The neuromuscular junction (NMJ) is a highly specialised cholinergic synapse formed between a motor axon terminal and its target skeletal muscle fibre. It acts to transmit signals from the central nervous system to the muscle to stimulate contraction for functions such as maintenance of posture, movement of limbs, and respiration. Many proteins contribute to the structure and function of the NMJ during development and throughout adult life, acting to maintain the safety factor of the NMJ, which ensures reliable and efficient signal transmission. In a rare group of disorders termed the Congenital Myasthenic Syndromes (CMS), there is genetic impairment to one of numerous genes encoding proteins that play a critical role at the NMJ [1
]. In patients, this manifests as fatigable muscle weakness, along with a number of other symptoms that vary between subtypes [2
In 2016, we identified missense mutations in MYO9A
, a gene encoding for the unconventional myosin IXA protein [3
], as the likely cause of CMS in three patients from two unrelated families. Subsequent research demonstrated an impact of MYO9A-loss on endo/exocytosis and vesicle trafficking in a nerve-cell line (NSC-34). MYO9A is a negative regulator of Ras Homolog Family Member A (RhoA) [4
], and the in vitro defects described could be partly ameliorated by blocking a downstream target of the RhoA pathway, Rho-associated protein kinase (ROCK), with Y-27632. This highlighted the potential involvement of the RhoGTPase domain of MYO9A in the pathophysiology of MYO9A-CMS. As trafficking and endo/exocytosis were impaired in the absence of MYO9A, we hypothesised that this may in turn affect release of proteins at the NMJ, and thus secretomics was performed on MYO9A-depleted NSC-34 cells [5
]. This revealed a significant decrease in secretion of agrin, a molecule crucial for post-synaptic NMJ formation and maintenance [5
]. To further investigate the role of MYO9A, a morpholino (MO)-based model in zebrafish was generated targeting their two MYO9A
). The zebrafish knockdown embryos demonstrated impaired movement during development, and observational analysis indicated the presence of disrupted NMJ morphology. Application of an agrin fragment compound generated by Neurotune, NT1654, ameliorated the movement defects and NMJ phenotype [5
]. While these findings provided support regarding a role for MYO9A at the NMJ and the positive potential for partial agrin-replacement by exogenous application of NT1654, the specificity of MOs and phenotypes produced have been the topic of some concern within the scientific community. This is largely due to a disparity between phenotypes observed in morphants and genetic mutants generated using techniques such as CRISPR/Cas9, as well as the frequent off-target effects induced by MO injection [12
Therefore, the aims of this paper were to implement a CRISPR/Cas9-mediated approach for Myo9aa/ab knockdown in zebrafish, characterise the NMJ defects in greater detail, and to test two treatment strategies. As we hypothesised that disruption to secretion in the absence of MYO9A was related to interactions of this unconventional myosin protein with the RhoA/ROCK pathway, modulation of this pathway was also tested in the zebrafish model using the ROCK inhibitor compound fasudil. Evidence for the benefits of NT1654 application as a potential treatment and as a proof of principal for the proposed mode of action of MYO9A-loss were also obtained in greater detail, and over a longer time-period than previously performed.
2. Materials and Methods
2.1. Zebrafish Maintenance
Zebrafish (AB wild-type strain or Golden slc24a5b1/+ strain, Zebrafish International Resource Centre, Eugene, OR, USA) were maintained according to Home Office guidelines (Project License: 70/8038), with a continuous light-dark cycle (14 h light, 10 h dark). Age of zebrafish expressed as hours post fertilisation (hpf). Euthanasia of fish was performed using a 1:1 ratio of fresh system water: 4 mg/mL tricaine methanesulfonate.
2.2. CRISPR/Cas9-Mediated Knockdown in Zebrafish
The protocol was modified from Varshney et al. [14
]. Single guide RNAs (sgRNAs) for generating CRISPR/Cas9-mediated knockdown zebrafish were designed using CRISPR scan (http://www.crisprscan.org
/, date last accessed 20 February 2019 [15
]). sgRNAs with the highest CRISPRscan score (efficiency score) and no predicted off-target effects were selected, with a total of 2 sgRNAs for each gene (2 × myo9aa
exon 2, 1 × myo9ab
exon 1, 1 × myo9ab
exon 12). A published sgRNA against the Tyrosinase gene (tyr)
was also included as an injected control [16
]. DNA oligonucleotides that included the target region and surrounding T7 promoter sequence and tail sequence were annealed with a universal bottom strand ultramer (5′AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3′, Table S1
). Annealing was performed using a MyTaq DNA polymerase kit as outlined in Tables S2 and S3
. A Qiagen PCR purification kit was then used to purify samples according to the manufacturer’s instructions. The purified product was used as a template for a transcription reaction using a MEGAshortscript T7 kit (Invitrogen, Waltham, MA, USA) according to the manufacturer’s instructions. A total of 8 reactions for each sgRNA were pooled to increase yield prior to application to a mirVana spin column (mirVana miRNA isolation kit, ThermoFisher Scientific, Waltham, MA, USA), and subject to RNA purification according to the manufacturer’s instructions. Before injection into zebrafish embryos, the sgRNA and Cas9 reaction mix was prepared as outlined in Table S4
, and heated to 37 °C for 5 min to improve knockdown efficiency as described by Burger et al. [17
For generating ‘crispant fish’ (the F0 mosaic fish), sgRNA and Cas9 protein were injected directly into the cell at the single cell stage. Individual sgRNAs were injected to confirm ability to induce insertions/deletions, before co-injection of all 4 sgRNAs for the experimental protocol. Fish were incubated in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM MgSO4, 0.33 mM CaCl2, and 0.01% methylene blue) at 28.5 °C for a maximum of 5 days, with dead fish removed daily. For experiments requiring dechorionation, pronase (Streptomyces griseus, Roche, Basel, Switzerland) was added to the zebrafish embryos at a final concentration of 1 mg/mL in E3 medium.
2.3. NT1654 Treatment
The agrin compound NT1654 was used to treat myo9aa/ab
crispant zebrafish. This compound is a 44 kDa artificial agrin fragment developed by Neurotune AG (Switzerland, [11
]). NT1654 (0.15 ng) was delivered to the embryos at the same time as sgRNA/Cas9 injection, due to the size being incompatible with water-based diffusion delivery [18
], as performed previously [5
]. Embryos from the same pair of fish were split into each of the three categories (wild-type, myo9aa/ab crispant
, treated myo9aa/ab
crispant) for each injection session to ensure fair comparisons for survival and development.
2.4. Fasudil Treatment
Fasudil (5 mM, Millipore Sigma, Darmstadt, Germany), a ROCK inhibitor, was used to treat the myo9aa/ab crispant zebrafish. The zebrafish were housed in 12-well plates, with ten or less fish per well, in E3 medium. A range of concentrations of fasudil were trialled to optimise the dose, from 1 nM to 100 µm, starting from 7 hpf and final dose-finding assessments made at 48 hpf. Based on preliminary survival rate data and assessment of chorion movements in response to each dose, a final concentration of 10 µm was used.
Fish were split into 3 treatment groups, wild-type, myo9aa/ab crispant, and 10 µm fasudil treated myo9aa/ab crispant. Treatment started at 7 hpf and was continued to 5 dpf, with solution changes daily.
2.5. Chorion Movements and Tactile Response Assay
Chorion movements were assessed at 24 hpf as previously described [3
]. Briefly, embryos were recorded using a Leica stereomicroscope mounted with a Chameleon digital camera (CMLN-13s2M, FLIR Systems, Kent, UK). Recordings were made for one minute and the number of full twists performed by each fish was then manually counted from the recordings. At 48 hpf, response of zebrafish to tactile stimulation was analysed as previously described; fish classed as having a ‘severe’ phenotype were not used for movement analysis [5
]. Briefly, fish were placed individually in a petri dish containing E3 medium on top of an illuminated stage, and a Canon legria hfr76 camera was clamped 7 cm above the dish. A fine pipette was used to touch the zebrafish on the back of the head and the response recorded. Room temperature remained constant at 28 °C throughout the experiment. Videos were imported into Fiji ImageJ [19
] as FFmpeg movies, and then manually thresholded to allow visualisation of the zebrafish, before converting to a binary image. The Trackmate plugin [20
] was then used to measure the movement of the zebrafish, with manual editing of each frame to ensure that only the zebrafish was detected and the movement identified was accurate. Values for distance travelled and average speed were exported, from which the initial acceleration could be derived. In vivo experiments were blinded prior to live recording and for image acquisition.
2.6. Zebrafish Whole-Mount Immunofluorescence
Fixation and staining of zebrafish was performed as described previously; fish with a severe phenotype were excluded from analysis [21
]. Briefly, fish at 24, 48, and 120 hpf were dechorionated, euthanized with Tricaine, and fixed in 4% paraformaldehyde in phosphate buffered saline (PBS) overnight at 4 °C. Fish at 120 hpf were subject to Collagenase A (Millipore Sigma, 1mg/mL) treatment for 90 min prior to immunofluorescence. The presynaptic NMJ was incubated overnight at 4 °C with a mouse anti-synaptic vesicle protein 2 (SV2, 1:200, Developmental Studies Hybridoma Bank, Iowa City, IA, USA) antibody detected with secondary antibody (Alexa Fluor 594 IgG goat anti-mouse, 1:500, Life Technologies, Waltham, MA, USA). α-bungarotoxin (αBTx) conjugated to Alexa Fluor 488 was incubated with the secondary antibody for 2 h at room temperature to detect postsynaptic acetylcholine receptors (AChRs, 1:1000, ThermoFisher Scientific). Antibodies were diluted in 5% horse serum in PBS with 0.1% tween. Washes were performed using PBS with 0.1% tween. Fish were mounted in Vectashield fluorescent mounting medium (Vector Laboratories). Z-stack images encompassing the entire volume of the zebrafish tail and multiple somites around somite 15 were obtained using a 40× oil-immersion objective on a Nikon A1R confocal microscope.
2.7. NMJ Morphology Assessment
Fiji (ImageJ, Madison, WI, USA) was used to generate maximum intensity projections of acquired z-stack images, and measurements were obtained at the same somite level between treatment groups. At 24 hpf, myotome area was measured and the presence or absence of a central AChR cluster noted, referred to as the ‘choice point’. The distance that motor neuron axons travelled past the choice point was also manually measured, along with total and average AChR area per 100 µm2. At 48 and 120 hpf, myotome size was measured, along with the number of presynaptic and postsynaptic clusters per 100 µm2, average size of clusters and total area of clusters 100 µm2, as well as the number of large clusters (>20 µm2).
Counts and measurements of clusters were performed by automatic thresholding and conversion of images to binary before using the ‘analyse particles’ tool in Fiji. The length of the myosepta (displaying AChR-positive areas) was manually measured, along with the contact by motor neurons, giving a value for percentage of myosepta overlaid by motor neuron.
Colocalisation analysis between SV2 and αBTx-positive signals was performed on maximum intensity projections of 120 hpf zebrafish encompassing multiple somites in the field of view. The ‘EzColocalization’ Fiji plugin was used for analysis according to the protocol described by Stauffer, et al. [22
]. Briefly, each fluorophore channel was subject to automatic thresholding to remove background, and the Mander’s correlation coefficient calculated to give a value between 0 and 1, reflecting the degree of co-occurrence of signals between both SV2 and αBTx, and also αBTx and SV2.
2.8. Statistical Analysis
Statistical analysis was performed using GraphPad Prism software (v8.0.2, BD Biosciences, San Jose, CA, USA). Data sets were first tested for normal distribution and, from these results, either nonparametric (Kruskal-Wallis with Dunn’s post hoc test) or parametric tests (one-way ANOVA with Turkey post hoc test) were applied. Outliers were identified and removed according to the ROUT method. Statistical significance was taken as p < 0.05. In vivo experiments were blinded prior to live recording and for image acquisition.
The NMJ is a complex synapse, with many proteins important for its development and function. The safety factor of the NMJ generally maintains its signalling ability, resulting in a consistent postsynaptic response to presynaptic release of ACh. Disruption to the morphology of the NMJ can impair this safety factor by reducing the probability of release below an effective threshold, the efficacy of postsynaptic signal amplification, the ACh clearance, or the function of AChRs, among a number of other mechanisms [28
]. The role of MYO9A
, a novel CMS-associated gene, at the NMJ is as yet unclear. The results of this study provide further evidence for the importance of this protein at the NMJ, and also support our proposed mechanism of action, as outlined in Figure 6
: Loss of MYO9A reduces its action on the RhoA/ROCK pathway, downstream effects on secretion including agrin release, less clustering/stability of AChRs, resulting in impaired signal transmission and CMS. While further investigations using electrophysiology will be required to attribute physiological changes to MYO9A-loss at the NMJ, a detailed overview of NMJ morphology indicates there are at least structural changes to this cholinergic synapse that are associated with disruptions to motor behaviours in zebrafish.
Supporting our previous results obtained using a MO-mediated knockdown approach in zebrafish against the two fish orthologue, myo9aa/ab
, crispant fish lacking expression of both of these myosin proteins displayed similar phenotypes. This included not only a curved tail, but also the presence of oedema, initially thought to result from the off-target effects produced from MO injection. MYO9A knockout mice exhibit low molecular weight proteinuria, which has recently been shown to produce oedema in early development of zebrafish—thus, this phenotype will require further investigation, as well as examination of patients for potential kidney problems [29
Morphants treated with NT1654 displayed a reduced survival rate, which could have resulted in a sampling bias of surviving zebrafish for the subsequent studies. Crispant fish treated with NT1654 do not exhibit a significant reduction in survival rates, indicating the previous finding was likely due to toxic effects of co-injecting two morpholinos with the compound. Fasudil treatment of crispant fish was also performed to examine the in vivo effect of blocking the predicted over-activity of the RhoA/ROCK pathway in the absence of Myo9aa/ab. This treatment significantly increased survival, indicating a positive action at 48 hpf and no overall detrimental development effects, despite the drug’s widespread action.
While crispants are a mosaic model, co-injection of 4 sgRNAs against the two genes targeted increases the chance of decreasing protein expression. Functional assessments of zebrafish revealed impaired movement at 24 and 48 hpf, similar to those reported for zebrafish lacking agrin [31
]. Mice lacking agrin (Agrnnmf380/nmf380
) also display poor motor control phenotypes, reflecting the importance of agrin in performing tasks relying on NMJ integrity [32
]—and thus providing support for reduced agrin secretion, contributing to the pathomechanism of MYO9A-CMS. As impaired movements in zebrafish are not restricted to MYO9A or agrin knockdown and are present in a number of other zebrafish models of CMS, further evidence was obtained by the improvement of these behaviours by the application of NT1654. Of particular interest was the improvement in acceleration, which at this time point has been shown to be indicative of muscle contraction force [26
]. A sarcopenia mouse model previously generated by increasing cleavage of agrin at the NMJ (by overexpression of neurotrypsin) displayed a reduced grip strength, which was restored with treatment using NT1654 [11
Agrin secretion from the motor neuron during NMJ synaptogenesis is a critical step, without which AChR clusters do not form correctly [8
]. In the myo9aa/ab
crispant fish, there were fewer, larger AChR clusters, with an overall decrease in postsynaptic staining at 48 hpf. This corroborates observations of reduced myotomal AChR clustering in agrin-deficient zebrafish [31
]. By 120 hpf, zebrafish lacking Myo9aa/ab exhibited an increase in number but decrease in average size of AChR clusters, as compared to wild-type fish. This also reflects observations in the Agrn
KO mouse, which exhibits a number of small AChR clusters opposing nerve terminals during development in utero [8
]. In support of the hypothesis that MYO9A-CMS is based on a reduction in agrin secretion, a number of the defects in NMJ morphology can be rescued by NT1654 treatment. For example, rescue of the presence of a choice point cluster of AChRs, the number of large clusters, and average AChR cluster area at 24 hpf was achieved. Application of NT1654 also increased the number of αBTx-positive clusters at 48 hpf, and myotome size at 120 hpf. Many of these improvements corroborate reported benefits of this compound in the aforementioned mouse model of sarcopenia [11
]. The effect of NT1654 on neurons extending from the spinal cord or on number of terminals at later time points was unexpected, although application of agrin to in vitro cultures of PC12 cells and chick retinal neurons has been shown to mediate FGF2-induced neurite extension [33
]. Furthermore, in the opposite scenario, removal of agrin from zebrafish has been shown to cause truncated primary motor axons during development, as well as erratic trajectories of these axons across the myotome, consistent with our findings [31
]. Similar effects of abnormal axonal extensions are also observed in the diaphragm muscle of a murine model of agrin deficiency (AGZ) [8
]. In the future, it would be interesting to identify whether the effects on motor neurons are limited to primary motor neurons or also involve secondary motor neurons in the zebrafish, which coordinate different forms of swimming. As primary motor neurons are associated with fast swimming responses, it is likely that this subtype is affected in our model; reflected in our movement data, older fish would be required to assess the secondary motor neurons which mediate slow, rhythmic swimming [24
While the main hypothesis for the mode of action for MYO9A-loss is a reduction in agrin secretion from the nerve, previous analysis of MYO9A knockdown NSC-34 cells and the current study on zebrafish reveal effects on nerve and presynaptic nerve terminal morphology [5
]. This could be related to cytoskeletal disruption due to RhoA over-activity, as application of Y-27632 was able to ameliorate some cytoskeletal defects in NSC-34 cells, and therefore improve intracellular trafficking [5
]. Furthermore, treatment of zebrafish with the ROCK inhibitor, fasudil, was found to provide some improvements to the NMJ, including all disrupted features identified at 24 hpf—such as neurite extension—corroborating reports that ROCK inhibition can extend neurites in vitro [34
]. The size of myotomes in the fish was also returned to wild-type size by fasudil treatment, correlating with results demonstrating a positive effect of fasudil application on muscle fibre size in an SMA mouse model [35
]. These results highlight an important role for the RhoA/ROCK pathway in the phenotype of Myo9aa/ab-depletion in early zebrafish development, and it is possible that the improvements observed are due to fasudil in vivo act upstream of agrin release, as outlined in Figure 6
. The lack of continued improvement to NMJ phenotype by fasudil throughout the time period assessed could be attributed to a number of causes. There may be a requirement for fine modulation of dosage for fasudil throughout development due to its widespread action and the likely differing contribution of the RhoA/ROCK pathway at different developmental stages. It could also be that effects of Myo9aa/ab at the NMJ are not mediated by the ROCK pathway as development proceeds, and instead are linked to other functions of this protein such as cross-linking actin filaments, or as yet unidentified interactions with other pathways of impact on the NMJ [36
]. Nevertheless, the movement benefits due to fasudil treatment suggest there may be modest improvements in release/functionality at the NMJ that are not detectable by the morphological changes measured here.
Studying crispants has provided the benefit of assessing 120 hpf fish, which display innervation patterns similar to those observed in adult fish [27
], which was not possible with MOs. Agrin has been shown to induce ectopic, fully differentiated postsynaptic compartments anywhere on the muscle, including insertion of AChRs [37
]. While there is an increase in cluster number per 100 µm2
at 48 hpf in NT1654-treated crispants, this is only to the level observed in wild-types. However, at 120 hpf, there is a further increase in cluster number, which may represent ectopic AChRs. Colocalisation analysis is a useful metric, and as αBTx is very specific to postsynaptic AChRs, it is expected that the majority of the signal would be overlaid by SV2 after full innervation is complete (in the absence of ectopic clusters) and all prepatterned receptors are incorporated. Conversely, antibodies against SV2 can detect presynaptic vesicles that are localised along the nerve during development/transport, therefore may not necessarily all co-occur with post-synaptic αBTx [27
]. Colocalisation analysis between AChRs and SV2 revealed no significant differences in co-occurrence between genotypes/treatment groups, indicating that around 70% of the AChR clusters identified were overlaid by nerves in all groups, which is similar to the proportions of colocalisation reported elsewhere and highlights the absence of ectopic cluster formation by NT1654 [38
]. However, co-occurrence of SV2 with AChRs at 120 hpf is significantly reduced in crispant fish, as compared to wild-type, by around 30%. This could signify either an increase in presence of SV2 clusters or a decrease in AChRs. Analysis of NMJ morphology reveals a significant reduction in the average area of AChR clusters, thus supporting our hypothesis that Myo9aa/ab-depletion impairs secretion of agrin—and thus, clustering and stabilisation of AChRs. Treatment of 120 hpf fish with NT1654 significantly improved the impairment of postsynaptic colocalisation with the presynapse. An increase in the number of AChRs as compared to wild-type fish was also observed, but this was accompanied by a further decrease in the average area of clusters, thus indicating that the improvement in co-occurrence due to NT1654 may be a result of increased receptor insertion at novel sites rather than increased size of clusters present. NT1654 also significantly increased the average area of SV2-positive clusters which could contribute to the improved alignment of pre and postsynaptic NMJ components.
The evidence presented here supports a role for MYO9A at the NMJ, and the view that its absence or dysfunction is a cause of CMS. While many avenues for exploration of the precise function of MYO9A at the NMJ remain open, including a role in the postsynapse, an action on RhoA/ROCK and agrin secretion have been demonstrated. In the future, it will be important to confirm that the morphological defects identified in zebrafish manifest as impaired signal transmission at the NMJ and, thus, underpin a CMS phenotype. Furthermore, there is potential for the use of NT1654 in a clinical setting, including in MYO9A-CMS and other disorders in which agrin dysfunction is suspected.