Progression of axonal excitability abnormalities with increasing clinical severity of diabetic peripheral neuropathy

(cid:1) There are progressive changes in axonal excitability in DPN in parameters that reﬂect the activity of axonal K v 1.1 channels. (cid:1) Mathematical modelling shows that excitability changes in severe DPN are explained by an increase in K v 1.1 conductances. (cid:1) Blockade of K v 1.1 channels may be a suitable treatment target for diabetic peripheral neuropathy.


Introduction
Type 2 diabetes is a condition of high global prevalence, which affects over 500 million people worldwide (Magliano et al., 2021).Peripheral neuropathy is a disabling condition which occurs frequently in persons with type 2 diabetes and causes neuropathic pain, sensory loss and muscle weakness (Pop-Busui et al., 2017).
There is no pharmacological treatment for diabetic peripheral neuropathy (DPN) and previous observational studies in large cohorts have failed to demonstrate any significant impact of intensive glycaemic control on reversing neurological disability in DPN (Duckworth et al., 2009, Feldman et al., 2017, Gaede et al., 2003, Ismail-Beigi et al., 2010).
Nerve excitability studies are a neurophysiological technique that provide insight into the behaviour of voltage-gated ion channels, pump and exchangers involved in impulse conduction.Excitability studies have been undertaken in peripheral neuropathy due to toxic, metabolic and immune aetiologies and have demonstrated changes of pathophysiological and clinical significance (Kiernan et al., 2020).Studies of axonal excitability in DPN have demonstrated a diverse set of abnormalities involving voltage-gated sodium and potassium (K v 1.1, fast; K v 7.2/7.3,slow) channels as well as inwardly rectifying conductances (Horn et al., 1996, Issar et al., 2021, Misawa et al., 2009).These previous studies have however generally viewed DPN as a single clinical entity, without reference to disease severity.
The present study explored the hypothesis that there may be a progressive change in axonal ion channel abnormalities in DPN that varied according to disease stage.Excitability studies were undertaken in a large cohort of persons with DPN of different degrees of severity and mathematical modelling of the human motor axon was utilised to assess for progressive changes in ion channel physiology which may be used to define future ion channel targets for DPN treatment.

Methods
The study was approved by the local institutional review board and all participants were enrolled with written informed consent.178 participants with type 2 diabetes were recruited from the Diabetes Centre of Prince of Wales Hospital in Sydney, Australia.Exclusion criteria were age <18 years, lower limb amputation and peripheral neuropathy due to other causes, and a history consistent with carpal tunnel syndrome.All participants underwent clinical assessment of peripheral neuropathy severity, as well as nerve conduction studies and median motor nerve excitability studies.
The Total Neuropathy Score (TNS) (Cornblath et al., 1999) was used to group participants according to their peripheral neuropathy severity.The TNS subdivides into four grades of severity, with higher grades representing more severe peripheral neuropathy: TNS 0-1 = grade 0, 2-8 = grade 1, 9-16 = grade 2, 17-24 = grade 3, 25-32 = grade 4. Participant demographic data was recorded including age, sex, diabetes duration, BMI and waist circumference.Serum HbA1c, eGFR and potassium levels were retrieved from last serum sample taken as per routine clinical care, up to 3 months prior to peripheral neuropathy assessment.Studies were also conducted in 60 healthy control participants.

Nerve excitability and mathematical modelling
Median motor nerve excitability was assessed using the TROND protocol and Qtrac software for each participant (Digitimer, London, United Kingdom).The median nerve was stimulated proximal to the wrist (with skin temperature >32 °C), at the site of least resistance with the assistance of non-polarising surface electrodes (Ambu, Sydney, Australia) and a DS5 Isolated Bipolar Current Stimulator (Digitimer, London, United Kingdom).The abductor pollicis brevis muscle was used as the site of recording compound muscle action potentials (CMAPs).
A 1 ms test pulse was used to produce stimulus-response (SR) data and calculate a target response of 40% of maximum.The threshold current to achieve the target response was tracked in three paradigms, namely strength-duration behaviour, threshold electrotonus, and recovery cycle.
Strength-duration time constant (SDTC) was calculated using Weiss' law as a measure of persistent Na + conductances at the node of Ranvier (Burke et al., 2001).2. Threshold electrotonus was assessed as the response to subthreshold depolarising and hyperpolarising conditioning currents, providing information on voltage-gated nodal and internodal conductances.Changes in threshold were assessed following a 1 ms test pulse during or after subthreshold conditioning currents of 100 ms at +40% (depolarising) or À40% (hyperpolarising) of the control threshold, produced based on the initial SR curve.Percentage threshold change was tracked at 10 ms periods and change in depolarising threshold change every 10-20 ms, S2 accommodation and hyperpolarising threshold change for 10-20 ms, 20-40 ms and 90-100 ms.S2 accommodation examines the period of depolarising threshold electrotonus where threshold reduction is attenuated and returns to control measures.3. Following supramaximal stimulation, the recovery cycle assessed change in threshold over various conditioning intervals.In the relative refractory period (early period of the recovery cycle), reduced excitability occurs due to Na + channel inactivation.This is followed by superexcitability which is associated with a decrease in threshold for impulse generation.Subexcitability is the final period of recovery cycle where activation of nodal K v 7.2/7.3channels (i.e slow K + channels) cause an increase in threshold (Bostock et al., 1998).
Analysis of nerve excitability recordings used the Bostock model of axonal excitability (Kiernan et al., 2020).The Bostock model is a validated model of a node of Ranvier and the internode and its relevant pathways (Kiernan et al., 2020), and assists with the understanding the clinical and pathophysiological implications of axonal excitability findings in normal controls and disease participants.The model incorporates changes in maximal conductance and permeabilities of the various Na + and K + ion channels, alterations in Na + /K + pump and leak currents and biophysical properties.The model was initially adjusted based on the mean nerve excitability data obtained from the control group before objectively fitting the mean data of each TNS group through simulation using a least squares approach.Each stage of the disease was modelled accounting for pathophysiological factors identified in diabetes (Brismar et al., 1987, Hong and Wiley, 2006, Krishnan and Kiernan, 2005, Zenker et al., 2012).Transient and persistent Na + channel permeability was modelled at the node however fast and slow K + conductances were modelled in the node and internode.As the model does not contain a juxtaparanodal compartment, an approximation of this is made by having a greater density of fast K + channels located at the internode compared to the node.Pump and leak currents and axolemmal capacitances were assessed in both compartments.Only the internodal hyperpolarisationactivated cation conductance was modelled.Barrett-Barrett conductance, which represents current flow between the node and internode through and underneath the myelin sheath, were also investigated (Howells et al., 2012).
Modelling analyses utilised a least squares approach to objectively fit simulated data with the mean recorded excitability data, by way of changes in either single or a combination of parameters of interest in an iterative fashion.Apparent 'discrepancy' between this study's data and simulated data was generated by comparison between recorded and simulated data of strength-duration behaviour, threshold electrotonus, current-threshold relationship, and recovery cycle.Paradigm weighting factors were 0.5, 1, 1, and 2, respectively and were kept constant for each TNS grade.The minimum interstimulus interval for the recovery cycle was set at 3 ms.Analyses were run in unclamped mode to permit secondary changes in resting membrane potential caused by changes in conductances or pump currents.

Statistical analyses
SPSS Statistics Version 26.0 for Windows was used for data analysis.A coded-system was used to deidentify and analyse participant data.Normality of data was assessed with the Shapiro-Wilk test.Normally distributed data was analysed using independent t-tests.Correlation analysis was carried out for clinical, ultrasound and nerve excitability parameters using Spearman coefficients (rho) and specifically for each peripheral neuropathy grade.Where group comparisons were performed analysis of variance was used.Multiple regression analysis was performed to assess for the influence of clinical, laboratory and the relevant nerve excitability variables on peripheral neuropathy severity, using 'stepwise' function on SPSS.Statistical significance was defined as p < 0.05.Figures were generated using GraphPad Prism version 7.00 for Windows (GraphPad Software, La Jolla, California).

Participant demographics
Participant demographic data are summarised in Table 1.Groups were categorised according to Total Neuropathy Grade (TNG), as a marker of DPN severity (Cornblath et al., 1999), with grades ranging from 0-4 and higher TNG indicating more severe neuropathy.As expected, neuropathy severity groups differed by age, diabetes duration, HbA1c, serum potassium, eGFR and BMI.With respect to neurophysiological parameters, decrease in peak response and an increase in motor latency were noted with increasing grade of neuropathy severity (Table 2), reflecting greater axonal loss with more severe neuropathy grades.This was consistent with a reduction in sural sensory nerve action potential (SNAP) and tibial compound action potential (CMAP) amplitudes with increasing DPN severity (see Table 1).
Generalised changes were noted in other excitability parameters, as noted in other studies of DPN (Issar et al., 2021, Krishnan andKiernan, 2005), including stimulus-response slope, SDTC, TEh20-40, TEd40-60 and RRP.However, these changes did not demonstrate progression with increasing grades of neuropathy severity (see Table 2).

Mathematical modelling
To provide insight into the underlying changes in membrane pathology, mathematical modelling of nerve excitability findings against an established model of a human axon was undertaken for each disease stage, with the exception of the TNG 0 group which was not significantly different from controls (Kiernan et al., 2020).Various membrane parameters implicated in the pathophysiology of diabetic neuropathy, focussing on K v 1.1 channels, K v 7.2/7.3channels and voltage-gated Na + channels (Arnold et al., 2013, Issar et al., 2019, Kwai et al., 2016) were fitted to collected data.Analysis of the TNG 1 group indicated that alterations in nerve excitability recordings were best explained by a reduction in nodal Na + permeability (TNG 1: 3.6; Control: 4.2, cm 3 s À1 Â 10 -9 ).This single change in the model reduced the discrepancy between the TNG 1 and control groups by 82%.In this early grade of neuropathy, adjusting for K v 1.1 conductances was not a suitable explanation, with a discrepancy reduction of 66%.In the TNG 2 group, modelling showed the change in nerve excitability could be interpreted as a reduction in nodal Na + permeability (TNG 2: 3.3; Control: 4.2, cm 3 s À1 Â 10 -9 ), which accounted for 84% of the difference from control data or an increase in K v 1.1 conductances (TNG 2: 35; Control: 20, nanosiemens), which reduced the discrepancy by 75%.The combination of modelling K v 1.1 conductances with nodal Na + permeability improved the fit to 86% (Na + permeability: cm 3s À1 Â 10 -9 ; K v 1.1 conductance: 23, nanosiemens).Analysis of the most severe neuropathy group (i.e TNG 3 and 4), demonstrated that nerve excitability findings were not satisfactorily explained by altered nodal Na + permeability, but rather by an increase in K v 1.1 conductances as the primary adjustment to the model, which reduced the discrepancy by 76% (TNG 3 + 4: 34; Control: 20, nanosiemens).In severe neuropathy, the second-best parameter change, a reduction in nodal Na + permeability, could only explain 72% of the discrepancy and combining these two parameters did not improve the fit beyond 78%.This suggests the most likely explanation for the excitability findings obtained from the most severe neuropathy group was an increase in K v 1.1 conductances.Moreover, progressively adjusting for K v 1.1 conductances provided a more suitable explanation for the findings with each higher grade of neuropathy.
Given the progressive involvement of K v 1.1 conductances, it may be hypothesized that an increase in electrical access between the node and internode was a possible explanation for the findings.However, an increase in the Barrett-Barrett conductances, which reflects the flow of current between the node and internode, was not a sufficient model parameter variation either alone (never exceeding 10%) or in combination with alterations in K v 1.1.Given the progressive changes noted in S2 accommodation, modelling was also undertaken at each stage for alterations in K v 7.2/7.3channels, although was a less satisfactory explanation of the nerve excitability changes for each group, resulting in a discrepancy reduction of 54% for TNG 1, 40% for TNG 2 and 47% for TNG 3-4.

Discussion
The present study has demonstrated novel excitability findings, supported by mathematical modelling, of selective upregulation of K v 1.1 channels in DPN, most apparent in high grades of disease severity.This study is novel in that excitability studies have been performed with reference to groups categorised by DPN severity, allowing for investigation of axonal ion channel dysfunction according to the severity of clinical impairment.While there were changes noted in other excitability parameters in all neuropathy grades, a progressive reduction was seen in TEd10-20 ms and superexcitability which are reflective of upregulation of K v 1.1 conductances.Multiple regression analyses provided evidence for the role of TEd10-20 and superexcitability which are markers reflective of K v 1.1 conductances and their association with severe peripheral neuropathy.These findings underscore the potential importance of increasing involvement of the juxtaparanodal region of the axon in severe diabetic peripheral neuropathy, and possibly a modifiable target.While progressive changes were also noted in S2 accommodation, mathematical modelling failed to provide any evidence that the excitability changes were due to alterations in K v 7.2/7.3channels and it is therefore more likely that the changes in S2 accommodation were secondary to changes induced by alterations in K v 1.1 channels.A limitation of the study is that all participants were assessed in a cross-sectional manner, rather than a single group of patients being followed longitudinally over the natural course of the condition.A further limitation of the study was that serum K + levels were not obtained on the day of peripheral neuropathy assessment, limiting further analysis of the influence of serum electrolytes on nerve excitability results.It should also be noted while participants with symptoms of carpal tunnel syndrome were excluded, findings of prolongation of median distal motor latency on nerve excitability studies was not a reason for exclusion.
The axonal juxtaparanodal region contains a high density of K v 1.1 channels (Arroyo et al., 1999, Inouye et al., 2014).The main role of these channels is to dampen nodal excitability following action potential generation by preventing re-excitation and maintaining internodal excitability (Chiu and Ritchie, 1984, Rasband et al., 1998, Wang et al., 1993).However, absence of the normal axo-glial junction structure results in relocation of these channels into the paranode from the adjacent juxtaparanode and may impair saltatory conduction (Bhat et al., 2001).The axo-glial junction serves as an important barrier between nodal Na + and juxtaparanodal K + channels, and loss of this barrier has been described in DPN (Corfas et al., 2004, Sima et al., 1988).Axoglial dysjunction occurs due to Na + /K + pump dysfunction and subsequent osmotic axonal changes, leading to nodal and paranodal swelling and loss of the junctional complexes linking myelin loops and the paranodal axolemma (Sima et al., 1986).Changes in the axo-glial junction may therefore represent a potential cause for the upregulation of K v 1.1 channels, as demonstrated in this study.
Studies in animal models of multiple sclerosis have demonstrated that inflammatory demyelination may result in increased exposure of K v 1.1 channels (Bagchi et al., 2014).While demyelination in multiple sclerosis is considered to be immune-mediated, abnormal myelination in DPN may be secondary to other factors, such as Schwann cell pathology which is now considered to play an important role in DPN pathogenesis (Feldman et al., 2017, Fünfschilling et al., 2012, Viader et al., 2013, Zenker et al., 2013).A recent study demonstrated that regulated necrosis of Schwann cells may be triggered by the necroptosis protein, mixed-lineage kinase domain-like protein (MLKL) in DPN (Belavgeni et al., 2022).A further study in streptozotocin-induced diabetic mice demonstrated that the more MLKL-deficient mice had greater dysregulation of myelination and greater changes in nerve conduction velocity (Guo et al., 2022), suggestive of a more severe DPN phenotype.
Several studies have highlighted the role of K + channelopathies in neurological disease, particularly in epilepsy and episodic ataxia (Bonardi et al., 2021, Tomlinson et al., 2012, Tomlinson et al., 2010).Nerve excitability studies undertaken in patients with episodic ataxia type 1 who have a loss-of-function mutation in the KCNA1 gene that encodes K v 1.1 channels have shown markedly abnormal motor nerve excitability studies, with an increase in depolarizing threshold electrotonus and superexcitability (Tomlinson et al., 2010).The present study has shown a pattern of change that is the converse of these changes noted in episodic ataxia type 1, particularly in DPN of greater severity, with mathematical modelling providing further support that the changes in the high neuropathy severity groups were consistent with a selective upregulation of K v 1.1 channels.DPN is a major cause of neuropathic pain, and the role of juxtaparanodal fast K + channels in the pathophysiology of neuropathic pain in DPN was not specifically explored in the present study.Studies have shown that blocking or genetic deletion of hyperpolarisation-activated cyclic nucleotide-gated (HCN2) ion channels has been associated with a reduction in diabetic neuropathic pain (Themistocleous et al., 2022, Tsantoulas et al., 2017).There is likely also a centrallymediated mechanism for DPN pain, as a recent study has demonstrated altered ventrolateral periaqueductal grey functional connectivity (Segerdahl et al., 2018).
The findings of this study have important translational implications, as K v 1.1 channels represent a therapeutic target which may be of value in addressing the unmet need for DPN treatments.Fampridine is a sustained-release form of 4-aminopyridine, a potent blocker of K v 1.1 channels, and is currently used in the management of multiple sclerosis, particularly as a means of improving walking disability (Dietrich et al., 2020, Goodman et al., 2009, Huynh et al., 2016, Pickering et al., 2017).The pathophysiological basis for its use relies on blockade of axonal K v 1.1 channels leading to enhanced conduction across demyelinated axons (Dietrich et al., 2021).Application of 4-AP has been shown to increase action potential duration (Sherratt et al., 1980) which may potentially mitigate the effects of reduced nodal Na + that were noted in this study in participants with mild disease severity.Preclinical animal studies of nerve crush models have demonstrated that prophylactic and early use of 4-aminopyridine results in recovery of nerve conduction velocity as well as influencing the speed and extent of behavioural recovery (Tseng et al., 2016).The findings of the present study therefore raise the prospect that fampridine may represent a potential treatment to improve nerve conduction and clinical symptoms in DPN.
In conclusion, the present study has provided evidence based on axonal excitability studies and assisted by mathematical modelling, that there is a progressive and selective upregulation of K v 1.1 conductances in DPN.These findings may provide a basis for investigation K + channel blockade as a treatment for DPN.
, Wang https://doi.org/10.1016/j.clinph.2024.02.002 1388-2457/Ó 2024 International Federation of Clinical Neurophysiology.Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).⇑ Corresponding author at: School of Clinical Medicine, UNSW, Sydney.E-mail address: arun.krishnan@unsw.edu.au(A.V. Krishnan).Clinical Neurophysiology 160 (2024) 12-18 Contents lists available at ScienceDirect Clinical Neurophysiology j o u r n a l h o m e p a g e : w w w .e l s e v i e r .c o m / l o c a t e / c l i n p h

Fig. 1 .
Fig.1.Group comparison of nerve excitability recordings, demonstrating mean nerve excitability data for participants with neuropathy severity grades 0-4 and healthy controls for (A) depolarizing threshold electrotonus and (B) recovery cycle paradigms.TNG; Total neuropathy grade; ms, milliseconds.

Table 1
Participant characteristics.Nerve excitability parameters in type 2 diabetic participants according to peripheral neuropathy severity grade.All values given as mean ± standard error.Variables of threshold electrotonus and recovery cycle unless otherwise indicated are expressed as percentage change.TNG, Total Neuropathy Score Grade; mV, millivolts; mA, milliAmp; SR, stimulusresponse; SDTC, strength-duration time constant; ls, microseconds; TEd, depolarising threshold electrotonus; Accomm, Accommodation; TEh, hyperpolarising threshold electrotonus; RRP, relative refractory period.