Inflammatory cytokines disrupt astrocyte exosomal HepaCAM-mediated protection against neuronal excitotoxicity in the SOD1G93A ALS model

Abstract

Astrocyte secreted signals substantially affect disease pathology in neurodegenerative diseases. It remains little understood about how proinflammatory cytokines, such as interleukin-1α/tumor necrosis factor–α/C1q (ITC), often elevated in neurodegenerative diseases, alter astrocyte-secreted signals and their effects in disease pathogenesis. By selectively isolating astrocyte exosomes (A-Exo.) and employing cell type–specific exosome reporter mice, our current study showed that ITC cytokines significantly reduced A-Exo. secretion and decreased spreading of focally labeled A-Exo. in diseased SOD1G93A mice. Our results also found that A-Exo. were minimally associated with misfolded SOD1 and elicited no toxicity to mouse spinal and human iPSC–derived motor neurons. In contrast, A-Exo. were neuroprotective against excitotoxicity, which was completely diminished by ITC cytokines and partially abolished by SOD1G93A expression. Subsequent proteomic characterization of A-Exo. and genetic analysis identified that surface expression of glial-specific HepaCAM preferentially mediates A-Exo’s axon protection effect. Together, our study defines a cytokine-induced loss-of-function mechanism of A-Exo. in protecting neurons from excitotoxicity in amyotrophic lateral sclerosis.

Inflammatory cytokines diminish astrocyte protection in ALS.

INTRODUCTION

Amyotrophic Lateral Sclerosis (ALS) is a fast progressive neurodegenerative disease in which upper and lower motor neurons (MNs) are degenerated. In ALS, dysfunction of MN axons has been widely observed, including defective energy supply and protein transport, cytoskeleton impairment, and altered axon interaction with environmental cues (1, 2). These axonal changes often precede motor neuronal soma and axon degeneration, considerably contributing to disease pathology. During early development, extracellular matrix (ECM) signals, including cell adhesion molecules (CAMs), N-cadherin, and integrins (3, 4), can substantially promote axon growth and maintain axon integrity. Cell adhesion and ECM molecules are among the most dysregulated genes in motor cortex and are the most aberrantly spliced in microdissected spinal cord anterior horn of patients with sporadic ALS based on transcriptome analysis (5, 6). Altered expression of specific axon attraction/repulsion signals, i.e., semaphorins and ephrins (7, 8), has also been observed in SOD1 mouse models of ALS. Recently generated C9orf72 PolyGR and polyPR knock-in mice further revealed an increase in ECM proteins that is neuroprotective (9).

Previous studies from human postmortem tissues and animal models, especially transgenic SOD1 mutant models, have demonstrated that altered glia to neuron signaling pathways, such as induction of a proinflammatory cytokine environment (10), reduced glutamate uptake, and metabolic support (2, 11), contribute to the pathogenesis of ALS. Consistently, recent transcriptome analysis from postmortem frontal and motor cortex of patients with ALS further showed that the ALS subtype with more pronounced glial activation and neuroinflammation gene signature tends to have accelerated disease progression (12). In particular, a number of studies have demonstrated that astrocyte conditioned medium (ACM) from mouse astrocytes expressing human SOD1 (hSOD1) mutant or from human (familial and sporadic) ALS patient brain astrocytes is able to substantially affect health and survival of primary, embryonic stem cell, or induced pluripotent stem cells (iPSC)-derived human and mouse MNs (13–15). A number of factors, including cytokines, reactive oxygen species, adenosine triphosphate/adenosine, and recently excessive polyphosphate, have been implicated to mediate ACM’s effect on MNs (16–18).

Exosomes (50 to 150 nm in diameter), a major type of secreted extracellular vesicles (EVs), are derived from intraluminal vesicles (ILVs) in the endosomal compartment and are released from multivesicular bodies (MVBs) during endosome maturation (19). EVs and exosomes secreted from various central nervous system cell types have been shown to regulate activity-dependent translation (20) and glutamate transporter function (21), to promote axon myelination and transport (22), and to maintain brain vascular integrity (23). In ALS models, EVs/exosomes are reported to associate with TDP43 (24), mutant hSOD1 (25), and C9orf72-derived dipeptide repeat proteins (26), which has been proposed to potentially facilitate propagation of protein aggregates in neurodegeneration (27) and triggers neuroinflammatory responses and even toxicity (28).

In the current study, we investigated how proinflammatory cytokines affect secretion and spreading of astrocyte exosomes (A-Exo.) in vitro and in vivo, as well as A-Exo. properties to MNs. Our results also revealed astrocyte exosomal surface hepatic and glial cell adhesion molecule (HepaCAM)-mediated neuroprotection against neuronal excitotoxicity that is disrupted by inflammatory cytokines in the SOD1G93A ALS mouse model.

RESULTS

Proinflammatory cytokines reduce secretion of A-Exo. and contribute to decreased spreading of A-Exo. along the spinal cords of SOD1G93A mice

Reactive astrogliosis, manifested by drastic molecular and morphological changes of astrocytes, is often closely associated with elevated proinflammatory cytokine levels in neurodegenerative diseases including ALS. A combination of such cytokines, interleukin-1α (IL-1α)/tumor necrosis factor–α (TNFα)/C1q (ITC), has been shown to transform astrocytes into a “reactive” state that resembles reactive astrocytes in neurodegenerative diseases (29). It remains unclear whether these cytokines affect exosome secretion from astrocytes. We first isolated exosomes from ACM using the size exclusion chromatography (SEC) approach as we recently described (30), which much more effectively separates exosomes from secreted proteins than the conventional ultracentrifugation method (30). The identity of exosomes was confirmed by positive mouse (m)CD63 immuno-electron microscopy (EM) (Fig. 1A), as well as typical exosome size distribution (Fig. 1B), determined by ZetaView nanoparticle tracking analysis (NTA). Quantitative measurement of secreted vesicles in ACM showed that ITC cytokine treatment significantly reduced the number of secreted exosomes in both nontransgenic (NTg) wild-type (WT) control and SODG93A⁺ astrocytes (Fig. 1, B and C). On the other hand, overexpression of human SOD1G93A protein only had a tendency, yet not significant (P = 0.52), in decreasing exosome numbers from astrocytes (Fig. 1C). We previously generated hCD63–green fluorescent protein (GFP)^f/f mice to selectively and efficiently label secreted exosomes and their intracellular precursor ILVs in a cell type–specific manner by tagging the C terminus of exosome surface marker human (h) CD63 with the CopGFP reporter (31). GFP⁺ exosomes from ACM were clearly observed following the SEC procedure (fig. S1A), and the GFP tag and exosome markers CD81 and endogenous CD63 were also detected in A-Exo. (fig. S1B). Consistent with previous report that CD63 is constantly internalized from the plasma membrane (32), we observed primarily internal localization of hCD63-GFP (Fig. 1D) and endogenous mCD63 in astrocytes (fig. S1C). Instead, hCD63-GFP fluorescence was primarily observed within the RAB7⁺ (a selective MVB marker) vesicular structure inside astrocytes (Fig. 1D, ii, and the orthogonal view in Fig. 1D, v). Colocalization of hCD63-GFP with RAB7 was also observed at the edge of the MVB (Fig. 1D, v). These images, together with GFP^+-secreted vesicles (fig. S1, A and B), support the MVB localization of inwardly budded hCD63-GFP⁺ ILVs in astrocytes (Fig. 1D, v) and the endosomal origin of exosomes found in ACM.

Fig. 1. Proinflammatory cytokines reduce secretion of A-Exo. and in vivo spreading along spinal cords of SOD1G93A mice.

(A) Representative immunoEM images of mouse (m)CD63 labeling in A-Exo. White arrows, CD63⁺ A-Exo.; scale bar, 100 nm. (B) Representative size distribution of A-Exo from untreated and cytokine (ITC)–treated NTg astrocyte cultures. (C) Relative fold change of NTA measured exosomes from untreated control and ITC cytokine–treated NTg and SOD1G93A astrocyte cultures. A-Exo. were SEC purified from ACM (20 ml per sample) and analyzed using ZetaView NTA. Astrocytes were all ~90 to 95% confluent. n = 6 to 8 biological replicates/group. (D) Representative images [scale bars, (i) 20 μm and (ii to v) 10 μm] and orthogonal views (XZ and XY) of hCD63-GFP (green) and RAB7 (purple) in NTg astrocyte cultures. White arrows, colocalizations of hCD63-GFP and RAB7. (E) Schematic view of AAV5-mCherry-Gfap-Cre injection into spinal cords of hCD63-GFP^f/+SOD1G93A⁺ and hCD63-GFP^f/+ mice and representative images of induced hCD63-GFP and mCherry reporter signals (indication of AAV5 spreading) in proximal and distal sections from the injection site. Scale bar, 200 μm. (F) Schematic view of AAV5-mCherry-Gfap-Cre injection into spinal cords of hCD63-GFP^f/+Ai14-tdT^f/+ mice and a representative image of induced hCD63-GFP and tdTomato reporter signals. White arrows, tdT⁺ astrocytes; scale bar, 20 μm. Quantification of the spreading distance of hCD63-GFP⁺ signal along spinal cords in AAV5-mCherry-Gfap-Cre (G) or AAV8-mCherry-CaMKIIa-Cre (H) injected mice. n = 4 mice per group for disease progression stage (P105 to P115, injected at P90 to P100); n = 3 mice per group for pre-disease stage (P75 to P80, injected at P60). Error bars denote SEM, and P values were calculated using one-way analysis of variance (ANOVA) followed by a Tukey post hoc test in (C) and (G) and unpaired t test in (H).

To examine how elevated neuroinflammation and SOD1G93A expression affects A-Exo. dynamics in vivo in ALS models, we performed a single stereotaxic injection of AAV5-mCherry-Gfap-Cre virus [0.5 μl, 4 × 10¹² genome copy (gc)/ml] into lumbar spinal cords of hCD63-GFP^f/+SOD1G93A⁺ and control hCD63-GFP^f/+ mice [postnatal day 90 (P90) to P100 age] (Fig. 1E). Bright hCD63-GFP⁺ fluorescence was observed at the injection site (Fig. 1E), which was indicated by positive mCherry signals (white arrows) expressed by the AAV. The restricted mCherry signals observed only at the injection site and nearby (~250 μm; fig. S2A) indicated minimal spreading of AAV from the injection site. The number of mCherry⁺ astrocytes at the injection site of hCD63-GFP^f/+SOD1G93A⁺ and control hCD63-GFP^f/+ mice is also highly similar (fig. S2B). As a result, hCD63-GFP signals observed on distant sections from the injection site indicate the spreading of exosomes initially produced and secreted from astrocytes at the injection site. Extracellularly localized hCD63-GFP⁺ signals (outside of tdT⁺ astrocytes, highlighted in the dashed line; Fig. 1F) were clearly visualized in the lumbar spinal cord of AAV5-mCherry-Gfap-Cre–injected hCD63-GFP^f/+Ai14-tdT^f/+ mice. We then quantified and calculated the percentage of hCD63-GFP⁺ area (red dashed circle) out of the ventral horn gray matter (GM) area (white dashed circle) on spinal cord sections with various distance from the injection site (fig. S2C). Consistent with reduced secretion of A-Exo. in vitro, the overall spreading distance (in both directions) of hCD63-GFP⁺ A-Exo. from the injection site was much reduced in the diseased (P105 to P115 days) hCD63-GFP^f/+SOD1G93A⁺ mice compared to age-matched control hCD63-GFP^f/+ mice (Fig. 1G). In parallel, reactive A1 astrocytes, indicated by up-regulated C3 immunoreactivity (29) [colocalization with glial fibrillary acidic protein (GFAP), white arrows; fig. S2D, i], were commonly observed in spinal cord sections of hCD63-GFP^f/+SOD1G93A⁺ but not in control hCD63-GFP^f/+ mice (yellow arrows; fig. S2D, i′), supporting their involvement in reducing A-Exo. spreading in vivo. This reduction is specific to the disease pathology, as the reduced spreading distance was only observed in the disease progression but not presymptomatic (P60) stage (Fig. 1G) at which the neuroinflammation is less elevated in spinal cords. Comparable spreading distance of A-Exo. between hCD63-GFP^f/+ and hCD63-GFP^f/+SOD1G93A⁺ mice in the presymptomatic stage also indicates that SOD1G93A overexpression itself unlikely reduces A-Exo. spreading. To examine whether the reduced spreading of A-Exo. in the ALS model is cell type specific, we also performed a single stereotaxic injection of AAV8-mCherry-CaMKIIa-Cre virus into lumbar spinal cords of hCD63-GFP^f/+SOD1G93A⁺ and control hCD63-GFP^f/+ mice (fig. S2E), which induced selective expression of hCD63-GFP in neurons and spreading of hCD63-GFP⁺ neuronal exosomes (N-Exo.) along spinal cords. We found that the overall spreading distance of N-Exo. (~3000 μm; fig. S2F) is far shorter than A-Exo. (8000 μm) with no difference between diseased hCD63-GFP^f/+SOD1G93A⁺ and control hCD63-GFP^f/+ mice (Fig. 1H). These results suggest that the spreading of A-Exo. is particularly affected by the disease pathology in ALS.

Misfolded SOD1 proteins are minimally associated with SOD1G93A A-Exo. (and N-Exo.) in vitro and in vivo

Misfolded SOD1 proteins, either derived from human (h) SOD1G93A or conformation altered endogenous SOD1, can be secreted extracellularly and induce multiple neurotoxic pathways (33). Although secreted mutant hSOD1 proteins were previously detected in SOD1G93A A-Exo. in vitro or from brain samples and can be toxic to spinal MNs (25, 34), these A-Exo. were isolated through ultracentrifugation, which is often associated with secreted ECM proteins and protein aggregates, as we and others previously showed (30, 35). To examine whether misfolded SOD1 is indeed associated with SEC-purified SOD1G93A A-Exo., SOD1G93A A-Exo. (2 μg per lane, from 20 ml of ACM per sample) were analyzed by immunoblotting of various SOD1 proteins. We also isolated A-Exo. using the ultracentrifugation protocol for comparison. As previously shown (25), misfolded SOD1 proteins were detected in ultracentrifugation-isolated A-Exo. (fig. S3, A and B) using two well-characterized antibodies that specifically recognize misfolded SOD1, SOD1 exposed dimer interface (SEDI, conformation specific) and A5C3 clone (hSOD1 specific) (36, 37). However, hSOD1G93A and misfolded SOD1 were not detected in SOD1G93A A-Exo. fractions (Fig. 2A) even with overexposure (fig. S3C). Instead, they were only detected in exosome-free ACM fractions (Fig. 2A). Treatment of SOD1G93A astrocyte cultures with ITC cytokines also induced no exosomal expression of misfolded SOD1 (Fig. 2B). Similarly, hSOD1G93A and misfolded SOD1 proteins were only detected in exosome-free neuron-conditioned medium (NCM) fractions (Fig. 2C) but not in SEC-purified N-Exo.

Fig. 2. Misfolded SOD1 proteins are minimally associated with SOD1G93A A-Exo. and N-Exo. in vitro and in vivo.

(A) Representative immunoblots of hSOD1G93A and misfolded SOD1 in eluted SEC fractions (pooled as indicated) of SOD1G93A ACM (10 ml). Thirty microliters of pooled fractions per lane were loaded. (B) Representative immunoblots of hSOD1G93A and misfolded SOD1 on SOD1G93A A-Exo. (2 μg per lane) isolated from ACM of untreated and ITC cytokine–treated SOD1G93A astrocytes. (C) Representative immunoblots of hSOD1G93A and misfolded SOD1 in eluted SEC fractions (pooled as indicated) of SOD1G93A NCM. Thirty microliters of pooled fractions per lane were loaded. Red dashed box highlights exosome fractions as determined by the exosome marker CD81. Representative confocal microscopy images of hCD63-GFP⁺ and misfolded SOD1 detected by either SEDI (D) or A5C3 (E) antibody from spinal cords of AAV5-mCherry-Gfap-Cre–injected hCD63-GFP^f/+SOD1G93A⁺ mice. [(ii) and (iii)] Two magnified fields from (i); scale bars, 20 μm. Sections used were ~1500 μm from the injection site with no mCherry signals. (F) Percentage of colocalization of hCD63-GFP⁺ puncta signal with misfolded SOD1 detected by either A5C3 or SEDI antibody from spinal cords of AAV5-mCherry-Gfap-Cre–injected hCD63-GFP^f/+SOD1G93A⁺ mice. n = 12 to 21 sections from 4 mice per group. Representative confocal microscopy images of hCD63-GFP⁺ and misfolded SOD1 detected by either SEDI (G) or A5C3 (H) antibodies from spinal cords of AAV8-mCherry-CaMKIIa-Cre–injected hCD63-GFP^f/+SOD1G93A⁺ mice. [(ii) and (iii)] Two magnified fields from (i); scale bars, 20 μm. Sections used were ~1000 μm from the injection site with no mCherry signals. (I) Percentage of colocalization of hCD63-GFP⁺ puncta signal with misfolded SOD1 detected by either A5C3 or SEDI antibody from spinal cords of AAV8-mCherry-CaMKIIa-Cre–injected hCD63-GFP^f/+SOD1G93A⁺ mice. n = 17 sections per three mice.

To further examine whether misfolded SOD1 is associated with A-Exo. in spinal cords of diseased SOD1G93A mice, we performed misfolded SOD1 immunostaining with either SEDI or A5C3 antibodies on lumbar spinal cord sections (~1500 μm from the injection site, no mCherry signals) of AAV5-mCherry-Gfap-Cre–injected hCD63-GFP^f/+SOD1G93A⁺ mice (P105 to P115). Specific misfolded SOD1 immunoreactivity on these sections was clearly visualized, together with AAV5-mCherry-Gfap-Cre–induced hCD63-GFP signals (Fig. 2, Di and Ei). Only occasional colocalization of misfolded SOD1 with hCD63-GFP puncta was observed, as shown in magnified fields [Fig. 2, D (ii and iii) and E (ii and iii)]. Single focal plane confocal image–based quantification found only an average of 2% (with the A5C3 antibody) or <1% (with the SEDI antibody) colocalized hCD63-GFP signals (Fig. 2F). Additional size analysis further found that the vast majority of colocalized hCD63-GFP⁺ puncta were >1 μm (fig. S3D), typical of intracellular hCD63-GFP⁺ endosomes (due to ILV labeling) but not secreted small-sized exosomes (<300 nm). The colocalization of misfolded SOD1 immunoreactivity, revealed by either SEDI or A5C3 antibodies, with hCD63-GFP on lumbar cord sections (~1000 μm from the injection site, no mCherry signals) of AAV8-mCherry-CaMKIIa-Cre–injected hCD63-GFP^f/+SOD1G93A⁺ mice (P105 to P110) is also low at either 5% (with the A5C3 antibody) or 1% (with the SEDI antibody). These in vitro and in vivo results indicate a minimal association of misfolded SOD1 with both A-Exo. and N-Exo., therefore contesting the view that SOD1G93A A-Exo. and N-Exo. facilitate the transmission of misfolded SOD1 in ALS (38, 39).

A-Exo. elicit no toxicity to primary spinal MNs

As there is only a minimal association between SOD1G93A with SOD1G93A A-Exo., we next examined whether A-Exo. from SOD1G93A astrocytes, especially treated with ITC cytokines, are toxic to spinal MNs. Primary spinal neuron cultures were prepared from embryonic day 12 (E12) embryos and treated [at 7 days in vitro (DIV)] with equal quantities (1 μg) of NTg, SOD1G93A, ITC-NTg, and ITC-SOD1G93A A-Exo. for 72 hours. The overall neuronal death was assessed by terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining followed by choline acetyltransferase (ChAT) and Hoechst 33342 nuclear staining (Fig. 3A). Spinal MNs that underwent cell death were identified as TUNEL⁺ChAT⁺Hoechst⁺ (Fig. 3A), and spinal MN death rate was calculated by dividing TUNEL⁺ChAT⁺Hoechst⁺ neurons by total ChAT⁺Hoechst⁺ neurons. We found that A-Exo., isolated from ACM of NTg or SOD1G93A⁺ astrocytes with or without ITC cytokine treatment, have no obvious toxic effect on either SOD1G93A⁻ (Fig. 3B) or SOD1G93A⁺ (Fig. 3C) spinal MNs compared to untreated spinal MN cultures, indicated by highly comparable MN death rates. SOD1G93A⁺ spinal MNs tend to have a reduced yet not significant (P = 0.12) death rate with NTg A-Exo. treatment compared to untreated control (Fig. 3C). To further examine the continuous effect of NTg and SOD1G93A A-Exo. on spinal MN survival, spinal MN cultures were prepared from HB9–enhanced GFP (eGFP) transgenic mice in which spinal MNs are selectively labeled with eGFP driven by the promoter of the homeobox transcription factor HB9 (40), which were also confirmed by positive ChAT immunoreactivity (Fig. 3D, eGFP⁺ChAT⁺ spinal MNs, 10 DIV). By quantifying the survival of HB9-eGFP⁺ spinal MNs from over a 9- to 10-day live-cell imaging period (examples of live-cell imaging from a culture dish in fig. S4A), we further examined the protective effect of NTg and SOD1G93A A-Exo. on spinal MNs. While NTg A-Exo. had no clear protective effect on survival of HB9-eGFP⁺SOD1G93A⁻ spinal MNs (Fig. 3E), a significantly higher (P = 0.04) survival rate in NTg A-Exo.–treated HB9-eGFP⁺SOD1G93A⁺ spinal MNs compared to control was observed from days 6 to 10 after treatment (Fig. 3F). However, SOD1G93A A-Exo. had no significant effect in protecting either HB9-eGFP⁺SOD1G93A⁻ or HB9-eGFP⁺SOD1G93A⁺ spinal MNs (Fig. 3, E and F). Notably, the survival rate of both SOD1G93A A-Exo.–treated HB9-eGFP⁺SOD1G93A⁻ and HB9-eGFP⁺SOD1G93A⁺ spinal MNs was not significantly lower than untreated control, further supporting the notion that SOD1G93A A-Exo. are unlikely to elicit toxicity to spinal MNs.

Fig. 3. A-Exo. from untreated and ITC cytokine–treated NTg and SOD1G93A astrocyte cultures elicit no toxicity to primary spinal MNs.

(A) Representative images of cultured spinal MNs immunostained with TUNEL, ChAT, and Hoechst 33342. White arrows, TUNEL-positive MNs; scale bar, 50 μm. TUNEL assay–based quantification of apoptotic SOD1G93A⁻ (B) and SOD1G93A⁺ (C) spinal MNs following treatment with 1× PBS (control), NTg, SOD1G93A, ITC-NTg, or ITC-SOD1G93A A-Exo, respectively. Spinal MN death rate was calculated by dividing TUNEL⁺ChAT⁺Hoechst⁺ neurons by total ChAT⁺Hoechst⁺ neurons. n = 22 to 40 fields (20×) from two biological replicates per group. Data are presented in violin plots. The lines in each violin plot indicate 75% quantile, median, and 25% quantile (from top to bottom). One microgram of A-Exo. was used in each treatment in (B) and (C). P values were calculated using one-way ANOVA followed by a Tukey post hoc test. (D) Representative image of ChAT⁺HB9-eGFP⁺ MNs in spinal cord neuronal cultures. Scale bar, 100 μm. Survival rate of HB9-eGFP⁺SOD1G93A⁻ (E) and HB9-eGFP⁺SOD1G93A⁺ (F) spinal MNs following treatment with 1× PBS (control), NTg, or SOD1G93A A-Exo. n = 3 to 5 independent samples with ~300 MNs per sample at DIV 3. Error bars denote SEM, and P value determined by two-way ANOVA, F_1,30 = 4.36 and Sidak post hoc analysis.

ITC cytokine treatment diminishes NTg A-Exo.–mediated protection against neuronal excitotoxicity in cortical neurons and in human iPSC–derived MNs

Glutamate-induced excitotoxicity is considered one of the primary pathogenic mechanisms in neurodegenerative diseases that lead to both neuronal cell death and axon degeneration (41). Encouraged by the beneficial effect of NTg A-Exo. on SOD1G93A⁺ spinal MN survival as observed above, to further determine whether NTg A-Exo. are protective against excitotoxicity, cortical neurons were first treated with NTg A-Exo. (1 μg) for 72 hours followed by l-glutamate (l-Glu., 100 and 1000 μM) treatment for 24 hours. Given the limited availability of spinal MNs, we decided to use cortical neuronal cultures to test the effect of various exosome types, as described below, against neuronal excitotoxicity. As expected, l-Glu. treatment induced dose-dependent neuronal death, either 30% (100 μM) or 50% (1000 μM) above (P = 0.004 or 0.0004, respectively) the baseline (Fig. 4A), as assessed by the TUNEL assay (fig. S4B). l-Glu. treatment also induced dose-dependent neurite degeneration, indicated by significantly increased βIII-TUBULIN⁺ neurite “beading” morphology (42) [Fig. 4, B (ii and iii) and C] that is typically caused by microtubule disassembly (43). Excitingly, NTg A-Exo. pretreatment almost completely prevented glutamate-induced excitotoxic neuronal cell death with much reduced TUNEL signals (Fig. 4A and fig. S4B). In addition, overall βIII-TUBULIN⁺ neurites in NTg A-Exo.–treated neuronal cultures were much better preserved, especially at the 100 μM l-Glu. treatment than in l-Glu.–treated neuronal cultures alone (Fig. 4, B and C). These results demonstrate that NTg A-Exo. significantly protects neurons against glutamate excitotoxicity.

Fig. 4. ITC cytokine treatment diminishes NTg A-Exo.–mediated protection against neuronal excitotoxicity in cortical neurons and in human iPSC–derived MNs.

(A) Cortical neuronal death rate following treatment with l-Glu. (100 or 1000 μM) and l-Glu. + NTg A-Exo. (1 μg). n = 23 to 31 fields (20×) per three biological replicates per group; Control, untreated cultures. Neuronal death rate was calculated by TUNEL⁺Hoechst⁺ neurons divided by Hoechst⁺ neurons. Representative images (B) of βIII-TUBULIN staining and quantification (C) of neurite beading in cortical neurons following the same treatment. White arrows in (ii), (iii), and (vi) indicate neurite beading; scale bar, 50 μm. Neurite beading index was calculated by the total number of “beads” per field (20×) divided by the βIII-TUBULIN⁺ area in the same field. n = 23 to 31 fields (20×) per three biological replicates per group. (D) Schematic diagram of treatment of neuronal cultures with l-Glu. and A-Exo. (1 μg) collected from different genotype and ITC treatment combination. (E) Cortical neuronal death rate treated with l-Glu. (100 μM), l-Glu. + NTg or SOD1G93A A-Exo., l-Glu. + ITC-NTg, or ITC-SOD1G93A A-Exo. n = 11 to 26 fields (20×) per two biological replicates per group. All legends are shown (and the same as) in (F) due to limited space. (F) Quantification of neurite beading in cortical neurons following the same treatment as indicated in (E). n = 27 to 33 fields (20×) per three biological replicates per group. (G) Quantification of neurite beading in human iPSC–derived MNs treated with l-Glu. (200 μM) and combinations of l-Glu. with different A-Exo. (2 μg). n = 26 to 48 fields (20×) per three biological replicates per group. Data from different fields of the same biological replicate were averaged and presented in each panel. Error bars denote SEM, and P values in (E) to (G) were calculated using one-way ANOVA followed by a Tukey post hoc test. P values in (A) and (C) were calculated using two-way ANOVA followed by a Tukey post hoc test. A.U., arbitrary units.

As elevated ITC cytokine levels are commonly observed in neurodegenerative diseases including ALS (10) that affect astrocyte reactivity (29), we then examined how ITC cytokine treatment and human SOD1G93A overexpression alter the protective effect of A-Exo. against neuronal excitotoxicity (experimental scheme shown in Fig. 4D). Equal quantities (1 μg) of SEC-purified A-Exo. were isolated from ACM of untreated and ITC cytokine–treated NTg and SOD1G93A⁺ astrocyte cultures and added onto the cortical neuronal cultures followed by l-Glu. (100 μM) treatment. While NTg A-Exo. consistently protected neurons from l-Glu.–induced excitotoxic cell death (Fig. 4E and fig. S4C) and βIII-TUBULIN⁺ neurite degeneration (Fig. 4F), SOD1G93A A-Exo. were unable to protect against l-Glu.–induced neuronal cell death (Fig. 4E and fig. S4C) but were still able to reduce l-Glu.–induced neurite degeneration (P = 0.01; Fig. 4F). ITC cytokine treatment on NTg astrocytes completely abolished NTg A-Exo’s protective effect against excitotoxic cell death (Fig. 4E and fig. S4C) and l-Glu.–induced neurite degeneration (Fig. 4F). A-Exo. purified from ITC cytokine–treated SOD1G93A astrocyte cultures further exacerbated l-Glu.–induced neuronal excitotoxic cell death and neurite degeneration compared to SOD1G93A A-Exo. alone (Fig. 4, E and F, and fig. S4C).

To further examine whether different A-Exo. affect the survival of human MNs, we treated human iPSC–derived MN cultures we previously established (44) with equal quantities (2 μg) of NTg, SOD1G93A, ITC-NTg, and ITC-SOD1G93A A-Exo. and l-Glu. (200 μM). Our differentiation protocol results in a culture with 80% neurons as MNs as confirmed by ChAT and HB9 immunostaining (fig. S5, A and B). Although l-Glu. (200 μM) treatment induced clear βIII-TUBULIN⁺ neurite damage (white arrows; fig. S5C, ii), confirmed by increased neurite beading index quantification (Fig. 4G), no apparent additional neuronal cell death induced by l-Glu. was observed (fig. S5D). This is likely due to attenuated l-Glu. toxicity in a growth factor–enriched medium for culturing human MNs. Consistent with mouse spinal MN cultures, various exosomes alone had no toxic effect on MN survival and neurite integrity, indicated by comparable MN death rates (fig. S5D) and similar neurite beading index (fig. S5E). NTg and SOD1G93A A-Exo. also similarly attenuated l-Glu. (200 μM)–induced neurite beading on human MNs (yellow arrows; fig. S5C, iii, and Fig. 4G), while ITC-NTg and ITC-SOD1G93A A-Exo. lost such protective effects (Fig. 4G). Together, these results show that A-Exo. are able to provide important protection to neurons against excitotoxicity, which can be abolished by ITC cytokine treatment on astrocytes.

ITC cytokine treatment and expression of SOD1G93A differentially alters protein compositions in A-Exo.

How cytokine treatment and expression of SOD1G93A affect proteomic compositions of A-Exo. is unknown. We therefore performed proteomic analysis on NTg, SOD1G93A, ITC-NTg, and ITC-SOD1G93A A-Exo. (20 μg per sample), as summarized in Fig. 5A, with in-gel trypsin digestion and liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis. As both ITC-NTg and ITC-SOD1G93A A-Exo. almost completely diminished A-Exo.’s protection against excitotoxicity, we were particularly interested in A-Exo. proteins altered by ITC cytokine treatment. Differentially expressed proteins (fold change >1.5) between NTg and ITC-NTg A-Exo or between SOD1G93A and ITC-SOD1G93A A-Exo were further overlapped. We found 25 co–up-regulated (underlined; Fig. 5B) and 78 co–down-regulated (underlined; Fig. 5C) proteins, which were primarily plasma membrane, cytoplasm, and extracellular space proteins (Fig. 5D) based on Ingenuity Pathway Analysis (IPA) subcellular localization analysis. More than half (45 of 78) of plasma membrane proteins were co–down-regulated (Fig. 5D), including transporters (adenosine triphosphatase Na⁺/K⁺ subunits, GJA1, SLC1A3, etc.) and ECM and adhesion molecules (such as CSPG4, NCAM1, HepaCAM). We recently showed that HepaCAM mediates A-Exo.’s effect in stimulating axon growth during early development (30). Altered extracellular space and cytoplasm proteins are shown in the heatmaps (fig. S6, A and B). Additional IPA pathway analysis further identified the complement pathway among the co–up-regulated proteins (fig. S6C) and several adhesion molecules to axon guidance pathways among co–down-regulated proteins (fig. S6D). Expression changes of representative complement and Ephrin pathway proteins are further shown in fig. S6 (E and F). SOD1G93A expression alone differentially altered protein compositions more than ITC treatment did in A-Exo, especially leading to much more up-regulated proteins (114, Fig. 5B) but smaller number of down-regulated protein (29, Fig. 5C) compared to NTg A-Exo. These altered proteins were more overlapped with ITC cytokine effect on NTg A-Exo. (24 up- and 7 down-regulated proteins) than on SOD1G93A A-Exo (10 up- and 10 down-regulated proteins) protein compositions. IPA pathway analysis further found ECM organization, integrin cell surface interactions, and Toll-like receptor cascades as top pathways of SOD1G93A-induced A-Exo. protein changes (fig. S6, G and H). Specific proteins in proteomic analysis were summarized in tables S1 to S9.

Fig. 5. ITC cytokine treatment and expression of SOD1G93A differentially alters protein cargoes in A-Exo.

(A) Exosomes were isolated from untreated and ITC cytokine–treated primary NTg and SOD1G93A astrocyte cultures for LC-MS/MS proteomic analysis. Three-way Venn diagrams of proteins that were up-regulated (B) and down-regulated (C) in A-Exo. from astrocyte cultures treated with ITC cytokines and SOD1G93A expression. Fold change > or < 1.5 was used as the cutoff to identify changed proteins either by ITC treatment or SOD1G93A expression. The number of proteins that are overlapped in each comparison was highlighted in the diagram. Black arrows indicate either up- or down-regulated proteins in A-Exo from each comparison, presented as the numerator divided by the denominator. (D) Cellular location of proteins up-regulated with cytokine stimulation in NTg and SOD1G93A (co–up-regulated) and down-regulated with cytokine stimulation in NTg and SOD1G93A (co–down-regulated). (E) Heatmap of co–up-regulated and co–down-regulated plasma membrane proteins. Values are Log₂(averaged iBAQ values). Representative immunoblot (F) and quantification (G) of HepaCAM protein in NTg, ITC-NTg, SOD1G93A, and ITC-SOD1G93A A-Exo. based on total loaded protein (gel staining intensity, 2 μg per lane); n = 5 to 6 independent samples per group. Representative HepaCAM immunoblot (H) and quantification (I) from lumbar spinal cords of SOD1G93A (onset, P90 to P100; mid-disease, P110 to P120) and age-matched NTg mice. N = 5 to 9 mice per group. β-Actin was used as the control for normalization. Error bars denote SEM, and P values are calculated using one-way ANOVA followed by a Tukey post hoc test in (G) and (I).

Subsequent immunoblotting of HepaCAM [~70 KDa, the glycosylated functional form (45)] further confirmed that ITC cytokine treatment on both NTg and SOD1G93A astrocytes reduced more than 70% (normalized with total proteins) HepaCAM expression on equal quantities (2 μg) A-Exo. (Fig. 5, F and G), in line with the diminished effect of these exosomes on protecting neurons from excitotoxity. HepaCAM expression in SOD1G93A A-Exo. was also 30% reduced (P = 0.07) compared to NTg A-Exo. (Fig. 5G). On the other hand, ITC cytokine treatment had no significant (P = 0.25 for NTg, although trending lower) effect in reducing cellular HepaCAM expression in NTg and SOD1G93A astrocytes (fig. S7, A and B), suggesting that ITC cytokines may alter the internalization and sorting of HepaCAM into endosome compartments to reduce HepaCAM levels in A-Exo. How HepaCAM expression changes in ALS models is essentially unknown. We first examined its expression in lumbar spinal cord of SOD1G93A mice at onset (P90 to P100) and the mid-disease stage (P110 to P120). While HepaCAM expression levels were slightly reduced at the onset compared to NTg controls, a 40% reduction (P = 0.02) of HepaCAM was observed at the mid-disease stage (Fig. 5, H and I). To specifically examine expression changes of HepaCAM in A-Exo. in diseased SOD1G93A mice, we analyzed HepaCAM immunoreactivity that is colocalized with hCD63-GFP⁺ A-Exo. from gray matter spinal cord sections of AAV5-mCherry-Gfap-Cre–injected hCD63-GFP^f/+SOD1G93A⁺ mice (P110). As shown in fig. S7C, clear HepaCAM immunoreactivity was observed in gray matter spinal cord sections. HepaCAM immunoreactivity colocalized with CD63-GFP⁺ A-Exo. was 30% reduced in hCD63-GFP^f/+SOD1G93A⁺ mice compared to control hCD63-GFP^f/+ mice (fig. S7D). These in vitro and in vivo results consistently show reduced expression levels of HepaCAM especially in A-Exo. in the SOD1G93A ALS model.

Astrocyte exosomal HepaCAM is necessary and sufficient to protect neuronal axons from excitotoxic degeneration

HepaCAM has been shown to facilitate proper membrane targeting of astroglial functional proteins (46, 47) and help define astrocyte to astrocyte boundary (48). Given that HepaCAM is on the surface of astrocyte-secreted exosomes and A-Exo. protect neurons from excitotoxicity, we postulate that HepaCAM is involved in A-Exo’s protective effects on neurons. Cortical neurons were treated with equal amount (1 μg) of NTg A-Exo. or HepaCAM knockout (KO) A-Exo. followed by l-Glu. (100 μM) treatment. HepaCAM-deficient A-Exo. had very similar (P = 0.46) protective effect in reducing l-Glu.–induced neuronal cell death as NTg A-Exo. (Fig. 6A). However, HepaCAM-deficient A-Exo.–treated cortical neurons showed significantly (P = 0.03) increased neurite beading induced by l-Glu. compared to NTg A-Exo. + l-Glu.–treated neurons (Fig. 6B), suggesting that the lack of HepaCAM preferentially diminishes the protective effect of A-Exo. against neurite degeneration but not neuronal cell death. Meanwhile, increased HepaCAM levels in A-Exo. (red arrow; fig. S7E) by lentivirus-mediated overexpression of HepaCAM in astrocytes significantly (P = 0.04) rescued HepaCAM-deficient A-Exo. effect on l-Glu.–induced neurite beading (Fig. 6B) but had no effect on neuronal survival (Fig. 6A). HepaCAM has a typical extracellular immunoglobulin G (IgG) domain (ECD) to potentially interact with itself and other surface proteins (49). We then directly cultured cortical neurons on poly-d-lysine (PDL) + bovine serum albumin (BSA) or PDL + HepaCAM ECD coating followed by l-Glu. treatment. As shown in representative images in Fig. 6C (ii, iii, and v), neurons grown on the PDL + HepaCAM coating showed much healthier neurites (yellow arrows) and much lower TUNEL signals (white circles) compared to neuronal processes (white arrows) with phosphate-buffered saline (PBS) + BSA coating. The TUNEL and neurite beading index quantification confirmed that HepaCAM ECD completely protected against l-Glu.–induced neuronal cell death (Fig. 6D) and reduced 60% of l-Glu.–induced neurite beading (Fig. 6E).

Fig. 6. HepaCAM preferentially mediates A-Exo.’s protection against glutamate-induced neurite degeneration.

Cortical neuronal death rate (A) and neurite beading index (B) in primary cortical neurons treated with l-Glu (100 μM), NTg A-Exo. + l-Glu, HepaCAM KO A-Exo. + l-Glu, HepaCAM KO + HepaCAM overexpression (OE) A-Exo. + l-Glu. Control, 1× PBS; n = 24 to 31 fields (20×) per three biological replicates per group for neuronal death rate. n = 24 to 38 fields (20×) per three biological replicates per group for neurite beading analysis. (C) Representative images of βIII-TUBULIN and TUNEL staining in primary cortical neurons grown on PDL + BSA or PDL + HepaCAM ECD and treated with l-Glu. (100 μM). White arrows, neurites with beading; yellow arrows, preserved axons with minimal beading. (ii) A magnified view from the box in (i); (v) a magnified view from the box in (iv); circles indicate TUNEL⁺ neurons in (iii). Scale bar, 30 μm [(i), (iii), (iv), and (vi)]; 20 μm [(ii) and (v)]. Cortical neuronal death rate (D) and neurite beading index (E) in primary cortical neurons grown on PDL + BSA or PDL + HepaCAM ECD and treated with l-Glu. (100 μM). n = 29 to 32 fields (20×) per three biological replicates per group for neuronal death rate. n = 18 to 26 fields (20×) per two biological replicates per group for neurite beading analysis. Data from different fields of the same biological replicate were averaged and presented in each panel. Error bars denote SEM, and P values in (A), (B), (D), and (E) were calculated using one-way ANOVA followed by a Tukey post hoc test.

To specifically examine whether A-Exo., especially exosomal HepaCAM, acts locally and directly at axons in protecting axons from l-Glu.–induced degeneration, we prepared neuronal cultures in microfluidic chambers by seeding neurons at one side (soma side) of the chamber so that axons grew out of the central groove to the other side (axon side) of the chamber (Fig. 7A, i and ii). AAV8-CAG-tdT was also added to the soma side to label neurons and subsequently axons (Fig. 7A, iii and iv), which allowed time-lapse imaging to follow l-Glu.–induced axon degeneration over time. We first tested whether l-Glu. locally induced axon degeneration by adding l-Glu. into the axon side after the axons were labeled with the tdT reporter. tdT⁺ axons showed minimal axon degeneration over time (yellow arrows; Fig. 7, B and C). In contrast, l-Glu. added into the soma side induced time-dependent and substantial axon degeneration (white arrows; Fig. 7, B and C), quantified by reduced tdT fluorescent intensity on axons (Fig. 7C), suggesting that l-Glu.–induced axon degeneration is primarily resulted from neuronal cell death but not locally induced at axons. Soma-added l-Glu.–induced axon degeneration was also confirmed with reduced TAU immunoreactivity (fig. S7F). This is not unexpected given that more glutamate receptors are enriched on plasma membrane near soma than in axons (50).

Fig. 7. Local and direct A-Exo. HepaCAM signaling to axons contributes to A-Exo.’s protection against excitotoxic axon degeneration.

(A) Schematic representation of the microfluidic chamber system and representative images of the soma and axonal compartments. (i) Differential interference contrast (DIC) image of neurons at the soma side; (ii) DIC image of axons at the axon side; (iii) tdT⁺-labeled neurons by adding AAV8-CAG-tdT onto the soma side of the chamber at 10 DIV; (iv) tdT⁺-labeled axons at the axon side of the chamber; white lines and gray arrows indicate the ending of the soma side (i) and the beginning of the axon side (ii) of the chamber. Scale bar, 100 μm. Representative images (B) of tdT⁺ axon degeneration and quantification (C) of axonal tdT⁺ intensity over time following addition of l-Glu. into either soma or axon sides. White arrows, degenerating axons indicated by beading following soma side of l-Glu. treatment; yellow arrows, healthy axons even after l-Glu. treatment at the axon side; scale bar, 50 μm; n = 19 to 30 fields (10×) per three chambers per two biological replicates per group. Diagram of the experiment and representative tdT⁺ axon images (D) and tdT⁺ intensity quantification (E) at the axon side of the microfluidic chambers. l-Glu was added onto the soma side and all A-Exo. (NTg and HepaCAM A-Exo.) were added onto the axon side. White lines and gray arrows indicate the beginning of the axon side of the chamber. HepaCAM ECD was coated on microfluidic chambers. Time-lapse tdT⁺ images were taken at 0-, 6-, and 22-hours after l-Glu. treatment, and the tdT fluorescent intensity was quantified. n = 19 to 39 fields (10×) per four chambers per two biological replicates per group; white arrows, degenerating axons; yellow arrows, healthy axons; scale bar, 50 μm. Error bars denote SEM. P values were calculated using two-way ANOVA followed by a Tukey post hoc test.

On the basis of our observations that HepaCAM deficiency preferentially diminishes the protective effect of NTg A-Exo. against neurite degeneration but not neuronal cell death (Fig. 6, A and B), we then decided to add NTg or HepaCAM KO A-Exo. into the axon side and l-Glu. into the soma side to examine the local and direct effect of these exosomes on l-Glu.–induced axon degeneration (Fig. 7D, diagram). Time-lapse imaging was performed on outgrowing tdT⁺ neuronal axons following A-Exo. (72 hours before) and l-Glu. treatments. We found that axons were better preserved with NTg A-Exo. pretreatment (yellow arrows; Fig. 7D, i′ and ii″) at both 6 and 24 hours after l-Glu. treatment (Fig. 7E) than with l-Glu. treatment alone, which induced clear axon beading and loss of tdT fluorescence signals (white arrows; Fig. 7D, i and ii). In contrast, pretreatment of HepaCAM-deficient A-Exo. was unable to protect axons from l-Glu.–induced degeneration (white arrows; Fig. 7D, i″ and ii″) at both time points (Fig. 7E), while HepaCAM ECD coating on the axon side is able to provide comparable protective effect on axons as NTg A-Exo (Fig. 7D, i‴ and ii‴), especially at 24 hours after l-Glu. treatment (Fig. 7E). These results further demonstrated a direct interaction between HepaCAM in A-Exo. and axons in protecting axons from l-Glu.–induced degeneration.

DISCUSSION

Our current study showed that proinflammatory ITC cytokines significantly reduced secretion of exosomes from astrocytes regardless of the astrocyte genotype. In parallel, focally labeled A-Exo. spread less distance along spinal cords in diseased SOD1G93A mice with elevated neuroinflammation and were minimally associated with misfolded SOD1 (including human SOD1G93A) proteins. In addition, our result showed that A-Exo. elicited no toxicity to mouse spinal and human iPSC–derived MNs. Instead, NTg A-Exo. were neuroprotective against excitotoxic neuronal cell death and axon degeneration. However, these protective effects of NTg A-Exo. were completely diminished by ITC cytokine treatment or partially reduced by SOD1G93A expression in astrocytes. These results define a previously unknown mechanism about how proinflammatory cytokines (and SOD1G93A expression) affect astrocyte-secreted signals in modulating neuronal survival and axon integrity, potentially enabling therapeutic developments by considering exosome-based neuroprotection against excitotoxicity in ALS and other neurodegenerative diseases.

Although misfolded SOD1 was previously reported to be associated with A-Exo. isolated using the ultracentrifugation method and induces toxicity to MNs (25), our study showed that misfolded SOD1 was not detected in SEC-isolated A-Exo. (and N-Exo.) but in exosome-free ACM (and NCM) fractions, supporting the notion that misfolded SOD1 is co-pelleted with exosomes during the ultracentrifugation. We showed that misfolded SOD1 proteins, as well as several other astrocyte-secreted synapse-modulating proteins (30), can be detected in ultracentrifugation- but not in SEC-isolated A-Exo (fig. S3). Given that astrocytes commonly secrete a number of protein signals (51), it is particularly important to separate astrocyte secreted proteins versus exosomes (or other vesicles). By using cell type–specific exosome reporter mice, we further observed very limited colocalization of misfolded SOD1 with A-Exo. and N-Exo. in vivo. Although it has been proposed that exosomes help propagate mutant hSOD1 in vitro (39), we found reduced A-Exo. spreading in diseased spinal cords of SOD1G93A mice, possibly due to reduced secretion of A-Exo. as we observed in vitro as a result of increased cytokines including ITCs in disease. Disease-induced ECM changes or altered phagocytosis of A-Exo. by activated microglia may also reduce A-Exo. spreading in spinal cords of ALS. Nonetheless, our results indicate that A-Exo. are unlikely to promote propagation of misfolded SOD1 proteins in vivo in ALS.

Previous studies in astrocyte/neuron cocultures or ACM treatment in vitro have mostly focused on the impact of mutant ALS genes, such as SOD1G93A (13, 14), but less on the influence of proinflammatory cytokines on astrocyte-mediated toxicity to (motor) neurons. In agreement with previous observations that ITC cytokines transform astrocytes into a reactive state and induce astrocyte dysfunction and even toxicity to neurons (29), our results here showed that A-Exo. isolated from ITC cytokine–treated astrocyte cultures were unable to protect neurons from excitotoxic cell death and axon degeneration as did by NTg A-Exo. On the other hand, A-Exo., either isolated from SOD1G93A or ITC cytokine–treated astrocytes, had no toxicity to spinal and human iPSC–derived MNs. Thus, our results support a loss of function of ITC (NTg and SOD1G93A) A-Exo. in protecting neurons from excitotoxicity in neurodegeneration but not gain of toxicity to neurons. As this loss of function of A-Exo. is particularly induced by ITC cytokines but less influenced by the SOD1G93A overexpression, it likely represents a common astrocyte-mediated pathogenic mechanism that contributes to both familial and sporadic ALS. Although hSOD1 mutant expressing astrocytes or astrocytes derived from sporadic ALS brain tissue have previously been reported to secrete toxic factors to induce MN cell death (13, 15), given that ACM contains highly heterogeneous factors, it is conceivable that astroglial secretion elicits both loss of function and toxicity mechanisms to affect MN health and survival in ALS. Neonatal cortical astrocytes were used in our current study to be able to isolate sufficient amount of A-Exo, as it is not technically feasible to isolate exosomes from adult spinal cord astrocytes. With overexpression of ALS-causing mutant proteins or cytokine treatment, neonatal cortical astrocytes have been well established as an in vitro model to study astrocyte impact on MN survival in ALS models (13, 14).

Our proteomic analysis showed that SOD1G93A overexpression and ITC cytokine treatment induced drastically differential changes of protein compositions in A-Exo. SOD1G93A overexpression mainly increased protein levels in A-Exo. compared to NTg A-Exo., but ITC cytokines had a profound effect in reducing protein levels in A-Exo. regardless of astrocyte genotype. In particular, a number of plasma membrane proteins were down-regulated by ITC cytokines, including the glia-enriched adhesion molecule HepaCAM. Consistent with that, ITC cytokines had much stronger effect than SOD1G93A overexpression on diminishing the protective effect of A-Exo. against excitotoxic neuronal cell death and neurite degeneration. Moreover, HepaCAM deficiency abolished protective effect of A-Exo. against excitotoxic axon degeneration. By using the microfluidic chamber to separate axons from neuronal cell bodies and applying A-Exo. locally into the axon side, our results further showed that local and direct interactions between A-Exo. HepaCAM and axons were required to mediate A-Exo.’s protection against excitotoxic axon degeneration. Meanwhile, this by no means excludes other axon-protective mechanisms that come from healthier neuronal cell bodies as a result of A-Exo.–mediated protection. Although cortical neonatal but not spinal cord astrocytes were used in our study, we previously showed that HepaCAM is expressed in both cortical and spinal cord A-Exo that similarly stimulates developmental axon growth (30), suggesting that astrocyte exosomal HepaCAM expression and its effect can be conserved between spinal cord and cortical astrocytes. This is further supported by our results here that cytokine treatment on astrocytes significantly reduced HepaCAM expression in A-Exo, consistent with HepaCAM decrease in spinal cords of diseased SOD1G93A mice.

How HepaCAM mediates local protective effects along axons and downstream neuronal pathways remain unclear. Several CAM molecules are known to activate receptor tyrosine kinase signaling in neurons to promote neurite outgrowth (52). In addition, necroptotic machinery, especially receptor-interacting kinase 1 (RIPK1) and RIPK3, have been particularly implicated in mediating excitotoxic axon degeneration (53). Whether and how A-Exo. HepaCAM interacts with kinase signaling pathways to be neuroprotective will be investigated in the future. HepaCAM is not required in A-Exo.’s protection against excitotoxic neuronal cell death as HepaCAM-deficient A-Exo. are able to equally protect against excitotoxic neuronal cell death as NTg A-Exo. despite the observation that HepaCAM ECD is sufficient to protect against excitotoxic neuronal cell death. It is possible that HepaCAM provides a protective “CAM” extracellular environment to depress glutamate stimulation on its receptors to induce excitotoxicity. Specific and distinct mechanisms how A-Exo. protect against excitotoxic cell death will be investigated in the future, which may provide additional avenues to develop neuroprotective strategies.

MATERIALS AND METHODS

Mice

hCD63-GFP–floxed mice were generated in the laboratory by homologous recombination, as previously described (31). The WT mice (C57BL/6J), Ai14-tdT^f/f (#007914), B6.C-Tg(CMV-cre)1Cgn/J (#006054), B6SJL-Tg (SOD1*G93A)1Gur/J (#002726), and B6.Cg-Tg(Hlxb9-GFP)1Tmj/J (#005029) were obtained from the Jackson Laboratory. HepaCAM KO mice were generated by breeding HepaCAM-floxed mice (48) (a gift from C. Eroglu at Duke University) with CMV-Cre mice. Both male and female mice were used in all experiments. All mice were maintained on a 12-hour light/dark cycle with food and water ad libitum. Care and treatment of animals in all procedures strictly followed the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Guidelines for the Use of Animals in Neuroscience Research. Animal protocols used in this study have been approved (B2022-50 and B2022-51) by Tufts University IACUC committee.

Mouse primary astrocyte culture

P0 to P3 WT B6 and HepaCAM-KO mouse pups were decapitated, and cerebral cortices were removed and transferred into astrocyte growth medium [Dulbecco’s minimum essential medium (DMEM) supplemented with 10% FBS (fetal bovine serum, Sigma-Aldrich) and 1% penicillin-streptomycin (Pen-Strep, Thermo Fisher Scientific)] for dissection on ice. Meninges were removed, and cortices were minced and placed into 0.05% trypsin-EDTA solution (T9324, Sigma-Aldrich) for 10 min in a 37°C water bath. The enzymatic reaction was stopped by addition of astrocyte growth medium. The tissue was washed twice with astrocyte growth medium and then gently dissociated by trituration with a fire-polished Pasteur pipette. Dissociated cells were filtered through a 70-μm strainer to collect a clear astrocyte cell suspension. For cytokine treatment, the astrocytes were treated for 72 hours with IL-1α (3 ng/ml, Sigma-Aldrich), TNFα (30 ng/ml, Cell Signaling Technology), and C1q (400 ng/ml, MyBioSource).

Lenti-HepaCAM-mGFP–mediated overexpression of HepaCAM in primary astrocytes

For HepaCAM overexpression in mouse primary astrocytes, the cells were seeded on a 10-cm dish and transduced with 5 μl (1.3 × 10¹¹ gc/ml) of Lenti-HepaCAM-mGFP virus. This virus was packaged by the Boston Children’s Hospital Viral Core based on a human HepaCAM mGFP-tagged HepaCAM lentiviral vector (Origene Technologies, #RC223693L4V). Three days after transduction, the medium was replaced with fresh astrocyte growth medium. The transduction efficiency was evaluated using the Keyence ALL-IN-ONE fluorescence microscope BZ-X710, which showed approximately 70% GFP-positive cells by day 10 of the culture.

Mouse primary cortical neuronal culture

Primary cortical neuron cultures were prepared from E14 to E16 mouse brain cortices. In brief, the cortices were dissected and dissociated using 0.05% trypsin-EDTA solution for 10 min at 37°C. The cells were seeded (1 ~ 2 × 10⁴ per well) on PDL and laminin (LN)–coated coverslips (GG-12-Laminin, Neuvitro) in 24-well culture dish with 1 ml of neuron plating medium containing DMEM, 10% of FBS, and 1% Pen-Strep at 37°C in a humidified chamber of 95% air and 5% CO₂. After a 12-hour seeding period, neuron plating medium was replaced by 1 ml of neuron culture medium composed of neurobasal medium (Invitrogen), 2% B27 neurobasal supplement (Thermo Fisher Scientific), 2 mM glutamine (1% of 100× GlutaMAX, Thermo Fisher Scientific), and 1% Pen-Strep (Thermo Fisher Scientific), and half volume of the medium was changed every 2 days.

Mouse primary spinal cord MN culture

Embryo spinal cords were obtained from E12 to E14 WT, HB9-GFP, or HB9-GFP⁺SOD1G93A⁺ mice. Embryo spinal cords were extracted in sterile conditions under a dissecting microscope using small forceps into cold Leibovitz’s L-15 medium (Sigma-Aldrich) with 1% Pen-Strep. Dorsal root ganglia were cut off from embryonic spinal cord and individual spinal cords were transferred into papain (2 mg/ml; Sigma-Aldrich) dissolved into 1× Hanks’ balanced salt solution (GIBCO) and digested for 20 min at 37°C. Cold L-15 medium was then added to the digested tissues and centrifuged at 280g for 10 min at 4°C. MN pellets were resuspended in 6 ml of cold L-15 medium and laid over 4 ml of Nycoprep density solution (1.06 g/ml; Accurate Chemical & Scientific Corporation) and centrifuged at 900g for 20 min at 4°C in a swinging bucket centrifuge. MNs were collected at the interface of the Nycoprep solution, washed with 10 ml of cold L-15 medium, and then concentrated by centrifuging at 425g for 10 min at 4°C. The MN pellets were gently resuspended with neuron plating medium containing DMEM, 10% of FBS (Sigma-Aldrich), and 1% Pen-Strep and plated on PDL + LN coated coverslips at 200,000 cells per well. MNs were incubated a 37°C for 12 hours, washed twice with warm neurobasal medium (Invitrogen), and cultured in 1 ml of MN medium: neurobasal medium supplemented with 2% B-27 (Thermo Fisher Scientific), 1% of 100× GlutaMAX, hydrocortisone (0.2 μg/ml; Sigma-Aldrich), insulin (2.5 μg/ml; Sigma-Aldrich), human recombinant NT3 (10 ng/ml; ProspecBio), human recombinant glial cell line–derived neurotrophic factor (GDNF, 10 ng/ml; Peprotech), human recombinant brain-derived neurotrophic factor (BDNF, 10 ng/ml; R&D Systems), and human recombinant ciliary neurotrophic factor (25 ng/ml; Novus Biologicals). Half the volume of MN medium was replaced every 2 days. To eliminate any residual glial cells, MN cultures were treated with 10 μM 5-Fluoro-2′-deoxyuridine (Sigma-Aldrich) for 24 hours after initial plating.

MN differentiation from iPSC lines

Control iPSC lines, previously characterized in Lopez-Gonzalez et al. (44), were differentiated into MNs as described in Du et al. (54, 55). Briefly, iPSCs were expanded in mTESR1 medium in Matrigel-coated wells. The next day, the culture medium was replaced with neural medium, KO DMEM/F12, and Neurobasal at 1:1, 0.5× N2, 0.5× B27, 0.1 mM ascorbic acid, and 1× Glutamax, supplemented with 3 μM CHIR99021, 2 μM DMH1, and 2 μM SB431542 to induce neuroepithelial progenitor cells. The medium was changed every other day. After 6 days, the neuroepithelial progenitors were dissociated with Accutase and split into Matrigel-coated wells. The cells were cultured in MN progenitor induction medium, above neural medium supplemented with 1 μM CHIR99021, 2 μM DMH1, 2 μM SB431542, 0.1 μM retinoic acid, and 0.5 μM purmorphamine. The medium was changed every other day for 6 days. Next, MN progenitors were dissociated with Accutase and cultured in suspension in above neural medium supplemented with 0.1 μM retinoic acid and 0.5 μM purmorphamine to induce MN differentiation. The medium was changed every other day. After 6 days, the cells were dissociated into single cells and 25 × 10³ cells were plated onto PDL/LN-coated coverslips in MN medium, above neural medium supplemented with 0.5 μM retinoic acid, 0.1 μM purmorphamine, 0.1 μM Compound E, BDNF (10 μg/ml), GDNF (10 μg/ml), and LN (1 μg/ml). Three days after plating, the MNs were treated with A-Exo. for 2 days.

Reagents and neuronal culture treatment

l-glutamic acid (Thermo Fisher Scientific) and human HepaCAM protein ECD (amino acid sequence 1 to 240) (Sino Biological Inc.) were used in this study. l-glutamic acid was freshly prepared before the treatment and added into neuronal growth medium for treatment (final concentration: 100 μM). HepaCAM ECD coating is described below. Neuronal treatment with various drugs and/or exosomes was generally at DIV 7 to 8 for 48 to 72 hours unless specifically described in the Results or figure legends.

Microfluidic chamber cortical neuron culture and axon labeling with AAV8-CAG-tdTomato (tdT)

Microfluidic chambers (model XC450) were procured from Xona Microfluidics and prepared as per the supplied guidelines with minor modifications. Briefly, the devices were coated in XC Pre-Coat solution and then washed twice with PBS. Next, the devices were coated with PDL (500 μg/ml; XONAPDL, Xona Microfluidics) solution or 1:1 volume PDL with HepaCam-ECD peptide (200 μg/ml). The microfluidic chamber was incubated at 37°C for 1 hour and then the wells were washed twice with PBS. Neurons were carefully added into the soma sides (as indicated in the diagram) at a density of 160,000 ~ 200,000 cells/20 μl in neuronal plating medium. After 5 min, 150 μl of the plating medium was added to all wells, and the cells were maintained at 37°C and 5% CO₂. After overnight incubation, all plating media were removed and 150 μl of neuronal growth medium was added to all wells. Additional media were added to the device every 1 to 2 days to account for media loss caused by evaporation. For neuronal and axonal labeling, the AAV8-CAG-tdTomato virus (1 μl, 7 × 10¹² gc/ml, #59462-AAVrg, Addgene) was added on soma sides at 5 DIV.

Time-lapse imaging and image analysis

Time-lapse images of live HB9-eGFP⁺ MNs on each coverslip were captured using the Keyence Fluorescence Microscope BZ-X700 at designated time points as indicated in Fig. 3. Images were taken at room temperature within 10 min, after which the cultured MNs were immediately returned to the incubator to continue culture at 37°C and 5% CO₂. The images were analyzed using Fiji ImageJ software with the Analyze Particles plugin, which was used to automatically detect and count HB9-eGFP⁺ MNs. Time-lapse imaging of AAV8-CAG-tdT–labeled axons within microfluidic chambers was performed using a Nikon Eclipse Ts2 inverted microscope with a 10× objective lens. Images were taken at designated time points as indicated in Fig. 7 after treatment with glutamate. Time-lapse images were captured at room temperature within 10 min, after which the microfluidic chambers were immediately returned to the incubator at 37°C and 5% CO₂.

Stereotaxic injections of mouse spinal cord with AAV5-mCherry-Gfap-Cre virus

AAV5-mCherry-Gfap-Cre was obtained from the University of North Carolina Vector Core (Chapel Hill, NC). Spinal cord ventral horn injections were performed with a Hamilton microliter syringe with gauge 33, point size 4, 45° bevel needle on a stereotaxic apparatus (Stoelting). Single dose of AAV5-mCherry-Gfap-Cre (0.5 μl, 4.3 × 10¹² gc/ml) was injected in vertebra L1 posterior median sulcus 0.3 to 0.4 mm laterally, 1.2 to 1.4 mm deep into either CD63-GFP^f/+ or CD63-GFP^f/+SOD1G93A mice. The injections were performed at a rate of 0.1 μl/min. Postoperative care included injections of buprenorphine-SR according to the IACUC requirement.

Exosome purification and ZetaView NTA analysis

Exosomes were prepared from ACM in primary astrocyte culture (initial seeding: 4 × 10⁶ cells/10-cm dish). After astrocytes became fully confluent, astrocyte growth medium was replaced with exosome-depleted astrocyte growth medium composed of DMEM, 10% exosome-depleted FBS (Thermo Fisher Scientific), and 1% Pen-Strep. ACM was replaced and collected every 3 days for up to three times (10 ml per 10-cm dish). ACM was first spun at 300g for 10 min at room temperature to remove suspension cells, 2000g for 10 min at 4°C remove cell debris and then underwent following the purification steps or was stored in −80°C until processing. For ultracentrifugation-based purification, ACM was spun at 10,000g for 60 min at 4°C. The supernatant was passed through a 0.22-μm polyether sulfone (PES) filter (Merck Millipore, MA, USA) followed by ultracentrifugation at 100,000g for 60 min at 4°C (SW 41 Ti Rotor, Beckman Coulter Inc). For SEC-based isolation, the supernatant was first concentrated (to 500 μl) by centrifugation at 3500g for 30 min at 4°C using Centricon Plus-70 Centrifugal Filter Devices 10 k (MilliporeSigma). The concentrated supernatant was passed through a 0.22-μm PES filter. The qEV original 35-nm columns (Izon Science, MA, USA) were then used according to the instructions of the manufacturer. Briefly, the column was rinsed with filtered PBS, and then 500 μl of concentrated and filtered supernatant from ACM was layered onto the top and each eluted fraction (500 μl per fraction) was collected. The eluted fractions were combined, as indicated in the Results and figure legend, and further concentrated using the Amicon Ultra-4 Centrifugal Filter Units (MilliporeSigma) in certain experiments.

Examination of exosome sample size and concentration was conducted using a ZetaView x30 Nanoparticle Tracking Analyzer (Particle Metrix, USA). Measurements were undertaken at ambient temperature (23°C) with samples diluted in freshly filtered (0.22 μm) water. The procedure encompassed two measurement cycles, each scanning across 11 distinct positions within the sample cell and recording 60 frames per position. The operational settings were standardized with a camera sensitivity of 70 and a shutter speed set to 150. Calibration of the instrument was confirmed before sample analysis by using 100-nm polystyrene beads as a reference. Data acquisition was facilitated by ZetaView Software 8.05 SP3, with exosome concentrations specifically quantified for particles ranging from 50 to 200 nm in diameter. Quantity of A-Exo. is often indicated by its protein quantity. Exosomes (1 to 1.5 μg) are typically isolated from 10 ml of ACM.

Exosome coating coverslips

Sterile PDL or PDL/LN-coated coverslips (Neuvitro) were rinsed twice with 1× PBS, and then A-Exo. (1 μg) purified from 10 ml of ACM were evenly added onto the top of the coverslips and incubated for 1 hour in 37°C cell culture incubator. The coverslips were then washed twice with 1× PBS before use.

Immunocytochemistry, immunohistochemistry, live-cell, and confocal imaging

For immunocytochemistry, cultured neurons were fixed in 4% paraformaldehyde for 15 min and permeabilized with 0.2% Triton X-100 for 5 min. The cells were blocked in 3% BSA for 30 min and incubated with the following primary antibodies overnight at 4°C: anti-ChAT (1:200, Millipore, AB144P), anti–βIII-TUBULIN (1:1000, MAB1195, R&D system), anti-RAB7 (1:100, catalog no. D9FF2, Cell Signaling Technology), anti-mouse TAU (1:500, GeneTex), anti-mouse anti-HB9 (1:50, DSHB 81.5C10), anti-Map2 (1:1000, Sigma-Aldrich, M9942), rat anti-GFAP (1:5000, zymed, 273756), and rabbit anti-GFAP (1:1000, Dako). After incubation with the primary antibodies, the neurons were washed three times with 1× PBS and then incubated with the following secondary antibodies for 1 hour at room temperature: anti-mouse Alexa Fluor 488, anti-rabbit Alexa Fluor 555, and anti-goat Alexa Fluor 633 (1:1000, Invitrogen), and mounted with Prolong Glass Antifade Mountant with NucBlue Stain (Invitrogen).

For immunohistochemistry, mice were anesthetized with a ketamine/xylazine cocktail and perfused with ice-cold PBS followed by ice-cold 4% paraformaldehyde. Dissected brains were postfixed overnight in 4% paraformaldehyde at 4°C for 24 hours and cryoprotected in 30% sucrose until the tissue sank. The tissue was embedded in OCT compound (Tissue-Tek), and 20-μm tissue sections were cut with a Leica cryostat. The following antibodies were used: GFAP (1:5000, Dako, #Z0334), hSOD1-EDI (1:300, StressMarq), misfolded SOD1 (1:500, MediMabs clone A5C3), and HepaCAM (1:500, R&D Systems, clone #419305). Primary antibodies were visualized with appropriate secondary antibodies conjugated with Alexa fluorophores (1:1000 Invitrogen) and mounted with Prolong Gold Antifade Mountant with 4′,6-diamidino-2-phenylindole (Invitrogen). Low magnification images were taken using the Zeiss Axio fluorescence microscope, and ZEN2 software was used to acquire and process images. Confocal images were taken using the Fast Lifetime CONtrast (Leica SP8 FALCON) confocal laser scanning microscope (15 to 20 μm Z stack with 0.5-μm step) magnified with 63× (numerical aperture 1.0) objectives, and images were processed with LAS X software.

TUNEL assay

TUNEL assay was performed with the EZClick TUNEL – in situ DNA Fragmentation/Apoptosis Assay Kit (Biovision) according to the manufacturer’s recommendations. NucBlue (Hoechst 33342, Invitrogen) was used as the nuclear counterstain.

Immunoblotting

Mouse spinal cord, primary astrocyte pellets, and exosome fractions were homogenized with lysis buffer [tris-HCL (pH 7.4), 20 mM; NaCl, 140 mM; EDTA, 1 mM; SDS, 0.1%; Triton X-100 1%; and glycerol 10%]. Protein inhibitor cocktail (P8340, Sigma-Aldrich) and phosphatase inhibitor cocktail 3 (P0044, Sigma-Aldrich) were added in a 1/100 dilution to lysis buffer before tissue homogenization. Total protein amount was determined by DC Protein Assay Kit II (Bio-Rad), and then the lysates were loaded on 4 to 15% Mini-PROTEAN TGX Stain-Free Protein Gels (Bio-Rad). Separated proteins were transferred onto a polyvinylidene difluoride membrane (Bio-Rad) with the Trans-Blot Turbo Transfer System (Bio-Rad). The membrane was blocked with SUPERBLOCK T20 (tris-buffered saline) blocking buffer (Thermo Fisher Scientific) and then incubated with the appropriate primary antibody overnight at 4°C. The following primary antibodies were used: CD81(1:1000, Santa Cruz Biotechnology, clone B-11), hSOD1 (1:1000, MBL, clone 1G2), hSOD1-EDI (1:300, StressMarq, #SPC-206), misfolded hSOD1 (1:500, MediMabs, clone A5C3), mouse CD63 (1:1000, Abcam, clone EPR21151), TurboGFP (1:2000, Evrogen, #AB513), β-actin (1:1000, Sigma-Aldrich, #A1978), and HepaCAM (1:500, Proteintech, #18177-1-AP). Secondary antibodies, including enhanced chemiluminescence (ECL) anti-mouse IgG (1:10000, GE Health care NA931V), anti-rabbit IgG–horseradish peroxidase (HRP) (1:5000, GE Health care NA934V), mouse anti-goat IgG-HRP (1:1000, Santa Cruz Biotechnology), and anti-rat IgG-HRP (1:5000, Thermo Fisher Scientific, 62-9520) are diluted with Super Blocking Buffer. Bands were visualized using the Chemidoc MP imaging system (Bio-Rad) with ECL Plus chemiluminescent substrate (Thermo Fisher Scientific) or Clarity Max Western ECL Substrate (Bio-Rad).

LC-MS/MS proteomics and data analysis

Exosome samples (20 μg per sample, three biological replicates per group) were separated on 4 to 15% mini-protein TGX precast protein gels (Bio-Rad) and subsequently stained with Coomassie Blue staining. Each sample lane was excised, digested with trypsin, and spiked with 0.2 pmol of ADH peptides (yeast alcohol dehydrogenase 1) as internal standard at the Mass Spectrometry Facility in University of Massachusetts Medical School. The samples were then injected into Orbitrap Fusion Lumos Mass Spectrometer (Thermo Fisher Scientific) in technical triplicates for label-free quantitation analysis. The data were searched against Swiss-Prot mouse protein database using Mascot search engine through Proteome Discoverer software. The data were exported in normalized as intensity-based absolute quantification (iBAQ) quantitative values in Scaffold software (version Scaffold 5.1.0, Proteome Software, OR, USA). iBAQ value (1 ×10⁴) was empirically set as the detection limit. Proteins with lower than 1 × 10⁴ iBAQ value in all samples were deemed not detected. Proteins were identified using the following cutoffs: protein threshold, 95%; number of peptides, 3. The resulting data were exported to Microsoft Excel for further analysis.

Only proteins detected in all three biological replicates (n = 278 proteins) were analyzed. Raw iBAQ values within each batch were normalized to yeast ADH1 protein iBAQ values, which was used as the reference protein to normalize batch effect. Normalized iBAQ values were expressed as fold change compared to the NTg (or SOD1G93A) no-treatment condition. Proteins with consistent up- or down-regulation between NTg versus ITC-NTg and SOD1G93A versus ITC-SOD1G93A across three batches were included in the analysis. The fold change values of individual proteins were then averaged across three batches. Proteins with an average fold change greater (or smaller) than 1.5 were considered significant for IPA pathway and function analysis. Differentially expressed proteins between NTg versus ITC-NTg and SOD1G93A versus ITC-SOD1G93A were further compared to identify the co–up-regulated and co–down-regulated proteins (n = 108 proteins). This list of proteins was then analyzed using QIAGEN IPA software (QIAGEN Inc.) to determine protein cellular localizations and associated pathways.

Immuno-EM imaging

EM imaging was performed in the Harvard Medical School Electron Microscopy Facility. For exosome samples from SEC columns, negative staining was performed. Briefly, 5 μl of the sample was adsorbed to a hydrophilic carbon–coated grid. Rat anti-mouse CD63 (1:20 BioLegend, #143901) was used as the primary antibody on the grid, and the grid was then incubated with rabbit anti-rat bridging antibody (1:50, Abcam, ab6703) and Protein A-gold 10 nm (University Medical Center Utrecht, the Netherlands). Excess liquid was removed with a filter paper (Whatman #1), and the samples were stained with 1% uranyl acetate. The grids were examined in a JEOL 1200EX transmission electron microscope, and images were recorded with an AMT 2 k CCD camera.

Imaging analysis

For colocalization analysis, colocalization of hCD63-GFP⁺ puncta with misfolded SOD1 were determined using Fiji ImageJ. Using the maximum intensity projection image, the hCD63-GFP channel image was first thresholded to create a binary black and white image. Then, the Particle Analyzer tool was used to count all hCD63-GFP⁺ puncta, measure their Feret diameter, and generate regions of interest (ROIs) of all the hCD63-GFP⁺ puncta. The misfolded SOD1 channel image was then thresholded (threshold was set on the basis of a negative control image), and the ROIs of hCD63-GFP⁺ puncta were overlaid on the misfolded SOD1 image. Misfolded SOD1 staining was then measured inside of hCD63-GFP–based ROIs. hCD63-GFP⁺ puncta with any misfolded SOD1 staining inside them were considered as hCD63-GFP⁺SOD1⁺ puncta, which was divided by the total number of hCD63-GFP⁺ puncta to obtain the colocalization ratio. The size distribution of CD63-GFP⁺SOD1⁺ puncta was also analyzed on the basis of the Feret’s diameter of each puncta. The colocalization of hCD63-GFP⁺ puncta with HepaCAM were measured in a similar way in Fiji ImageJ using the Particle Analyzer tool with a thresholded maximum projection image of the CD63-GFP channel to count, measure, and create ROIs of all CD63-GFP⁺ puncta. Any puncta with diameter >2000 nm were excluded, the remaining ROIs of the CD63-GFP⁺ puncta were then overlaid on the HepaCAM channel image, and the integrated density of HepaCAM staining intensity was measured inside CD63-GFP⁺ puncta.

The neurite beading index was used to quantify neurite damages based on βIII-TUBULIN immunofluorescence in 20× images. All images from a given experiment were processed concurrently via ImageJ by using its particle analysis feature. Parameters for identifying neurite beading were set as follows: particle size threshold between 0.3 and 25 μm² and the circularity index constrained to 0.8 to 1.0. Particles exceeding 25 μm² in size or having a circularity index below 0.8 were considered as non-degenerating neurites. For each image, total particles were counted and then divided by βIII-TUBULIN⁺ area.

Statistical analysis

All statistical analyses were performed using GraphPad Prism 8. Group differences in each assay at each time point were analyzed by Student’s t test (two-group comparison), one-way analysis of variance (ANOVA) (three or more group comparison, one independent variable), or two-way ANOVA (three or more group comparison, two independent variables). Statistical test(s) used are specified in figure legends. When present, error bars are presented as SEM. No custom code was used in the analysis. Statistical significance was tested at a 95% (P < 0.05) confidence level, and P values are shown in each graph.

Supplementary Materials

The PDF file includes:

Figs. S1 to S7

Legends for Tables S1 to S9

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S9