegenerative diseases.^24,25 Considering that different factors may execute functions via distinct signaling pathways, a combination of multiple factors may be more fruitful.

Neuron-derived neurotrophic factor (NDNF), also named as epidermacan, is originally found in a transcriptome-based systematic identification of extracellular matrix proteins and later found to be mainly expressed in neurons in the brain and spinal cord with enrichment in particular cell types.^26,27 In cultured mouse hippocampal neurons, NDNF promotes neurite outgrowth and protects neurons from glutamate or Aβ-induced toxicity.²⁶ Interestingly, a recent study through single cell sequencing of retinal ganglion cells (RGCs) after optic nerve injury, has shown that cells with high-level expression of NDNF are more likely to survive, and overexpression of NDNF significantly promotes RGC survival and axon regeneration after optic nerve crush.²⁸ Another study has shown that NDNF administration protects against dexamethasone-induced skeletal muscle atrophy.²⁹ Nevertheless, the role of NDNF in neurodegenerative diseases has not been explored.

Here, we demonstrate an effect of AAV-NDNF on phenotypic progression in SOD^G93A mice, the most commonly used preclinical ALS animal model.^30,31 By combining the advantages of engineered AAV-PHP.eB capsid³² with the CBh promoter (cytomegalvoris enhancer/chicken beta-actin hybrid intron and promoter),³³ we achieved efficient delivery and wide-scale gene expression in the brain and spinal cord after intrathecal injection. The AAV-NDNF transduction at various post-symptomatic stages markedly ameliorates disease progression or prolonged survival of ALS mice. We also investigated the mechanisms of NDNF action by measuring the histological and signaling pathway changes.

Results

NDNF expression in spinal cord motor neurons

NDNF is conserved among different species with homology in protein sequences above 85% (Figure S1). In the mouse spinal cord, the expression level of NDNF gradually increased after birth, reaching plateau at approximately postnatal month 5 (Figures 1A and 1B). In the widely used ALS animal model SOD1^G93A mice, the level of NDNF in the spinal cord had no significant change compared with wild-type (WT) mice and during disease progression from early or middle stages (9 weeks) to late stages (17 weeks) (Figure S2). Immunostaining results showed that almost all NDNF positive cells were co-labelled by neuronal marker NeuN in the spinal cord of adult mice (Figure 1C), in line with the conclusion that NDNF is derived from neurons in the nervous system. Notably, motor neurons in the ventral horn displayed evident expression of the Ndnf gene, as reflected from fluorescence in situ hybridization, which showed co-labeling of Ndnf and cholinergic motor neuron marker Chat (Figure 1D). Given the role of NDNF in promoting RGC survival after optic injury and its expression in motor neurons, we determined its effect on ALS progression.

Figure 1.

Spatial and temporal expression patterns of NDNF in the spinal cord

(A and B) NDNF expression level in the lumbar spinal cord was detected by immunoblotting at P0, P7, and P14, and months (M) 1, 2, 5, and 8. SC, spinal cord. The relative level of NDNF was quantified using GAPDH as control with the value at P7 set as 1 (B). (C) Immunostaining for the expression of NDNF in the spinal cord of 2-month-old adult mice. DAPI was used to label nucleus. Scale bars, 200 μm for the whole image and 50 μm for the insets. (D) Expression of Ndnf and Chat transcripts detected using in situ hybridization in the adult mouse spinal cord. Scale bars, 200 μm for the whole image and 50 μm for the insets. See also Figure S2.

Intrathecal delivery of AAV-PHP.eB enables efficient transduction of neurons in the brain and spinal cord

Neurotrophic factors have been used to treat neurodegenerative diseases for decades, but the results of clinical trials have been disappointing, mainly because the systemic delivered molecules have difficulty reaching the areas where neurons are lost.²⁴ Recent advances in the engineering of AAVs have enabled efficient and wide-scale transduction of target genes into various cell types in the CNS, and thus broadened the avenues to treat neurological disorders.^34,35 For example, AAV-PHP.eB, a capsid variant of AAV serotype 9, predominantly infects neurons with higher transduction efficiency compared with its prototype.^32,36 Thus, we chose the AAV-PHP.eB capsid for effective neuronal transduction. To minimize the transduction of peripheral organs, we chose intrathecal injection to deliver virus into the cerebrospinal fluid by lumbar puncture injection into the space caudal of the vertebral column (Figure 2A). To test the transduction efficiency, we generated target plasmid for the expression of GFP under the control of human synapsin (hSyn) or CBh promoter. Notably, intrathecal injection of AAV-hSyn-GFP or AAV-CBh-GFP at the same dose (5 × 10¹¹ vg per mouse) led to comparable wide-range transduction of cells in the brain (Figure 2B). However, in the spinal cord motor neurons, AAV-CBh injection displayed greater transduction efficiency compared with AAV-hSyn group (Figure 2C). Indeed, almost all ChAT-positive motor neurons were labeled by GFP in AAV-CBh-GFP injected mice (Figure 2C). Based on these results, we generated AAV-CBh construct encoding mouse NDNF with 3×Myc tags fused at the C-terminus, and hereafter shortened as AAV-NDNF (Figure 2D). The level of NDNF over-expression was determined by immunoblot analysis for transfected 293T cells (Figure 2E). Intrathecal delivery of AAV vectors did not affect the healthy state and body weight of treated WT mice (Figure S3).

Figure 2.

Efficient transduction in the brain and spinal cord by intrathecal AAV injection

(A) Schematic diagram of the intrathecal injection. Virus was injected via the intervertebral foramen into the CSF at indicated lumber (L) position of spinal cord (SC). (B) GFP signals in the brain of mice intrathecally injected with AAV-hSyn-GFP or AAV-CBh-GFP. Scale bars, 500 μm (left) or 50 μm (insets in right). (C) GFP signals in the spinal cord lumber segment of adult mice intrathecally injected with AAV-hSyn-GFP or AAV-CBh-GFP (PHB.eB capsid). Samples were collected 3 weeks post injection and longitudinal sections were stained with anti-ChAT antibody to mark spinal motor neurons. Note that some astrocytes (indicated by white arrows) were also transduced in AAV-CBh-GFP mice. Scale bars, 100 μm. (D) Construct design of AAV-GFP and AAV-NDNF vectors, with promoter and expression elements inserted between ITRs. (E) Cell lysates of HEK293T cells transfected with AAV-GFP or AAV-NDNF were subjected to immunoblotting for the expression of NDNF. See also Figures S4 and S5.

Remarkably, more than 80% of neurons in the cortex or spinal cord, ranging from lumbar, thoracic, to cervical levels, were transduced in mice with lumbar puncture injection of AAV-NDNF (Figure S4). In addition, a smaller percentage of astrocytes was also transduced (Figures S4A–S4C). This transduction efficacy is comparable with that achieved via intravenous infusion.^32,36 Interestingly, a band labeled by Myc antibody with predicted size of NDNF-3xMyc fusion protein was present in the cerebrospinal fluid (CSF), suggesting the secretion of NDNF into the CSF (Figure S5A). In line with this notion, the NDNF level was increased in the CSF of AAV-NDNF injected mice (Figure S5B). The modest increase of NDNF in the CSF was likely due to limited diffusion of secreted NDNF from original infected cells.

AAV-NDNF administration improves motor behavior and survival of ALS mice

AAV vectors have the advantage of conferring sustained expression of target genes in transduced cells after a single treatment. We determined behavior consequences after AAV-NDNF or control virus (5 × 10¹¹ vg per mouse) administration at different stages of disease in the SOD1^G93A mice. First, we intrathecally injected viruses into SOD1^G93A mice in early post-symptomatic stage at post-natal day 61 (P61) or in middle post-symptomatic stage at P85, and measured changes in the body weight and motor performance at later stages (Figure 3). As shown in Figure 3A, the SOD1^G93A mice gradually lost body weight from late post-symptomatic stages. Notably, this tendency was markedly dampened in AAV-NDNF P61 injected group (Figure 3A), but not in P85 injected group (Figure 3B). Furthermore, Catwalk analysis for the motor behaviors showed that P120 SOD1^G93A mice treated with AAV-NDNF at P61 exhibited more regular pace stride (Figures 3C and 3D), faster movement speed (Figures 3E and 3F), greater landing intensity of hind legs (Figure 3H), longer stride of hind legs (Figure 3I), more diagonal standing representing running posture (Figure 3J), and reduced distance of print positions related to leg dragging (Figure 3K). AAV-NDNF treatment of SOD1^G93A mice at P85, a middle post-symptomatic stage, also improved motor performance as reflected from changes in some parameters in Catwalk analysis (Figures 3L–3T). Thus, AAV-NDNF delivery has improved motor functions in ALS mice.

Figure 3.

Effects of AAV-NDNF administration on the body weight and motor behaviors in SOD1^G93A mice

(A and B) SOD1^G93A mice received a single intrathecal injection of AAV at P61 (A) or P85 (B). The body weight change relative to that at P100 in each group was presented (∗p < 0.05, ∗∗∗p < 0.001). ns, no significant difference, AAV-NDNF vs. AAV-GFP injected SOD1^G93A mice, ANOVA with Tukey’s post hoc test. (C) Representative footprints and frequency of WT or SOD1^G93A mice injected with AAV-GFP or AAV-NDNF at P61 collected by Catwalk analysis. LF, left front paw; LH, left hind paw; RF, right front paw; RH, right hind paw. (D–K) SOD1^G93A mice received a single intrathecal injection of AAV at P61 and gait analysis was measured at P120, for the time duration needed for crossing the collection area (E), average speed (F), mean intensity of the front paw (G), mean intensity of the hind paw (H), the stride length of the hind paw (I), the percentage of mice in diagonal stance posture that reflects running state (J), and the print positions between the hind paw and the previously placed front paw on the same side of the body and in the same step cycle (K). Data are shown as mean ± SEM of nine mice in each group per independent experiment. (L–T) SOD1^G93A mice received a single intrathecal injection of AAV at P85 and gait analysis was measured at P120 for motor performance. Statistical parameters are consistent with that of P61 injection. Data are shown as mean ± SEM (n = 8 for WT, n = 9 for SOD1^G93A mice injected with AAV-GFP or AAV-NDNF). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns, no significant difference, ANOVA with Tukey’s post hoc test.

AAV-NDNF administration preserves motor neurons and neuromuscular junctions

The degeneration of upper and lower motor neurons and the denervation of skeletal muscles are the common features responsible for motor deficits in ALS. In line with this, SOD1^G93A mice at P135 showed marked reduction in the number and percentage of survival spinal motor neurons marked by ChAT (Figure 4) and upper motor neurons in the cortex (Figure S6). AAV-NDNF delivery at P61 (Figures 4A–4C) or P85 (Figures 4D–4F) markedly increased the survival of spinal motor neurons, and had mild but not significant effect on cortical motor neurons (Figure S6). Compared with WT mice, only approximately 50% motor neurons were retained in P135 SOD1^G93A mice treated with control AAV; in AAV-NDNF-treated ALS mice, this percentage increased to 70.73% in the P61 AAV-NDNF injected group and 63.28% in the P85 injected group (Figures 4C and 4F) (p < 0.01, AAV-NDNF vs. AAV-GFP). These results indicate that NDNF administration prevents the degeneration of motor neurons in ALS mice.

Figure 4.

AAV-NDNF administration increases the number of survival motor neurons in SOD1^G93A mice

(A and D) SOD1^G93A mice received a single intrathecal injection of AAV at P61 (A) or P85 (D). (B and E) Representative confocal microscopy images of lumbar spinal cord longitudinal sections (L3–L5) stained with ChAT antibody to label motor neurons together with GFP or Myc antibody to mark AAV-transduction and NDNF expression in P135 WT or SOD1^G93A mice with AAV-GFP or AAV-NDNF injection at P61 (B) or P85 (E). Scale bars, 500 μm (top row) or 250 μm (boxed areas). (C and F) Quantification for the survival of motor neurons in WT or SOD1^G93A mice with AAV-GFP or AAV-NDNF injection at P61 (C) and P85 (F). Data are shown as mean ± SEM of five mice in each group injected at P61 and six to seven mice injected at P85. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, Student’s t test. See also Figure S6.

Consistent with increased neuronal degeneration in ALS mice, only around 8.27% neuromuscular junctions (NMJs) at the endplate in SOD1^G93A mice at P135 were fully innervated and others were partially innervated or un-innervated (Figures 5B and 5C). Notably, in SOD1^G93A mice treated with AAV-NDNF at P85, the percentage of innervated NMJs was markedly increased and that of un-innervated NMJs markedly decreased (Figures 5A–5C). At late symptomatic stages, the loss of neuronal control usually leads to muscle atrophy.³⁷ Notably, SOD1^G93A mice at P135 showed a marked reduction in muscle mass compared with WT mice and AAV-NDNF treatment at P105 mildly but significantly prevented the muscle loss (Figures 5E and 5F). In line with the attenuation of neuromuscular pathology, AAV-NDNF injection at late post-symptomatic stages mildly extended lifespan in SOD1^G93A mice compared with AAV-GFP group (Figure 5G).

Figure 5.

NDNF ameliorates muscle atrophy and denervation in SOD1^G93A mice and extends life span

(A and D) Schematic diagram of the intrathecal injection. (B) Representative images of gastrocnemius muscles stained with antibodies against neurofilament (NF) and synapsin (Syn) to label presynaptic nerves, together with fluorophore-conjugated α-BTX to mark postsynaptic acetylcholine receptor (AChR) in P135 WT or SOD1^G93A mice injected with AAV-GFP or AAV-NDNF at P85. Scale bars, 50 μm. (C) Quantitative analysis for the percentage of NMJs that were fully innervated (100%), denervated (0%), or partially innervated. Data are shown as mean ± SEM of three mice in each group with approximately 35 muscle fibers counted in each mouse. ∗∗∗p < 0.001, Student’s t test. (E) Representative images of the gastrocnemius muscle in P135 WT or SOD1^G93A mice injected with the AAV-NDNF or AAV-GFP at P105. Scale bars, 500 μm. (F) Quantitation of the muscle weight. Data are shown as mean ± SEM (n = 9 for WT mice, n = 9 for SOD1^G93A mice injected with AAV-GFP, n = 6 for SOD1^G93A mice injected with AAV-NDNF). ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, Student’s t-test. (G) Effects of AAV-NDNF administration on the lifespan SOD1^G93A mice. SOD1^G93A mice received a single intrathecal injection of AAV at P100-110 and then life span were determined. Log rank test was used for life span analysis (n = 22–24 mice in each group, ∗∗p < 0.01, AAV-NDNF vs. AAV-GFP injected SOD1^G93A mice). See also Figure S7.

Similar to other neurodegenerative diseases, protein aggregation caused by mutation-induced conformation changes and misfolding of mutated or interacting proteins is believed to confer toxicity to affected cells, such as motor neurons in ALS.^9,38 In SOD1^G93A mice at P135, extensive ubiquitin-positive inclusions were observed in the ventral horn of spinal cord with concurrent motor neuron loss (Figures 6A and 6B). Interestingly, the intensity of ubiquitin signals was markedly decreased in SOD1^G93A mice treated with AAV-NDNF (Figures 6A and 6B). It is unclear how NDNF counteracts the proteostasis impairment and whether the decrease in ubiquitin signals is secondary to increased motor neuron survival. Neuroinflammation has been shown to play important roles in ALS pathogenesis that is characterized by neurotoxic microglia and astrocyte activation.^39,40,41 The SOD1^G93A mice also exhibited increases in the amount of activated GFAP-positive astrocytes and Iba1-positive microglia (Figures S7A–S7D). However, AAV-NDNF administration had no effect on microglia and astrocyte activation (Figures S7A–S7D). Nevertheless, these results are not able to rule out the possibility of the involvement of these cell types in NDNF-mediated attenuation of ALS progression.

Figure 6.

AAV- NDNF suppresses abnormal accumulation of ubiquitinated protein aggregation in SOD1^G93A mice

(A) Representative images of immunostaining with indicated antibodies in spinal cord transections from P135 WT mice or SOD1^G93A mice injected with AAV-GFP or AAV-NDNF at P85. Ubiquitin signals (red) mark misfolded aggregated proteins, ChAT (white) marks motor neurons, GFP or Myc (green) marks AAV infected cells. Scale bars, 100 μm (left) or 20 μm (insets in right). (B) Quantification of the ubiquitin intensity in the ventral horn (VH) in each group. Data are shown as mean ± SEM of ubiquitin intensity presented as arbitrary units (au) from 3 mice each group with 30 areas quantified with the value of WT mice set as 1. ∗p < 0.05, ∗∗∗∗p < 0.0001, Student’s t test.

AAV-NDNF administration increases cell survival signaling and reduces apoptotic signaling

To further investigate the mechanisms by which AAV-NDNF administration attenuates motor neuron degeneration, we determined gene expression changes in the ventral spinal cord of SOD1^G93A mice treated with AAV-NDNF or control virus. Compared with control group, AAV-NDNF treated SOD1^G93A mice at middle post-symptomatic stage showed 87 up-regulated genes and 207 down-regulated genes (fold change>2; p < 0.05) (Figure 7A; Table S1, GSE226291). The Gene Ontology (GO) analysis showed that the up-regulated genes were enriched in pathways regulating spermatogenesis, skeleton muscle contraction, and dendritic spine morphogenesis, as well as cell division (Figures 7B and 7C). Down-regulated genes were enriched in immune response, apoptotic signaling mediated by P53, and tumor necrosis factor (TNF) production (Figures 7B and 7D). Actually, P53 has been shown to be involved in C9orf72-mediated pathogenesis of ALS and frontotemporal dementia,⁴² and over-production of inflammatory cytokines like TNF is associated with neuronal loss in ALS.^40,41

Figure 7.

Gene expression changes in SOD1^G93A mice with AAV-NDNF administration

(A) AAV-GFP or AAV-NDNF were intrathecally administrated into SOD1^G93A mice at P74 and samples of spinal cord ventral part of P104 mice were subjected to bulk RNA-sequence analysis. The sequencing results were compared with the mouse reference transcriptome (mm10), and 28,162 genes were detected in total. In AAV-NDNF group, 87 genes were up-regulated and 207 genes were down-regulated (fold change >2) (p value <0.05). (B) Heatmap of the top 30 DEGs overlapped in 3 replicate samples (excluding the genes beginning with Gm, which did not have a given name). The expression level of individual gene represented by the color bar is relative to the mean value of all samples (see Table S1 for details). (C and D) Pathway enrichment analysis of up-regulated (C) or down-regulated genes (D). (E) Venn diagram showing the number of overlapping genes up-regulated in SOD1 mutant mice and down-regulated after AAV-NDNF injection (see Tables S2 and S3 for details). (F) Pathway enrichment analysis of genes up-regulated in SOD1 mutant mice and reversed after AAV-NDNF injection. See also Figure S8.

It has been shown that ALS mice displayed significant increases in pro-inflammatory pathways responsible for neurotoxicity⁴³ and immunosuppressive T lymphocytes become reduced and dysfunction in ALS patients.⁴⁴ In line with these findings, the analysis of a published datasets (GSE106364) about differentially expressed genes (DEGs) in the spinal cords of SOD^G86R mice compared with control mice⁴⁵ identified 593 up-regulated DEGs (Figure 7E; Table S2). Among them, 70 genes were observed to be down-regulated in AAV-NDNF-treated SOD1^G93A mice (Figure 7E; Table S3). Interestingly, the genes up-regulated in ALS mice and reversed by AAV-NDNF were enriched in the pathways promoting TNF production, immune response, and T cell apoptotic process, or intrinsic apoptotic signaling pathway by p53 class mediator (Figure 7F).

The mammalian target of rapamycin (mTOR) signaling plays central role in integrating nutrients, growth factors, and cellular energy inputs in biological processes, such as gene transcription, protein synthesis, inhibition of apoptosis, promotion of cell growth, and inhibition of apoptosis, and thus is considered a novel target to arrest neurodegeneration.⁴⁶ We found that AAV-NDNF administration markedly increased the intensity of phosphorylated ribosomal protein S6 (pS6), suggesting the increased activation of mTOR signaling pathway, in spinal cord motor neurons (Figures S8A–S8D). This result is in line with the conclusion that NDNF activates cell survival signaling.

Discussion

ALS is a fatal disease featured by selective deterioration of upper cortical and lower spinal motor neurons, but the etiology remains elusive. Because of the paucity of therapeutic targets, current treatments are based on symptom control and respiratory support with limited available small molecular medicines that only provide modest benefits.^14,47 Several gene therapy clinical trials targeting a specific ALS variant gene or sporadic disease without identified genetic cue providing common protective factors are on the way, opening a new avenue of therapeutic possibilities.^17,19,23 In this study, we provide evidence showing that AAV-mediated delivery of NDNF attenuates the disease progression of SOD1^G93A mice, the most widely used ALS model in preclinical studies.

Because of the heterogeneity in genetic traits and disease mechanisms, supplementing trophic or protective factors may provide a common disease-modifying approach for both genetic and sporadic patients. Alongside this idea, several trophic factors have been tested for their effects on disease progression in ALS mice. For example, AAV9-mediated intramuscular delivery of insulin-like growth factor 1 has been shown effective in preserving spinal motor neurons and mildly extending life span in SOD1^G93A mice.^48,49 Other trophic factors including glial-derived neurotrophic factor (GDNF), granulocyte-colony stimulating factor, and hepatocyte growth factor delivered by various AAV vectors intramuscularly or intra-spinally also have shown protective effects, leading to improvement in motor strength or prolonged survival.^50,51,52 Notably, systemic administration of AAV9-GDNF into SOD1^G93A rats has shown adverse side effects, albeit with only modest improvement.⁵³ In addition to trophic factors, AAV delivery of chaperone-like factors that directly affect accumulation of disease-related misfolded proteins has also been tested; e.g., AAV9-mediated intraspinal delivery of migration inhibitory factor showed significant delay in disease onset and prolonged survival most likely via inhibiting SOD1 misfolding and aggregates formation.⁵⁴ Nevertheless, its pro-inflammatory effects are a concern because neuroinflammation is believed to contribute to ALS pathogenesis. The results described in this study about a role for NDNF in ameliorating ALS progression demonstrates an attractive potential in treating all forms of ALS and even other neurodegenerative diseases.

Unlike classical neurotrophic factors that have been extensively studied like above mentioned GDNF and brain-derived neurotrophic factors, the mechanisms by which NDNF exerts its functions remain unclear. Recently, Nord, an NDNF-like protein in Drosophila, has been shown to control wing development via fine-tuning bone morphogenetic protein signaling.⁵⁵ NDNF is also expressed in cultured human endothelial cells and has been shown to promote endothelial cell function and revascularization in ischemic mice, likely via the activation of Akt/endothelial nitric oxide synthase signaling.⁵⁶ We found that, in SOD1^G93A mice spinal cord, over-expression of NDNF caused the up-regulation of signal pathways governing biosynthetic processes and cell growth and the down-regulation of cell death signaling (Figures 7C and 7D). In line with this notion, mTOR signaling was activated in spinal motor neurons with NDNF over-expression (Figure S8). All these results support the conclusion that NDNF protects motor neurons from degeneration most likely via activation of cell survival and growth signaling. Because the CBh promoter used to drive NDNF expression is not neuron selective, we cannot rule out the contribution of non-neuronal factors such as neuroinflammation and the vasculature system in the role of NDNF in ALS. Indeed, it has been suggested that neurovascular signaling is closely associated with neurodegenerative diseases, including ALS.⁵⁷

A genetic study has shown that Ndnf is mutated in hypogonadotropic hypogonadism, a rare genetic disorder characterized by infertility and the absence of puberty, and mice lacking Ndnf showed delayed gonadotropin-releasing hormone neuron migration and altered olfactory axonal projections to the olfactory bulb.⁵⁸ Interestingly, the top up-regulated genes induced by NDNF are in the category of spermatogenesis (Figure 7C). The effects of ALS and NDNF on fertility would be an interesting topic.

Because both upper and lower motor neurons are affected in ALS and contribute to disease progression, modifying strategies need to consider the challenge of crossing the brain-blood barrier (BBB) and long-range transport. In decades, many studies have tried to optimize the AAV vectors, including the promoter strength, selective tropism, and dosing, to enhance transduction efficiency. The discovery for the ability of nature occurred AAV9 in targeting motor neurons and crossing the BBB has made it realized the aim of CNS targeting via intravenous administration.⁵⁹ Recent modified AAV-PHP.eB based on AAV9 has further enhanced the efficiency of non-invasive gene delivery to the wide-scale CNS system via the vasculature in mice.³² To minimize potential side effects on other organs as that seen in a previous GDNF study for SOD1^G93A mice,⁵³ we chose the route via direct intraspinal injection of AAV-PHP.eB, which has achieved wide-scale expression of target genes in the whole brain and spinal cord. Although AAV-NDNF has displayed marked protective effects on spinal motor neurons, the effect on cortical motor neurons is minimal (Figure S6). The beneficial effects of AAV-NDNF intrathecal administration might mainly attribute to its effect on spinal motor neurons.

Taken together, the identification of NDNF as a neurotropic factor that is protective for motor neurons in ALS mice suggests a potential to be considered in clinical therapies of ALS and other neurodegenerative diseases. Whether the combination of multiple factors exhibits synergistic effects merits further investigation.

Materials and methods

Animals

All animal experiments including mouse housing, breeding, and surgical procedures were executed in compliance with the ethical guidelines of the Institutional Animal Care and Use Committee of ShanghaiTech University. Both male and female animals were used for the analysis (see Table S4 for gender distribution in related experiments). All mice were housed under a 12-h light-dark cycle in the institutional animal care facility. SOD1^G93A transgenic mouse line (stock number 004435; Jackson Laboratories) was bred on a pure C57BL6 background to eliminate confounding genetic influences. The early post-symptomatic disease stage is defined as the age at which the animal’s motor deficiency is detected in gait analyses (approximately 60 days).⁶⁰ The middle disease period was determined as the age at which the animal’s hindlimb began tremors (approximately 90 days).⁶¹ At the middle post-symptomatic stage, significant death of somatic motor neurons innervating limb muscles has occurred. Mice at the end-stage disease show up to 50% loss of lumbar motor neurons (approximately 130–140 days).⁶²

Immunoblotting

Mice under deep anesthesia were decapitated and tissues were quickly harvested, snap-frozen in liquid N₂, and stored at −80°C. For protein preparation, frozen tissues or cultured cells were lysed in RIPA buffer (Beyotime, P0013B) with protease inhibitor cocktail (B14001, Bimake) added. After centrifugation (12,000 rpm for 15 min), supernatants were collected and mixed with loading buffer (Tsingke, TSJ010) and then heated at 95°C for 5 min. The proteins were separated by SDS-PAGE, and transferred to polyvinylidene difluoride membranes, which were then blocked with 5% BSA in 0.1% tween in PBS for 1 h followed by incubation with primary antibodies overnight at 4°C, and visualization with horseradish peroxidase (HRP)-conjugated secondary antibodies. Optical densities of bands were analyzed by Fiji. The source of antibodies and dilutions were as follows: GAPDH (Proteintech, 60004-1-Ig, 1:1,000), NDNF (Abcam, ab177456, 1:500), Myc (Sigma, 05-419, 1:1,000), pS6 (Cell signaling, 5364, 1:1,000), S6 (Cell signaling, 2317, 1:1,000), Actin (Millipore, MAB1501, 1:1,000), HRP-conjugated goat anti-mouse (Beyotime, A0216), and HRP-conjugated goat anti-rabbit (Beyotime, A0208).

AAV-NDNF construct and virus production, purification, and titration

To generate AAV-NDNF construct, total RNA was extracted from mouse lung tissue with Trizol (Invitrogen, 15596026) and then full-length cDNA of NDNF was obtained by reverse transcription followed by PCR using forward and reverse primers (5′-ATGGAGCTGTTCTACTGGTGTTTGCT -3′ and reverse primer 5′-CTAACAGACCTTCCTTGTTTTCACAATTTTACTTTGATAC -3′), and then inserted between inverted terminal repeats (ITRs) under the control of CBh or hSyn promoter. AAV production, purification, and titration followed the procedures described in Addgene protocols (Addgene, AAV Production). Briefly, AAV was produced in HEK293T cells by triple transfection with the AAV helper plasmid pHelper, pUCmini-iCAP-PHP.eB (Addgene, #103005), and target plasmid encoding GFP or NDNF. Poly ethylene imine (PEI 40000, Yeasen, #40816ES02) was used for transfection of plasmids into HEK293 cells, and the ratio of PEI to plasmid was 3:1. Virion particles were purified on an iodixanol gradient (Sigma-Aldrich, #D1556-250ML) with ultracentrifugation for 2.5 h at 4, 8,000×g at 4°C, and concentrated in PBS using Amicon Ultra-100 centrifugation filters (Millipore, #UFC910096), and then stored with single-use aliquots at −80°C until injection. Virus titer was determined as virus genomes per milliliter, by measuring the levels of ITR with qPCR. Before releasing the viral DNA from the particles, all extra DNA was removed by digestion with DNase I (Thermo Fisher Scientific, #EN0525). The qPCR was performed using the SYBR Green (Bimake, #B21702), with the following ITR primers: 5′-GGAACCCCTAGTGATGGAGTT-3′ (forward), 5′-CGGCCTCAGTGAGCGA-3′ (reverse); and the results were analyzed by QuantStudio Real-Time PCR Software (Thermo Fisher Scientific) and R program (http://www.r-project.org).

Intrathecal injection

Intrathecal injection followed the procedure described in a previous study.⁶³ Briefly, animals are anesthetized using avertin (500 mg/kg, intraperioneally) until no pedal withdrawal reflex was observed. The fur on the back of the anesthetized mice were shaved from the tail to the caudal thoracic spine. AAV buffered with PBS (5 × 10¹¹ vg per mouse, 50 μL) was injected into the intrathecal space between the L3 and L4 or L4 and L5 vertebrae with a 30G needle. Indications of proper entry include a reflexive flick of the tail or of a hind leg. Five minutes after injection, the needle was withdrawn without fluid leaking. The animals were placed in a recovery cage to recover, and monitored until normal activity resumed.

Behavior analysis

Animal gaits were analyzed using a computer-assisted Catwalk that objectively quantified gait parameters, including duration, speed, different phases of the step cycle, pressure applied during locomotion, and distance between paws. Mice need to be trained in advance at a fixed time every day, with each training repeated three times lasting for 1 week. To motivate animals to cross the glass plate, an attractive goal box with its own bedding was attached to the walkway. During the recording process, run maximum variation (%) reflected whether the mice passed through the data collection area at a uniform speed.⁶⁴ If this value was less than 30%, the data were considered valid. Each mouse was recorded three times and all the eligible tests were used for motor behavior analyses.

Immunostaining and RNAscope

Animals are euthanized using avertin (500 mg/kg, intraperitoneally) and then trans-cardially perfused with PBS, followed by 4% paraformaldehyde fixation. Subsequently, perfused tissues were post-fixed overnight, dehydrated in 30% sucrose for 1 day, and then sectioned at 30 μm using a freezing microtome (Leica, CryoStar NX50). Sections of control and experimental groups were pasted on the same glass slide to maintain uniform conditions during the staining and image collection processes.

For immunostaining, fixed tissue sections were washed with PBS three times and subjected to antigen retrieval by citrate and then permeabilized in 0.3% Triton X-100 in PBS for 30 min. After blocking with 10% BSA (Sigma, #V900933) for 2 h at room temperature, the slices were incubated with various primary antibodies at 4°C overnight, washed with PBS three times, incubated with secondary antibodies for 2 h at room temperature in the dark, and mounted with mounting reagent (DAKO, S3023) for observation. Primary antibodies or reagents for immunostaining were as follows: DAPI (Beyotime, C1002), ChAT (goat, Millipore, Ab1449), NeuN (mouse, Millipore, MAB377), NeuN (rat, oasis Biofarm, OB-PRT013-01), NDNF (rabbit, Abcam, ab177456), GFP (chicken, Aves, GFP-1020), Myc (rabbit, Cell Signaling Technology, 2278S), Myc (mouse, Sigma, 05–419), S100 (rabbit, Dako, Z0311), Neurofilament (rabbit, Biolegend, 837904), synapsin-1 (rabbit, Cell signaling, 5297S), ubiquitin (rabbit, Proteintech, 10201-2-AP), GFAP (rabbit, Dako, Z0334), Iba1 (rat, Abcam, Ab283346), and pS6 (rabbit, Cell signaling, 5364). Secondary antibodies for staining were as follows: Alexa Fluor 488-conjugated donkey anti-mouse (Invitrogen, A21202), Alexa Fluor 488-conjugated donkey anti-chicken (Jackson Laboratories, 703-546-155), Alexa Fluor 555-conjugated donkey anti-rabbit (Invitrogen, A31572), Alexa Fluor 555-conjugated donkey anti-rat (Invitrogen, A48270), Alexa Fluor 555-conjugated donkey anti-goat (Invitrogen, A21432), Alexa Fluor 647-conjugated donkey anti-rabbit (Invitrogen, A32795), and Alexa Fluor 647-conjugated donkey anti-goat (Invitrogen, A21447).

For the quantification of motor neurons in the brain, sections were stained with NeuN to label the nuclei of neurons. Coronal sections from bregma 1.18–0.14 were used for quantitative analysis. The cortical layers (I–VI) were visualized by DAPI staining. Two adjacent brain slices were taken from each mouse, and two square fields (250 μm height × 250 μm) in layer V were selected for each slice to count NeuN positive cells.

For the quantification of NMJs, gastrocnemius muscles were teased into filaments composed of two to three muscle fibers, and stained with antibodies against neurofilament and synaptic vesical protein synapsin-1, and fluorophore-conjugated α-bungarotoxin (α-BTX, Biotium, 00005) to label presynaptic nerves and postsynaptic acetylcholine receptors, respectively. Images were acquired on an Olympus confocal microscope using a 20× objective and superimposed onto maximum-intensity projections by Fiji for better visualization. The graphical abstract created with BioRender.com.

RNAscope analysis was performed by using reagents from Advanced Cell Diagnostics Company following protocol introduced previously,⁶⁵ with Ndnf (ACD, 447471-C1) and Chat (ACD, 40831-C2) probes.

Analysis for NDNF in the CSF

The CSF of the control or AAV-NDNF infected mice was collected following the procedure described previously.⁶⁶ The level of NDNF in CSF were measured using commercial ELISA kits (JN823680, Jining Shiye) and the presence of NDNF-Myc fusion protein was determined by immunoblotting with anti-Myc antibody.

RNA sequencing

High-throughput sequencing of total RNA isolated from lumbar spinal cord was performed on Illumina NovaSeq PE150 system with average length of 150 nucleotides for every read of paired end. Raw data were filtered by Cutadapt to generate clean reads and then mapped to mouse reference genome (mm10) using HISAT2 (version 2.2.1). Counts tables were generated with HTSeq (version 2.0). Statistical analysis was performed in R (version 4.1.1). Transcript levels were normalized using DESeq2 package (version 1.32.0). Transcripts with zero counts across all samples were removed, and mathematical artifacts (e.g., negative infinites) were replaced with “NA,” DEGs were identified using DESeq2 package. The Benjamini-Hochberg procedure was used to control the false-discovery rate of the multiple tests and q < 0.05 accepted as significant. Heatmap analysis was performed in heatmap package (version 1.0.12). GO analysis was performed using DAVID website (https://david.ncifcrf.gov/tools.jsp) with p < 0.05 and log₂ (fold change) of >1 or <−1. RNA sequencing data are presented in Table S1 and have been deposited in GEO (GEO: GSE226291).

Statistical analyses

Data satisfied to Gaussian distribution were quantified with the Student’s t test for comparison between the two groups. ANOVA was used to compare differences between groups over time and among multiple groups. Kaplan-Meier log rank test for survival was used to compare lifespan between groups. All results are expressed as means ± SEM. The statistical significance was indicated by ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, and ns (no significant difference). The numbers of experimental replicates were included in related figure legends. All statistical analyses were conducted using GraphPad Prism7 for Windows.

Data and code availability

The published article includes main datasets generated during this study and is freely available for research use only. The raw RNA-sequencing data of lumber spinal cord in mice treated with AAV-NDNF and control virus have been deposited to the NCBI GEO Archive under the accession number GSE226291. The processed data was shown in Table S1.

Supplemental information

Table S1. TEffects of NDNF on gene expression in the spinal cord of SOD1^G93A mice

AAV-GFP or AAV-NDNF were intrathecally administrated into SOD1^G93A mice at P74, and after 30 days, samples of spinal cord ventral part were collected and subjected to bulk RNA-sequence analysis. In AAVNDNF group, 207 genes were up-regulated and 87 genes were down-regulated (fold change >2, p-value < 0.05). Raw data has been deposited under the accession number GSE226291.

Table S2. The expression levels of genes up-regulated in SODG^86R mice

Raw data was extracted from GSE106364.

Table S3. The expression levels of genes down-regulated in AAV-NDNF treated SOD1^G93A mice

The list shows 70 overlapping genes that were up-regulated in SODG^86R mice (GSE106364) and downregulated after AAV-NDNF treatment of SOD1^G93A mice (GSE226291).