Abstract
Altered synaptic function is thought to play a role in many neurodegenerative diseases, but little is known about the underlying mechanisms for synaptic dysfunction. The squid giant synapse (SGS) is a classical model for studying synaptic electrophysiology and ultrastructure, as well as molecular mechanisms of neurotransmission. Here, we conduct a multidisciplinary study of synaptic actions of misfolded human G85R-SOD1 causing familial amyotrophic lateral sclerosis (ALS). G85R-SOD1, but not WT-SOD1, inhibited synaptic transmission, altered presynaptic ultrastructure, and reduced both the size of the readily releasable pool (RRP) of synaptic vesicles and mobility from the reserved pool (RP) to the RRP. Unexpectedly, intermittent high-frequency stimulation (iHFS) blocked inhibitory effects of G85R-SOD1 on synaptic transmission, suggesting aberrant Ca2+ signaling may underlie G85R-SOD1 toxicity. Ratiometric Ca2+ imaging showed significantly increased presynaptic Ca2+ induced by G85R-SOD1 that preceded synaptic dysfunction. Chelating Ca2+ using EGTA prevented synaptic inhibition by G85R-SOD1, confirming the role of aberrant Ca2+ in mediating G85R-SOD1 toxicity. These results extended earlier findings in mammalian motor neurons and advanced our understanding by providing possible molecular mechanisms and therapeutic targets for synaptic dysfunctions in ALS as well as a unique model for further studies.
Significance Statement
The squid giant synapse (SGS) presents one of the few mature nervous systems in situ that mimics mammalian neuromuscular junctions, while allowing precise experimental manipulations and live measurement with superior spatial and temporal resolution. Applying these unique features to studying the molecular mechanisms of amyotrophic lateral sclerosis (ALS), a devastating adult-onset neurodegenerative disease without cure, offers clues to understand the pathogenesis of the disease. Our results demonstrating synaptic dysfunction caused by ALS-associated mutant SOD1 protein and its underlying molecular pathways may suggest a novel approach to an effective therapeutic intervention as well as identify biomarkers for early diagnosis. Furthermore, the altered synaptic vesicle behavior and Ca2+ dynamics revealed through the perturbation of neurotransmission by ALS extends our understanding of fundamental synaptic physiology at both molecular and cellular levels.
Introduction
Amyotrophic lateral sclerosis (ALS) is a fatal adult-onset neuromuscular disease with dysfunction and loss in both upper and lower motor neurons, resulting in progressive muscle weakness and atrophy, eventual paralysis, and death usually within three to five years after initial diagnosis. Inherited forms of ALS represent roughly 10% of cases, with similar clinical symptoms to sporadic ALS patients. Unfortunately, there is still no effective cure for ALS. To enable therapeutic intervention at an early stage of the disease, the mechanisms underlying ALS and presymptomatic biomarkers for early diagnosis must be identified.
Mutations in Cu/Zn superoxide dismutase (SOD1) are responsible for ∼25% of familial ALS cases, but SOD1 is expressed in many cell types where it removes superoxide radicals and is a component of redox signaling pathways. ALS-associated mutations induce a conformational change within SOD1 due to protein misfolding (Bruijn et al., 1998) and aberrant interactions with other proteins (Pasinelli et al., 2004; Urushitani et al., 2006). Genetic knock-out of SOD1 does not produce motor deficits seen in ALS, indicating that SOD1 mutations confer a toxic gain-of-function (Bunton-Stasyshyn et al., 2015; Taylor et al., 2016; Ghasemi and Brown, 2018). A glycine-to-arginine substitution at residue 85 (G85R) is one highly studied ALS-associated mutation: G85R-SOD1 is incapable of dimerizing normally and instead misfolds into soluble monomers or oligomers, as well as insoluble aggregates that are toxic to motor neurons (Wang et al., 2009a,b).
To characterize G85R-SOD1 toxicity specific to neurons, various forms of misfolded G85R-SOD1-YFP were perfused into isolated axoplasm from squid giant axons (Song et al., 2013), and both soluble monomers and oligomers inhibited axonal transport from cell body to synapse through aberrant activation of signaling pathways (Song et al., 2013). Although G85R-SOD1 inhibition of anterograde transport of synaptic components could lead to synaptic dysfunction, questions remained as to whether G85R-SOD1 has a direct effect on synaptic transmission. Impairment of neuromuscular junctions is detected early in patients and animal models (Moloney et al., 2014), consistent with a “dying-back” neuropathy where synaptic dysfunction and loss is a primary pathology eventually leading to neuronal death. Decreased synaptic vesicle numbers and reduced presynaptic proteins in presynaptic terminals of several ALS animal models further suggest a presynaptic effect (Wang et al., 2009a). These changes are seen before motor deficits are detectable, suggesting that synapses could be a primary target of G85R-SOD1 (Frey et al., 2000; Fischer et al., 2004; Hegedus et al., 2007; Sakowski et al., 2012).
Synaptic transmission is altered in ALS mouse models and induced Pluripotent Stem Cells (iPSCs)-derived motor neurons, raising key questions: how do ALS-associated proteins alter synaptic functions? Can synaptic function be restored? Unfortunately, the complexity of mouse motor systems, chronic nature of disease pathology, small size of presynaptic domains, and difficulty of targeting synapses limit utility of mouse and cell culture models in addressing questions about presynaptic mechanisms.
The squid giant synapse (SGS), which mimics neuromuscular junctions with similar ion channel compositions and neurotransmission machinery (Katz and Miledi, 1967, 1977), was used to evaluate presynaptic and postsynaptic actions of G85R-SOD1 here. The large size of squid giant presynaptic terminals permits experimental manipulations such as protein microinjection, direct electrophysiological measurements of transmission, current and voltage clamping, and Ca2+ imaging, facilitating detailed studies of presynaptic molecular mechanisms (Bloedel et al., 1966; Llinas et al., 1980, 1994; Augustine and Charlton, 1986; Llinás et al., 1985; Augustine et al., 1988, 2006; Augustine, 1990; Lin et al., 1990; Fukuda et al., 1995; Sugimori et al., 1998). The squid synapse is particularly suitable for studying ALS, because it is (1) fast-transmitting with high release capacity, similar to synapses affected in ALS; and (2) enriched in ion channels (e.g., Ca2+), neurotransmitters (e.g., glutamate) and receptors (e.g., AMPAR) implicated in ALS pathology.
Here, we microinjected G85R-SOD1 in presynaptic terminals, recorded both presynaptic and postsynaptic membrane potentials with and without high-frequency stimulations (HFSs), performed live ratiometric Ca2+ imaging, and examined synaptic vesicle morphology/distribution by electron microscopy (EM). Correlated functional and structural studies demonstrated the acute synaptic inhibition by G85R-SOD1 as evidenced by diminished postsynaptic potentials (PSPs), decreased synaptic vesicle number at active zones (AZs), and reduced synaptic mobilization from reserved (RP) to readily releasable pool (RRP) within 30 min after G85R-SOD1 injection. Unexpectedly, while continuous trains of HFS depleted synaptic vesicle availability and enhanced synaptic inhibition by G85R-SOD1, intermittent HFS (iHFS) applied every 30 min before and during G85R-SOD1 infusion kept synapses firing normally for 7 h. Given the effects of HFS on presynaptic Ca2+, we tested for aberrant Ca2+ signaling by ratiometric Ca2+ imaging with presynaptic fura-2. Ca2+ increased throughout terminals including in “palm” regions before presynaptic axons branch, which lack plasmalemmal Ca2+ channels. Reducing intracellular Ca2+ by EGTA rescued synaptic function and restored normal structure in the presence of G85R-SOD1 as did iHFS. These results are consistent with and extend findings in mammalian nervous systems where neurons with lower Ca2+ buffering capacity are selectively affected in ALS and these neurons showed Ca2+ imbalances presymptomatically (Grosskreutz et al., 2010). Understanding sources of increased Ca2+ and molecular mechanisms underlying Ca2+ activation as well as how iHFS rescues synaptic function will provide unique insights into synaptic dynamics and a possible molecular basis for synaptic pathology in human ALS patients.
Materials and Methods
Proteins and reagents
G85R-SOD1-His, WT-SOD1-His, G85R-SOD1-His-YFP, and WT-SOD1-His-YFP were produced in E. coli BL21/DE3 cells under pET vectors, and purified on Talon resin first and then by chromatography on MonoQ 10/10, in the Horwich Laboratory as described previously (Song et al., 2013). All chemicals used were American Chemical Society quality or better, from Sigma and Invitrogen (now Fisher Scientific). For microinjection, proteins and reagents were dissolved in 200 mM KCl, 100 mM taurine, 250 mM K-isethionate, and 50 mM K-HEPES, pH 7.4). Tetramethylrhodamine-dextran (3 kDa, Invitrogen) was used for co-injection with non-fluorescent reagents to visualize and monitor the injection.
SGS preparation and electrophysiology setup
The giant squid synapse is formed between the terminal finger of second order axons of the pre-nerve and the third order axon of the last stellate nerve, which is the giant axon used for axonal transport studies as well as for classical voltage-clamp studies of ion channels (Young and Keynes, 2005). Following a standard protocol (Llinas et al., 1980, 1981, 1994; Llinás et al., 1985; Augustine and Charlton, 1986; Augustine et al., 1988, 2006; Augustine, 1990; Smith et al., 1993), stellate ganglia of small female Loligo pealeii (RRID: SCR_002864, Marine Biological Laboratory) were removed from the mantle carefully and rapidly under running seawater, tied off at each end of the presynaptic and postsynaptic axons, isolated from the sheath and connective tissues, pinned with fine cactus needles to a thin-layer of Sylgard on the bottom of a 35-mm Petri dish chamber, superfused continuously with oxygenated squid saline (455 mM NaCl, 54 mM MgCl2, 11 mM CaCl2, 10 mM KCl, 3 mM NaHCO3, and 10 mM HEPES, pH 7.2) at 10–15°C.
The most distal digit of the presynaptic axons and the most medial fiber, also known as the giant axon, form the giant synapse (Fig. 1A). One microelectrode for injecting was inserted in the presynaptic axon to inject current at 0.033 Hz for basal stimulation and at 50 Hz for HFS (each pulse is 2 μA for 2 ms for basal stimulation and 2 μA for 1 ms for HFS), near the palm where the second order axon enters the ganglion to branch and form multiple synapses with postsynaptic axons. At the terminal of the most medial branch, a second microelectrode was inserted to presynaptically inject proteins and reagents of interest as well as recording the presynaptic membrane potentials. Finally, at the postsynaptic terminal of the giant synapse, a third microelectrode was inserted near the medial presynaptic digit to record the postsynaptic membrane potentials. Electrodes 1 and 3 were filled with 3 M KCl while the second microelectrode allowed the injection of 50 μM SOD1 proteins or reagents of interest in 100 mM KCl (Smith et al., 1993) at 0.1 Hz (each injection was 50 psi for 250 ms). The injection efficiency of SOD1 proteins was monitored by a fluorescence microscope. All synapses were injected with the same pressure for the same amount of time during data acquisition, and the final mean fluorescence intensities in the presynaptic terminals were measured and compared between samples to ensure comparable levels of infused proteins.
Extended Data Figure 1-1.
No effects of fluorescent dextran on synaptic transmission. Synapses infused with rhodamine-dextran alone kept firing for >2 h, without significant changes in synaptic strength as evidenced by constant postsynaptic membrane potential waveform (A) and EPSP slope (B). Download Figure 1-1, TIF file.
Extended Data Figure 1-2
Recovery of synaptic transmission after the removal of G85R-SOD1-YFP. The synapse was injected with G85R-SOD1-YFP continuously for 30 min until the significant reduction in EPSP slope and the failure to elicit a postsynaptic action potential. As the injection stopped, G85R-SOD1-YFP slowly diffused away from the presynaptic terminal to the axon, leading to a full recovery of EPSP after 150 min, suggesting that synaptic machinery remained intact after treatment. After the synapse and neurotransmission were stabilized for more than 1 h, G85R-SOD1-YFP was injected again and a similar inhibitory effect was observed. Download Figure 1-2, TIF file.
Electrophysiology data acquisition and analysis
Both presynaptic potential and PSP were recorded using sharp microelectrodes with an Axoclamp-2A amplifier (Axon Instrument) and data analyzed using Labview (National Instruments) software (Yulong Li). Raw waveforms were used to calculate PSP slope using the same parameters for all experiments. For data acquired under HFS, PSP slopes were analyzed, integrated, plotted, and fitted linearly using the last 50 time points to derive the vesicle mobilization rate (slope of the linear fit) and the RRP size (intersection with the y-axis).
Ratiometric Ca2+ imaging
We dissolved 1 mM fura-2 (pentapotassium salt, ThermoFisher F1200, CHEBI:52 081), a Ca2+ indicator, in 100 mM KCl and infused it into the presynaptic terminal by electrophoresis (100 nA current) through the first presynaptic electrode inserted in the palm. Fura-2 infusion continued for 10 min, and the synapse was allowed to equilibrate for 30 min before presynaptic injection of SOD1 proteins through the second microelectrode to ensure fura-2 was diffused evenly throughout the whole presynaptic terminal at roughly 100 μM. Ratiometric live Ca2+ imaging was performed every 30 s with ratiometric images taken at Ex360 nm and Ex390 nm assisted by an ultra-high-speed wavelength switching λ DG-4/5 xenon arc lamp system. Ratios at various locations (A) were calculated as (fluorescent intensity360nm)/(fluorescent intensity390nm) as a measure of intracellular Ca2+ concentration (Smith et al., 1993). In vitro calibration was performed by mixing squid axoplasm with equal volumes of buffer containing 5 μM fura-2, 400 mM KCl, and 40 mM Na-HEPES (pH 7.2). Three doses of CaCl2 were set by adding 20 mM EGTA (low Ca2+, <10−8 M), or 20 mM EGTA + 13.3 mM CaCl2 (intermediate Ca2+, ∼6.7 × 10−7 M), or 20 mm CaCl2 (high Ca2+). Various ratios under each condition were measured using the same optics and the background noise was subtracted. The dose response curve was obtained to allow the determination of Kd′ and the two ratios. Rmax was defined as the ratio value at saturating Ca2+ level, and Rmin was defined as the ratio value at the minimal Ca2+ concentration.
EM
After recordings, synapses were fixed in 4% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer for 12 h at 4°C, then washed in 0.1 M cacodylate buffer three times and postfixed in 1% osmium tetroxide for 2 h at 4°C, followed by blocking impregnation with 2% uranyl acetate in 0.1 M sodium acetate, pH 5.0 for 24 h. After washing, dehydration was performed in ethanol with increasing concentration: 50% for 10 min, 75% for 10 min, 80% for 10 min, 85% for 10 min, 95% for 10 min, and 100% for 5 min twice, followed by propylene oxide for 5 min three times. Resin and propylene (1:1 for 24 h, 2:1 overnight, and 3:1 for 2–4 h, and finally 100% resin overnight) infiltration was conducted before embedding in silicone molds in the oven at 62°C for 72 h. Sectioning and imaging were performed following standard EM protocols (Augustine et al., 2006; Moreno et al., 2009). Vesicle density at the active zones (AZs) was determined as the number of vesicles per square micrometer. Distance between vesicles and AZs was measured and plotted as described previously (Morgan et al., 1999). All EM samples were numbered, processed, and analyzed under blinded conditions until the last step when data had to be pooled.
Statistical information
All experiments were repeated at least five times. The data were analyzed by one-way ANOVA followed by the Tukey post hoc test (or nonparametric multiple t tests, without assuming consistent SD) and plotted in Prism 7 (GraphPad software). Quantitative data were plotted as mean ± SEM, p values were calculated and four statistical thresholds were marked: p < 0.00,001, p < 0.0001, p < 0.001, and p < 0.005 (p < 0.05 indicated statistical significance). For EM analysis, nested one-way ANOVA was performed to compare WT and G85R-infused synapses (102 synapses across three synapses for each group).
Results
G85R-SOD1-YFP but not WT-SOD1-YFP inhibits synaptic transmission
To assess whether the mutant SOD1 is directly toxic to the synapse, as opposed to indirectly affecting it, e.g., by blocking axonal anterograde vesicular trafficking (Song et al., 2013), G85R-SOD1-YFP (n = 8) or WT-SOD1-YFP control (n = 5) was microinjected in the presynaptic terminal site (Fig. 1A) through the micropipette extending in from lower right; note that black color indicates the extent of YFP fluorescence after 40 min of infusion. Another microelectrode was used to pass current pulses for basal stimulation (2 μA × 2 ms, every 33 s; Fig. 1A, micropipette extending in from lower left) to evoke single action potentials, detected by intracellular voltage recording simultaneously from both presynaptic and postsynaptic sites (Fig. 1A for sites of recording and Fig. 1B, middle and bottom traces). This protocol is referred to hereafter as “basal stimulation.” Measurement of voltage changes at both presynaptic and postsynaptic sites allows for assessment of the efficiency of synaptic transmission. Strikingly, after 40 min of G85R-SOD1-YFP infusion, the postsynaptic action potential was no longer detected, with only a subthreshold PSP observable (Fig. 1B, Vpost). By comparison, there was no change in the PSP with a WT-SOD1-YFP infusion (Fig. 1B, control) or with fluorescent dextran alone (Extended Data Fig. 1-1). A time course study of WT-SOD1-YFP and G85R-SOD1-YFP was conducted to observe the kinetic behavior of EPSPs from basal stimulation during the respective infusions (Fig. 1C; note that points are shown only for every 10th pulse, i.e., at 5-min intervals). With infusion of the mutant protein, the EPSP slope began to reduce by ∼10 min, and was completely lost by ∼60–70 min. By contrast, no effect on EPSP was observed with WT-SOD1-YFP infusion for >100 min (Fig. 1C). Thus, in contrast with the absence of an observable effect of WT-SOD1 infusion on postsynaptic EPSP, injection of monomeric G85R-SOD1-YFP produced a substantial time-dependent inhibition of synaptic transmission.
To rule out the possibility that synaptic defects were caused by physical damage induced by electrode impalement and microinjection, injection was stopped when postsynaptic action potentials were abolished. This allowed the diffusion of G85R-SOD1-YFP away from the synaptic terminal through the presynaptic axon. The EPSP slope started to increase 90 min later as the presynaptic fluorescent intensity of G85R-SOD1-YFP significantly decreased and EPSPs were fully restored in ∼2.5 h after most mutant SOD1 had diffused out of the terminal (Extended Data Fig. 1-2). Normal synaptic transmission continued for >1 h until a second injection of G85R-SOD1-YFP significantly inhibited synaptic transmission in the same fashion as the first injection. These results suggested that G85R-SOD1-YFP inhibits synaptic transmission at least in part by directly affecting neurotransmitter release from the synaptic terminal and this effect was reversible on the removal of G85R-SOD1-YFP from the synaptic terminal.
G85R-SOD1-YFP impairs synaptic vesicle availability at the AZ
To examine whether G85R-SOD1-YFP inhibited synaptic transmission by affecting synaptic vesicle pool dynamics, we microinjected G85R-SOD1-YFP or WT-SOD1-YFP for 15 min and then applied trains of HFS presynaptically (50 Hz for 5 s, with 5 s intervals between trains) and measured postsynaptic depression of EPSPs, which are proportional to the underlying EPSCs and are uncontaminated by action potentials, to determine the size of the RRP and the rate of vesicle mobilization from the reserved pool (RP). EPSP slopes during both the first train of a series trains of HFS (Fig. 2A) and the sixth train (Extended Data Fig. 2-1) were compared in WT and G85R synapses. For controls, there was a two-phase diminution of EPSP slope across the 250 spikes, whereas with G85R synapses the first phase was barely detectable during the sixth train of HFS and was significantly inhibited during the first train. The first phase is generally associated with utilization of the RRP, which appears to be smaller with the mutant, whereas the second phase is generally associated with utilization of the RP, which appears to be present in both with significant reduction by G85R-SOD1-YFP. To further measure these pools, the EPSP slope values were continuously integrated across one train of the same synapse before infusion and one train after 30 min of WT- or G85R-SOD1-YFP infusion. The EPSPs of each condition were integrated and averaged to generate traces (Fig. 2B). Notably, the WT-SOD1 trace resembled that of the preinfusion trace (not shown), indicating that WT-SOD1 has little effect on synaptic vesicle availability and mobilization, whereas G85R-SOD1 trace exhibited integrated values that were substantially lower than those of WT-SOD1. The relative size of the RRP was determined for each condition by intersection of slope with the y-axis, and the mobilization rate from the RP was directly determined from the slopes (dotted lines; linear fit of the integrated EPSP slopes from the last 50 time points). The values and statistics for independent experiments are shown (n = 11 for WT and n = 6 for G85R; Fig. 2C). G85R-SOD1-YFP drastically reduced both the RRP size and the vesicle mobilization rate from RP to RRP as compared with WT-SOD1-YFP. Therefore, reduced RRP size and slower RP mobilization rate, could lead to the loss of synaptic transmission.
Extended Data Figure 2-1
Significant inhibition of EPSP by continuous HFS in G85R-SOD1-YFP-infused synapse. EPSP slopes taken during the 6th train in a set of continuous trains of HFS showed even more dramatic decreases by G85R-SOD1-YFP, compared with the WT-SOD1-YFP, particularly at the beginning, suggesting limited synaptic vesicle availability, consistent with the morphological changes found in EM. Note that in panel A, the EPSP slope with G85R-SOD1-YFP in the 1st train was initially >150, while in the 6th train, it never got above 40. In contrast, the initial EPSP slopes in the 1st train and the 6th train were almost identical after perfusion with WT-SOD1-YFP. Download Figure 2-1, TIF file.
EM of WT-SOD1-YFP and G85R-SOD1-YFP-infused synapses
To visualize the synaptic vesicles directly, we conducted EM analyses of synapses infused for 40 min under basal stimulation with the respective proteins (n = 3 for each group). The WT-SOD1-YFP-infused synapses exhibited normal morphology with abundant packing of synaptic vesicles against the presynaptic membrane, comparable to published EM data (Augustine et al., 2006) of normal SGS (Fig. 3A, red * indicates an AZ), whereas the G85R-SOD1-YFP-infused synapses reproducibly exhibited synaptic vesicles that did not abut the membrane at AZs, lying in a zone distant from it (Fig. 3B). The vesicles present in this zone also appeared generally larger, less compact in morphology and reduced in number relative to WT-SOD1 synapses. We also consistently observed additional structures, e.g., in this image, what appears to be a pre-autophagosome (labeled by blue #) and two adjoining prelysosomes (labeled by purple ^). By contrast we did not observe such structures in WT-SOD1-infused presynaptic AZs. Notably, many of the pre-autophagosomes appeared to contain vesicles. This potentially reflects a local mechanism to remove G85R-SOD1-YFP-affected synaptic vesicles in the RRP and a subpopulation of vesicles in the RP, whereas, interestingly, the most remote RP pools appear to be spared, suggesting a potential source for synaptic vesicle supply, which may help restoring synaptic function if released on proper stimulations. More detailed characterization of affected vesicles was conducted by counting the number of electron lucent vesicles, clathrin-coated vesicles, and large electron lucent vesicles per AZ as described previously (Morgan et al., 1999; Fig. 3C–F). Nested ANOVAs suggested that while clathrin-coated vesicles did not seem to be affected by G85R-SOD1, the total number of available synaptic vesicles was dramatically reduced. A cumulative measurement of AZs (102 across three biological repeats per group) infused with WT-SOD1 or G85R-SOD1 (binned by 50 μm in distance) showed that such reductions seem to be evenly distributed in the RRP regardless of the distance from the membrane (Fig. 3G–I). Because of the limited number of vesicles further away from the AZs (>675 μm) in both groups, the increase in the relative vesicle distribution in that vicinity may not be statistically different between the G85R-SOD1-infused synapses and the WT-SOD1-infused synapses. This is further confirmed by the cumulative vesicle distribution curves (without binning) in WT and G85R-SOD1 synapses (Fig. 3I).
Prevention of G85R-SOD1-YFP-associated synaptic deficits by iHFS
Unexpectedly, when a single train of HFS (50 Hz for 5 s) was applied presynaptically every 30 min, G85R-SOD1-YFP-infused synapses, otherwise receiving continuous basal stimulation, maintained constant EPSPs for over 8 h (Fig. 4A–C). This contrasts with the finding (Fig. 1) that in the setting of G85R-SOD1-YFP infusion, continuous basal stimulation leads to steady decline and then complete loss of EPSPs within 40–60 min. This also contrasts with the finding (Fig. 2) that continuous HFS trains with 5-s intervals also inhibited synaptic transmission by depletion of the RRP. Correlating with maintenance of healthy physiology, synaptic vesicle dynamics appeared to be normal, as demonstrated by presence of the first phase of EPSP slope during a single 50-Hz train, reflecting presence of an RRP resembling wild type (Fig. 4D). Additionally, the integrated EPSP slope averaged across multiple 50-Hz trains now resembled that of WT-SOD1-YFP synapses (n = 6 for each; Fig. 4E). Finally, in EM analyses (Fig. 4F), consistent with presence of a normal RRP, vesicles were observed abutting the AZ after iHFS was applied to a synapse infused with G85R-SOD1-YFP for 6 h. These results indicate that iHFS is able to preserve the RRP in the face of G85R-SOD1-YFP infusion. Interestingly, in G85R-SOD1-injected synapses (three out of five) with dramatic decreases in the slope of the PSP but remaining >50, application of one HFS train restored EPSP slope and the RRP size, without rescuing vesicle mobilization rate (Extended Data Fig. 4-1). However, in the other two synapses where PSP slopes dropped to below 50, adding one HFS further reduced them to almost 0. This suggests that to achieve the rescuing effects, the timing of the HFS application with regard to existing synaptic strength is important. However, the underlying molecular mechanism of this rescuing effect is unclear and requires further study. Altogether, the intermittent stimulation reverses toxicity of the misfolded protein by restoring synaptic vesicle availability at the AZ.
Extended Data Figure 4-1
Rescuing effects of HFS in dying synapses. In three out of five synapses where the EPSP slopes were significant inhibited by G85R-SOD1-YFP but were still above 50, one train of HFS surprisingly restored neurotransmission as indicated by the rescue of EPSP slope. Synaptic membrane potentials were measured in presynaptic and postsynaptic terminals before G85R-SOD1 infusion (A), after inhibition (B), and after one train of HFS (C). HFS partially restored the RRP size without rescuing vesicle mobilization rate in these synapses. In the other two synapses where EPSP slopes had dropped below 50, a train of HFS further decreased the EPSP slope to almost 0, consistent with its role in depleting RRP. Download Figure 4-1, TIF file.
G85R-SOD1 induces Ca2+ increase in the presynaptic terminal
Intrigued by the rescuing effect of iHFS in G85R-SOD1-infused synapses, we examined potential mechanisms. Since iHFS is known to alter Ca2+ dynamics, we wondered if aberrant Ca2+ signaling may underlie G85R-SOD1-YFP toxicity. To test this possibility, we performed ratiometric Ca2+ imaging using fura-2, a fluorescent Ca2+ indicator dye, which was delivered to the presynaptic axon electrophoretically and allowed to diffuse into the palm and presynaptic branches [Fig. 5A, regions of interest (ROIs) were selected to show 1: palm, 2–6: PreG, presynaptic axon which forms the giant synapse, and 7: PreS, another small presynaptic branch]. In parallel, the presynaptic axon was subjected to basal stimulation while both presynaptic and postsynaptic membrane potentials were recorded. The presence of fura-2 provides an indication of changes in Ca2+ levels and location during stimulation. To avoid interference with Ca2+ imaging, WT-SOD1 and G85R-SOD1 constructs used in these studies lacked the YFP tag and injections were monitored by co-perfusion of tetramethylrhodamine-dextran. While overall characteristics of synaptic function are similar with the introduction of fura-2, the impairment of synaptic transmission by G85R-SOD1 was delayed in the presence of fura-2 (>60 min, previously <30 min after G85R-SOD1 injection), probably due to the buffering of Ca2+ by indicator dye. This impairment was correlated with a concurrent rise in Ca2+ levels at 60 min. Synapses injected with fura-2 alone or tetramethylrhodamine-dextran alone kept constant EPSP normally for >2 h as those injected with WT. Furthermore, synapses injected with WT-SOD1 behaved similarly to those with WT-SOD1-YFP, excluding the potential role of YFP tag on synaptic transmission.
Extended Data Figure 5-1
Increases in Ca2+ levels under HFS. Synapses infused with WT-SOD1 (A) of G85R-SOD1 (B) exhibited increased Ca2+ levels upon HFS, mainly in the presynaptic terminals, with comparable changes at the protein infusion site and the rest of the terminal (PreG). The palm, which lacks Ca2+ channels did not show changes in Ca2+ influx, consistent with previous findings. G85R-SOD1 did not seem to alter the overall pattern, except for larger fluctuations particularly at the palm and the protein infusion site, which may indicate changes in Ca2+ homeostasis independent of Ca2+ channels. Download Figure 5-1, TIF file.
G85R-SOD1 infusion (n = 19) caused increases of Ca2+ throughout the presynaptic terminal compared with WT controls (n = 12), as derived from the ratios of (fluorescent intensity360nm)/(fluorescent intensity390nm) averaged across ROIs 2–6 (Fig. 5B). While WT-SOD1 under basal stimulation showed no change in Ca2+ concentration after 1 h protein infusion (Fig. 5C), G85R-SOD1 increased global Ca2+ concentration in the palm, the presynaptic giant axon (PreG), as well the other small presynaptic branch (PreS), with the most significant elevation at the site of G85R-SOD1 injection (Fig. 5D) perhaps due to the highest G85R-SOD1 concentration at the injection site. Both the palm and PreS regions, further away from the injection site, would contain lower concentrations of mutant protein due to diffusion. Interestingly, the palm which lacks plasma membrane Ca2+ channels and does not normally show Ca2+ increases during HFS did not exhibit increased Ca2+ with WT-SOD1 (Extended Data Fig. 5-1A), consistent with earlier findings (Smith et al., 1993). However, in the presence of G85R-SOD1, the palm also displayed aberrant Ca2+ increases with HFS (Extended Data Fig. 5-1B), suggesting that sources of Ca2+ other than plasma membrane Ca2+ channels may contribute to this Ca2+ misregulation. Taken together, the evidence suggests that G85R-SOD1-induced Ca2+ influx may mediate its inhibitory effects on synaptic transmission and the effect may depend on intracellular Ca2+ stores.
Given that iHFS rescued G85R-SOD1-associated synaptic defects, we wondered if iHFS could correct the aberrant localization of Ca2+ increases. Fura-2 ratiometric Ca2+ imaging demonstrated Ca2+ increases during HFS in both WT- and G85R-SOD1-injected synapses (Extended Data Fig. 5-1), confirming previous findings of HFS on Ca2+ dynamics by others (Smith et al., 1993). In the presence of G85R-SOD1, Ca2+ concentrations in the palm appeared more variable across the synapses, with some palm regions showing Ca2+ increases as high as those in the synaptic terminal, suggesting high baseline Ca2+ concentration. However, there seemed to be no significant difference between WT-SOD1-injected (n = 12) and G85R-SOD1-injected (n = 19) synapses at the injection site and the other locations within the presynaptic terminals during iHFS, indicating normal Ca2+ channel functions at the injection site, thereby excluding the possibility of physical membrane damage caused by the injection. Consistent with iHFS rescue in synaptic physiology and in contrast to G85R-SOD1-induced Ca2+ influx in the absence of iHFS, aberrant Ca2+ activation was not observed after iHFS in G85R-SOD1-injected synapses (Fig. 5E,F). These results suggest that iHFS may reset Ca2+ dynamics in the terminal and prevent the local Ca2+ overloads that contribute to synaptic dysfunction induced by G85R-SOD1.
EGTA prevents toxic effects of G85R-SOD1-YFP on synaptic physiology and morphology
The iHFS rescue experiments in combination with Ca2+ imaging and the delayed inhibition produced by fura-2 suggested that aberrant Ca2+ activation by G85R-SOD1 may be the cause of synaptic inhibition. To further test this hypothesis, we chelated free Ca2+ in the presynaptic terminal by co-injecting 50 mM EGTA (final concentration at 5–10 mM in the synapse). Previous studies had shown that EGTA levels <80 mM do not affect synaptic transmission (Adler et al., 1991). EGTA was sufficient to block the inhibitory effects of G85R-SOD1-YFP on synaptic transmission and G85R-SOD1-YFP-injected synapses maintained steady EPSPs for over 2 h without iHFS (n = 6), which was comparable to WT-SOD1-YFP-injected synapses (Fig. 6A; Extended Data Fig. 6-1; five WT-SOD1-infused and eight G85R-SOD1-infused synapses were also shown in Fig. 1C). Not surprisingly, ratiometric Ca2+ imaging using fura-2 did not show any local Ca2+ increases in synapses co-injected with G85R-SOD1 and EGTA (Fig. 6B,C). Defects in synaptic vesicle pools, number and distribution induced by G85R-SOD1 were also corrected by EGTA, as evidenced by the usual RRP size, and normal mobilization rate (Fig. 6D,E). Furthermore, EM showed comparable numbers of synaptic vesicles at the AZs (Fig. 6F). The numbers of docked and electron lucent synaptic vesicles were similar between the WT-SOD1 and G85R-SOD1+EGTA-infused synapses, while EGTA might slightly increase clathrin coating (Fig. 6G; 102 synapses for each group). These results suggest that Ca2+ dysregulation may underlie G85R-SOD1-associated synaptic dysfunctions and that limiting increases in free cytoplasmic Ca2+ with EGTA may restore normal synaptic morphology and function.
Extended Data Figure 6-1
Unnormalized EPSP slopes. EPSP slopes were measured and plotted without normalization for synapses infused with WT-SOD1-YFP (A), G85R-SOD1-YFP (B), and G85R-SOD1-YFP + EGTA (C). Download Figure 6-1, TIF file.
Discussion
The discovery of giant fibers in the squid nervous system by J. Z. Young (Young and Keynes, 2005) >80 years ago led to several milestone discoveries in modern neuroscience, including the Nobel-prize-winning work on ionic mechanisms for initiating and propagating action potentials by Andrew Huxley and Alan Hodgkin (Schwiening, 2012). Although advanced imaging and genetic tools as well as new model systems have been developed in recent years, these classical mature synapse and axon models continue to be used to understand fundamental neurobiology and provide a unique model to study molecular mechanisms of human neurologic diseases. For example, the original discovery of the molecular motor kinesin in isolated axoplasm from the squid led to studies of axonal trafficking deficits resulting from altered motor function in several neurodegenerative diseases (Song et al., 2016; Brady and Morfini, 2017). In a similar way, studies of molecular mechanisms underlying basic synaptic physiology in the SGS have led to exploration of disease mechanisms involving synaptic dysfunction (Moreno et al., 2009). Here, we use the giant axo-axonic synapse in the squid stellate ganglion to study synaptic actions of mutant human SOD1 (G85R-SOD1) protein, which causes a familial form of ALS.
Presynaptic injection of G85R SOD1 inhibits neurotransmission by disturbing synaptic vesicle physiology
Synaptic loss at affected neuromuscular junctions leading to a dying-back neuropathy appears to be a primary defect associated with pathology in ALS patients and mouse models bearing G85R-SOD1 mutations. Consistent with these observations, we found that infusion of ALS-associated human G85R mutant SOD1 protein but not the WT-SOD1 into the presynaptic terminal of the SGS significantly inhibited EPSPs, leading to a failure in eliciting postsynaptic action potentials. This occurred within 30 min of continuous infusion into the presynaptic terminal without diminishing the presynaptic action potentials, while infusion of the WT-SOD1 had no effect on synaptic transmission as evidenced by sustained firing of synapses with constant EPSPs (Fig. 1).
High-frequency stimulus protocols designed to deplete the RRP of presynaptic vesicles showed that vesicle dynamics were significantly inhibited by G85R-SOD1, reducing both the size of the RRP and the mobilization rate from the RP to the RRP (Fig. 2). This finding was further validated by EM data that demonstrated significant reductions of synaptic vesicles at AZs in the presence of G85R-SOD1 protein (Fig. 3). Ultrastructural studies provided additional evidence of altered vesicle trafficking as membranous structures resembling multivesicular bodies and autophagosomes were identified in presynaptic regions infused with G85R-SOD1, but not WT-SOD1. Inhibition of synaptic transmission was reversed by stopping presynaptic infusion of G85R-SOD1 and allowing sufficient time for the protein to diffuse out of the synapse via the presynaptic giant axon (Extended Data Fig. 1-2), which suggested that the synaptic machinery was not physically and permanently damaged.
Nature of synaptic vesicle disturbance
A clue to the mechanism for G85R-SOD1-induced synaptic inhibition came from an experiment where the synapse was exposed to iHFS at 30-min intervals before and during G85R-SOD1 infusion. This was expected to inhibit synaptic transmission based on earlier results showing that continuous HFS depleted presynaptic RRP. Unexpectedly, G85R-SOD1 synapses exposed to iHFS maintained constant EPSPs for over 7 h, exhibiting normal synaptic vesicle dynamics and normal numbers of synaptic vesicles at AZs despite the presence of pathogenic SOD1 protein (Fig. 4). This finding led us to examine Ca2+ dynamics as HFS is known to redistribute Ca2+ both temporally and spatially (Smith et al., 1993).
Live cell ratiometric Ca2+ imaging studies showed that G85R-SOD1 consistently increased Ca2+ levels in presynaptic terminals with the largest increases at the injection site, the site with highest G85R-SOD1 concentration. Surprisingly, Ca2+ levels were also increased in the palm region, which lacks plasmalemmal Ca2+ channels and normally does not exhibit Ca2+ increases in response to HFS (Fig. 5). These data suggested that the source of Ca2+ may not be restricted to or dependent on local plasma membrane Ca2+ channels. This agrees with previous findings that Ca2+ influx solely through voltage-gated Ca2+ channels is not able to induce motor neuron death (Van den Bosch et al., 2002). Instead, release of Ca2+ from endoplasmic reticulum and/or mitochondrial stores is required, and these organelles may release Ca2+ in response to misfolded G85R-SOD1 protein.
Co-injection of EGTA and application of iHFS prevent physiopathology
If increased free Ca2+ levels are responsible for synaptic failure, co-injection of EGTA, a Ca2+ chelator, at levels that do not affect synaptic transmission at the SGS should prevent synaptic inhibition by G85R-SOD1 by buffering free Ca2+. Consistent with the conclusion that presynaptic increases in free Ca2+ levels were responsible for the observed synaptic dysfunction, EGTA prevented inhibition of synaptic transmission when co-injected into the presynaptic terminal with G85R-SOD1, as documented by (1) EPSPs remained constant for over 2 h; (2) sufficient numbers of synaptic vesicle were present at the AZs with normal mobilization from RP to RRP; and (3) cytoplasmic free Ca2+ levels did not increase (Fig. 6). Similar effects were also achieved by iHFS applied every 30 min for over 7 h (Fig. 4). Interestingly, iHFS also abolished the Ca2+ increases induced by G85R-SOD1 (Fig. 5) perhaps by redistributing Ca2+ and restoring homeostasis. This process may alter regulation of other organelles such as ER, mitochondria, and autophagosomes in the synapse to restore overall synaptic health. However, determining the exact molecular mechanisms requires more in-depth studies.
Questions remain as to how increased cytoplasmic free Ca2+ alters synaptic vesicle dynamics in the presynaptic terminal. Ca2+ is a key player in many cellular and subcellular processes, is required for regular synaptic function, and is normally tightly regulated. In normal synapses, free Ca2+ increases are largely restricted to the AZ through local plasmalemmal channels during synaptic transmission (Sudhof, 2004). The levels of EGTA used in these studies do not affect Ca2+ influx at AZs but do buffer cytoplasmic free Ca2+. There are multiple Ca2+ sensors with a wide range of affinities in the presynaptic compartment, all of which may contribute to one or more of the complex steps involved in synaptic transmission and synaptic vesicle dynamics. While the role of some of these sensors is established, such as evidence that synaptotagmin is the Ca2+ sensor for synaptic vesicle fusion and neurotransmitter fast release, the physiological roles for many others remain unclear (Sudhof, 2004).
The existing literature has suggested a variety of potential mechanisms for Ca2+ related pathology in both human and mouse ALS-vulnerable neurons (Grosskreutz et al., 2010), such as the subgroups of motor neurons in spinal cord and brainstem expressing low levels of Ca2+ buffering proteins, even at the presymptomatic stages. Similarly, Ca2+ dysregulation has also been observed in several other neurodegenerative diseases and may lead to unfolded protein responses (UPRs) and ER stress-related signaling activation (Grosskreutz et al., 2010; Walker and Atkin, 2011; Hetz and Saxena, 2017; Remondelli and Renna, 2017), Ca2+ permeable AMPA receptor dysregulation (Tateno et al., 2004; Tortarolo et al., 2006; Van Den Bosch et al., 2006), ER/mitochondria membrane collapse (Tortarolo et al., 2006; Watanabe et al., 2016; Bernard-Marissal et al., 2018), and aberrant Ca2+ influx in affected neurons and glia. In addition, changes in protein expression in affected motor neurons (e.g., Ca2+ channels and regulators) may also contribute to dysregulation of Ca2+ levels in disease. While altered transcription may play a role in the slow progression of the disease, this is unlikely to contribute here given the time scale and reversibility of these effects as well as the lack of strong evidence of local transcription in the SGS.
In conclusion, our results in the SGS indicate that mutant SOD1 affects the regulation of free Ca2+ levels in the presynaptic terminal, leading to a disruption of normal vesicle trafficking. These effects were not restricted to regions near the AZs but were more extensive and affected regions of the presynaptic compartment that lack plasmalemmal Ca2+ channels. The effects of mutant SOD1 may occur without affecting the activity of local plasmalemmal Ca2+ channels required for neurotransmitter release at the AZ, given that (1) presynaptic regions lacking Ca2+ channels also showed increased levels of free Ca2+, (2) low levels of EGTA can reduce free Ca2+ levels in the presence of G85R-SOD1 without affecting Ca2+ influx through plasmalemmal Ca2+ channels responsible for normal synaptic transmission, and (3) the discovery that iHFS could rescue synaptic transmission. The effect of iHFS was unexpected and further argues for normal function of AZ Ca2+ channels in the plasmalemma independent of intracellular stores of Ca2+. The effect of iHFS suggests that further study is needed of how patterns of activity can preserve synaptic function and this may be a basis for novel therapeutic interventions in ALS.
The effects observed here may relate to those observed earlier in the squid giant axon innervating muscles in the mantle. There we found that G85R-SOD1-YFP inhibited axonal trafficking of presynaptic vesicles by activating a MAPK stress pathway, which impaired kinesin-based transport responsible for anterograde cargo transport (Song et al., 2013). In addition, we observed that G85R-SOD1 was associated with several synaptic proteins critical for synaptic transmission, such as synapsin and syntaxin-binding protein 5, at a much higher level than WT-SOD1-YFP. HSP110, an HSC70 co-chaperone which prevented the neurotoxicity of G85R-SOD1 in the axon by potentially sequestering or refolding G85R-SOD1, also abolished the association of Synapse-associated protein and syntaxin-binding protein 5 with G85R-SOD1-YFP (Song et al., 2013). The roles that G85R-SOD1 interactions with synaptic proteins may play in affecting synaptic vesicle dynamics as well as regulation of these interactions via Ca2+ dependent or independent pathways remain to be determined. Various neuropathogenic changes have been proposed to be a consequence of Ca2+ dysregulation, such as UPR and autophagy, in ALS and several other neurodegenerative diseases. Our EM data (Fig. 6) showed that EGTA, when co-injected with G85R-SDO1, restored synaptic morphology and prevented the formation of pre-autophagosome and adjoining pre-lysosomes, which were observed when presynaptic G85R-SOD1 was injected alone.
Implication of acute toxic effects of G85R-SOD1 on SGS physiology as relates to longer term ALS pathology in mammalian synapses
Like many other adult-onset neurodegenerative diseases, one big puzzle in the field is “why does it take 40–50 years to develop ALS?” Several disease-related processes occur during aging which can be overcome in young neurons due to robust compensatory mechanisms. In this study, we focus on acute effects of pathogenic SOD1, but a number of age-related factors may explain the delayed onset of clinical symptoms.
Unlike the acute injection of pathogenic SOD1 in this study where we observed the presence of pre-autophagosome and pre-lysosome formation within the time frame of both electrophysiological recordings and morphologic studies, ALS-associated mutant SOD1 in patients misfolds and accumulates gradually to pathologic levels over time, allowing adaptive changes in cellular and molecular mechanisms underlying synaptic transmission. This is due to, at least partially, mechanisms for actively removing pathogenic proteins in early stages of disease, which may be sufficient to slow the accumulation of toxic materials. However, these clearance mechanisms may eventually be unable to keep up with misfolded protein accumulations. By directly injecting defined levels of mutant SOD1 into an isolated SGS system, we ensured accumulation of misfolded protein locally in the synapses within a short period of time which limits compensatory effects in the isolated synapse with little protein degradation and synthesis in this time frame.
Similar to robust protein homeostatic pathways during development and maturation, Ca2+ buffering mechanisms and synaptic transmission may be redundant early on in the complex system such as the mammalian models where neuronal circuits (including the other cell types) are intact and dynamic. They can adapt to slowly increased stresses created by the presence of mutant SOD1, much as synaptic functions exhibit adaptive changes during normal aging. The use of isolated acute preparations of giant synapse also eliminates the possibility that transcriptional changes of synaptic proteins can compensate for the effects of toxic proteins on synaptic transmission machinery. For example, the question of whether ALS-affected neurons show hyper- or hypo- excitability has been debated for some time. As a matter of fact, several groups have reported hyperexcitability of motor neurons at early stages followed by hypoexcitability at later stages of ALS in both iPSC culture models and in mouse models. These changes may be due to various compensatory mechanisms at the molecular, cellular and systemic levels which also may contribute to the delayed onset of symptoms. Here, we can isolate inhibitory effects due to the simplicity of the squid synapse model system where many other factors, such as compensation from other neurons, changes in gene expression or axonal transport, or the impact of a complex environment can no longer play a role.
Furthermore, it is worth keeping in mind that synaptic loss, as a result of gradual decrease in synaptic strength as demonstrated here, occurs long before cell death and can be observed in presymptomatic stages of both ALS mice and patients. By measuring EPSP slopes and vesicle dynamics in the SGS, we have provided a sensitive readout of neurotransmission in real time and revealed changes in synaptic strength which are more subtle in pre-symptomatic patients and mouse models but are detectable long before clinically relevant behavioral abnormalities appear. Correlating Ca2+ dynamics and HFS effects with synaptic health in real time illuminates the kinetic changes in the presynaptic terminal with high temporal and spatial resolution, which suggests the importance of identifying potential therapeutic time windows. The acute effects demonstrated here may illustrate molecular mechanisms underlying disease pathology that are manifested over time in patients.
Relevant to mammalian physiology and pathology, the SGS also shares fundamental synaptic machineries with neuromuscular junctions in human, making it a unique ALS-disease model with a potential for answering important but otherwise difficult questions about pathogenic mechanisms and synaptic transmission in motor neuron diseases. The SGS provides a unique platform for (1) performing live measurements of both presynaptic and postsynaptic activities in response to disease-associated proteins with temporal and spatial resolutions surpassing those in most other model systems; (2) conducting morphologic studies (e.g. EM) in the same preparations characterized by electrophysiological studies and live-imaging of cellular and molecular pathways; (3) evaluating pharmacological inhibitors, biochemical tools, and electrophysiological protocols that may rescue disease phenotypes in a time dependent manner; and (4) identifying translatable mechanistic findings, due to the highly conserved synaptic machinery, that will provide better understanding and treatment of multiple neurodegenerative diseases where misfolded proteins cause disease pathology.
In conclusion, the development of novel models for study of neurodegenerative diseases such as those using the squid nervous system has the potential to provide unique insights into disease mechanisms. For example, the combination of our electrophysiological studies in the giant synapse with studies of axonal transport in isolated axoplasm from the postsynaptic giant axon of the squid, allowed us to identify signaling pathways altered by these pathogenic proteins (Song et al., 2013). These results were then developed further in mouse models of ALS and patient tissues or iPSCs. Our results here are also consistent with previous findings of Ca2+ dysregulation and synaptic dysfunction in both ALS mice and patient derived neurons (Grosskreutz et al., 2010; Tradewell et al., 2011; Leal and Gomes, 2015). For example, affected neurons exhibit aberrant Ca2+ dynamics, which leads to hyperexcitability or hypoexcitability at various stages (Wainger et al., 2014; Devlin et al., 2015; Martínez-Silva et al., 2018). Further, oculomotor neurons, which are relatively spared in ALS, have a higher buffering capacity for Ca2+ (Vanselow and Keller, 2000; Obál et al., 2006), which was thought to contribute to their overall resistance to pathology in patients and mouse models. Those data were unclear, however, as to whether these changes involved plasmalemmal Ca2+ channels or intracellular Ca2+ stores or both, an issue that could be addressed in the SGS.
Given recent work on the squid genome and improved mariculture protocols that have the potential to allow systematic long-term studies of squid nervous systems, there will be opportunities to develop novel transgenic ALS squid models. Therefore, studies using the SGS will continue to complement studies using human iPSCs and animal models of ALS as we seek to understand how multiple disease related pathways can independently affect the pleiotropic neuronal functions that interact to produce the phenotype of ALS, a multifactorial disease with a complex disease pathology.
Acknowledgments
Acknowledgements: The author thanks Dr. Arthur Horwich, Dr. George Augustine, Dr. Scott Brady, Dr. Rodolfo Llinas, and Dr. Wayne Fenton for their valuable suggestions and critical evaluation of the experiments and the manuscript. SOD1 proteins were gifts from the Horwich laboratory, prepared by Dr. Weiming Ni and Dr. Wayne Fenton. The author also received help from Dr. Jorge E. Moreira and Dr. Pri Comassio with EM data acquisition and analysis. Finally, the author would like to sincerely thank all the 2014 Grass Fellows, Dr. Alberto Pereda (director), Dr. Matthew Kittelberger (associate director), Dr. Felix Schweizer (president), the Marine Resources Center (MRC) at the Marine Biological Laboratory, and all Grass Trustees, without whose support and encouragement the project wouldn't have been possible.
Footnotes
The authors declare no competing financial interests.
This work was supported by the Grass Foundation, the Howard Hughes Medical Institute, the Massachusetts General Hospital Jack Satter Foundation, the Harvard University ALS and Alzheimer’s Endowed Research Fund, and Harvard Brain Science Initiative.
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.