The Antidiabetic Drug Metformin Regulates Voltage-Gated Sodium Channel NaV1.7 via the Ubiquitin-Ligase NEDD4-2

Visual Abstract


Introduction
Metformin is a antidiabetic drug with pleiotropic effects. Although it has been used for many decades as a first-line treatment to lower hyperglycemia and glycosylated hemoglobin in type 2 diabetes (Sterne, 1957;Hundal et al., 2000), its primary mechanism of action and effect on pain were found later. Zhou et al. (2001) showed that metformin induces the activation of adenosine 59monophosphate protein kinase (AMPK) in hepatocytes, a phenomenon that in turn regulates transcription factors and downstream signaling pathways, which was confirmed in humans (Musi et al., 2002). Patients with diabetes treated with metformin show less radiculopathy pain (Taylor et al., 2013) and musculoskeletal pain (Carvalho-e-Silva et al., 2021). In patients with fibromyalgia, metformin treatment significantly improved clinical symptoms such as pain, fatigue, depression, disturbed sleep, and tender points (Bullón et al., 2016).
One of the first studies indicating a potential use of metformin in experimental neuropathic pain showed reduction of tactile allodynia-like behavior in mice after spared nerve injury (SNI) and spinal nerve ligation (SNL; Melemedjian et al., 2011). In parallel, there was inhibition of the protein kinase complex mechanistic target of rapamycin (mTOR) complex 1 and activation of the ERK (extracellular signal-regulated protein kinase) pathways, ultimately leading to the dysregulation of translational control (Melemedjian et al., 2013). In rats, metformin reversed the chronic constriction injury-induced changes in phosphorylated signal transducer and activator of transcription 3 in spinal dorsal horn (SDH) and activation of microglia and astrocytes in the SDH, as well as decreased pain-like hypersensitivity (Ge et al., 2018). Metformin also prevented cisplatin-induced or spinal cord injury mechanical and thermal hypersensitivity (Mao-Ying et al., 2014), and attenuated the levels of tumor necrosis a and interleukin 1b in the spinal cord (Afshari et al., 2018).
Metformin acts via a multitude of intracellular signaling pathways in which the master kinase seems to be AMPK. This kinase is activated by an increase in the cytosolic AMP/ATP ratio, playing a crucial role in cellular energy homeostasis. A downstream ubiquitin protein ligase activated by AMPK is neuronal precursor cell expressed developmentally downregulated-4 type 2 (NEDD4-2), a ligase that acts on the epithelial sodium channel(ENaC) by downregulating its currents via ENaC-NEDD4-2-induced internalization (Bhalla et al., 2006). This mechanism of action has also been demonstrated for potassium channels such as K V 1.5 (Mia et al., 2012), K V 1.7 (Andersen et al., 2012), and K V 7.1 (Alzamora et al., 2010). AMPK activation could similarly enhance NEDD4-2 function and favor sodium channel internalization in sensory neurons. In a previous study, researchers showed that downregulation of NEDD4-2 is associated with an increase in total sodium current (I Na ) and voltage-gated sodium channel 1.7 (Na V 1.7) currents in the SNI model of neuropathic pain. Restoring normal NEDD4-2 expression after SNI prevented full development of hypersensitivity (Laedermann et al., 2013). This leaves open the question as to whether metformin can pharmacologically modulate NEDD4-2 function to decrease sodium current expression, in particular Na V 1.7, which in turn might explain its effect in various experimental models of chronic pain.
In this study, we test the hypothesis that metformin inhibits I Na and Na V 1.7 currents by decreasing their expression at the plasma cell membrane via NEDD4-2. In a heterologous system, we demonstrate that cotransfection of Na V 1.7 and NEDD4-2 results in a significant decrease in the expression of the functional form of this channel coupled with lower current densities on metformin treatment. In Nedd4-2 fl/fl mice, we show that metformin reduces both current densities and membrane expression of Na V 1.7 compared with the SNS-Nedd4L À/À knock-out animals. Furthermore, the excitability of dorsal root ganglion (DRG) neurons decreases after metformin treatment, there are altered parameters of single action potentials (APs), and there is hyperpolarization of the resting membrane potential (RMP). Our results show that metformin reduces both the expression and current densities of Na V 1.7 in a NEDD4-2-dependent manner, whereas neuronal excitability decreases in both Nedd4-2 fl/fl and SNS-Nedd4L À/À knock-out animals. These findings provide new avenues for research investigating the effect of metformin on neuronal excitability independent of NEDD4-2.

Study approval
All experiments involving animals were performed according to the regulations of the University of Lausanne animal care committee. This investigation conformed to the Swiss Federal Laws on Animal welfare, the Guide for the Care and Use of Laboratory Animals (National Institutes of Health, 1996), and to the International Association for the Study of Pain guidelines for the use of animal in research (Zimmermann, 1983).

Animals
Animals were grouped housed, in a standard environment (litter, paper roll, and tissue), with free access to food and water, and a 12 h photoperiod. The transgenic mouse line SNS-Nedd4L À/À and their control littermates Nedd4L fl/fl were used. Briefly, the floxed Nedd4L mouse line (Nedd4l tm1.1Blyg ; catalog #3846430, MGI; RRID:MGI:3846430 provided by Prof. O. Staub, University of Lausanne, Lausanne, Switzerland) was crossed with the SNS-Cre line [Tg (Scn10a-cre)1Rkun; catalog #3042874, MGI; RRID: MGI:3042874; provided by R. Kuner, University of Heidelberg, Heidelberg, Germany) expressing the Cre recombinase under the promoter of Na V 1.8 (SNS). This breeding leads to a conditional knockout of NEDD4-2 in sensory Na V 1.8-positive DRG neurons. Five-to-8week-old animals were used for experiments. Males and females were used for the experiments.

Primary neuronal culture
Mice were killed with cervical dislocation/lethal injection of pentobarbital (50 mg/kg). For electrophysiological recordings, L4 and L5 DRGs were dissected and collected in oxygenated complete saline solution (CSS; composition: 137 mM NaCl, 5.3 mM KCl, MgCl 2 ·, 6H 2 O, 25 mM sorbitol, 10 mM HEPES, and 3 mM CaCl 2 ; pH adjusted to 7.2 with NaOH). DRGs were then harvested and digested in 5 ml of solution containing the following: liberase blendzyme TH (catalog #5401151001, Roche) at a concentration of 0.5 U/DRG, 12 mM EDTA (catalog #E5134, Sigma-Aldrich) in oxygenated CSS for 20 min at 37°C. Neurons were further digested with Liberase blendzyme TM (CSS #5401127001, Roche) in 5 ml of solution (0.5 U/DRG, 12 mM EDTA in 5 ml CSS) plus 30 U/ml papain (catalog #P3125, Sigma-Aldrich) for 10 min. Then, neurons were suspended in 1 ml of DRG medium containing DMEM/F12 (catalog #21331-020, Thermo Fisher Scientific) with 10% FBS and 1% P/S, supplemented with 1.5 mg of trypsin inhibitor (catalog #T6522, Sigma-Aldrich) and 1.5 mg of purified bovine serum albumin (BSA; catalog #A9647, Sigma-Aldrich). Mechanical dissociation (12 strokes) was performed to triturate the DRGs gently. Finally, isolated neurons were plated on 12 mm coverslips (bovine serum albumin #631-1577, VWR) coated with 0.1 mg/ml poly-D-lysine (bovine serum albumin #P7886, Sigma-Aldrich). Neurons were only recorded at 12 h to prevent neurite outgrowth that degrades the space clamp. For cell surface biotinylation experiments, L2-L6 DRGs were dissected from each side of the spinal cord and pooled to obtain 10 DRGs per sample. The dissociated neurons were split in two, for the control and the metformin conditions; plated on six-well plates (catalog #353224, Corning); and kept in culture for 6 d before metformin treatment.
Cell surface biotinylation assay HEK293 cells transiently cotransfected or DRG neurons in culture were washed with cold 1Â PBS, pH 7.4 (catalog #10010, Thermo Fisher Scientific) and then treated with 0.5 mg/ml EZ-link Sulfo-NHS-SS-Biotin (catalog #21331, Thermo Fisher Scientific) in cold 1Â PBS for 30 min at 4°C. The cells were then washed three times with 200 mM glycine (catalog #A1067, AppliChem) in cold 1Â PBS to inactivate biotin, and twice with cold 1Â PBS to remove excess biotin. The cells were scraped and lysed with 1Â lysis buffer that contained 50 mM HEPES, pH 7.4, 100 mM NaCl, 1 mM EGTA, pH 8, 10% glycerol, 1% Triton X-100, 10 mM N-ethylmaleimide, complete protease inhibitor cocktail (catalog #11697498001, Roche), and phosphatase inhibitor cocktail (catalog #04906837001, Roche) for 1 h at 4°C on a wheel. Whole-cell lysates were centrifuged at 16,000 Â g for 15 min at 4°C. The protein concentration of supernatants was measured by using a Bradford-based assay (catalog #500-0006, BIO-RAD). Subsequently, 2 mg (500 mg for neurons) of the supernatant was incubated with 30 ml of Streptavidin Sepharose High Performance Beads (catalog #17-5113-01, GE Healthcare) for 2 h at 4°C on the wheel, and the remaining supernatants were kept as input fractions (S0 fractions). The beads were subsequently washed five times with 1Â lysis buffer plus 1 mM phenylmethanesulfonyl fluoride (catalog #P7626, Sigma-Aldrich) with centrifugation steps (1000 Â g for 2 min at 4°C) between washes (the supernatant was discarded after each centrifugation). Elution was performed with 40 ml of 5Â sample buffer containing 1.5 M sucrose, 10% SDS (catalog #L4390, Sigma-Aldrich), 12.5 mM EDTA, 0.3 M Tris, pH 8.8, 0.25% bromophenol blue, and 150 mM dithiothreitol (catalog #A1101, AppliChem), at 37°C for 1 h with gentle shaking. A last centrifugation at 2500 Â g for 1 min was performed to separate the beads from the supernatant fractions. These biotinylated fractions (S2 fractions) were analyzed by Western blot for Na V 1.7 expression at the cell surface. The input fractions were diluted 1:5 in 5Â sample buffer, incubated at 37°C for 30 min, and used to evaluate the total expression of Na V 1.7 and the other proteins of interest by Western blot.

Solutions
Transfected HEK293 cells and DRG neurons in culture were treated with 20 mM metformin (catalog #LKT-M2076-G025, LKT Labs) solubilized in their respective culture media, for 12 h, before the assessment of electrophysiological recordings or cell surface biotinylation assays. The concentration of metformin applied was the optimum for cell survival according to various dose-response assays (data not shown). Control groups were incubated with their respective culture media only.

Electrophysiology
Data were acquired with a Multiclamp 700B Microelectrode Amplifier and pClamp 10 software (Molecular Devices; RRID:SCR_011323) or with an EPC-9 Amplifier and Patchmaster software (HEKA Electronics; RRID:SCR_000034). Data were analyzed Research Article: New Research with KaleidaGraph 4.03 (Synergy Software; RRID:SCR_ 014980) and GraphPad Prism 7 (GraphPad Software; RRID:SCR_002798). Low-pass filtering was set to 5.0 kHz. Resistance of the borosilicate pipettes (catalog #BF150-86-7.5, Sutter Instrument) was 1.5-3 MV. After opening, HEK293 cells were kept at À70 mV for at least 2 min to dialyze and equilibrate the cell. DRG neurons were kept at À60 mV for 5 min to dialyze the cell and to allow Na V 1.8 voltage dependencies to stabilize and inactivate Na V 1.9 currents. Cells were then clamped at À80 mV for 2 min before starting the recordings. The I Na current densities (pA/pF) were obtained by dividing the peak I Na by the cell capacitance.
The current-clamp recordings were made on cultured primary neurons from Nedd4L fl/fl and SNS-Nedd4L À/À animals. Recordings were made following a 500 ms ramp current ranging from 100 to 500 pA, and the APs were counted and represented at each ramp. Small step pulses of 5 ms with 25 pA increments were done, and the first AP triggered was analyzed for different parameters. Peak amplitude represents the maximum amplitude of the AP measured from the RMP. The maximum rise time is the time from the start of the stimulation until the maximum AP amplitude was reached. The baseline is the voltage amplitude starting from the AP threshold until the start of the AP overshoot, and the rheobase is the minimum current needed to elicit an AP.
The voltage-dependent steady-state activation (SSA) curves were determined from I-V curves, where the Na 1 current was evoked from a holding potential of À100 mV to test pulses of 100 ms ranging from À80 to 140 mV in increments of 5 mV. For DRG neuron recordings, each step was preceded by a 3 s prepulse at À120 mV. The leakage current was subtracted by using the P/4 procedure in Clampex software. The sodium channel conductance was calculated following the formula G Na = I Na /(V m -V rev ), where I Na is the peak current amplitude, V m is the current potential, and V rev is the reversal potential for the current determined by a linear fit. The activation curve for each individual cell was fitted with a Boltzmann equation with the formula where G max is the maximum conductance, V 1/2 is the potential at which half of the sodium channels are activated, V m is the current potential, and k is the slope factor.
On the other hand, the voltage-dependent steady-state inactivation (SSI) curves were measured from a holding potential of À100 mV using 500 ms prepulses ranging from À130 to 15 mV, in increments of 5 mV, followed by a test pulse to 0 mV. The inactivation curves were fitted with the Boltzmann relationship as follows: where I max is the maximum value for the sodium current, V 1/2 is the potential at which I Na is half-inactivated, V m is the membrane potential achieved using a prepulse step, and k is the slope factor. Five nanomolar Protoxin II (ProTxII; catalog #P0033, Sigma-Aldrich) was used to isolate Na V 1.7-mediated currents from DRG primary culture (Schmalhofer et al., 2008).

Statistical analyses
For statistical analysis, we used GraphPad Prism and R software. Data are presented as a box plot with whiskers extending from the 10th to 90th percentile (minimum to maximum, including every individual point), or the mean 6 SEM. Unpaired t test with Welch's correction was used to compare the means between two groups, and two-way repeated-measures ANOVA, followed by a post hoc analysis using a Sidak's multiple-comparison test, was used to compare the effect of metformin treatment in different conditions; p , 0.05 was considered to be significant. The statistical tests, t, F, and p values, and the effect size calculated with Cohen's factors are provided in Extended Data Table 1-2.

Results
In vitro, the effect of metformin on the Na V 1.7 current density depends on the presence of NEDD4-2 We performed whole-cell patch-clamp experiments in HEK293 cells cotransfected with Na V 1.7 and different concentrations of NEDD4-2 plasmids. Figure 1, A and B, shows, respectively, representative traces of the Na V 1.7 current and the quantification of the maximum Na V 1.7 current density, from a holding potential of À100 mV to test pulses of 100 ms ranging from À80 to 140 mV in increments of 5 mV, in the different NEDD4-2 concentration conditions (Extended Data Fig. 1-2).
Metformin downregulates total I Na and Na V 1.7 current densities in DRG neurons via NEDD4-2 We used freshly dissociated L4 and L5 mouse small DRG neurons with a membrane capacitance ,30 pF, considered to be nociceptive neurons (Lopez-Santiago et al., 2006), and performed whole-cell patch-clamp recordings of total I Na and Na V 1.7 currents ( Fig. 2A,C, Extended Data Fig. 2-1). From total I Na currents, we pharmacologically isolated Na V 1.7 current by using 5 nM ProTxII, a concentration at which the toxin blocks Na V 1.7 selectively (Schmalhofer et al., 2008). To investigate the contribution of NEDD4-2 expression in the metformin-mediated downregulation of the Na V 1.7 current, we used a DRG neuronspecific NEDD4-2 knock-out mouse line (SNS-Nedd4L À/À ). Mice heterozygously expressing Cre recombinase under the control of the Na V 1.8 promoter (SNS-Cre; Agarwal et al., 2004) were crossed with homozygous mice carrying the Nedd4-2 floxed allele (Nedd4L fl/fl ; Shi et al., 2008). In the Nedd4L fl/fl control mice, the total I Na was significantly decreased by the metformin treatment (total I Na Nedd4L fl/fl , control vs metformin: p = 0.0139, t (47.72) = 2.554, n = 22 or 28/group; Fig. 2B, left). Conversely, the total I Na current density was not significantly changed in the DRG neurons from the SNS-Nedd4L À/À mice (total I Na SNS-Nedd4L À/À , control vs metformin: p = 0.3637, t (48.92) = 0.9169, n = 23 or 27/ group; Fig. 2B, right). Similarly, a .50% decrease in Na V 1.7 current density was observed on metformin incubation in the control mice (Na V 1.7 current density in Nedd4L fl/fl , control vs metformin: p = 0.0184, t (39.09) = 2.461, n = 19 or 23/ group; Fig. 2D, left), but not in the DRG neurons from the SNS-Nedd4L À/À mice (Na V 1.7 current density in SNS-Nedd4L À/À , control vs metformin: p = 0.4717, t (37.18) = 0.7271, n = 20/group; Fig. 2D, right). Furthermore, similar to the results in HEK293 cells, metformin did not modify the V 1/2 of activation of either total I Na or Na V 1.7 in DRG neurons from the Nedd4L fl/fl mice ( Table 2).

Control Metformin
Control Metformin Figure 1. The effect of metformin on the Na V 1.7 peak current depends on the presence of NEDD4-2 in HEK293 cells. A, Representative traces of Na V 1.7 recordings, at the maximum peak current, from Na V 1.7-transfected HEK293 cells, cotransfected with different concentrations of NEDD4-2 plasmid. These data are supported by further data presenting the multiple traces, obtained at each step of the protocol, in Extended Data Fig. 1-2. Cells were incubated with culture medium for control conditions (top) or 20 mM metformin (bottom) over 12 h. High [NEDD4-2] (ratio 1:1), Low [NEDD4-2] (ratio 1:0.3), and NEDD4-2CS (mutated form of NEDD4-2 with inactivated catalytic site) conditions are presented. B, Bar graphs presenting the Na V 1.7 peak current in the four conditions with control medium or 20 mM metformin (n = 12-47 cells recorded/group). C, Representative blots of surface biotinylation of HEK293 cells cotransfected with Na V 1.7 and Low [NEDD4-2]. The upper band corresponds to the fully glycosylated form of Na V 1.7, while the lower band corresponds to the core glycosylated form. D, Respective quantification of Na V 1.7 expression in total input fraction (S0) and biotinylated fraction (S2), with control medium or metformin treatment (n = 10/condition). Values were normalized to tubulin expression. These data are supported by further data presenting Na V 1.7 expression, with other concentration of NEDD4-2, in Extended Data Fig. 1-1. Data are represented as a box plot with whiskers extending from the 10th to the 90th percentile, and each dot represents one recorded cell (A) or one sample (D). Statistical analysis was performed using an unpaired t test with Welch's correction for both patch-clamp and Western blot datasets (Extended Data Table 1-2). *p , 0.05, **p , 0.01, ***p , 0.001. Detailed references for the resources used are presented in Extended Data Table 1-1. Table 2: Biophysical properties of total Na V and Na V 1.7-mediated currents in DRG neurons from Nedd4L fl/fl or SNS-Nedd4L 2/2 mice, with (1) or without (-) metformin treatment Nedd4-2 fl/fl mice SNS-Nedd4-2 -/mice Na V total Na V 1.7 Na V total Na V 1.7 Metformin -1 -1 -1 -1 SSA V 1/2 (mV) À37.1 6 2.3 À35.6 6 3.4 À42.8 6 3.2 À40.4 6 4.7 À37.0 6 2.7 À39.9 6 4.8 À36.5 6 14.0 À43.2 6 4.9 Slope 6.5 6 0.5 6.6 6 0.6 5.9 6 1.0 5.2 6 1.2 6.2 6 0.4 **8.2 6 0.4 7.0 6 1.5 5.9 6 0.6 n 15 10 6 5 12 8 3 6 The SSA was measured on total I Na current density or Na V 1.7-mediated currents for each genotype and treatment condition. The V 1/2 (mV) is the voltage at which half of the available channels are activated with the SSA protocol. Data are expressed as the mean 6 SEM. Statistical analysis was performed using an unpaired t test with Welch's correction (Extended Data efficiently to its substrate (e.g., Na V 1.7), promoting more internalization and reducing I Na and the Na V 1.7 current.

Discussion
Despite the recent development of Na V 1.7-specific inhibitors, small molecules, anti-Na V 1.7 antibodies, or other  Figure 2. Metformin reduced the total I Na and Na V 1.7 current densities in DRG neurons in a NEDD4-2-dependent manner. DRG neuronal primary cultures were incubated for 12 h with culture medium or 20 mM metformin. A, Representative traces of total I Na , at the maximum peak current, recorded with an I-V protocol in DRG neurons from Nedd4-L fl/fl mice or SNS-Nedd4-L À/À mice. The traces were normalized to the capacitance of the recorded cell and expressed in pA/pF over time. These data are supported by further data presenting the multiple traces, obtained at each step of the protocol, in Extended Data Figure 2-1. B, Quantified total I Na current density in the four conditions (n = 22-28 cells recorded/condition). C, Representative traces of Na V 1.7 current density, at the maximum peak current, recorded with a I-V protocol. Na V 1.7 currents were isolated from the total I Na by acute application of ProTxII at 5 nM during the recordings. The traces were normalized to the capacitance of the recorded cell and expressed in pA/pF over time. D, Quantified Na V 1.7 current density in the four similar conditions (n = 19-23 cells recorded/condition). Data are represented as a box plot with whiskers extending from the 10th to the 90th percentile (minimum to maximum), with each dot representing a single recorded cell. Statistical analysis was performed using an unpaired t test with Welch's correction (Extended Data Table  1-2). *p , 0.05, **p , 0.01, ***p , 0.001. Detailed references for the resources used are presented in Extended Data Table 1-1. new Na V isoform inhibitors (Alsaloum et al., 2020), alternative strategies could be represented by post-translational regulation of the channel activity for the treatment of chronic pain (Laedermann et al., 2015). Metformin has this potential and offers the advantage of being used in clinical settings (Asiedu et al., 2016;Demaré et al., 2021). We demonstrate here that metformin induces a decrease in Na V 1.7 channel expression through NEDD4-2 and ultimately reduces the excitability of the nociceptive neuron. Metformin can enhance the NEDD4-2-mediated epithelial Na 1 channel ENaC internalization, in an AMPK signaling-dependent manner (Bhalla et al., 2006). AMPK induces ENaC inhibition through the stabilization of the protein complex including NEDD4-2, thus producing ENaC retrieval from the plasma membrane, in which the binding motif in the cytoplasmic tail of the b -ENaC subunit is essential for the interaction with the ubiquitin ligase NEDD4-2 (Bhalla et al., 2006;Asher et al., 2003). Similarly, we found increased AMPK phosphorylation with metformin in HEK293 cells, as well as in cultured DRG neurons, and we observed that NEDD4-2 is essential for the metformin effect on Na V 1.7 expression at the plasma membrane and Na V 1.7 current density. Ultimately, restoring NEDD4-2 expression-for example, in experimental neuropathic pain-with a recombinant adeno-associated viral vector reduced the I Na current densities in DRG sensory neurons and repressed the development of mechanical allodynia following SNI (Laedermann et al., 2013).
Although the mechanisms linking metformin to AMPK phosphorylation have been well investigated, this drug is a promiscuous compound that affects multiple pathways. For its efficacy, metformin requires the protein-threonine kinase LKB1, which phosphorylates and activates AMPK (Woods et al., 2003). In addition, another upstream kinase of AMPK is Ca 21 /calmodulin-dependent protein kinase (CaMKK), which, when inhibited pharmacologically or downregulated by using RNA interference, abolished AMPK activation (Woods et al., 2005). Thus, metformin may act on different intracellular pathways, independent of AMPK, inducing changes in transcription factors and ion channel activity. For example, the pleiotropic effect of metformin was investigated in relation to the phosphorylation of different amino acids, but also with its interaction with the putative kinase targets such as SGK1 (serine/ threonine-protein kinase), MAP2K2 (dual-specificity mitogen-activated kinase kinase 2), MAPK14 (mitogen activated protein kinase 14), CDK7 (cyclin-dependent kinase 7), and EGFR (epidermal growth factor receptor), raising questions related to their downstream chain of events (Hart et al., 2016). In our study, we observed a decreased level of NEDD4-2 phosphorylated at Ser-328. This site has been found to be phosphorylated by  Table 1-2). *p , 0.05, **p , 0.01, ***p , 0.001. Detailed references for the resources used are presented in Extended Data Table 1-1. SGK1 (Debonneville et al., 2001;Flores et al., 2005). Flores et al. (2005) found that the phosphorylation of NEDD4-2 via SGK1 led to an impairment of ENaC-NEDD4-2 interaction and consequently to more channels at the cell surface. Interestingly, Bhalla et al. (2006) showed an AMPK-dependent modulation of NEDD4-2 controlling ENaC activity; however, they found that the SGK1 pathway was not involved in the AMPK regulation of ENaC. Because the levels of phosphorylated Ser-328 were altered, our results suggest that the SGK-1 pathway could be involved in this effect. It would be an interesting pathway to explore in parallel to AMPK signaling, especially because the sites of NEDD4-2 phosphorylated by AMPK are not well identified (Arévalo, 2015). In cultured sensory trigeminal (TG) neurons, metformin induces an increase in AMPK activation and the inhibition of the mTOR pathway (Melemedjian et al., 2011). Metformin, as well as AMPK activators, can abrogate the phosphorylation of mTOR, 4EBP, and rS6 (for which phosphorylation is increased after nerve injury). The authors also suggested that pharmacological AMPK activation (including by using metformin) negatively normalizes aberrant translational control after peripheral nerve injury, a phenomenon that contributes to reduced pain hypersensitivity in experimental models of neuropathic pain (SNI and SNL).
Metformin could act transcriptionally and post-translationally to decrease neuronal excitability. Interestingly, AMPK activators could potentially have similar mechanisms, as for the allosteric modulator A769662: acute application of this compound blocked in a dose-dependent manner voltage-gated sodium channels in rat TG neurons primary cultures (Asiedu et al., 2017). The effect of metformin via NEDD4-2 could also occur through the modification of protein expression. We found that the amount of Na V 1.7 in whole-cell lysates of cultured DRG neurons was decreased after metformin treatment, and this regulation was dependent on NEDD4-2. Thus, NEDD4-2 acts on other regulatory machinery of membrane protein internalization, and this could happen directly through an enzymatic post-translational ubiquitination or indirectly. Interestingly, Grimsey et al. (2015) described an atypical role for NEDD4-2 E3 ubiquitin ligase. They showed that different G-protein-coupled receptors could stimulate p38-MAPK activation via NEDD4-2-mediated ubiquitination (Grimsey et al., 2015). Thus, NEDD4-2 may also regulate ion channel protein expression via p38-MAPK signaling. Consequently, it was shown that metformin regulates the remodeling of the small conductance calcium-activated potassium channels (SK2 and SK3) in the atrial tissues of a rat model of type 2 diabetes mellitus through the inhibition of NOX4 expression (constitutive NAPDH oxidase), but also by significantly suppressing the p38-MAPK signaling pathway (Liu et al., 2018).
Regulation of I Na was also investigated in relation with metformin treatment and interactions with cytoskeleton binding proteins. In the human breast cancer MCF-7 cell line, metformin decreases phosphorylation of cofilin, an actin-binding protein involved in cytoskeleton dynamics,  which is upstream regulated by LIM-kinase and Rho-kinase, suggesting a possible interaction of metformin with Rho-kinase (Özdem _ Ir and Ark, 2021). Furthermore, in chick DRG neurons Rho-kinase phosphorylates the collapsin response mediator protein 2 (CRMP2), and inactivates the ability of CRMP2 to promote microtubule assembly and Numb-mediated endocytosis during growth cone collapse (Arimura et al., 2005). Beside the complex with Numb, CRMP2 interacts also with Eps15 and NEDD4-2 to promote the endocytosis of Na V 1.7 as a specific silencing of these targets changed the sodium currents (Dustrude et al., 2016;Cai et al., 2021;Gomez et al., 2021). But this appears only in relation with the CRMP2 post-translational modification state of SUMOylation. Therefore, we do not exclude this pathway by which metformin, mediated by Rho-kinase and CRMP2, may interact with NEDD4-2 to promote Na V 1.7 internalization and loss of sodium currents. The recordings are acquired from at least three different patch-clamp sessions. A, Representative traces of APs in different conditions elicited by ramp protocols of 500 ms (inset). B, Box plots showing the analysis of APs number in Nedd4L fl/fl mice (top; n = 14-16/treatment) and SNS-Nedd4L À/À mice (bottom; n = 11-16/treatment); data are represented as a box plot with whiskers extending from the 10th to the 90th percentile (minimum to maximum), with each dot representing a single recorded cell. Statistical analysis was performed using a two-way repeated-measures ANOVA, followed by Sidak's multiple-comparisons post hoc test (*p , 0.05, **p , 0.01, ***p , 0.001; Extended Data Table 1-2). C, Representative single AP and its parameters in different conditions, as illustrated by the color code. D, Analysis of AP parameters, which are significantly changed on metformin treatment in Nedd4L fl/fl mice (n = 14 cells/treatment) or SNS-Nedd4L À/À mice (n = 10-16 cells/treatment); data are represented as a box plot with whiskers extending from the 10th to the 90th percentile (minimum to maximum). These data are supported by further data presenting the AP characteristics, obtained from DRG neurons of Nedd4L fl/fl or SNS-Nedd4L À/À mice treated with or without metformin, in Extended Data Figure 5-1. Statistical analysis was performed using an unpaired t test with Welch's correction (Extended Data Table 1-2). *p , 0.05, **p , 0.01, ***p , 0.001. Detailed references for the resources used are presented in Extended Data Table 1-1. AMPK acts on NEDD4-2 and consequently on the ion channels expressing the intracellular binding motif. This fact confirms our data on sodium current densities in both the heterologous and homologous primary systems in which treatment with metformin decreased the total I Na and Na V 1.7 currents with only a marginal effect on the biophysical properties of the channel. The modulation by metformin determined by NEDD4-2, however, is not only restricted at sodium channels. Treatment of mouse polarized kidney cortical cells with metformin inhibits K V 7.1 activity by promoting NEDD4-2-dependent channel ubiquitination (Alzamora et al., 2010). Interestingly, the expression and current levels of K V 1.4 are downregulated in an AMPK-PKC manner independent of NEDD4-2 or the related ubiquitin ligase NEDD4-1, and independent of the phosphoinositide 3kinase (PI3K)-SGK1 signaling pathway, suggesting that NEDD4-2 probably exerts different functions depending on the binding motif of the cytoplasmic tail of ion channels (Andersen et al., 2018).
Our current-clamp data are in line with previously published studies indicating that metformin can reduce neuronal excitability, lowering the frequency of APs (Melemedjian et al., 2011). Na V 1.7 is one of the most important members of the fast kinetics tetrodotoxin-sensitive sodium channels that contributes to the rising phase of APs (Bennett et al., 2019). Therefore, metformin treatment, lowering Na V 1.7 expression, reduces the amplitude and maximum rise time of single APs, whereas the baseline is higher. We have shown that metformin is also able to partially reduce neuronal excitability in NEDD4-2 knock-out animals, alongside changes in the rheobase and RMP, indicating a possible effect through different pathways and channels.
Finally, a recent study showed that the effect of metformin was not limited to peripheral neurons, but metformin treatment also altered the activation of spinal microglia and astrocytes in male mice after SNI (Inyang et al., 2019). The authors described several mechanisms of action of metformin on the androgen-dependent regulation of OCT2 (organic cation transporter-2), but the possibility that metformin could also regulate microglial ion channels should not be excluded. Indeed, different potassium currents are increased in activated microglia (e.g., Kir2.1mediated currents; Madry et al., 2018;Gattlen et al., 2020), which are also regulated by NEDD4-2, and could be the target of metformin.
Our study was not conceived to answer the question of a gender effect during metformin treatment, but the issue is of particular interest. The link between the cytoskeleton binding protein CRMP2 SUMOylation and Na V 1.7 showed a reduction of expression and currents in female mice compared with males (Moutal et al., 2020). Moreover, knocking down the expression of NEDD4-2 with silencing RNA in DRG sensory neurons restored the loss of sodium currents, in CRMP2 SUMO-null knock-in (CRMP2K374A/ K374A) female mice . Apart from the link among CRMP2, NEDD4-2, and Na V 1.7 that is gender different, the treatment with metformin reverses SNIinduced mechanical and cold hypersensitivity in male but not in female mice. This effect might be due, at least in part, to how metformin differently activated microglial cells between the two genders, whereas AMPK activity increases in DRG neurons in both male and female (Inyang et al., 2019). Whether changes into AMPK activity, NEDD4-2, and Na V 1.7 currents after metformin administration is gender dependent need to be further investigated.
In conclusion, we have provided novel data regarding the mechanism of action of metformin on sodium channels, which play a crucial role in physiological and pathologic pain. Comprehension of those mechanisms opens new alternatives for the diminution of hyperexcitability related to neuropathic pain, in particular by targeting posttranslational regulation of sodium channels.