Early interneuron dysfunction in ALS: Insights from a mutant sod1 zebrafish model

Abstract

Objectives

To determine, when, how and which neurons initiate the onset of pathophysiology in ALS using a transgenic mutant sod1 zebrafish model and identify neuroprotective drugs.

Methods

Proteinopathies like ALS involve mutant proteins that misfold and activate the heatshock stress response (HSR). The HSR is indicative of neuronal stress and we use a fluorescent hsp70-DsRed reporter in our transgenic zebrafish to track neuronal stress and to measure functional changes in neurons and muscle over the course of the disease.

Results

We show that mutant sod1 fish first exhibit the HSR in glycinergic interneurons at 24 hours post fertilization (hpf). By 96 hpf, we observe a significant reduction in spontaneous glycinergic currents induced in spinal motor neurons. The loss of inhibition is followed by increased stress in the motor neurons of symptomatic adults and concurrent morphological changes at the neuromuscular junction (NMJ) indicative of denervation. Riluzole, the only approved ALS drug and apomorphine, an NRF2 activator, reduce the observed early neuronal stress response.

Interpretation

The earliest event in the pathophysiology of ALS in the mutant sod1 zebrafish model involves neuronal stress in inhibitory interneurons, resulting from mutant Sod1 expression. This is followed by a reduction in inhibitory input to motor neurons. The loss of inhibitory input may contribute to the later development of neuronal stress in motor neurons and concurrent inability to maintain the NMJ. Riluzole, the approved drug for use in ALS, modulates neuronal stress in interneurons, indicating a novel mechanism of riluzole action.

Introduction

Neurodegenerative diseases including amyotrophic lateral sclerosis (ALS) are characterized by the presence of protein inclusions in the affected neurons. Emerging data indicate that protein misfolding may be of mechanistic importance in these diseases¹. Mutations in the ubiquitously expressed superoxide dismutase (SOD1) gene account for 20% of cases of the familial form of ALS. Over 150 mutations in the SOD1 gene have been discovered, including the point mutations G93R and G85R². Recent studies also implicate SOD1 in the sporadic form of ALS and suggest a prion like propagation of misfolded SOD1^3–5. Interestingly, some of the newly identified genes implicated in ALS, such as TARDBP and FUS are also proteins that show a high propensity to misfold and prion like activity⁶. However, we still do not know the precise mechanism by which mutant proteins cause toxicity^{5, 7}. The emerging consensus view is that multiple interacting pathophysiological factors, including protein misfolding, contribute to the neuronal toxicity in ALS.^{8, 9}

Despite progress in revealing multiple molecular processes involved in disease pathology, relatively little is known about when and how the disease, which starts focally, spreads throughout the motor network^10–12. Interestingly even in the subtypes of ALS caused by SOD1 mutations, there is considerable phenotypic heterogeneity. Ravits and Spada, 2009¹² hypothesised that despite disease heterogeneity, the disease poses common themes that may involve common mechanisms. They propose that ALS may in fact be an orderly, actively propagating process and that fundamental molecular mechanisms may be uniform.

The zebrafish is emerging as a useful tool for studying neurological diseases relevant to humans. Previously, we had shown that mutant sod1 transgenic fish show the hallmarks of adult onset neurodegenerative ALS, including defective motor performance, motor neuron loss, a loss of neuromuscular connectivity and muscle atrophy¹³. The aforementioned observations demonstrate the usefulness of the zebrafish as a model for this disease.

However, one of the current limitations when working with in vivo models of ALS is the lack of a good readout for the pre-symptomatic course of the disease. The zebrafish offer great advantages in studying early disease processes as they develop rapidly, reaching postembryonic life at around 3 days post-fertilization, which is developmentally similar to the neonatal mouse (for a comparison of developmental stages in human, mouse and zebrafish, see Table 1). Moreover, the embryonic and larval zebrafish spinal cord is functionally and anatomically similar to that of humans, yet it is also optically transparent and experimentally accessible, making it ideal for the study of spinal circuits in normal and pathophysiological conditions¹⁴.

Table-1. Comparison of neural developmental stages in human, mice and zebrafish.

Spontaneous body movement indicative of neural activity occurs soon after neural tube closure in mammals and development of the neural keel in zebrafish. 24hpf (embryo), 72hpf-9dpf (larvae), 6months-1 year (adult) are the time points examined in this manuscript.

	Embryo		Fetus
	Neural groove	Neural tube	Spontaneous limb movement	Free living
Human	23 dpf	4 wpf	9–10 wpf	40 wpf
Mouse	8.5 dpf	9.5 dpf	12 dpf	21 dpf
	Embryo			Larvae
	Neural thickening	Neural keel
Zebrafish	10.3 hpf	11.6–16 hpf	17–24 hpf	56–72 hpf

dpf: days post fertilization; hpf: hours post fertilization; wpf: weeks post fertilization; dpp: days post partum. Larval zebrafish: 72hpf-30dpf; juvenile zebrafish: 30 dpf-3months; breeding adult: 3 months-2 years.

In the current study we monitored in vivo, early neurological changes caused by mutant sod1 gene. The sod1 zebrafish ALS model harbours a fluorescent heatshock stress response (HSR), reporter gene (hsp70-DsRed). The HSR is an endogenous cellular pathway that attempts to refold the damaged proteins in stressed cells, although this response is not always sufficient or beneficial¹⁵. Thus, the HSR mediated DsRed fluorescence in the sod1 zebrafish model of ALS represents a useful tool for monitoring perturbations in cellular homeostasis caused by sod1 mutation. This facilitates the mapping of disease focality and spread through the CNS, by the spatiotemporal readout of the neuronal stress response in the spinal cord of mutant zebrafish and provides an understanding of the cells and networks involved in disease propagation in ALS.

We present evidence that the HSR is an indicator of early pathogenic processes occurring in neurons. The HSR is first observed at embryonic stages, in discrete populations of inhibitory interneurons in the spinal cord and is followed by dysregulation of glycine release from these inhibitory interneurons. Further, we observe that following interneuron dysfunction, motor neurons start exhibiting neuronal stress. More interestingly, we show that motor neurons showing the HSR also show dysfunctional neuromuscular junctions (NMJs). Taken together, our observations suggest that the mutant Sod1 induced HSR is a robust predictor of neuronal dysfunction and thus is a reliable marker of disease pathogenesis. Finally, we also show that the neuronal stress readout can be used to identify neuroprotective compounds such as riluzole and identify biological targets that may ameliorate early pathophysiological disease processes that are currently not well explored. Although the sod1 zebrafish model may by itself not be sufficient in developing new therapies for ALS, this model system would provide a rapid way to triage compounds for screening in higher vertebrate models, with the potential for more rapid identification of promising compounds for translation into human clinical trials.,

Materials and methods

Details given in the supplementary methods section

Generation of transgenic zebrafish

The zebrafish sod1 transgenic fish lines were created according to protocols described previously by Ramesh et al.¹³. For all the transgenic strains, a suffix Sh was added (to imply Sheffield strain). The transgenic lines utilized for this study included the Tg(sod1: sod1^WT;hsp70:DsRed)os4-Sh4, the line expressing the highest level of WTSod1 (3.3x as compared to non-transgenics), referred to as WTos4-Sh4 line; Tg(sod1: sod1^G93R;hsp70:DsRed)os10-Sh1 referred to as G93Ros10-Sh1 (High expressor with Sod1 expression increased 3X and comparable to WTos4-Sh4), Tg(sod1: sod1^G93R;hsp70:DsRed)os10-Sh2 referred to as G93Ros6-Sh2 (Moderate expressor with Sod1 expression increased 2.5X) and; Tg(sod1: sod1^G85R;hsp70:DsRed) os6-Sh3 line, referred to as G85Ros6-Sh3 (Low expressor with Sod1 expression increased 1.5X). When both G93R and G85R lines are discussed they are referred to as MUTsod1 lines.

Electrophysiology

Whole cell voltage clamp recordings were conducted in 4 dpf larvae as previously described ¹⁶. The fish were perfused with Evans physiological saline containing the neuromuscular blocker D-tubocurarine (10µM), the sodium channel blocker tetrodotoxin (TTX, to synaptically isolate neurons), kynurenic acid (2.5mM, to block spontaneous glutamatergic currents) and bicuculline (25µM, to block spontaneous GABAergic currents). Cells were voltage clamped at -75 mV, a potential at which the chloride-conducting glycine receptors generate inward currents. Sulforhodamine (0.1%) was included in the electrode solution to visually identify the cell type. The frequency of glycinergic miniature postsynaptic currents (mPSC) were determined by averaging the number of events in a 300 second period. To examine rise time, decay and amplitude of mPSCs, the first 50 mPSCs were selected from each recording and averaged across each experimental condition.

Fluorescent RNA in situ hybridisation

Fluorescent RNA in situ hybridisation was performed on 24 hpf G93Ros10 embryos, as previously described¹⁷. A mix of equal concentrations of the probes (glyt2a, glyt2b and DsRed), (gad65,67 and DsRed) or (vglut2 and DsRed) was used as previously described¹⁸. Quantification of cells was done by counting DsRed/glycine, DsRed/GABA and DsRed/vglut2 positive cells in the mid-trunk region and the percentage of DsRed cells showing glycine, GABA or glutamate staining, as well as the percentage of each cell type showing DsRed staining was calculated from an average of 10 embryos (a total of 429, 505 and 634 DsRed cells were counted respectively for each riboprobe pair).

Immunofluorescence

Immunofluorescence, was performed as described by Ramesh et al ¹³. Quantitative analysis of confocal images was performed on image stacks of 16–20 µm thickness (0.5-1µm /section).

Drug testing

The G93Ros10-Sh4 line was used to identify drugs that inhibit neuronal stress. 24 hpf transgenic embryos (25 embryos/treatment) were incubated with sterile embryo media with the optimised concentration of the test compound and changed daily and maintained for 5 days. At 5 dpf, the lysates from transgenic or non-transgenic embryos were obtained by sonication (Misonix Sonicator 4000) on ice followed by centrifugation. 50 μL of supernatant was analyzed in a 96-well plate (Corning-3880, Corning Life Sciences) at DsRed wavelength.

Image analysis

Image analysis for NMJ analysis was performed using NIH ImageJ software and quantitative analysis of the NMJ morphology was performed using a colocalization analysis plugin.^{19, 20}

Statistical analysis

Statistics were performed using GraphPad Prism 5. Unpaired t-tests or ANOVA with post-hoc Bonferroni testing were used to compare groups. Electrophysiological data were compared using the two way Kolmogorov–Smirnov test.

Results

Induction of neuronal stress in transgenic sod1 zebrafish

When developing the sod1 transgenic zebrafish, we hypothesized that misfolding of the Sod1 protein in vulnerable cell populations would cause cellular stress and activate the HSR allowing identification of potentially dysfunctional neurons. The HSR responsive, hsp70-DsRed construct that is inserted adjacent to the sod1 gene is driven by the hsp70 minimal promoter (1.5Kb fragment), and is not driven by the adjacent sod1 gene. The basal hsp70 promoter will allow induction of the DsRed reporter only in the presence of heat shock or cellular stress. In the absence of stress or heat shock, the promoter is silent and no DsRed is synthesized. This method of reporter expression is commonly used in zebrafish research and multiple lines within our lab using other transgenes (eg SMN) made with this linked expression construct behave similarly²¹.

We generated multiple G93R (High expressor: G93Ros10-Sh1, Moderate expressor: G93Ros6-Sh2), G85R (Low expressor: G85Ros6-Sh3) and WT (High expressor: WTos4-Sh4) sod1 transgenic zebrafish carrying the hsp70-DsRed stress reporter gene. As expected the sod1 transgenic fish showed red fluorescence throughout the body and also in the spinal cord upon heat shock (Figure 1A, +HS). Interestingly, we found that in the absence of heat shock the embryos expressing Sod1 exhibited specific DsRed expression in the CNS indicating the presence of neuronal stress (Figure-1A,-HS). The high (G93Ros10-Sh1) and moderate expressor (G93Ros6-Sh2) mutant sod1 lines showed DsRed induction in distinct neuronal groups in the dorsal spinal cord (Figure-1B, top and middle left panel). The highest expressor WTos4-Sh4 line showed a minimal level of DsRed induction, despite having high transgene copy numbers and 3 fold increased Sod1protein expression; levels that are identical to the G93Ros10-Sh1 mutant line (Figure-1B, bottom left panel, Figure-1C). Most importantly, endogenous hsp70 was also up-regulated in cells that expressed DsRed (Figure-1B, top row, middle panel), evident in the neurons that show high DsRed expression levels. The moderate expressor mutant and the high expressor wild type lines did not show strong DsRed or detectable hsp70 induction (Figure-1B, middle and bottom row, middle panel). DsRed is a very stable protein and this stability greatly amplifies the hsp70 induction signal. This was shown by a strong DsRed signal persisting for several days after heatshocking (Supplementary figure1 E,F) in the transgenic fish, while endogenous hsp70 which is tightly regulated ²² returned to background levels at 15 hours post-heatshock(Supplementary Figure 1A,B Vs Figure 1C,D).

Figure 1. sod1 transgenic fish induce the heat shock response (HSR) without exposure to heat stress.

A) Live G93Ros10-Sh1 showing induction of hsp70 measured by DsRed fluorescence at 7 dpf in whole embryos (Top panel) and 30 hpf spinal cord (bottom panel). Larvae were heat shocked (+HS) or left unexposed to heat shock (−HS) and images of HSR induction were compared. When exposed to heat, the larvae show global induction of the HSR (Left panel), while in the absence of heatshock only neuronal, neuroepithelial and occasionally muscles show induction of the HSR (Right panel). B) Multiple mutant lines G93Ros10-Sh1, G93Ros6-Sh2 show HSR induction in the absence of heatshocking. Confocal image of spinal neurons showing induction of endogenous hsp70 (middle column) in high expressor G93Ros10 line (top panel) in the same cells that show strong DsRed expression (Top row, Left Column). Moderate expressor G93Ros6-Sh2 line (middle row) and high expressor WTos4-Sh4 line (bottom row) do not show elevated hsp70 levels above background. C) Quantitation of the DsRed fluorescence in individual neurons in the zebrafish embryonic spinal cord (30 hpf) by average fluorescence intensity. Average fluorescence of individual DsRed positive neurons was measured and analysed by ANOVA (*p < 0.05 for G93Ros6-Sh2 and WTos4-Sh4, and ***p<0.0001 for the G93R os10-Sh1 and WTos4-Sh4). Size bar=10µM

Stressed neurons in the embryonic spinal cord are predominantly inhibitory glycinergic interneurons

The zebrafish spinal cord is composed of neurons and glia²³. The neurons include a variety of interneurons and motor neurons whilst the glial population consist of oligodendrocytes that myelinate axons and radial glial cells (similar to astrocytes) that provide support to neurons²³. In order to determine the spatiotemporal onset of the HSR in the spinal cord, we used in situ hybridization and antibody staining to determine the identity of DsRed expressing cells in mutant embryos and larvae. Interestingly, DsRed expression was never observed in oligodendrocytes, radial glia or in motor neurons (olig2 or hb9 positive early differentiating and ChAT positive mature) in the mutant sod1 larvae (Supplementary Figure-2).

Thus, we reasoned that the first spinal neurons to activate HSR were likely to be interneurons. We next sought to determine which interneuron population exhibit the HSR. The zebrafish spinal cord has been characterized in detail and the different spinal interneurons have stereotyped neuroanatomy, with predictable anatomical positions and axonal trajectories²⁴. Moreover, each class expresses one of only three neurotransmitters: glycine, GABA and glutamate²⁴. We therefore used a combination of anatomical and transmitter expression characteristics to identify the cell types in which the HSR was present.

To determine if the stressed neurons of sod1 mutant embryos were glycinergic we used dual color fluorescence in situ hybridization with riboprobes targeted against glycine transporter (glyt2a, b) and DsRed RNA. We observed numerous cells in which glyt2a,b and DsRed co-localized (Figure-2, top panel). Similarly, DsRed expression often co-localized with cells positive for anti-glycine antibodies (Supplementary figure-3B). Quantitation of the percentage of glycinergic interneurons which showed the neuronal stress response revealed that almost half of the glycinergic interneurons (49.23±12.8 percent) showed DsRed expression at 24 hpf. The percentage of DsRed positive neurons that were glycinergic was 44.2 ± 9.5 percent.

Figure 2. Embryonic mutant sod1 zebrafish show induction of neuronal stress predominantly in the spinal inhibitory glycinergic interneurons.

In situ hybridization with probes for inhibitory and excitatory neurons (left panels) and DsRed riboprobes (middle panel). (Top Panel) glyt2 positive glycinergic inhibitory interneurons; (Middle panel) gad65,67 positive GABAergic inhibitory interneurons (DoLA and some glycinergic interneurons); (Bottom panel): vglut2 positive excitatory interneurons (CoPA and CoSA ascending interneurons) that cross modulate the spinal locomotor circuitry. Time: 24 hpf.

The majority of DsRed positive cells that co-expressed glycine had dorsally located perikarya and axons that projected rostrally through the dorso-medial aspect of the spinal cord towards the brain. Such axonal projections are observed in Commissural Secondary Ascending-glycinergic (CoSA-gly) (mammalian V0 like), Commissural Bifurcating Longitudinal (CoBL) (mammalian dl6 like) and Circumferential Ascending (CiA) (mammalian V1 like) which comprise the three glycinergic interneurons of the zebrafish spinal cord at this stage of development.^24,25

Of these, the CoBL and CoSA interneurons are pax2 positive^18,26,27. We therefore used anti-pax2 antibodies to label CoBL and CoSA interneurons. However, as our DsRed and Pax2 antibodies were both rabbit derived, we were limited to monitoring pax2 colocalization within strongly DsRed positive cells unenhanced by immunolabeling. Multiple fields with interneurons that showed strong DsRed expression showed colocalization with pax2 antibody staining (Supplementary Figure-3A).

To determine whether GABAergic interneurons expressed the HSR, we performed in situ hybridization with riboprobes targeted against GABA biosynthetic enzymes gad65 and gad67. We found that 14.2 ± 6.8 percent of DsRed positive neurons had a GABAergic transmitter phenotype (Figure-2, middle panel) whilst the percentage of GABA positive neurons that showed HSR was 8.18±4.95 %.

We next asked whether the HSR occurred in the glutamatergic interneurons of the mutant embryonic spinal cord. Investigation of the expression of the glutamatergic excitatory neurotransmitter using vglut2 riboprobes showed that 23 + 5.4 percent of the DsRed positive neurons were vglut2 positive, (Figure-2, bottom panel). Analysis of the percentage of vglut2 positive neurons that showed the HSR indicated that 10.8+1.94 % of vglut2 neurons were DsRed positive.

Glutamatergic interneurons in zebrafish have primary ascending or descending axons. Among them the commissural primary ascending (COPA) and commissural secondary ascending (CoSA-glu) interneuronal subtypes have ascending axons²⁴. After careful examination of over 25 DsRed labeled embryos, we were unable to see any interneurons with primary descending axons. Hence, we believe that the vglut2 positive interneurons that show the HSR are either CoPA or CoSA-glu interneurons. Thus, the populations of neurons that show the HSR in the embryonic spinal cord were primarily inhibitory glycinergic interneurons and some GABAergic and glutamatergic neurons that modulate the local reciprocal activation and inhibitory circuits required for swimming.

Reduced glycinergic input onto motor neurons following induction of interneuron stress

To this point, our findings suggested that glycinergic inhibitory interneurons comprise the majority of stressed neurons in the early mutant sod1 zebrafish spinal cord. Hence, we used in vivo patch clamp electrophysiology to monitor glycinergic inputs onto motor neurons at 2dpf, a time when the stress response is just manifesting and at 4 dpf, a time when the stress response is pervasive. Glycinergic transmission was monitored during voltage clamp recordings by synaptically isolating neurons with the sodium channel blocker tetrodotoxin (TTX) and blocking glutamate and GABA transmission with kynurenic acid (2.5mM) and bicuculline (25µM) respectively. Under these conditions, spontaneous miniature postsynaptic currents (mPSCs) were observed in motor neurons which represented quantal release of glycine from glycinergic interneurons (Figure 3A). The frequency of these events did not differ significantly in non-transgenic and WTos4-Sh4 fish (control = 0.68 ± 0.09 Hz, WTos4-Sh4 = 0.84 ± 0.13 Hz, p >0.05) but was approximately 50% lower in motor neurons of sod1 mutant fish (0.32 ± 0.05 Hz, p <0.001, Figure 3A, C). In addition, cumulative probability plots revealed a 15% increase in the half-life (control = 2.21 ± 0.08 ms cf. sod1 mutant = 2.55 ± 0.08 ms, p<0.05) but no change in rise time (control = 0.58 ± 0.02 cf. sod1 mutant = 0.60 ± 0.04 ms, p>0.05) or amplitude (control = 20.27 ± 0.63 pA cf. sod1 mutant = 18.9 ± 0.5 pA, p>0.05) of mPSCs (Figure 3D). However, at 2dpf, there were no differences in these parameters between the WT and mutants (data not shown). Taken together, these data idicate that the stress response in glycinergic interneurons is predictive of their impaired function.

Figure 3. Reduced glycinergic transmission onto motor neurons of sod1 zebrafish larvae.

A. Representative traces depicting voltage clamp (holding potential = −75 mV) recordings of spontaneous glycinergic mPSCs in motor neurons of wild type (WT), WT sod1 overexpresser (WTos4-Sh4) and sod1 mutant (G93Ros10-Sh1) fish at 4 dpf. Downward deflections represent occasional quantal release of glycine from presynaptic terminals. B. Average of 30 consecutive glycinergic mPSCs from each experimental condition. C. Bar chart depicting mean mPSC frequency for each experimental condition. (**WT Vs G93R p <0.001) D. Cumulative probability plots of mPSC amplitude, rise time and half-life (p<0.05) in WT (black lines) and G93Ros10-Sh1 (grey lines) motor neurons.

Chronic loss of inhibitory input may contribute to motor neuron stress

As the motor neurons show dysregulation of inhibitory input within a few days after birth, potentially due to pathophysiological changes within inhibitory interneurons, we hypothesized that loss of inhibitory input may contribute to the onset of motor neuron stress, due to chronic loss of inhibitory interneuron input.

We had earlier shown that staining for pre- and post-synaptic NMJ markers was reduced in the muscles of 11dpf sod1 transgenic zebrafish ¹³. However, at these larval stages NMJs were not grossly perturbed, indicating that motor neuron loss was unlikely to have occurred at the stage when DsRed expression is first observed in the interneuron population¹³. Analysis of DsRed expression the high expressor G93Ros10-Sh1 line in 9dpf larvae (which represent a similar developmental stage to the pre-weaning stages in mice) showed widespread stress in interneurons, but an absence of stress in motor neurons, suggesting that motor neurons do not show the HSR at the larval stage of development (Supplementary Figure-4).

We examined whether symptomatic adult zebrafish, that were 12 months old, show evidence of motor neuron stress. Interestingly, spinal motor neurons from mutant sod1 adult fish aged 12 -18 months that exhibit reduced motor function showed evidence of motor neuron stress (Figure-4A, top and middle panel). Greater DsRed induction in the spinal cord of the high expressor transgenic line as compared to the low expressor was observed. However, unlike the three independent mutant sod1 lines, no DsRed expression was observed in the spinal motor neurons of the high expressor WTos4-Sh4 line (Fig-4A, bottom panel). This mutant sod1 specific HSR could also be observed in the motor neurons of young adult zebrafish at 6 months of age (Supplementary Figure-5). Thus, it appears that the HSR in motor neurons is induced between 9 dpf and 6 months of age.

Figure 4. Mutant sod1 zebrafish show stress in the large spinal motor neurons of the adult spinal cord and concurrent loss of NMJ.

A. Spinal cord cross sections from 1–1.5 year old adults zebrafish stained with DAPI, DsRed antibody and ChAT antibody show robust induction of the HSR in spinal cord motor neurons. DsRed colocalised with ChAT in high expressor (3X) G93Ros10-Sh1 line (Top panel), and the moderate expressor (2X) G85Ros6-Sh3 line (Middle panel). The high expressor (3X) WTos4-Sh4 line shows little DsRed expression and DsRed did not co-localize with the large ChAT positive motor neurons (Bottom panel).B. Muscle sections labeled with synaptic vescicle-2 (SV2) antibody (blue), α-bungarotoxin(α-Btx) (green), and DsRed (red) in high expressor G93Ros10-Sh1 (Left panel), low expressor G85Ros6-Sh3 (Middle panel), and high expressor WTos4-Sh4 (Right panel). Normal NMJs’ are indicated by arrows. Abnormal NMJs (arrowheads) where pre and post-synaptic markers are absent or small and punctate are detected in the muscle sections from the mutant lines (Left and Middle panel) but not in the high expressor WT line (Right panel).C. 113 NMJs from multiple sections were measured for NMJ volume from confocal stacks across multiple planes in SV2 positive-DsRed negative axons and SV2 positive-DsRed positive axons using colocalization software from NIH Image J and analyzed by unpaired t-test. Significant reduction in NMJ volume was observed associated with stressed motor axons as compared to the non-stressed axons. The mean is represented as a line over the distribution. Each dot represents the volume of an individual NMJ. P<0.00001.

Stressed motor neurons show reduced neuromuscular synaptic volume and loss of muscle innervation

How muscle denervation occurs in ALS is a fundamental question that remains unanswered and conflicting data indicate that both muscle and motor neurons play an important role in disease pathogenesis^28–30. We had previously shown that, similar to both mice and humans, the number of NMJs in the G93Ros10 mutants was reduced and the few that remained were abnormally small and punctate in appearance¹³. However, it was unclear as to whether these perturbations occurred only at sites innervated by stressed motor neurons. To determine if this was the case, we compared NMJs innervated by stressed (DsRed positive) and non-stressed (DsRed negative) motor neurons. DsRed positive axons were detected in the musculature of the mutant (Figure-4B, left and middle, DsRed panel) but not in the WT sod1 transgenic zebrafish (Figure-4B, right, DsRed panel).

We also observed that the DsRed positive axons in the high expressor G93Ros10-Sh1 showed no identifiable NMJ structures and were completely devoid of α-bungarotoxin labeled post-synaptic structures (Figure-4B, left, α-Btx panel, arrow). However, in the low expressor G85Ros6-Sh3 muscle sections, we observed that the DsRed positive axons showed some recognizable NMJs which were however, abnormally small in size and often punctate in appearance (Figure-4B, middle, α-Btx panel, arrow head). In contrast, the non-stressed axons in the same section that did not show DsRed expression showed large well developed NMJs (Figure-4B, left and middle, α-Btx panel, arrow). The high expressor G93Ros10-Sh1 line showed almost no NMJs arising from the DsRed-expressing stressed axons, thus limiting quantitative analysis of the NMJ in this line. Hence, we chose to use the low expressor G85Ros6-Sh3 line for quantitation of NMJ volume in axons derived from stressed and unstressed neurons. The G85Ros6-Sh3 line showed some NMJ structures associated with the stressed axons and allowed appropriate measurement of the effects of stress within the axonal compartment on NMJ integrity. NMJs were identified at the distal ends of stressed and non-stressed axons and the NMJ volume within the region of interest (ROI) was quantitated using three dimensional images from confocal stacks. We observed an almost 5.5 fold decrease in NMJ volume associated with stressed axons as compared to non-stressed axons (1097±477 Vs. 6245±864, p<0.00001)(Figure 4C). The NMJs in the WTsod1 transgenic line were normal and showed large and well developed synapses (Figure-4B, Right, α-Btx panel, arrow).

Riluzole and nuclear factor (erythroid-derived 2)-like 2 (NRF2) activators reduce neuronal stress

The major impetus for developing the zebrafish ALS model is their suitability for high throughput drug screening, thereby facilitating discovery of drugs that ameliorate ALS. To determine whether the zebrafish model has the potential to identify novel ALS therapies, we tested the ability of the anti-excitotoxic drug riluzole to modify the neuronal stress in zebrafish larvae. Riluzole was chosen for this study as it is the only drug shown to have disease modifying effect in ALS patients. We subjected 24 hpf G93Ros10-Sh1 embryos to a 4 day incubation in 1, 3, 5, 7 and10 µM riluzole and used DsRed fluorescence as a marker for progressive changes in the neuronal stress response. We observed that riluzole caused a dose dependent reduction in DsRed fluorescence with an IC50 of approximately 7 µM (Figure-5a,b). Tricaine, a local anesthetic that inhibits neuronal sodium channels³¹ like riluzole and also reduces excitotoxicity, produced a significant decrease in DsRed fluorescence (Figure 5b).

Figure 5. Inhibition of the stress response in sod1 G93Ros10-Sh1 zebrafish embryos by riluzole and an NRF2 activator R-apomorphine.

A: Dose response curve showing dose dependent inhibition of the stress response by riluzole in sod1 G93Ros10-Sh1 embryos treated for 4 days with 1, 3, 5, 7 and 10 µM riluzole, p<0.001. B: Percent inhibition of the stress response expressed by reduction in DsRed fluorescence in embryos treated with 610µM tricaine (p<0.000001), 10µM riluzole (p<0.00001), and 10µM R-apomorphine (p<0.001) as compared to 0.1% DMSO treated embryos. Mean±SEM.

The NRF2 pathway plays an important role in regulating oxidative stress, the cellular handling of misfolded proteins and in mounting the autophagy response^{32, 33}. Thus, we hypothesized that drugs that upregulate this pathway may potentially reduce neuronal stress in this model system. We used R-apomorphine which is known to be an activator of the NRF2 anti-oxidant response (ARE)³⁴ to determine the effects of ARE induction on the stress response observed in this model system. Treatment of embryos with R-apomorphine produced a significant reduction in the neuronal stress response (Figure-5b). Thus, two neuroprotective drugs acting through diverse mechanisms were able to reduce the readout of neuronal stress in this model system.

Discussion

Zebrafish have become a powerful model for the study of degenerative diseases as their experimental accessibility, small size and genetic similarity to mammals facilitates detailed analysis of disease mechanisms and drug screens. Zebrafish expressing mutant sod1 develop hallmark features of ALS commonly associated with both murine models and the human disease¹³ (a comparison of pathophysiological changes in our sod1 zebrafish model, sod1 mouse models and human ALS are summarized in Table-2).

Table-2. Summary of ALS disease pathophysiology in human, mice and zebrafish.

Comparison of pathophysiological changes in ALS in human, mouse [SOD1 G93A high copy; B6SJL-Tg(SOD1*G93A)1Gur/J strain that shows 13X increase in sod1 activity] and zebrafish [transgenic sod1G93R high expressor; Tg (sod1:sod1G93Ros10) that shows 3X increase in sod1 expression levels].

	Neuronal stress	Electrophysiology	NMJ denervation	Golgi changes, insoluble aggregates	Astrocytosis/ Microgliosis/ Mitochondria/ Inclusion/UPS	Gross motor symptoms
Human	?	Prior to onset in adults†	Prior to onset in adults†	In post-mortem specimens	In post-mortem specimens	40–50 years
	Embryo	Fetal	Post-partum
Mouse (Sod1G93AHiGur)	?	12 dpf* to 6 dpp** (G85R, G93Alow)	~30dpp	~40–50dpp	~70–90dpp	~100dpp
Zebrafish (G93R)	24 hpf ** (This manuscript)	96hpf** (This manuscript)	Normal NMJ but reduced staining at 11dpf. 6 months and 1 year (tested) Observed denervation in stressed motor neuron (This manuscript)	UPR at 1 month***.	?/- No gross inflammation (unpublished) Motor neuron stress. 6-months (This manuscript)	1 year

dpf: days post fertilization; hpf: hours post fertilization; dpp: days postpartum. UPR: unfolded protein response

^†earliest tested,

^*earliest tested in organotypic culture;

^**Earliest tested in intact spinal cord,

^***Personal communication (C.Beattie)

In the current study we have extended our analysis of mutant zebrafish to track early, pre-symptomatic perturbations associated with the sod1 mutation. Using the hsp70-DsRed marker gene as a novel readout of neuronal stress we have characterized perturbations in the spinal cord from early embryonic stages through to adult life, establishing a correlation between neuronal stress and pathophysiological changes in stressed neurons. We find that in early stages of development, inhibitory neurons are sensitive to the presence of mutant Sod1, activating the stress response upon its expression. Moreover, using electrophysiological approaches we show that glycinergic neurotransmission onto motor neurons is impaired in mutant sod1 fish. These perturbations precede the onset of pathophysiological defects in the motor neurons and at the neuromuscular junction, which occur later in life.

SOD1 is a ubiquitously expressed protein but mutation in the SOD1 gene produces a disease affecting the CNS. How this gene selectively perturbs neurons during ALS is not well understood, although the accepted view is that neurons are selectively sensitive to mutant SOD1. Our observation suggests that neuronal stress as measured by the HSR response occurs in spinal interneurons long before it is observed in motor neurons. This implicates interneurons as important components of ALS disease progression in this mutant sod1 zebrafish model. More importantly, we provide for the first time evidence that this process occurs in vivo.

A possible role for inhibitory interneurons in ALS has been hypothesized previously (see Turner and Kiernan (2012)³⁵ for a comprehensive review on this subject). Both human ALS patients and mouse Sod1 models have been reported to show a loss of spinal cord interneurons, aberrant recurrent inhibition and motor neuron hyperexcitability, observations that suggest that dysregulation of inhibitory influences on motor neurons may represent an early aspect of disease pathophysiology. Moreover, recent in vitro studies show that motor neurons derived from Sod1 mice have abnormal glycine receptor expression and reduced responses to glycinergic input⁴⁰. As shown in Table-2, electrophysiological abnormalities are the earliest changes observed in mice suggesting that early pre- and post-natal development is an area important for investigation in relation to the mechanisms of Sod1 toxicity. Together these observations suggest that defective inhibitory transmission may promote motor neuron stress and accelerate disease progression. Indeed, it is perhaps worth noting that that human patients carrying the SOD1D90A homozygous mutations develop an atypical slowly progressing form of ALS and in these patients, inhibitory interneurons are spared⁴¹. However, Hossaini et al, 2011⁴², reported that interneuron pathology occurred following motor neuron death in the low expressor SOD1G93A mice. This study was limited to examining adult mice, a few weeks prior to the onset of symptoms and did not look at early embryonic or neonatal animals. Interestingly, mice from this line show evidence of hyperexcitability as early as postnatal day 6–10, long before the onset of symptoms⁴³. The source of this hyperexcitability is still unclear, but the data presented in this paper raise the possibility that dysfunctional inhibitory interneurons may be responsible for these changes. Future studies looking at different neuronal populations during embryonic, neonatal and adult stages of development in the mutant Sod1 mouse model would be necessary to establish the exact timing of damage to various cell types.

Interestingly, overexpression of wild type Sod1 also induced neuronal stress in interneurons, albeit at a much lower level than in the mutant sod1 zebrafish lines. This observation stands in agreement with previous studies that show high levels of wild type Sod1 are toxic, although to a much lesser extent than mutant Sod1^{3, 5, 44}. For example, transgenic mice that overexpress wild type SOD1 also show motor neuron loss, although it occurs far later (2 years) ⁴⁵. It is important to note that the high expressor G93Ros10 line that expresses similar levels of Sod1 to the WTos4 line showed a 10 fold greater HSR induction compared to the wild-type expressing line. Similar to the mouse over expressing WTSOD1 where the disease was limited and not progressive, in WTSod1 over expressing transgenic zebrafish, the HSR failed to spread to motor neurons, to produce NMJ denervation or to cause muscle atrophy. Our observations are in keeping with those seen in WT SOD1 over-expressing transgenic mice. Another interesting finding from our study is that motor neurons did not show induction of the HSR at the early embryonic and larval stages, a time when inhibitory interneurons exhibited clear dysfunction. Rather the onset of HSR induction in motor neurons occurred long after, as the fish progressed towards adult life. These findings suggest that the pathophysiological changes are not static but spread progressively through the neural network controlling motor system function. The findings in this zebrafish model indicate that inhibitory interneurons may be the cell type within the spinal cord most susceptible to neuronal stress induced by the presence of mutant Sod1. One route for the propagation of the pathophysiology from inhibitory interneurons to motor neurons is through dysregulation of the inhibitory input to motor neurons as discussed earlier. The zebrafish model we have generated offers new approaches to test this hypothesis and identify the circuitry involved in the pathophysiological cascade.

Dying back axonal pathology is one of the proposed mechanisms leading to motor neuron death, although we do not know whether this occurs as a primary event or secondary to changes in the motor neuron perikaryon^{46, 47}. We had earlier shown that NMJs in mutant sod1 transgenics were reduced and the few that were present were morphologically abnormal (small and punctate). This HSR readout in our zebrafish model allows us to track the axons of stressed motor neurons into the muscle and evaluate whether stress in the motor neuron perikaryon leads to detectable pathology at the NMJ. We show that NMJs from stressed motor neurons were abnormal as they were absent or were small and punctate in appearance as compared to non-stressed axons. This effect was specific to the mutant sod1 lines and was not observed in the Sod1 over- expressing line. This indicates that the stress we observe in the motor neuron perikarya is predictive of synaptic degeneration at the NMJ. The sod1 zebrafish model will allow more detailed dissection of the processes involved in denervation at the NMJ as now we can distinguish motor axons arising from non-stressed and stressed motor neurons.

Based on the combination of novel findings in the sod1 zebrafish model of ALS, we propose a draft model of ALS disease propagation in which neuronal stress begins very early in life initially affecting the inhibitory interneuron pool whose dysfunction may then contribute to pathophysiological changes in motor neurons later in life (Figure-6b). In this proposed model, it is also possible that interneurons are more vulnerable to the toxic effects of mutant Sod1, while motor neurons may require two hits (Sod1 toxicity and loss of inhibition) to drive pathology. Future cell specific transgene expression and/or tissue transplant studies, which can be performed in zebrafish embryos, will allow further exploration of this hypothesis. Nonetheless, stressed motor neurons are unable to maintain normal NMJ which may in turn lead to lack of the trophic support necessary for maintenance of functional neuromuscular contacts, thereby leading to denervation. Such a model has been previously suggested⁴⁷, but here we propose that stressed motor neurons are dysfunctional and induce a dying back phenotype, while NMJs of non-stressed neurons remain intact. The evidence we provide here shows a direct link between the level of chronic neuronal stress and neuromuscular degeneration.

Figure 6. Model of neuronal stress propagation in zebrafish model of ALS.

A: While wild type sod1 over expressor zebrafish show some interneuron stress in early development, the stress levels are low and there is no propagation of the stress response to motor neurons. B: Mutant sod1 transgenic zebrafish show stress initially in the inhibitory interneurons that cause dysfunction of glycinergic inhibitory interneurons with reduced glycinergic input to motor neurons. This lack of inhibition may be one factor contributing to the stress response developing in motor neurons in the adult zebrafish. Stressed motor neurons are unable to maintain synaptic function with resulting synaptic withdrawal at the NMJ. The retraction of the pre-synaptic input at the NMJ may lead to a loss of trophic support from muscle resulting in a vicious cycle of injury to the motor neuron, eventually leading to motor neuron degeneration and muscle atrophy. Shades of grey indicate stress levels.

While, we note that the early disease features seen in this zebrafish model (electrophysiological changes and NMJ denervation) are also observed in the pre-symptomatic rodent models of SOD1-related ALS and to a limited extent in pre-symptomatic human patients, further work is necessary to confirm if these early changes reflect the initiation of a progressive pathophysiological cascade that eventually culminates in motor neuron loss. However, the similarities in the pathophysiological changes occurring across these diverse species and the gradual nature of these changes that precede motor neuron loss suggest that they are likely to represent an integral component of the disease process. Nevertheless, the ability to use the zebrafish to monitor these changes in real time in vivo due to their small size, transparency and easily accessibility, provides us with a valuable tool for studying the cascade of motor neuron injury in ALS from early pathophysiological changes to motor neuron cell death.

The sod1 zebrafish model also holds promise as a link between cell based assays and rodent models, thus providing a new platform for high-throughput screening of neuroprotective compounds. In the current study we tested the effects of riluzole, the only therapeutic compound that has shown benefit in both the SOD1^G93A murine models and human ALS patients, on neuronal stress in the mutant sod1 zebrafish larvae. We found that riluzole reduced the HSR in zebrafish, indicating that it may slow progression of disease. Although riluzole shows efficacy in ALS, the mechanism by which it mediates neuroprotection is still unclear. Riluzole is known to reduce neuronal excitability by stabilizing the inactive state of voltage gated sodium channels and also by acting as a non-competitive NMDA receptor antagonist⁴⁸. Interestingly, tricaine, another drug that modulates voltage gated sodium channels showed a similar inhibition of the HSR. Together these observations suggest that modulation of neuronal sodium channel activation may affect disease progression, possibly in an activity-dependent manner. Dissecting the molecular mechanisms by which riluzole reduces the neuronal stress response will be useful and may allow the development of compounds with greater neuroprotective efficacy. We also identified that activators of the NRF2 transcription factor, which plays an important role in neuroprotection and operates as a master regulator of the anti-oxidant and stress response pathways, can also modulate the early neuronal stress response observed in this model. Although, this early stage screening allowed identification of neuroprotective compounds, further validation of hits from this assay will be necessary in higher vertebrate models of ALS. The in vivo high throughput zebrafish model fills an important gap in ALS drug discovery and may provide lead candidates for drug screening in rodent models, thus accelerating the drug discovery process in ALS with the potential for more rapid identification of promising compounds for clinical translation.

Abbreviations

G93R os10-sh1

Tg(sod1:sod1G93R; hsp70:DsRed)ohio state 10-sheffield strain 1

G93R os6-sh2

Tg(sod1:sod1G93R; hsp70:DsRed)ohio state 6-sheffield strain 2

G85R os6-sh3

Tg(sod1:sod1G85R; hsp70:DsRed)ohio state 6-sheffield strain 3

WTos4-sh4

Tg(sod1:sod1; hsp70:DsRed)ohio state 4-sheffield strain 4