Change of voltage-gated sodium channel repertoire in skeletal muscle of a MuSK myasthenia gravis mouse model

Muscle-speciﬁc kinase myasthenia gravis (MuSK MG) is caused by autoantibodies against MuSK in the neuromuscular junction (NMJ). MuSK MG patients have ﬂuctuating, fatigable weakness, in particular of bulbar muscles. Severity diﬀers greatly between patients, in spite of comparable autoantibody levels. One explanation for inter-patient and inter-muscle variability in sensitivity might be variations in compensatory muscle responses. Previously, we developed a passive transfer mouse model for MuSK MG. In preliminary ex vivo experiments we observed that muscle contraction, in particular of mice with milder myasthenia, had become partially insensitive to μ-Conotoxin-GIIIB, a blocker of skeletal muscle NaV1.4 voltage-gated sodium channels. We hypothesized that changes in NaV channel expression proﬁle, possibly co-expression of ( μ-Conotoxin-GIIIB insensitive) NaV1.5 type channels, might lower the muscle ﬁbre’s ﬁring threshold and facilitate neuromuscular synaptic transmission. To test this, we here performed passive transfer in mice, using ‘high’, ‘intermediate’ and ‘low’ dosing regimens of puriﬁed MuSK MG patient IgG4 and compared myasthenia levels, μ-Conotoxin-GIIIB resistance, muscle ﬁbre action potential characteristics and ﬁring thresholds. High-and intermediate-dosed mice showed clear, progressive myasthenia, not seen in low-dosed animals. However, diaphragm NMJ electrophysiology demonstrated almost equal myasthenic severities amongst all regimens. Nonetheless, low-dosed mouse diaphragms showed a much higher degree of μ-Conotoxin-GIIIB resistance. This was not explained by upregulation of Scn5a (the NaV1.5 gene), lowered muscle ﬁbre ﬁring thresholds or histologically detectable upregulated NaV1.5 channels. It remains to be established which factors are responsible for the μ-Conotoxin-GIIIB insensitivity and whether the NaV repertoire change is compensatory beneﬁcial, or a bystander eﬀect.

resistance.This was not explained by upregulation of Scn5a (the Na V 1.5 gene), lowered muscle fibre firing thresholds or histologically detectable upregulated Na V 1.5 channels.It remains to be established which factors are responsible for the observed μ-Conotoxin-GIIIB insensitivity and whether the Na V repertoire change is compensatory beneficial or a bystander effect.
K E Y W O R D S homeostasis, MuSK, myasthenia gravis, Na V channels, neuromuscular junction, passive transfer

| INTRODUCTION
Myasthenia gravis (MG) is an autoimmune disease clinically characterised by fluctuating, fatigable muscle weakness and caused by autoantibodies against key synaptic membrane proteins at the neuromuscular junction (NMJ) (Huijbers et al., 2022;Phillips & Vincent, 2016).The majority of patients ($85%) has predominant IgG1 and IgG3 autoantibodies directed against the nicotinic acetylcholine receptor (AChR).A second group of patients ($5%) has predominant IgG4 autoantibodies against muscle-specific kinase (MuSK), a transmembrane tyrosine kinase receptor which is essential for initiation and maintenance of postsynaptic AChR clustering at the NMJ (DeChiara et al., 1996).MuSK IgG4 autoantibodies block the assembly of the agrin-LRP4-MuSK molecular machinery and induce disruption of postsynaptic AChR clustering, ultimately resulting in myasthenic muscle weakness (Borges & Richman, 2020;Klooster et al., 2012;Phillips & Vincent, 2016).Autoantibody titers roughly correlate with disease severity on a group level and within individual patients but are not precise indicators of disease severity in individual patients (Bartoccioni et al., 2006;Huijbers et al., 2016;Niks et al., 2008).In addition, not all skeletal muscles are equally affected.MuSK MG patients differ greatly in symptoms but typically present with (transient) ocular muscle weakness, progress with bulbar muscle weakness and often generalise thereafter.The inter-patient and inter-muscle group variations may indicate differences in sensitivity to MuSK autoantibody attack.Possibly, differences between individuals and muscle types in the ability to homeostatically compensate for the NMJ dysfunction may be underlying this clinical variation.
Synapses, including the NMJ, are dynamic structures that display plasticity upon changing conditions (Davis & Muller, 2015;Maynard et al., 2023).Hypothetically, several homeostatic mechanisms directly related to the process of synaptic transmission at the myasthenic NMJ could counteract the effect of reduced postsynaptic AChR density: (1) upregulation of presynaptic acetylcholine (ACh) release, (2) upregulation of AChR expression, (3) downregulation of acetylcholinesterase (the enzyme that breaks down ACh) or (4) alternative expression of muscle voltage-gated sodium channels, lowering the firing threshold.In the case of AChR MG, several compensatory mechanisms have indeed been shown at the myasthenic NMJ.Especially, there is increase of quantal content (i.e. the number of ACh quanta released per nerve impulse), which correlates with the severity of AChR density reduction at individual NMJs (for review, see Plomp, 2017).Furthermore, AChR subunit mRNA transcripts have been shown upregulated in muscles of severe MG patients (Guyon et al., 1994(Guyon et al., , 1998)), and downregulation of acetylcholinesterase mRNA was found in muscles of a MuSK MG mouse model (Punga et al., 2011).Na V 1.4 skeletal muscle membrane sodium channels are extra concentrated at NMJs and become reduced in density in AChR MG, because of local complement-mediated membrane damage (Ruff & Lennon, 1998).This causes an increase in the muscle fibre's action potential firing threshold (Ruff & Lennon, 2008).As far as we know, no compensatory changes at the level of Na V channels have been reported.
Previously, we passively immunised mice with purified IgG4 from plasmapheresis material of patients with MuSK MG (Klooster et al., 2012).These mice developed a myasthenic phenotype with regional differences in muscle weakness, and we showed electrophysiological NMJ defects resulting from a markedly reduced AChR density and cluster fragmentation (Klooster et al., 2012;Verschuuren et al., 2018).Interestingly, in our and other experimental MuSK MG models, a compensatory upregulation of MuSK and AChR α-subunit mRNA was seen but, in contrast to the situation in AChR MG, nervestimulation-evoked ACh release was not found to be increased (Klooster et al., 2012;Mori et al., 2012;Morsch et al., 2013;Punga et al., 2011;Viegas et al., 2012).This implies that compensatory responses in MuSK MG differ from those in AChR MG.In further (unpublished) exploratory ex vivo experiments with muscles exposed to our patient MuSK MG IgG4 fraction in a passive transfer model, we noted that the muscle contraction of some mice, particularly those with milder myasthenic phenotypes, had become partially insensitive to inhibition by μ-Conotoxin-GIIIB.This is a selective blocker of skeletal muscle Na V 1.4 voltage-gated sodium channels and a pharmacological tool used in NMJ electrophysiology analysis (Plomp et al., 1992).We hypothesised changes in Na V channel expression profile, possibly involving coexpression of (μ-Conotoxin-GIIIB insensitive) Na V 1.5 type channels.Such extra Na V channels at the NMJ may potentially lower the muscle fibre's action potential firing threshold, thereby facilitating neuromuscular synaptic transmission.Thus, the present study was undertaken to identify a potentially novel homeostatic response at NMJs challenged by MuSK autoantibodies.

| IgG purification
IgG4 was purified by U-protein express BV (Utrecht, The Netherlands) from two pooled batches of therapeutic plasmapheresis material obtained from a MuSK MG patient.The IgG4 was dialysed against phosphatebuffered saline (PBS), concentrated to 15.5 mg/mL with a Vivaspin20 concentrator (Sartorius), filter sterilised (Millipore, Amsterdam, The Netherlands) and stored in aliquots at À20 C.

| Mice
We used 2-4 months-old female NOD.CB17-Prkdcscid/J (Nod/scid) mice, which produce no IgG and lack a functional complement system because of C5 deficiency (Shultz et al., 1995).Original breeders were from Jackson Laboratory (Bar Harbor, ME, USA) (Stock #001303).Mice were housed in sterile, individually ventilated cages.Sterilised food and drinking water were provided ad libitum.All procedures involving the use of laboratory animals were performed in accordance with Dutch law and Leiden University guidelines, including approval by the National and Local Animal Experiments Committees.

| Passive transfer regimens
On the basis of our preliminary (unpublished) observation that especially muscles from the milder disease MuSK MG model mice showed indications of changes in Na V channel profile, we hypothesised that NMJs activate protective/ compensatory responses most optimally in a mild disease setting.Therefore, we decided to generate a MuSK MG model with a threshold dosing of MuSK MG IgG4, leading to a subclinical phenotype.We compared its neuromuscular functional aspects with that of conventionally, higher dosed mice with an overt myasthenic phenotype.An outline of the experimental setup is given in Figure 1a.Mice were daily injected intraperitoneally with .1 mg/g IgG4, that is, 'high-dosed' (HD; n = 12), .04mg/g, that is, 'intermediatedosed' (ID; n = 3) or .02mg/g, that is, 'low-dosed' (LD; n = 12).Control mice were injected with PBS, pH = 7.4 (n = 7).Five days prior to the start of injections (day À5 to day 0), the mice were daily trained to get used to handling and set a baseline for the muscle strength tests (see below).Starting from day 0, animals received a daily IgG4 injection, after completing muscle strength tests.Blood samples were collected according to the scheme in Figure 1a by puncture of the tail vein.After clotting at room temperature, serum was collected through centrifugation and stored at À20 C until further use.The mice were kept in the experiment until day 15 or were euthanised earlier if they reached the humane endpoint of losing >20% of their day 0 body weight.The investigator was blinded for the injected material and the dosing throughout all experiments and unblinded only after completing all further analyses.

| In vivo muscle strength tests
In vivo muscle strength and endurance of mice was assessed daily, as described before (Klooster et al., 2012).Forelimb/abdominal muscle strength was tested using a grip strength meter (type 303500, Technical and Scientific Equipment GmbH).The average force value was calculated from a session of 10 consecutive pulls with a few seconds pause in between.
To evaluate fatigability of limb and abdominal muscles, the inverted mesh hanging test was performed, as described previously (Kaja et al., 2007;Klooster et al., 2012).The test ended upon completing the maximum hanging time, which was set at 180 s, or otherwise after three attempts with pauses of a few seconds.

| Electromyography
To determine compound muscle action potential (CMAP) decrement, repetitive nerve stimulation electromyography under anaesthesia was performed in calf muscle at the endpoint day, as described previously (Kaja et al., 2007;Klooster et al., 2012;van der Pijl et al., 2016).After recordings, mice were euthanised by CO 2 inhalation; a blood sample was taken via vena cava puncture; and several muscles (diaphragm, epitrochleoanconeus (ETA), extensor digitorum longus (EDL), gastrocnemius and triceps) were dissected for ex vivo electrophysiology, contraction experiments, microscopical study and mRNA expression analysis, as described in the succeeding texts.

| Visual assessment of μ-Conotoxin-GIIIB resistant contraction
Directly after dissection, the most dorsal part ($25%) of the right hemidiaphragm was separated and used for NMJ histology and RNA analysis (see succeeding texts).The remaining part, with the innervating phrenic nerve, was used for assessment of a potential residual contraction of diaphragm muscle after 20 min μ-Conotoxin-GIIIB (3 μM; PeptaNova, Sandhausen, Germany) incubation.During .3Hz nerve stimulation, a visual scoring upon low magnification (5X objective) microscopic inspection was made (0 = no contraction, 1 = weak contraction in few fibres, 2 = weak contraction in larger part of the hemidiaphragm, 3 = clear contraction in many parts of hemidiaphragm.These residual muscle contractions were video recorded with a ThorLabs USB 3.0 CMOS camera.

| Ex vivo diaphragm muscle contraction
Contraction force of left hemidiaphragms was measured upon single and tetanic (40 Hz for 7 s) phrenic nerve stimulations, as described previously (Plomp et al., 2022).

| Ex vivo NMJ electrophysiology
Directly after the residual contraction assessment in right hemidiaphragms, with action potentials still blocked by μ-Conotoxin-GIIIB, micro-electrode recordings of miniature endplate potentials (MEPPs) and (upon .3 and 40-Hz nerve stimulation) endplate potentials (EPPs) were made at NMJs, as described previously (Klooster et al., 2012).NMJspecific postsynaptic membrane action potential firing threshold was determined in muscle fibres of left nervehemidiaphragms, as described previously (van der Pijl et al., 2016).From these recordings, we also determined the duration and height of muscle fibre action potentials.

| Immunohistology and morphometrical analysis
For AChR histological staining, a 3-mm-wide strip from the dorsal side of the right hemidiaphragm as well as the whole ETA muscle of the mice were used.The muscles were fixed in 1% paraformaldehyde in PBS at room temperature for $30 min.After three 10-min washes with PBS, α-bungarotoxin conjugated to Alexa Fluor 488 (AF488-BTx, 1 μg/mL; Invitrogen) was applied, and muscle preparations were incubated for 2 h at room temperature.Non-bound AF488-BTx was removed by six 10-min washes with PBS, after which specimens were mounted with Prolong Gold antifade reagent (Life technologies).The tissue was imaged under a Leica CTR5000 spectral confocal microscope (Leica Microsystems).AChR staining area was quantified using the thresholding feature of the Image J v1.48k programme (http://rsbweb.nih.gov/ij/) at 20 randomly chosen NMJs within maximal projections of z-stacks.
To establish the levels of Na V 1.4 and Na V 1.5 protein expression at NMJs, we performed immunohistology on cross sections of gastrocnemius and triceps muscles of controls (n = 6) and mice showing residual muscle contraction in LD (n = 6) and HD (n = 6).After dissection, muscles were embedded in OCT medium (Tissue-Tek) and immediately frozen in small embedding moulds (Sigma-Aldrich) using liquid nitrogen to preserve optimal skeletal muscle morphology.Muscles were then sectioned in the transversal plane (10-μm thickness) with a cryotome (Leica CM3050 S Research Cryostat), mounted onto SuperFrost Plus adhesion slides (Thermo Scientific) and stored at À20 C. Unspecific binding of antibodies was blocked using 5% normal goat serum (Millipore) for Na V 1.4 staining or 5% donkey serum for Na V 1.5 staining and .2%Triton X-100 (Sigma-Aldrich) in PBS for 2 h at room temperature.Monoclonal mouse anti-Na V 1.4 clone N255/38 (1:1000; Neuro-Mab) or polyclonal rabbit anti-Na V 1.5 (1:50; Abcam) were diluted in the appropriate blocking solution, and sections were incubated overnight at 4 C.After incubation, three 10-min washes with PBS were used to remove unbound antibodies.The sections were incubated with secondary antibodies, that is, either goat anti-mouse Alexa Fluor 488 (1:1000, Invitrogen) or Alexa Fluor 488-conjugated donkey anti-rabbit IgG (1∶1000; DAR-488, Invitrogen) and Alexa Fluor 594 α-BTx (1:500) for 2 h at room temperature.Sections were then washed in PBS and mounted with Prolong Gold antifade reagent, and imaged using a Leica DM5500 B microscope (Leica Microsystems).
For quantification of Na V 1.4 and Na V 1.5 signals, Image J software was used.First, TIFF images were converted to 8-bit pictures and thresholded by Li 0 s method to segment AChR area from background.Na V 1.4 and Na V 1.5 staining intensities were quantified within the thresholded AChR area.A background segment was selected and measured in the cytoplasm area in each picture.To calculate specific Na V staining intensity, the intensity of this background was subtracted from the intensity of Na V staining within the BTx-detected AChR area.

| qPCR analysis
Gene expression analysis was performed on a whole gastrocnemius and a dorsal strip of diaphragm dissected from LD and HD mice which showed residual muscle contractions in the ex vivo μ-Conotoxin-GIIIB sensitivity assessment, as well as from PBS-control mice.Muscles were snap frozen in liquid nitrogen-cooled isopentane directly after dissection.RNA was extracted with QIAzol lysis reagent (Qiagen) and miRNeasy Mini Kit (Qiagen) according to the manufacturer's instructions.RNA concentration was determined using a Nanodrop device (Thermo Scientific).First strand cDNA was synthesised from total RNA using RevertAid H Minus First Strand cDNA Synthesis kit (Thermo Fisher Scientific).For realtime qPCR, all samples were run in triplicate, and Gapdh was used as housekeeping gene.Technical replicates that differed >.5 in Cq from the others in the triplicate were excluded.The reactions were carried out with iQ SYBR Green Supermix (Bio-Rad) and a CFX384 Touch Real-Time PCR Detection System (Bio-Rad).Relative expression data were calculated with the 2 -ΔΔCT method.Primers are listed in Table 1.

| Statistical analyses
Data are presented as mean ± standard error (S.E.M.), with n representing the number of animals per group.Student's t-test was used to assess statistically significant differences between two experimental groups, and multiple unpaired t-tests or one-way ANOVA with Tukey's post hoc tests were performed to assess statistically significant differences between more experimental groups.Statistical significance was calculated with GraphPad Prism 9.3.1 software (San Diego, CA, United States).Values of p < .05were considered statistically significant.

| In vivo neuromuscular analyses of the different MuSK MG IgG4 mouse model dosing groups show clear myasthenia in HD but not in LD mice
In vivo analysis revealed a dose-dependent development of a myasthenic phenotype in the different MuSK MG IgG4 dose groups (Figure 1).HD mice showed a clear, progressive decrease in body weight, starting around day 6-7 after the start of the daily injections, reaching the humane endpoint ($20% loss) by day 10 (n = 12, Figure 1b).This replicates the body weight time course in earlier passive transfer experiments with the same dosing regimen with this IgG4 material (Huijbers et al., 2019).ID mice (n = 3) showed body weight loss, too.However, it progressed more slowly, starting from day 10 after the onset of injections, reaching the humane endpoint around day 13-15.LD (n = 12) and PBS-control (n = 7) mice did not show body weight loss.In HD and ID mice, pulling force in the forelimb grip strength test showed progressive decline, starting around days 6 and 9, respectively, reaching $80% loss at their respective endpoint days (Figure 1c).LD mice started to show a tendency of loss of strength from day 7 onwards.The maximum hanging time in the inverted mesh test became progressively reduced in HD and ID mice, starting around day 6, dropping to 18.7 ± 3.2 and 13.3 ± 3.0 s, respectively, on their day of endpoint (Figure 1d).Only two of the 12 LD mice did not reach the maximum hanging time of 180 s anymore, starting from days 10-11, leading to a tendency of decrease in the mean group values.All PBScontrol mice reached 180 s on all days.Thus, the myasthenia level of the LD mice seemed on the brink of being detectable in these in vivo tests.

| Disproportional difference in serum concentrations of MuSK IgG4 in HD and LD mice
To assess antibody exposure, MuSK MG IgG4 antibody titers were determined in serum samples of the mice throughout the experiment.MuSK IgG4 in serum of the groups accumulated to a plateau level reached around day 8 (Figure 1e).HD, ID and LD mouse endpoint serum contained 3740 ± 352, 1043 ± 71 and 320 ± 27 μg/mL, respectively.Thus, LD mouse serum levels were $12 times lower than HD mouse levels, whereas the injected daily dose of MuSK IgG4 was only 5 times lower (i.e..02 vs. .1 mg/g per day).This might suggest that in LD mice, a relatively high proportion of the total intraperitoneally T A B L E 1 qPCR primer pairs for Na V channels Scn4a, Scn5a, AChR α1, ε, and γ subunits, atrophy gene Trim63 and the housekeeping gene Gapdh.

| Repetitive nerve stimulation electromyography shows clear myasthenia in HD and ID mice
To evaluate myasthenic NMJ transmission failure, repetitive nerve stimulation electromyography in calf muscle was performed.Of the 12 HD mice, seven died shortly after administration of the anaesthetics, presumably from some degree of respiratory depression in combination with the severity of the myasthenia.Although this complication precluded the electromyographic measurement of CMAPs, muscles were quickly dissected and included in further analyses.Anaesthesia-related death did not happen in any of the tested ID, LD and PBS-control mice.
The five surviving HD mice, as well as all the ID mice, showed a clear decrement ($20%) of the CMAP amplitude during the 20 nerve stimulations at 40 Hz.LD and PBS-control mice did not show any CMAP decrement (Figure 1f).

| Ex vivo muscle and NMJ analyses
The in vivo analyses showed that HD and ID mice had comparable, overt myasthenic features, albeit with a slightly slower onset and progression for the ID group.LD mice showed hardly any myasthenic phenotype.We therefore decided to study detailed ex vivo NMJ parameters related to possible compensatory mechanisms only in HD and LD mice, and compare them with PBScontrols.

| Diaphragm muscles of LD mice show residual contraction after μ-Conotoxin-GIIIB incubation
μ-Conotoxin-GIIIB selectively blocks skeletal muscle Na V 1.4 channels, thereby inhibiting muscle contractions and thus allowing for undisturbed micro-electrode recording of EPPs during phrenic nerve stimulation (Plomp et al., 1995).In the PBS-control group, μ-Conotoxin-GIIIB completely blocked ex vivo muscle contraction in hemidiaphragm upon .3Hz nerve stimulation (all scored 0 in the visual assessment, n = 7; Figure 2).In contrast, the great majority of LD mice hemidiaphragms clearly showed residual muscle contractions.Only two of the 12 muscles showed complete inhibition (score 0); the rest scored 1 (5/12), 2 (4/12) or 3 (1/12), according to our scale for contraction responses (Figure 2).In hemidiaphragms of HD mice, such pronounced residual contractions were not observed.Only three of the 12 muscles showed minor residual contraction (score 1), whereas nine showed normal μ-Conotoxin-GIIIB sensitivity (score 0).See supporting information video for video recordings of these residual contractions.

| Quantification reveals only small residual contraction force after μ-Conotoxin-GIIIB incubation
In ex vivo muscle contraction experiments, we quantified the force delivered by hemidiaphragm muscles upon single and tetanic (40 Hz for 7 s) phrenic nerve stimulations.Twitch contractions did not differ among LD, HD and PBS-control groups, either in amplitude, half-width or area-under-the-curve (Figure 3a-d).Tetanic contraction of muscles from LD and HD mice showed comparable myasthenic features, that is, a clear tetanic fade in combination with lower peak amplitude and 40-44% reductions of the area-under-the-curve, as compared to the muscles of PBS-control mice (Figure 3d,e).We also attempted to quantify the visually observed residual contraction in diaphragms of LD mice in the presence of 3 μM μ-Conotoxin-GIIIB.However, after 20-30 min toxin incubation, there was no detectable muscle contraction in all groups (Figure 3e).Apparently, the visually observed movement of the LD mice muscles (Figure 2 and supporting information video) in fact represented only a very low contraction force that remained under the detection limit of the ex vivo contraction measurement system.

| NMJ electrophysiology reveals similar myasthenic levels at LD and HD diaphragms
Surprisingly, in view of the clearly different in vivo myasthenia levels of LD and HD mice, ex vivo microelectrode recordings at NMJs of their diaphragms showed a similar severity of myasthenic features, all in the range as reported earlier with this and other patient IgG4s (Klooster et al., 2012;Huijbers et al., 2019).We found equally reduced amplitudes and frequencies of MEPPs, and equal reductions of EPP amplitudes in diaphragm (all by$40-50%, as compared to the PBS-control, Figure 4a,c,d).Interestingly, the resting membrane potential showed a tendency of a few mV depolarisation, statistically significant only in diaphragm muscle fibres of HD mice (Figure 4e).The quantal content (i.e. the number of ACh quanta released per nerve impulse) in NMJs of HD mice was decreased by $22%, compared to PBS controls (PBS controls 41.10 ± 2.65; LD 35.72 ± 1.28 and HD 31.90 ± 1.59; Figure 4g).The rundown of EPP amplitudes during high-frequency (40 Hz) nerve stimulation was similar at NMJs of LD and HD mice (both by $40%), which was more pronounced than the rundown observed in PBS controls (Figure 4h,i).Thus, diaphragm NMJ electrophysiology of LD and HD mice showed a comparable myasthenic level, in spite of LD mice receiving 5 times lower MuSK MG IgG4 doses and showing no outspoken in vivo myasthenia.An important factor in this discrepancy may be the intraperitoneal injection route, which might have caused a relatively high MuSK binding at diaphragm NMJs because of the direct diffusional exposure after the injection.To investigate this further, we recorded MEPPs in EDL muscles from a limited number of mice.EDL is a forelimb muscle which is only indirectly exposed, via the bloodstream, to the intraperitoneally injected IgG4.Indeed, EDL NMJs from LD mice showed a milder myasthenic electrophysiological phenotype than EDL NMJs from HD mice.MEPP amplitudes were .55± .07,.45± .05 and .27± .02mV in PBS control, LD and HD NMJs, respectively (Figure 4b).In EDL muscle fibres, the resting membrane potential was similar for LD, HD and PBS controls (Figure 4f).

| No change in the firing thresholds of muscle fibres
Differences in Na V density/composition at the postsynaptic NMJ might lead to changes in firing threshold or  3d) and reductions of 44% (LD) and 40% (HD) of the (e) area-under-the-curve, as compared to the muscles of PBS-control mice.No detectable muscle contraction in neither group after 3 μM μ-Conotoxin-GIIIB incubation (Figure 3e).Bars represent mean ± SEM.N = 4-6 mice per group.kinetics of muscle fibre action potentials.However, firing thresholds determined in fibres from LD, HD and PBScontrol mice were similar, that the maximal amplitude of EPPs that just did not trigger an action potential in this assessment procedure was $8.6 mV in all groups (Figure 5a,b).Action potentials were somewhat reduced in amplitude and prolonged in duration in muscle fibres of HD mice compared with LD mice, and showed a tendency of increased rise times (Figure 5c-h).

| Histology shows clear reduction of AChRs at LD and HD mouse NMJs without clear changes in Na V expression
AChR staining of whole-mount diaphragm and EDL NMJs corroborated the electrophysiological results.Diaphragm NMJs from LD and HD mice showed comparable reductions ($80%) of the total AF488-BTx signal (i.e.staining area multiplied by intensity), compared to that of PBS controls (Figure 6a,b).In contrast, the reduction at EDL NMJs from LD mice ($15%, not statistically significant) was much less outspoken than that at EDL NMJs from HD mice ($70%, Figure 6c).
Staining of gastrocnemius and triceps cross sections for NMJ-localised Na V 1.4 channels did not reveal significant changes in immunoreactivity in muscles isolated from LD and HD mice, compared to PBS controls (Figures 7a,b and 8a,b).Na V 1.5 channels were detected only at very low density at NMJs in both muscles, with a slight tendency of an increase at triceps NMJs of both LD and HD mice, compared to PBS control (Figures 7a,c and  8a,c).NMJ area, as defined by AChR staining area, was reduced in HD triceps but not in HD gastrocnemius, compared to PBS control (Figure 7d and 8d).Furthermore, NMJ AChR intensities were decreased in HD and, to a lesser extent, LD gastrocnemius and triceps (Figure 7e and 8e).

gene expression in diaphragm muscle
To evaluate whether the observed partial μ-Conotoxin-GIIIB insensitivity in LD mouse diaphragm was caused by a shift in Na gene expression as a possible result of (partial) denervation and atrophy, we evaluated with RT-qPCR the gene expression level of AChR subunits, the atrophy marker muscle-specific RING finger protein 1 (MuRF-1, referred to as tripartite motif-containing 63 [Trim63] here), and Scn4a and Scn5a, encoding Na V 1.4 and Na V 1.5, respectively.The level of Chrna was clearly upregulated in LD and HD when compared with PBS controls.Interestingly, the fetal AChR γ subunit gene Chrng was increased up to 3-fold in LD and 1.5-fold in HD when compared with PBS controls, whereas the level of adult AChR ε subunit gene Chrne was downregulated in HD diaphragm.Trim63 level was markedly upregulated ($4-fold) in HD, compared to PBS controls (Figure 9a).
In gastrocnemius, the level of Chrna was upregulated ($2.2 fold) only in HD mice (Figure 9b).In addition, we observed increases in expression of Chrng and Trim63 in gastrocnemius of HD mice only.All these observations match with previous reports describing increased Chrng, Trim63 and trends to increased Chrna expression in MuSK MG mice (Punga et al., 2011).
No changes in Scn4a and Scn5a gene expression were observed in either diaphragm or gastrocnemius (Figure 9a,b).This is in line with the unchanged NMJ immunostaining, as described above.

| DISCUSSION
Changes in Na V density, electrophysiological characteristics, or isotype expression that enhance/modulate the excitability of skeletal muscle fibres may in principle counteract the synaptic transmission failure at myasthenic NMJs.In the present study, we characterised such a possible compensatory response in a passive transfer MuSK MG IgG4 mouse model.
We used a previously developed passive transfer model for MuSK MG, which showed patient-specific IgG4 potency-dependency as well as a dose-dependent course of the disease (Huijbers et al., 2019;Klooster et al., 2012).In these and other (unpublished) studies, we observed partial resistance of ex vivo muscle contraction to Na V 1.4 blocker μ-Conotoxin-GIIIB and got the impression that it was more outspoken in muscles from mice that had shown a relatively mild myasthenic phenotype in vivo.Therefore, we formed several dosing groups (LD, ID and HD) in our current study to systematically investigate this matter.Indeed, visual scoring of the nerve stimulation-evoked contraction of diaphragm muscles from LD mice (i.e.without an overt myasthenic phenotype) revealed more pronounced μ-Conotoxin-GIIIB resistance, as compared to diaphragms from HD mice which had shown a clear myasthenic in vivo phenotype.On the other hand, the magnitude of this residual contraction was only small; it remained under the detection limit of our whole hemidiaphragm contraction measuring system.Apparently, only few fibres or only small (e.g.perisynaptic) parts of fibres contributed to the visually detected residual contraction.
Our subsequent analyses of Na V channels at functional, morphological and genetic level did not reveal clear differences between LD and HD diaphragms.Importantly, we found no difference in muscle fibre firing threshold between LD and HD groups.This indicates that the specific change in Na V density or characteristics underlying the observed μ-Conotoxin-GIIIB resistance is not a compensatory mechanism that facilitates neuromuscular transmission by lowering the firing threshold.Interestingly, firing thresholds of LD and HD fibres were found similar to those of PBS controls.This differs from the situation in AChR MG, where there is a clearly higher threshold compared to healthy controls, caused by removal of Na V 1.4-containing postsynaptic membrane because of complement activation (Ruff & Lennon, 1998, 2008).The absence of complement activation (because of the use of IgG4 and Nod/scid mice) in our MuSK MG IgG4 model most likely explains the absence of a change in firing threshold but may to some extent differ from the human situation.This finding of unchanged firing thresholds was corroborated by our histological analyses which showed no clear loss of synaptic Na V 1.4 channels at either LD or HD MuSK MG IgG4 NMJs.The unchanged firing thresholds furthermore suggest unchanged activation/gating properties of the Na V 1.4 channels.In addition, rise times and amplitudes of action potentials, indicators of Na V 1.4 behaviour, did not differ between PBS control and LD fibres.However, our results do not exclude the (perhaps more distant) possibility that firing thresholds may increase during long-term intense activity, because of use-dependent behavioural changes of Na V channels in the MuSK MG muscle fibres.
A well-known effect of muscle inactivity after denervation is upregulation of Na V 1.5 channels, especially at synaptic regions, without alterations in Na V 1.4 level (Awad et al., 2001;Goldin, 1999).The muscle weakness induced by MuSK MG IgG4 might trigger this phenomenon too, especially in fibres with severely dysfunctional NMJs.Na V 1.5 channels are insensitive to μ-Conotoxin-GIIIB and could thus underlie the residual contractions we observed visually after Na V 1.4 inhibition.Unfortunately, no specific Na V 1.5 inhibitors exist that could be used to test this in ex vivo preparations.Alternatively, we immunostained muscle sections for Na V 1.5 channels at the NMJ.Although we detected very low-level expression, no clear upregulation was found in LD and HD muscles.Together with the observed lack of scna5 upregulation in our gene expression analysis, this indicates that residual contractions most likely do not involve Na V 1.5.Thus, our results do not support the hypothesis that changes at the level of Na V 1.5 channels form a compensatory mechanism in MuSK MG NMJs to counteract the suboptimal synaptic transmission.It remains to be seen what the exact mechanism is which underlies the observed residual contraction after block of Na V 1.4 by μ-Conotoxin-GIIIB.It remains unclear if how and which Na V channels are involved, and how it relates to the myasthenia induced by the MuSK MG IgG4.The lack of well-established selective Na V 1.5 blockers limits these studies.The clear demonstration that the phenomenon is more outspoken in LD mice is intriguing.Perhaps, the effects of more pronounced myasthenia in HD mice in some way causes reversal or repression of this phenomenon.One factor is that LD mice remained in the in vivo experiment for 4-5 days longer than HD mice, which had their ex vivo analyses around day#10, forced by reaching the humane endpoint.It might thus be possible that the Na V repertoire change needs some more days and developed over this period.
A matter which complicates the interpretations and comparisons between the LD and HD MuSK MG IgG4 mice is that diaphragm muscles most likely were disproportionally exposed to the intraperitoneally injected IgG4.The concentration of IgG4 in LD serum at plateau level was 12 times lower than in HD serum while the dosing was only 5 times lower.Although this may suggest a saturable compartment of systemic binding sites (MuSK anywhere and perhaps unspecific binding sites), it may also suggest that NMJs at diaphragm (and muscles of abdominal cavity) absorbed a large proportion of the MuSK MG IgG4 before entering the systemic circulation.This was confirmed by our ex vivo synaptic electrophysiological analysis and the contraction experiments, which showed almost equal myasthenic levels in LD and HD diaphragm muscles, whereas myasthenic features at NMJs of EDL, that is, a distant muscle only exposed to IgG4 via the bloodstream, were clearly less outspoken in LD mice.The upregulation of Chrna1 and Chrng in LD diaphragm but not LD gastrocnemius is in line with this.
On the other hand, the absence of an increase in atrophy marker Trim63 in both diaphragm and gastrocnemius of LD mice, while being upregulated in both these muscles of HD mice, is more difficult to reconcile.Furthermore, a relatively high degree of denervation in HD as compared to LD diaphragm might have biased our electrophysiological results.However, we did not observe spontaneous fibrillations, and 'silent' NMJs were not encountered during the micro-electrode impalements of fibres for MEPP/ EPP measurement.This suggests that if functional denervation was present, it was not of large scale.Disproportionally affected diaphragm muscle because of the intraperitoneal injection route of IgG4 in the MuSK MG IgG4 model was already suggested and discussed to some extent by us previously (Huijbers et al., 2019).Older studies comparing differential tissue distribution of antibodies after either i.p. or i.v.administration route also showed relatively high levels in diaphragm muscle after i.p. injection (Flessner & Dedrick, 1998;Wahl et al., 1988).Peritoneal fluid may preferentially enter the diaphragm via specialised stomata (Abu-Hijleh et al., 1995).In the light of this, it might be interesting to perform a more comprehensive ex vivo physiological study, including EPP analyses and investigation of μ-Conotoxin-GIIIB resistance, at NMJs of a distal, non-F I G U R E 9 Gene expression analysis shows no changes in Na V 1.4 and Na V 1.5 gene expression.(a) Diaphragm and (b) gastrocnemius muscles, isolated from low-dosed (LD), high-dosed (HD) and PBS-control (Ctrl) mice.Expression levels of genes of interest were normalised to the PBScontrol group and the housekeeping gene Gapdh.Data represent means ± SEM.N = 4-6 mice per group.Statistical significance was tested with one-way ANOVA.diaphragm muscle.However, such studies are technically more difficult than in diaphragm.
Further studies might also include exploration of the role of ClC-1.These Cl À ion channels are present in high density at the skeletal muscle fibre membrane and are important determinants of the resting membrane potential (Pedersen et al., 2016).They might (indirectly) affect Na V channel expression and/or behaviour.Possibly, the minor resting membrane potential changes observed in diaphragm muscle fibres in the current study might be the result of ClC-1 changes.Furthermore, it might explain the observed slight reduction in action potential amplitude in HD mouse muscle fibres.Additionally, different splice variants or post-translational modifications of the Na V channels might influence channel expression, interactions or behaviour in MuSK MG muscle fibres.Although Na V 1.4 splice variants have not been described in mice, these are known for Na V 1.5 (Kerr et al., 2004;Loussouarn et al., 2015;Schroeter et al., 2010;Wang et al., 2017).Other potentially relevant factors that may be interesting to study are ankyrins.These are scaffolding proteins that are responsible for clustering of Na V channels at neuronal axonal nodes of Ranvier and, importantly, in muscle play a crucial role in Na V 1.4 clustering at the postsynaptic NMJ membrane (Zhang et al., 2021).More distant/speculative options would be expression of neuronal type Na V channels (i.e.non-1.4 or À1.5), or perhaps some type of voltage-gated Ca 2+ channels which would underlie the (action) potentials that cause residual contractions after μ-Conotoxin-GIIIB block of Na V 1.4.

| CONCLUSION
We exposed a phenomenon of Na V 1.4-independent muscle contraction ex vivo, preferentially in mice experiencing subclinical MuSK MG.We show this is unlikely caused by a compensatory upregulation of Na V 1.5 or other factors that could potentially lower the muscle fibre firing threshold.The exact nature of the Na V repertoire change and whether it is taking place at all muscles remains to be establish, just as whether it is compensatory beneficial for the suboptimal neuromuscular transmission at the myasthenic NMJ.

F
I G U R E 1 In vivo analysis and clinical characteristics of MuSK MG in the three different dose groups of MuSK MG IgG4 treated Nod/scid mice.Experimental scheme of the induction of (a) MuSK myasthenia gravis.High-dosed (HD, n = 12) and intermediate-dosed (ID, n = 3) mice clearly showed progressive loss of (b) body weight, (c) grip strength and (d) hanging time in the inverted mesh test, whereas PBS control (ctrl) mice remained constant (n = 7).Low-dosed (LD, n = 12) mice did not show significant reductions.(e) The anti-MuSK IgG4 concentration in serum increased to plateau values reached after $8 days and was $12 fold higher in HD serum, as compared to LD. (f) Repetitive nerve stimulation electromyography at the endpoint day showed a decrement of CMAP amplitude during 40-Hz nerve stimulation in the ID and HD groups.Data represents mean ± SEM.Statistical significance was tested with one-way ANOVA.

F
I G U R E 2 Visual assessment of residual muscle contraction after incubation with μ-Conotoxin-GIIIB in ex vivo diaphragm.A blinded observer scoring of muscle contractility: 0 = normal sensitivity (i.e.complete paralysis after 15 min incubation with 3 μM μ-Conotoxin-GIIIB), 1 = very little contraction left; only few fibres contract, 2 = more widespread contraction left, 3 = vigorous contraction left.Y-axis indicates percentage of mice, absolute numbers of mice noted in the bars.X-axis represents visual scores from 0 to 3.

F
I G U R E 3 Ex vivo diaphragm muscle contraction.Quantification of the force upon single and tetanic (40 Hz for 7 s) nerve stimulations.Twitch contractions were similar in LD, HD and PBS-control groups, either in (a) amplitude, (b) half-width or (c) area-under-the-curve.(d) Twitch contraction profile examples.Muscles from LD and HD mice clearly showed tetanic fade and lower peak amplitudes (Figure

F
I G U R E 4 NMJ electrophysiology.Micro-electrode recordings at NMJs from PBS-controls (Ctrl), low-dosed (LD) and high-dosed (HD) MuSK MG IgG4 mice.(a) MEPP amplitude in diaphragm NMJs was equally reduced in LD and HD NMJs, compared with Ctrl.(b) MEPP amplitude in HD EDL was reduced, compared with LD.(c) MEPP frequency in LD and HD diaphragm was clearly reduced, compared with Ctrl.(d) EPP amplitude during .3-Hznerve stimulation was similarly decreased in LD and HD, compared with Ctrl.(e) Diaphragm LD and HD muscle fibres were slightly depolarised, compared with Ctrl.(f) Resting membrane potential in LD and HD EDL muscle fibres was unchanged, in LD, HD mice compared with PBS controls.(g) Quantal content at .3-Hz nerve stimulation was decreased in HD mice.(h) Pronounced EPP rundown at 40-Hz nerve stimulation in LD and HD mice.(i) The mean amplitude of the 21st-35th EPP presented as percentage of the first EPP in the train.Bars represent mean ± SEM.Statistical significance was calculated using oneway ANOVA.

F
I G U R E 5 Analysis of nerve stimulation-evoked muscle fibre action potential (AP) firing threshold and AP basic parameters in diaphragms from PBS-control (Ctrl), low dose (LD) and high dose (HD) MuSK MG IgG4 mice.(a) Representative example of a justsubthreshold EPP and a delayed muscle action potential recorded in the presence of d-tubocurarine in response to phrenic nerve stimulation at variable frequencies (.2-20 Hz).(b) Unchanged AP firing threshold in muscle fibres from LD and HD mice.(c-e) Representative examples of muscle fibre APs recorded.(f) AP peak amplitude was somewhat reduced in HD fibres and (g) slightly increased in duration.(h) Tendency for increased AP rise time in LD muscle fibres.Arrows in Figure 5a,c,d, e indicate nerve stimulation artifacts.Bars represent mean ± SEM.N = 4-6 mice per group.Statistical significance was calculated using one-way ANOVA.

F
I G U R E 6 Reduction of postsynaptic neuromuscular junction (NMJ) area in diaphragm and EDL in low-dosed (LD) and high-dosed (HD) MuSK MG IgG4 treated mice.(a) Representative pictures of AF488-BTx stained AChRs at NMJs in diaphragm and EDL.(b) Total AChR intensity quantification (i.e.area Â intensity) was severely reduced to an equal extent in LD and HD diaphragm NMJs, compared to PBS controls.(c) In EDL NMJs, HD AChR intensity was more reduced than in LD NMJs.Data represents mean ± SEM.N = 4-6 mice per group.Statistical significance was tested with one-way ANOVA.
4 and Na V 1.5 immunostaining of neuromuscular junctions (NMJs) in gastrocnemius cross sections.Representative images from cross sections of PBS-controls, LD and HD mice, immunostained for (a, upper panel) Na V 1.4 and (a, lower panel) Na V 1.5.AChRs were labelled with AF488-BTx.Merge images indicate overlap of Na V 1.4 and Na V 1.5 with AChR area.(b) Na V 1.4 and (c) Na V 1.5 immunofluorescence intensities in LD and HD gastrocnemius NMJs did not differ statistically significantly from PBS controls.(d) AChR area and (e) staining intensity were reduced in NMJs from HD mice.The values are means ± SEM.N = 6 mice per group.Statistical significance was tested with one-way ANOVA.
4 and Na V 1.5 immunostaining of neuromuscular junctions (NMJs) in triceps cross sections.Representative images from cross sections of PBS-controls, LD and HD mice, immunostained for (a, upper panel) Na V 1.4 and (a, lower panel) Na V 1.5 in PBS-controls, LD and HD mice.AChRs were labeled with AF488-BTx.Merge images indicate overlap of Na V 1.4 or Na V 1.5 with AChR area.The few bright fluorescent spots in LD and HD Na V 1.5 staining are considered artefacts.(b) Na V 1.4 and (c) Na V 1.5 fluorescence intensities in LD and HD NMJs did not differ statistically significantly from PBS controls.(d) AChR area was similar in all groups.(e) AChR staining intensity was reduced in NMJs from LD and HD mice.The values are means ± SEM.N = 6 mice per group.Statistical significance was tested with one-way ANOVA.