The endogenous calpain inhibitor calpastatin attenuates axon degeneration in murine Guillain‐Barré syndrome

Abstract Axon degeneration accounts for the poor clinical outcome in Guillain‐Barré syndrome (GBS), yet no treatments target this key pathogenic stage. Animal models demonstrate anti‐ganglioside antibodies (AGAb) induce axolemmal complement pore formation through which calcium flux activates the intra‐axonal calcium‐dependent proteases, calpains. We previously showed protection of axonal components using soluble calpain inhibitors in ex vivo GBS mouse models, and herein, we assess the potential of axonally‐restricted calpain inhibition as a neuroprotective therapy operating in vivo. Using transgenic mice that over‐express the endogenous human calpain inhibitor calpastatin (hCAST) neuronally, we assessed distal motor nerve integrity in our established GBS models. We induced immune‐mediated injury with monoclonal AGAb plus a source of human complement. The calpain substrates neurofilament and AnkyrinG, nerve structural proteins, were assessed by immunolabelling and in the case of neurofilament, by single‐molecule arrays (Simoa). As the distal intramuscular portion of the phrenic nerve is prominently targeted in our in vivo model, respiratory function was assessed by whole‐body plethysmography as the functional output in the acute and extended models. hCAST expression protects distal nerve structural integrity both ex and in vivo, as shown by attenuation of neurofilament breakdown by immunolabelling and Simoa. In an extended in vivo model, while mice still initially undergo respiratory distress owing to acute conduction failure, the recovery phase was accelerated by hCAST expression. Axonal calpain inhibition can protect the axonal integrity of the nerve in an in vivo GBS paradigm and hasten recovery. These studies reinforce the strong justification for developing further animal and human clinical studies using exogenous calpain inhibitors.


| INTRODUCTION
Guillain-Barré syndrome (GBS) is an autoimmune inflammatory peripheral neuropathy associated with varying degrees of paralysis and recovery. 1 A poor long-term prognosis, principally manifested by incomplete motor recovery, is due to irreversible axon degeneration, for which there is currently no treatment. In the axonal variant of GBS, acute motor axonal neuropathy (AMAN), the motor axon is directly injured leading to acute conduction failure. Clinical outcome in AMAN segregates into two extremes: reversible conduction failure (RCF) with rapid recovery or primary axon degeneration with poor recovery, although in many individual cases this dichotomisation may be mixed across nerve territories and along individual axons. RCF is understood to be a consequence of axonal conduction block at the nodes of Ranvier (NoR) and distal motor nerves in the absence of axon transection and degeneration. 2,3 In contrast, axon degeneration may be extensively distributed along the axon and, when proximally sited in the nerve roots, is associated with regeneration failure, permanent denervation and disability. 4 The factors that dictate axonal stability and survival are currently unknown. In this regard, the concept of the "metastable state" 5 usefully describes the tipping point beyond which an injured axon is incapable of local repair and recovery and thus undergoes the process of axon transection with ensuing Wallerian degeneration of the distal fragment. Shifting the metastable state in favour of local repair provides a useful hypothetical framework for considering axon-protective therapeutic interventions that could improve outcomes.
Mechanisms underlying the pathogenesis of AMAN have been revealed through human autopsy studies and in animal models that attempt to recapitulate human disease pathology. [6][7][8][9][10] Clearly, many complex factors need to be considered when interpreting both human and animal pathology. 11 With this caveat, the prevailing view of AMAN pathogenesis is a disease mediated by complement fixing anti-ganglioside antibodies (AGAb) induced by a preceding infection (generated by molecular mimicry in the case of Campylobacter jejuni infections) that bind to the peripheral nerves where gangliosides are highly enriched. 1 Human tissue, and rabbit and mouse models of AMAN have shown complement deposition over the motor nerve terminals (MNT) and both proximal and distal motor axonal NoR. Activation of the complement cascade culminates in the formation of the membrane attack complex (MAC) pore, leading to bi-directional movement of water and ions, including Ca 2+ ions, 12 which coincides with pre-synaptic and nodal dysfunction and axonal conduction block. 8,9,13,14 From these models, it has been shown that at the MNT and NoR, neurofilament immunostaining, an indicator of disruption to axonal integrity preceding axon degeneration, is lost following AGAb complement-mediated injury. Nodal and paranodal disturbances also include mislocalisation of voltage-gated sodium (Nav) channels, cytoskeletal anchoring proteins and cell-adhesion molecules, indicating catastrophic disruption to the nodal architecture. These effects have been successfully attenuated in animal models through classical pathway complement inhibitors 8,10,15,16 and multiple clinical trials in humans are underway. [17][18][19] The calcium-dependent proteases calpain I (μ-calpain) and II (m-calpain) are expressed in axons and selectively cleave proteins in response to physiological (micromolar, μ) or pathogenic (millimolar, m) calcium ion levels, respectively. Calpain I has a role in limited and specific proteolysis for regulatory processes, while calpain II has a role in digestive or pathologic processes. Calpain II has been implicated in neurodegeneration and axon disintegration in a variety of diseases and injury models; through the use of calpain inhibitors or calpainmodulated transgenic mice, a role for calpains in Wallerian degeneration, Taxol-induced sensory neuropathy and traumatic brain injury (TBI) has been revealed. [20][21][22][23] As we have shown a loss of known calpain substrates neurofilament, AnkyrinG and Nav channels [24][25][26] in our AMAN model, we proposed that calpain activation is one major mechanism underlying axon degeneration in AMAN. We previously showed in our ex vivo mouse model, through the exogenous application of soluble calpain inhibitors, that the observed structural disturbances are mediated by activation of calpain due to the influx of Ca 2+ ions through MAC pores. 8,12,27 To explore the benefit of calpain inhibition in vivo, we assessed axonal protection using a transgenic calpastatin over-expression paradigm. Calpastatin is a calcium-dependent endogenous inhibitor specific for calpains 28 which are widely expressed in the mammalian nervous system. Transgenic mice that over-express the gene encoding human calpastatin (hCAST), show high expression in CNS and PNS tissue homogenates compared to wild type (WT) littermates, and by immunohistochemistry display strong calpastatin expression in neurons and axons. 20,21 Therefore, we used these mice to assess axonal structural and functional recovery in acute and extended sub-acute AMAN mouse models.

| Materials
Mouse monoclonal IgG3 anti-GD1b ganglioside antibody MOG1 and anti-GM1 ganglioside antibody DG2, (AGAbs), were used for all experiments, as previously described and justified. [29][30][31] Briefly, AGAbs were generated by immunising ganglioside-deficient mice with ganglioside liposomes or ganglioside-mimicking Campylobacter jejuni lipo-oligosaccharide. Both AGAb bind strongly to neuronal tissue in mice, including distal motor nerves, without any notable binding to glial cells as described previously. 16,32 Normal human serum (NHS) was added as a source of complement to induce injury through AGAb-mediated complement fixation. NHS was collected from a single donor, rapidly frozen and stored in multiple aliquots at À70 C to preserve complement activity.
The following antibodies were used for immunostaining studies to identify proteins and complement: mouse anti-AnkG (Thermo

| Mice
Three strains of mice were used: hCAST, GalNAc-T À/À -Tg(neuronal) (henceforth referred to as Neuronal) and hCAST Â Neuronal. hCAST and Neuronal transgenic mice have previously been described. Briefly, hCAST mice express human calpastatin (hCAST) cDNA driven by the mouse prion protein (Prp) promoter, resulting in axonal expression throughout the nervous system. 21 Calpastatin is a calcium-dependent endogenous inhibitor specific for calpains. 28 Neuronal mice express the full-length cDNA encoding GalNAc-T under the control of the human Thy1.2 promoter, restricting complex ganglioside expression (including GM1) to mature neurons. 16 All mice were bred on a C57BL/6J background (Harlan, UK) and backcrossed for 7 generations. hCAST negative littermates were used as WT controls.
hCAST were maintained as heterozygotes by breeding WT females with male hCAST heterozygotes. Male and female mice, ranging from 10 to 20 g, were used at 4 to 6 weeks of age, and no notable differences between genders were observed. The number of mice per treatment is reported per experiment. Mice were maintained under a 12 h light/dark cycle in controlled temperature and humidity with ad libitum access to food and water. For each study, mice were killed by rising CO 2 inhalation.
All procedures were conducted in accordance with a licence approved and granted by the United Kingdom Home Office (POC6B3485).

| Injury
Ex vivo triangularis sterni (TS) nerve-muscle preparations were used as described previously. 14 Briefly, TS from each mouse provide two preparations treated as follows: "Control" (AGAb only), "Injury" (AGAb plus NHS). TS were incubated for 4 h at 32 C in Ringer's with 100 μg/mL AGAb (MOG1), with the addition of 40% NHS for the injured preparation. TS were washed 3x in Ringer's followed by fixation with 4% PFA at 4 C for 20 min. Washes with PBS, 0.1 M glycine and PBS followed. TS were transferred to 100% EtOH for 10 min at À20 C then thoroughly washed in PBS. Tissue was incubated overnight at 4 C in blocking solution plus primary antibodies. TS were rinsed 3x in PBS, followed by 2 h in secondary antibodies at RT in the dark. TS were mounted in Citifluor mounting medium (Citifluor Products, UK) following final PBS washes. For each neural marker, n = 3 mice per genotype were used.
Briefly, TS were dissected into two halves (0 min and 60 min groups) from each animal in Ringer's solution and then incubated with 100 μg/mL AGAb (MOG1) and BTx-555 for 2 h at 4 C. All TS were washed 3x with Ringer's. One half was immediately fixed (0 min) and the other half moved to 37 C for 1 h (60 min) to promote AGAb internalisation, followed by further washes. TS were washed and fixed as above. Tissue was incubated overnight at 4 C in PBS + 1% NGS plus secondary antibody. TS were mounted in Citifluor mounting medium following final PBS washes. For "Surface AGAb" measurements, imaging for IgG3 intensity at the MNT was performed as described below.
TS were then re-washed and incubated with 0.5% Triton X-100 + 5% NGS for 30 min at RT to permeabilise the membrane. PBS washes, a further incubation in the same secondary Ab for 2 h at RT, washes and remounting followed. Imaging for "Total AGAb" IgG3 (surface + internalised) intensity to measure internalisation was performed.

| Acute in vivo injury paradigm
The in vivo model used here is based on the axonal injury model previously described 10

| Sub-acute in vivo injury paradigm
The sub-acute injury model, described previously, 32 was used to assess recovery. Here, we delivered AGAb targeting GM1 (DG2) to Neuronal mice to represent a model of AMAN. 14 In addition, this combination of AGAb and genotype tends to have a less severe outcome and allowed the experimental window to be extended to assess recovery. Therefore, Neuronal and hCAST Â Neuronal mice were injured with anti-GM1 ganglioside antibody. This model was performed as described above with the following modifications.

| Serum Simoa
The Simoa NF-light assay is a digital immunoassay used to quantify NF-L in serum, which signifies axonal damage. 34 Simoa was performed as described previously using the Simoa Nf-light kit in a Simoa HD-1 Analyser (Quanterix, Lexington, MA, USA). 35

| Experimental design
Mice were randomly allocated to treatment groups using a random number generator. A power analysis was performed using G*Power software (3.0.10 G*Power, RRID:SCR_013726) to determine group size; n = 3 (ex vivo) and n = 4 (in vivo) were selected for each treatment group. The effect size for behavioural output was based on previous experiments, with calculations made on a basis of 80% power, and a significance criteria of 0.05. All tissue was coded prior to imaging and analysis to prevent researcher bias.

| Statistical analysis
The numbers of independent animals are described in the Materials and Methods and indicated in the figure legends. Statistical differences were determined using GraphPad Prism 6 software (GraphPad Prism, RRID:SCR_002798). Unpaired, one-tailed t-tests were performed when comparing treatment without comparing genotype.
Two-way ANOVA were used for ex vivo immunoanalysis and was followed by Tukey's post-hoc tests for multiple comparisons to compare either genotype or treatment effect. One way-ANOVA followed by Tukey's post-hoc test for multiple comparisons was used for in vivo immunoanalysis and Simoa data. Repeated measures two-way ANOVA were used to analyse respiratory data followed by Sidak's or Tukey's tests for multiple comparisons to compare either genotype or time effect. Parametric testing was used, and differences were considered significant at P < .05. Data was plotted as the mean ± S.E.M using dotplots.

| Endogenous hCAST expression protects axonal and nodal integrity in an acute ex vivo injury model
First, given that calpains can have a role in membrane remodelling, we assessed the membrane dynamics of hCAST compared to WT mouse tissue. As we have previously shown AGAb internalisation occurs at MNT when incubated at 37 C for 60 min compared to 0 min, 33 we investigated membrane internalisation properties at this site. Here, WT MNT showed a significant reduction in surface AGAb intensity at 60 min compared to 0 min as expected ( Figure 1A, unpaired onetailed t-test, P < .05). hCAST MNT surface AGAb intensity level also showed a significant reduction at 60 min compared to 0 min ( Figure 1A, unpaired one-tailed t-test, P < .05). For both WT and hCAST, total staining intensity was recovered to 0 min surface levels by Triton X-100 permeabilization, which allows secondary antibody access and visualisation of intracellular AGAb ( Figure 1B). These results confirm normal internalisation and suggest no detrimental effects of endogenous calpain inhibition on AGAb binding.
F I G U R E 1 Legend on next page.
In our acute ex vivo model of AMAN, complement deposition at the distal nerve MNT and NoR significantly increased in injured tissue from both WT and hCAST mice compared to genotype control ( Figure 1C,E, unpaired one-tailed t-tests, P < .05). In line with the AGAb binding results, this suggests hCAST expression does not interfere with complement activation or deposition, and therefore the initiation of injury in our model is not impaired. Next we studied MNT and NoR occupied with NF-H immunostaining under injured conditions compared to control ( Figure 1D,E, two-way ANOVA for treatment (control vs. injury), F To assess the protection of structural integrity at the NoR, we next assessed AnkyrinG (Figure 2A

| Endogenous hCAST expression protects axonal and nodal integrity in an acute in vivo injury model, yet respiratory function remains impaired
In our in vivo acute mouse model of AMAN, we previously reported that following initiation of injury through intraperitoneal injection of AGAbs and complement, the diaphragm is severely compromised due to distal conduction failure, accompanied by respiratory dysfunction. 14,16 This is presumed to be due to the loss of regulation of electrophysiological function resultant from MAC pores allowing uncontrolled trans-axolemmal salt and water fluxes, rather than axon degeneration per se. Here we report that WT and hCAST mice dosed with AGAb and a source of complement, NHS, display a similar wasplike abdomen phenotype, and severe respiratory dysfunction. Tidal  However, from 48 h, hCAST Â Neuronal mice recover to both baseline F I G U R E 2 Endogenous hCAST expression protects nodal integrity in an acute ex vivo injury model. Triangularis sterni (TS) nerve-muscle preparations from WT (WT, n = 3) and hCAST (n = 3) mice were treated ex vivo with anti-ganglioside Ab (AGAb) with (Injured) or without (Control) normal human serum (NHS) as a source of complement. Representative images show the site of expected nodal protein immunostaining (magenta) indicated by arrowheads in the gap between myelin basic protein (MBP, orange) immunostaining. A) Significantly fewer AnkyrinG (AnkG) positive NoR were observed in WT injured tissue compared to control. This was improved by hCAST expression, but not to control levels. (B) NoR with Nav channel (Nav1.6) immunostaining was significantly reduced in WT injured TS compared to control; however, significant restoration occurred with hCAST expression. Scale bar = 5 μm. Dotplots = average ± S.E.M. Two-way ANOVA comparing treatment (control vs. injury) or genotype (WT vs hCAST) followed by Tukey's post-hoc multiple comparison tests were performed; *signifies P < .05, ** signifies P < .01, *** signifies P < .001 F I G U R E 3 Endogenous hCAST expression protects axonal integrity but not function in an acute in vivo injury model. WT and hCAST mice were dosed i.p. with 50 mg/kg anti-GD1b antibody (MOG1) followed 16 h later with 30 μL/g normal human serum (NHS). Naïve control littermates received no treatment. Respiratory function was monitored and diaphragm distal nerves assessed by immunoanalysis 6 h post NHS delivery. (A) Compared to baseline (black circles), at 6 h post-injury (white circle), WT and hCAST mice displayed a wasp-like abdomen (arrowheads) and a significantly reduced tidal volume (TV), measured using whole-body plethysmography (EMMS, UK). Naïve mice showed no change. Representative respiratory flow charts for each treatment group show reduced TV in WT and hCAST mice. (B) Complement deposition and axonal integrity (neurofilament, NF-H, occupancy) were compared at the diaphragm motor nerve terminals (MNT). Representative images illustrate complement deposits (green) overlying the MNT, identified by bungarotoxin (BTx, orange), in injured mice. (C) Simoa, used to measure levels of serum neurofilament light (NF-L), showed an increase in WT injury compared to both naïve control and hCAST injured mice. Scale bar = 10 μm. Results are represented as the mean ± SEM, n = 4/genotype/ treatment. Repeated measures two-way ANOVA comparing group (Naive vs WT injury vs hCAST injury) with time (baseline vs. 6 h) followed by Sidak's post-hoc multiple comparison tests were performed on respiratory function data. One-way ANOVA followed by Tukey's post-hoc multiple comparison tests were used for immuno-and serum analysis data. * signifies P < .05, ** signifies P < .01, *** signifies P < .001 and naïve levels, while Neuronal respiratory function remains impaired until 96 h. Simoa data shows that at 96 h the NF-L serum levels following injury in both genotypes are comparable to naïve ( Figure 5C Aberrant calpain activation is associated with a range of neurodegenerative disorders, including but not limited to cerebral ischemia, Alzheimer's disease, Parkinson's disease and muscular dystrophies. As such, calpain inhibition as a therapeutic strategy is an established concept. Many animal models have trialled exogenous calpain inhibitors as therapeutics, and, more recently, early clinical trials have F I G U R E 4 Endogenous hCAST expression protects nodal integrity in an acute in vivo injury model. WT and hCAST mice were dosed i.p. with 50 mg/kg anti-GD1b antibody followed 16 h later with 30 μL/g normal human serum. Naïve control littermates received no treatment. Representative images show the site of expected nodal protein immunostaining (magenta) indicated by arrowheads in the gap between myelin basic protein (MBP, orange) immunostaining at the node of Ranvier (NoR). A) Significantly fewer AnkyrinG (AnkG) positive NoR were observed in WT or hCAST injured tissue compared to Naïve controls. hCAST expression significantly improved frequency compared to WT injury. (B) NoR with Nav channel (Nav1.6) immunostaining was significantly reduced in WT injured tissue compared to Naïve controls and significant protection occurred with hCAST expression. Scale bar = 5 μm. Dotplots = average ± S.E.M., n = 4/genotype/treatment. One-way ANOVA followed by Tukey's post-hoc multiple comparison tests; * signifies P < .05, ** signifies P < .01, *** signifies P < .001 commenced, 36 with some promising results. These studies have helped reveal the benefits and pitfalls of calpain inhibition for potentially treating GBS.
Calpain II knockout mice and exogenous application of calpain inhibitors have been successfully trialled to attenuate nervous system injury in mouse models of TBI and taxol-induced sensory neuropathy. 22,23,37 We have previously reported the success of exogenous application of soluble calpain inhibitors in attenuating structural and nodal protein cleavage in ex vivo models of AMAN. 8,27 Here, we have extended these studies to assess the efficacy of calpain inhibition in vivo. The development of calpain inhibitors has been hampered by limited water solubility, cell permeability, metabolic stability and specificity. 36 Currently, the endogenous calpain inhibitor calpastatin is the only selective inhibitor for calpains, although strategies to improve the specificity of exogenous inhibitors are being developed. 36 Therefore, we elected to use transgenic mice that over-express the gene for Repeated measures two-way ANOVA comparing group (Naive vs Neuronal vs hCAST Â Neuronal) with time followed by Tukey's post-hoc multiple comparison tests were performed on respiratory function data. One-way ANOVA followed by Tukey's post-hoc multiple comparison tests was performed on Simoa data. * signifies P < .05, ** signifies P < .01, *** signifies P < .001. Black asterisks signify comparison between hCAST Â Neuronal and naïve mice, grey asterisks signify comparison between Neuronal and naïve mice, and # signifies comparison between Neuronal and hCAST Â Neuronal mice human calpastatin (hCAST) under control of the Prp promoter, resulting in widespread neuronal hCAST expression in both peripheral and central nervous systems, inclusive of axons, at levels as much as 80-fold higher than endogenous mouse calpastatin. 21 It has been reported that hCAST mice display a normal phenotype, and by immunolabelling with a calpastatin antibody, Ma et al. 20 showed calpastatin levels are increased in transgenic peripheral nerves and MNT compared to WT. In addition, it was reported that hCAST mouse sciatic nerves were structurally protected in a WD transection paradigm. The hCAST mouse benefits from site-specific calpain inhibition with no off-target toxicity from calpain inhibition at other sites and eliminates the need to deliver calpain inhibitors globally or implant invasive delivery pumps into our mice. Here we showed that we could recapitulate our ex vivo results using hCAST mice, and additionally that axonal neurofilament and nodal proteins AnkyrinG and Nav channels were protected in our acute in vivo AMAN model.

| Incomplete structural protection and mis-matched functional improvements
In this transgenic model, the dose of calpastatin is fixed and therefore cannot be controlled or altered in the way that soluble inhibitors could be. Calpastatin reversibly binds and blocks calpain, and levels are likely not enough to completely protect all calpain-mediated cleavage. 28 Indeed, we see a partial protection of immunostaining for the Nav channel and structural proteins NF-H and AnkyrinG in both our ex and in vivo models. NFH and AnkG are known calpain substrates 38 [Boivin, 1990 #222], and Nav channels are either cleaved directly 25 or mis-localised through cleavage of their AnkyrinG tether. The additional disturbances in protein localisation we observed could also be attributed to other cytotoxic mediators or physiological changes such as inflammatory oedema or membrane swelling caused by the presence of MAC pores and associated water influx. Calpain is involved in membrane remodelling under normal physiological conditions 28 ; therefore, it is possible that complement pore shedding from the axolemmal membrane could be impaired in the presence of calpain inhibition. Nevertheless, we did not find any changes to AGAb binding or internalisation, or indeed complement deposition, in our paradigms. In subjects with GBS, calpain therapeutics would be delivered after disease onset; even if membrane dynamics were altered in an unexpected way by calpain inhibition, the short-term effects would likely be minimal, and the benefits would more likely outweigh any subtle impairments. 36 Intriguingly, despite the structural protection observed in our acute model, nerve function remained impaired. MAC pore complement deposition remained present; therefore, we interpreted as a failure of membrane potential homeostasis since free movement of water and ions will occur bi-directionally through MAC pores and impair resting membrane potential and thereby function. 8 Here, we describe similar results in our acute in vivo model. In the initial stages of injury, physiological function is not protected, but early structural protection appears to ultimately promote a more rapid functional recovery. It is possible that this reduces the likelihood of axon degeneration and poor outcomes through a shift in the metastable state in favour of protection. 5 Critically, it is important to appreciate that inexcitable axons do not directly equate to degenerating axons; therefore, early axon-protective intervention could attenuate axon degeneration and improve axon fate, currently a major gap in knowledge.
Protection of structure but failure to prevent conduction loss was also observed after sciatic nerve transection in hCAST mice 20 and in optic nerves exposed to anoxia in the presence of pharmacologic calpain inhibitors. 39 Despite this lack of functional recovery following transection, normal endogenous calpastatin expression levels in WT mice are unable to prevent injury caused by calpain activity in an acute nerve transection model. Whereas the authors report successful reduction in NF-L and NF-H proteolysis and cytoskeletal preservation in the sciatic nerve from over-expressing hCAST mice up to 5 days. It was proposed for the latter models that other calcium-dependent processes could potentially mediate this persistent loss of function. In the same transgenic mice, Schoch et al. 21 found an attenuation of calpainmediated cleavage of structural protein α-spectrin, voltage gated sodium channels, and collapsin response mediator protein-2 in brain tissues after TBI. In a neuron-specific calpastatin overexpressing mouse, similar attenuation of calpain-mediated proteolysis coincided with behavioural improvements in a model of TBI. 40 Intriguingly, TBI models using exogenous calpain inhibitors demonstrate functional improvements, but with more limited success in lessening proteolysis or neuron death. [41][42][43] Variable responses in calpain inhibitor drug studies may be related to complexities in delivery in the correct timewindow, dosing frequency, or titration of dosages. 44 Therefore, the benefit of complement and calpain inhibitor delivery after the onset of injury is critical and remains to be assessed in GBS models. However, we propose that the "tipping point" of the axon in the metastable state hypothesis 5 could be shifted in favour of repair (rather than transection) by later delivery. Indeed, we report here the accelerated functional recovery of mice where axons are protected early. Additionally, a combinatorial therapeutic approach using both complement and calpain inhibition would likely enhance patient outcome and is an avenue worth exploring in future studies. Use of these transgenic mice is a simplified system, and thus the trial of exogenous calpain inhibitors in human GBS would be the crucial next step.

| Serum protein levels to predict therapeutic efficacy
It has been shown from patient samples that long-term prognosis in GBS patients can be predicted by serum levels of NF-L, with higher cleavage indicating a poor outcome. 35,45 Interestingly, serum biomarkers can also be associated with neuropathy-related inflammatory oedema and are a more sensitive marker than electrophysiological abnormalities. 11 Quantitative clinical assessment is a major unmet need in understanding treatment efficacy, and the use of a response biomarker would be beneficial for future trials. 34 For the first time in an animal model, corresponding with immunostaining loss, we show the axon breakdown product NF-L in the serum as a useful readout of axon injury and clinical disease. After a single acute insult, NF-L levels increased, and 96 h later, normalised to naïve levels, suggesting no ongoing injury. Most significantly, we used this method to show attenuation of axon damage with calpain inhibition. This occurred despite acute functional loss and coincided with a faster functional recovery in an extended injury paradigm. Due to blood volume collection restrictions in experimental mice, we were not able to measure serum NF-L at every time-point. However, we have previously reported using immunostaining that NF-H and MAC, along with function as assessed by WBP, return to normal 72 h after sub-acute injury in WT mice in vivo, suggesting some relationship between axon integrity and function exists. 32 Therefore, based on these initial findings, it is possible that serum levels could correspond with function and be used for clinical assessment. These results suggest serum NF-L levels could be used to assess the efficacy of drug treatment in future models and substantiate the use of serum assays to predict prognosis in patients. Indeed, biomarkers of axonal injury are likely to be more useful than clinical assessment, as electrophysiological function, and thus clinical performance, may be highly dissociated from structural integrity in the acute phase of illness, as discussed above.