Melatonin alters the excitability of mouse cerebellar granule neurons by inhibiting voltage‐gated sodium, potassium, and calcium channels

Besides its role in the circadian rhythm, the pineal gland hormone melatonin (MLT) also possesses antiepileptogenic, antineoplastic, and cardioprotective properties, among others. The dosages necessary to elicit beneficial effects in these diseases often far surpass physiological concentrations. Although even high doses of MLT are considered to be largely harmless to humans, the possible side effects of pharmacological concentrations are so far not well investigated. In the present study, we report that pharmacological doses of MLT (3 mM) strongly altered the electrophysiological characteristics of cultured primary mouse cerebellar granule cells (CGCs). Using whole‐cell patch clamp and ratiometric Ca2+ imaging, we observed that pharmacological concentrations of MLT inhibited several types of voltage‐gated Na+, K+, and Ca2+ channels in CGCs independently of known MLT‐receptors, altering the character and pattern of elicited action potentials (APs) significantly, quickly and reversibly. Specifically, MLT reduced AP frequency, afterhyperpolarization, and rheobase, whereas AP amplitude and threshold potential remained unchanged. The altered biophysical profile of the cells could constitute a possible mechanism underlying the proposed beneficial effects of MLT in brain‐related disorders, such as epilepsy. On the other hand, it suggests potential adverse effects of pharmacological MLT concentrations on neurons, which should be considered when using MLT as a pharmacological compound.


| INTRODUCTION
Although best known as a neurohormone released at night from the pineal gland, melatonin (MLT) has also been proposed to be an ancient molecule that first emerged in bacteria 3.0-2.5 billion years ago. 1 Its primary function in these cells likely were its antioxidant properties.Over the course of many million years, other functions of MLT evolved in multicellular animals, including its effects on the circadian rhythm.MLT's long evolutionary history may partially explain the plethora of effects that have been attributed to MLT in humans.It is being increasingly used in research and in clinical settings in an attempt to treat a variety of conditions such as depression,2 epilepsy, 3 hypertension, 4 cardiac ischemia, 5 osteoporosis, 6 and cancer. 7To induce therapeutic effects in those diseases, the necessary MLT dosage frequently reaches high pharmacological concentrations in the millimolar range, whereas its physiological plasma concentrations are in the pico-and nanomolar range.However, it should be noted that our knowledge of MLT's concentrations inside specific tissues is too limited to determine, which concentrations of MLT are "physiological."Furthermore, some studies indicate that compartmentalization in certain areas of the body, for instance, in the bile and cerebrospinal fluid, may permit the natural accumulation of up to millimolar concentrations of endogenous MLT. 8,9 Irrespective of these uncertainties, we will refer to micro-and millimolar concentrations of MLT as "pharmacological concentrations" in the present study.
The effects of pharmacological MLT doses on the human body have only been studied sparingly so far.In 1967, Barchas et al. 10 attempted to determine a lethal dose in mice, to no avail.The authors stopped at an administered concentration of 800 mg/kg, as they reached the limit of MLT's solubility in blood.Even at such high concentrations, the mice did not show any large changes in behavior, other than an increased amount of sleep.To this day, the lethal dose of MLT, or whether MLT is lethal at all, remains unknown.][13] In a current clinical trial (NCT04568863), physicians are administering 5 mg MLT per kg body weight and per day in an attempt to reduce coronavirus disease 2019 (COVID-19) mortality in an intensive care unit.A recent meta-analysis concludes that high doses of MLT are safe to intake, as they cause no severe side effects and only occasionally induce mild adverse effects mostly limited to drowsiness, dizziness and headaches. 14However, comprehensive randomized controlled studies addressing the safety of pharmacological doses of MLT are still lacking and skepticism remains towards the many reported effects of pharmacological MLT concentrations. 15,16Our knowledge of the concentrations that can be reached in humans through the administration of pharmacological MLT doses is also limited, in particular due to the firstpass metabolism, which would quickly reduce the dose of MLT applied to humans. 17,18n physiological concentrations, MLT acts via specific G protein-coupled receptors, MT1 and MT2, which have nanomolar affinity for MLT. 19The effects of pharmacological MLT concentrations on a plethora of pathologies are often attributed to its antioxidant effects.Although MLT has antioxidant properties, whether it can act as a reactive oxygen species scavenger in vivo and the pathways through which it exerts its many MLTreceptor-independent effects remain unclear.As its pharmacological efficacy is debated, MLT is sold as an over-the-counter dietary supplement in most countries and it is generally considered safe for short-term use.Therefore, more extensive studies on higher MLT doses are necessary so that we can better understand both the beneficial mechanisms of action of pharmacological doses, as well as predict possible side and long-term effects.
Regarding the effect of MLT on ion channels and the electrophysiological properties of neurons, pharmacological concentrations have been shown to significantly affect different voltage-gated ion channels.For example, 1-100 μM MLT inhibited high-voltage activated Ca 2+ channels in cultured rat dorsal root ganglion neurons. 20And micro-and millimolar levels of MLT inhibited voltage-gated Ca 2+ (Ca v ) channels and neurotransmitter release in PC12 cells and in rat hippocampal slices. 21In MCF-7 breast cancer cells, MLT greatly altered voltage-gated Ca 2+ and K + currents. 22Additionally, millimolar concentrations were able to inhibit K v 1.3 currents in T-lymphocytes and direct binding to the channels has been proposed. 23Most of these pharmacological effects appear to be largely independent of MT1-and MT2-receptors.Their exact nature remains unknown, but may in part rely on direct interactions of MLT with specific ion channels.Physiological MLT doses on the other hand affect ion channels primarily through receptor-dependent mechanisms and are better understood. 24,25he effects of both physiological and pharmacological doses of MLT on cerebellar granule cells (CGCs), the most numerous neurons in the brain, are currently unknown.In the present study, we discovered that pharmacological concentrations of MLT inhibit voltagegated Na + (Na v ), K + (K v ), and Ca 2+ (Ca v ) channels in primary mouse CGCs.Furthermore, we report how this inhibition of voltage-gated currents affects the excitability of CGCs.In summary, we present new data about how high MLT concentrations affect the biophysical properties of mature neurons.

| Animals
CGCs were isolated from 7-day-old C57BL/6 wild-type mice.Mice were housed and handled according to guidelines from the Federation for European Laboratory Animal Science Associations.They were housed in a temperature-controlled room (20°C-24°C) with a 12 h light/dark cycle and with food and water ad libitum in a pathogen-free facility.Mice were killed by decapitation.The use of organs for scientific purposes was approved by the committee of the animal welfare officers of the Institute for Laboratory Animal Science & Experimental Surgery and Central Laboratory for Laboratory Animal Science at the Uniklinik RWTH Aachen under the internal number AZ 80012/A4.

| Isolation and maintenance of primary mouse CGCs
To make the Trypsin/DNAse solution, the following compounds were mixed: 100 mg Trypsin (Sigma-Aldrich) + 10 mg DNAse I (Sigma-Aldrich) + 100 μL MgCl 2 (80 mM). 9 mL HBSS (PAN-Biotech) was added and the pH was adjusted to 7.8.Finally, the volume was adjusted to 10 ml with HBSS.It was filtered sterile and stored in 1 and 2 mL aliquots at −80°C.To make the DNAse I solution, the following compounds were mixed: 10 mg DNAse I (Sigma-Aldrich) + 50 mg glucose + 20 ml Neurobasal A medium (Gibco, Thermo Fisher Scientific).The solution was filtered sterile and stored in 1 and 2 mL aliquots at −80°C.
Isolation of primary mouse CGCs was performed as described by Girbes et al. 26 WT mice pups were killed at P7 via decapitation.The head was placed on ice.The skin and skull were removed with forceps and scissors to expose the undamaged brain.After careful removal of bone and skin fragments around the cerebellum, the latter was positioned between the opened forceps and gently detached from the rest of the brain.The isolated cerebellum was put in ice-cold HBSS and cleaned carefully under a stereomicroscope (Stemi DV4, Carl Zeiss AG) from any foreign structures such as connective tissues and other brain stem structures.Afterwards, it was cut into three smaller pieces.The isolated pieces of the cerebella from different mice were transferred from the ice-cold HBSS to a clean 15 cm Falcon tube (Sarstedt AG & Co. KG).The pieces were washed three times with 5 ml ice-cold HBSS per three cerebella.HBSS was removed and 1 mL per three cerebella of a Trypsin/ DNAse solution was added.The cerebella were incubated at room temperature (RT) for 15 min and then washed again with 5 mL ice-cold HBSS per three cerebella.HBSS was removed and 1 mL per three cerebella of a DNAse I solution was added.Now, the pieces of the cerebella were mechanically homogenized with glass Pasteur pipettes with a continuouisly decreasing diameter.To test whether the cerebella were appropriately homogenized, they were put on ice for 5 min.If no precipitate was visible after 5 min, they were homogenized successfully into single cells.If a precipitate was visible, the homogenization steps were repeated.Next, the homogenized samples were centrifuged for 15 min at 100 g and 4°C.The supernatant was aspirated and the neurons were resuspended in 5 ml BrainPhys TM medium/three cerebella (prewarmed to 37°C).Finally, the number of neurons was determined with a Neubauer chamber.For RT-qPCR and patch clamp experiments, CGCs were seeded at a density of 2 000 000 cells/mL on coated glass cover.Neurons were put in the incubator for at least 24 h before any consequent experiment was performed.
CGCs were cultured in BrainPhys TM neuronal medium supplemented with SM1 according to the manufacturer (STEMCELL Technologies Inc.).Before seeding, 96-well plates and glass cover slips were coated with 0.1% poly-L-lysine (Sigma-Aldrich) for at least 30 min at 37°C.

| Whole-cell patch clamp
CGCs were seeded on coated cover slips and incubated at least 7 days before recording.Next, the cover slips were mounted in a perfused bath on the stage of an inverted microscope (IX71, Olympus) and kept at RT. Patch-clamp experiments were performed in the whole-cell configuration.Patch pipettes had an input resistance of 8-10 MΩ.
Currents were recorded using a patch-clamp amplifier (Axopatch 200 B), the Axon-CNS (Digidata 1440A), and Clampex software (Molecular Devices).Data were filtered at 1 kHz with a low-pass filter and were analyzed with the PCLAMP software.The sampling rate was 20 kHz.

| Ratiometric Ca 2+ imaging
To determine intracellular Ca 2+ concentrations, CGCs were seeded on cover slips, mounted in a cell chamber and perfused as described in the Section 2.4.Fluorescence was measured every 2 s on an inverted microscope (IX71, Olympus, Chromaphor) using a Fluar ×20/0.75objective (Olympus) and Till Vision real-time imaging software (Till Photonics).Cells were loaded for 30 min at 37°C with 2 μM Fura-2-AM (Thermo Fisher Scientific) in the bath solution.Fura-2 was excited at 340/380 nm and the emission was recorded between 470 and 550 nm using a sensicam CCD camera (PCO Imaging).Acquisition and data analysis were done using the Till Vision software and Excel.
The bath solutions used for the Ca 2+ imaging experiments consisted of the same bath solution described in Section 2.4 that was used to measure APs and K v channels.The 30 mM K + solution contained the following compounds (in mM): NaCl 85, KH 2 PO 4 0.4, K 2 HPO 4 1.6, KCl 28, D-glucose 5, MgCl 2 1, HEPES 5, sodium gluconate 25, and calcium gluconate 3. pH was adjusted to 7.4 with NaOH and HCl.

| Statistical analysis
Whole-cell patch clamp experiments with n = 4 are shown as box plots with whiskers (Figures 1C and 2B) and were analyzed with a nonparametric Friedman test (Figure 2B).For whole-cell patch clamp experiments with n > 4, normal distribution was tested via Shapiro-Wilk tests.In case of normal distribution, the data are shown as mean ± SD and were analyzed with a paired Student's t test or repeated-measures analysis of variance, as appropriate.In case of nonnormal distribution, the data are shown as box plots with whiskers and were analyzed with a Wilcoxon signed-rank test or a Friedman test, as appropriate.Whiskers extend to the minimum and to the maximum value of the data set, respectively.For Ca 2+  For all whole-cell patch clamp experiments, MLT was preincubated for 30 s before measurements.We used CGCs from 7-day-old mice and performed current clamp in the whole-cell configuration to inject current into CGCs and to elicit APs.Although all CGCs were excitable, their AP frequency varied considerably.Therefore, we divided the cells into three different groups, depending on the number of APs that was elicited by the injection of a rheobase current (Figure 1A,B).Type I cells generated one AP, Type IIa cells two to four APs, and Type IIb cells >4 APs.Strikingly, 3 mM MLT dramatically reduced the number of APs in Type IIa and IIb cells to one AP (Figure 1A,C) and consequently also the number of spikes/s in both IIa and IIb cells (Figure 1D).In contrast, MLT had no effect on the number of APs in Type I cells, which produced only one AP also in the absence of MLT.MLT did not affect threshold potential and AP amplitude (Figure 1E,F).However, it did reduce both the amplitude of afterhyperpolarizations (AHP) and of the rheobase (Figure 1G,H).

| Pharmacological concentrations of MLT inhibit Na v currents in CGCs
To uncover the basis for the strongly reduced excitability of CGCs, we investigated how MLT affected different voltage-gated channels, starting with Na v channels, as they are responsible for forming APs.We isolated Na v currents in CGCs by depolarizing the cells from a holding potential of −120 mV to voltages ranging from −80 to +60 mV in 10 mV steps of 50 ms duration each; K + channels were blocked by TEA-Cl and 4-AP in the bath and by TEA-Cl and Cs + in the pipette solution.Ca v channels were blocked by Cd 2+ in the bath solution.We found that 3 mM MLT reduced the amplitude of Na v currents at −10 mV approximately twofold (Figure 2A,B).This inhibition was fully and quickly reversible.Application of luzindole, a commonly used MT1/2 inhibitor, did not counteract the Na v inhibition by MLT, suggesting that the inhibition was most likely receptor independent (Figure 2A,B).Using different concentrations of MLT, we found that millimolar concentrations of MLT are required to inhibit Na v s; concentrations up to 500 μM had no noticeable effect on Na v current amplitude (Figure 2C), confirming that the effect of MLT on Na v s was receptor independent.

| Pharmacological concentrations of MLT inhibit an outwardly rectifying K + (ORK) current
Next, we isolated an ORK current by depolarizing the cells from a holding potential of −90 mV to voltages ranging from −70 to +50 mV in 20 mV steps of 150 ms duration each; Na v channels were blocked by TTX in the bath.This current only slowly and partially inactivated (Figure 3A). 3 mM MLT reversibly inhibited the ORK current at +50 mV approximately twofold (Figure 3A).The ORK current had a reversal potential more negative than −70 mV, indicating that it was a K + current.Replacing K + with Cs + , which inhibits K + currents, in the pipette fully inhibited the ORK current, confirming that it was carried by K + channels (Figure 3B).To determine which K + channels were responsible for the ORK current, we first, considered that depolarization opened Ca v channels, which in turn activated Ca 2+activated K + channels.To address this possibility, we applied IbTx, a specific inhibitor of large conductance Ca 2+ -activated K + channels (BK channels).In another experiment, we removed all Ca 2+ ions from the extracellular solution.Both manipulations did not affect the ORK current, ruling out the involvement of BK channels (Figure 3C).In agreement, addition of ω-Aga IVA, a specific inhibitor of P/Q-type Ca v channels, also had no effect on the ORK current (Figure 3A).

| The ORK current of CGCs is carried by K v channels
We next used 20 μM nimodipine, which partially inhibited the ORK current, similar to MLT (Figure 4A).
Nimodipine is best known as an inhibitor of several Ca v channels, but it is a fairly unspecific drug that also inhibits several types of K v channels. 27The involvement of K v channels was confirmed by using three unselective inhibitors of K v channels: verapamil, quinidine, and TPA.9][30][31] Quinidine, better known as an inhibitor of Na + channels, unspecifically inhibits delayed rectifier K + channels and A-type K + channels. 32,335][36] All three inhibitors largely inhibited the ORK current similar to MLT (Figure 4B).Combined with our findings reported in Figure 3, this proves that it is indeed carried by K v channels.However, α-DTx, an inhibitor of K v 1.1, K v 1.2, and K v 1.6, had no effect on the ORK current, showing that these K v channels were not involved (Figure 4C).Finally, we used different concentrations of MLT and found that MLT inhibits K v channels at lower concentrations than Na v channels (Figure 4D); 500 μM MLT already significantly inhibited the ORK current.The insensitivity of the ORK current to 10 μM MLT (Figure 4D), however, again indicates a receptorindependent mechanism.

| Pharmacological concentrations of MLT inhibit Ca v currents in CGCs
As MLT inhibited Na v and K v channels, we addressed whether it can also affect Ca v channels.We isolated Ca v currents by depolarizing the cells from a holding potential of −60 mV to voltages ranging from −60 to +60 mV in 10 mV steps of 300 ms duration each; K + channels were blocked by TEA-Cl in the bath and by TEA-Cl and Cs + in the pipette solution.Na v channels were blocked by TTX in the bath.As expected for the tiny capacitance of CGCs, being among the smallest neurons in the CNS, they consistently produced very small, but significant Ca v currents (Figure 5A,B).Strikingly, 3 mM MLT inhibited the Ca v current almost completely (Figure 5A,B).Next, we determined whether MLT would shift the membrane potential necessary to inactivate Ca v channels with an inactivation protocol.Starting from a holding potential of −60 mV, the cells were first depolarized for 1250 ms with varying preconditioning pulses, ranging from −60 to +30 mV in 10 mV steps, to measure the voltage dependence of Ca v channel inactivation.Next, a short conditioning pulse was applied for 1.25 ms at −60 mV.Then, a 0 mV test pulse was applied for 42.5 ms and the resulting current was measured.Finally, the cells were returned to the holding potential of −60 mV to remove inactivation of Ca v channels.To avoid complete inhibition, we used 10 μM MLT, which significantly inhibited the Ca 2+ signal in Ca 2+ imaging experiments (see below, Figure 7C).However, there was no noticeable shift in the current-voltage curve by 10 μM MLT (Figure 5C).
Next, we studied how MLT affects Ca v channels in CGCs in greater detail using Ca 2+ imaging.Ionomycin, a calcium ionophore, was used a positive control at the end of each recording.Application of 100 nM MLT, a physiological concentration, or 3 mM MLT, a pharmacological concentration, to CGCs had no effect on the intracellular Ca 2+ concentration [Ca 2+ ] i (Figure 6A).In contrast, when we applied 30 mM K + to depolarize the cells and to open Ca v channels, [Ca 2+ ] i robustly increased (Figure 6B).Repeated stimulations with 30 mM K + slightly reduced the [Ca 2+ ] i increase by 10%-20%.To minimize the influence of the gradual reduction of the Ca 2+ signal during repeated stimulations, we randomly applied stimulating solutions in all subsequent experiments.Co-application of 30 mM K + with different concentrations of MLT showed that MLT inhibited the high K + -induced increase in [Ca 2+ ] i (Figure 6C).For all Ca 2+ imaging experiments, MLT was co-applied with 30 mM K + without any preincubation.As even 10 μM MLT inhibited the Ca 2+ signal by about 40%, Ca v channels of CGCs appeared to be more sensitive to MLT than Na v and K v channels.However, as 100 nM MLT did not cause inhibition (Figure 6C), also the effect on Ca v s was receptor independent.

| Inhibition of Ca v currents by MLT is MT1/2 receptor and G βγ independent
Considering the comparatively low MLT concentrations required for inhibiting Ca 2+ signals, we specifically assessed whether this inhibition was MT1/2 receptordependent.Co-application of 1 μM luzindole with MLT and 30 mM K + did not affect the reduction of [Ca 2+ ] i by MLT, indicating that the inhibition of Ca v s was indeed receptor-independent (Figure 7A,B).Even at a very high concentration (50 μM), luzindole did not affect the MLTmediated reduction of [Ca 2+ ] i (Figure 7C).However, at this concentration, luzindole elicited an unspecific, MLTindependent elevation of [Ca 2+ ] i (Figure 7C), which has previously observed by others in pancreatic acinar cells and had been attributed to luzindole-induced release of Ca 2+ from the endoplasmic reticulum. 37T1/2 receptors couple to G proteins and the G βγ subunit has previously been demonstrated to be important in many MLT-mediated effects, especially those associated with ion channels. 25,38,39Therefore, to assess the possibility of a G βγ -mediated effect, we pre-and coapplied gallein, a G βγ subunit inhibitor, with MLT and 30 mM K + .Gallein had no effect on the MLT-mediated inhibition of Ca v channels (Figure 7D), demonstrating that it was also independent of G βγ subunits.

| Pharmacological concentrations of MLT inhibit several types of Ca v channels in CGCs
1][42][43][44] Pre-application of 100 nM ω-Aga IVA followed by co-application of 1 μM ω-Aga IVA 42 with 30 mM K + reduced the Ca 2+ signal by approximately 54%, suggesting that about 50% of the Ca v -mediated Ca 2+ signal was carried by P-and Q-type channels (Figure 8A).Co-application of 100 nM ω-Aga IVA with 30 mM K + and 3 mM MLT further reduced the Ca 2+ signal, indicating that MLT inhibits other types of Ca v channels as well.
We then repeated the experimental protocol with a co-application of 1 μM nimodipine, which reduced the Ca 2+ signal more than ω-Aga IVA (Figure 8A), suggesting that another 15% of the Ca v -mediated Ca 2+ signal was carried by L-type channels and that the remaining Ca 2+ signal was probably a mixture of T-, N-and R-type currents.As the application of MLT plus nimodipine caused an even stronger inhibition than nimodipine alone, MLT probably inhibited several types of Ca v channels, including, but not limited to, P-/Q-/L types.Figure 8B summarizes these observations.
When comparing the sensitivity of MLT to different voltage-gated channels, we conclude that it appears most sensitive to Ca v , followed by K v and then Na v channels (Figure 8C).For a better interpretation of the summarizing chart, it should be noted that the Ca v currents were only measured indirectly through Ca 2+ imaging and might therefore not precisely represent the actual currents.

| DISCUSSION
In our study, we discovered that pharmacological concentrations of MLT quickly and reversibly inhibited Na v , K v , and Ca v channels.Moreover, we found that MLT reduced AP frequency, AHP, and rheobase.To our knowledge, this is the first study reporting that MLT affects the excitability of CGCs, which is among the most abundant neurons of the CNS.The inhibition of multiple voltage-gated channels was likely independent of MT1 and MT2 receptors.The evidence for this conclusion is severalfold.First and most importantly, whereas MT receptors have nanomolar affinity, at least 10 μM MLT were necessary to inhibit Ca v s; for inhibition of Na v s and K v s, concentrations needed to be even higher.Second, even for Ca v s luzindole did not reverse the inhibiton by MLT, also not when applied at comparable concentrations with MLT (1 μM applied with 10 μM MLT; Figure 7A,B).Further increasing luzindole concentrations was not possible due to unspecific effects (Figure 7C).Third, the inhibition was also not reversed by gallein, which would have been expected if MT1/G βγ signaling was involved, as has been demonstrated for the effects of MLT on BK channels and Ca v 2.2. 25,39Fourth, inhibition had rapid kinetics and was observed in some experiments without any apparent delay (Figures 6-8).We speculate that MLT inhibited these diverse ion channels via a common and unspecific mechanism, such as a membrane-mediated mechanism.Indeed, previous studies have shown that at pharmacological concentrations, MLT integrates into lipid head groups of lipid bilayers, thus reducing membrane thickness and increasing membrane fluidity, 45,46 and it is known that membrane properties influence the activity of ion channels. 47,48n primary mouse CGCs, we observed differences in the excitability of individual neurons and classified neurons that fired only one AP as "Type I" and neurons that fired APs repeatedly as "Type II."Previously, it has been found that immature CGCs produce low frequency APs, often being carried by Ca v instead of Na v channels. 49ature CGCs produce fast and high-frequency spikes instead. 50,51Cells with mature repetitive Na v spikes appear as early as P4, with their number increasing drastically up to P21. 49 Considering that the CGCs we used for patch clamp recordings were isolated from P7 mice and cultured for at least seven days, we propose that "Type I" cells represent an immature class of CGCs and that "Type II" neurons, with their fast-spiking nature, appear to better represent typical mature neurons.The most striking effect of 3 mM MLT on CGC excitability was a strong reduction of AP frequency.In most CGCs, MLT reduced it to just one AP, even in type IIB characterized by a continuous burst pattern.
3][54] Concerning the cerebellum, Zhang et al. 24 reported that in Purkinje neurons, nanomolar levels of MLT induced hyperexcitability by inhibiting P-type Ca v channels MT1 receptordependently.The reduced Ca v current in turn most likely resulted in a decreased Ca 2+ -activated K + current, hence increasing AP firing rate.
Reports on whether MLT can affect Na v currents remain sparse.The only study on this topic demonstrates that low concentrations of MLT potentiate Na v currents receptor-dependently in ocular tissue. 55Hence, to our knowledge, we are the first to observe that pharmacological MLT concentrations can partially inhibit Na v channels in neurons.As the inhibition was quick and reversible, it might have been caused by a direct interaction of MLT and Na v channels.It is possible that the inhibition of Na v currents by MLT contributes to the reduced excitability of CGCs. 56On the other hand, the AP amplitude of CGCs was not significantly affected by 3 mM MLT, suggesting that the effect of MLT on Na v channels was relatively small.Voltage-clamp experiments indeed suggested that MLT had the lowest affinity to Na v channels.High concentrations were required to partially inhibit currents.
It has previously been demonstrated that fast spiking CNS neurons such as CGCs owe their high frequency to an interplay between Na v and K v 3 channels. 57,58Mature CGCs show a strong expression of K v 3.1 and K v 3.3, and also express K v 1.1 and K v 1.2. 59Immature CGCs also have a robust expression of K v 3.1, but only a weak expression of K v 1.1 and K v 1.2. 60The relative immature nature of the CGCs in our study might explain why α-DTx, a specific inhibitor of K v 1.1, K v 1.2 and K v 1.6, had no effect on the total K v current.MLT-mediated inhibition of both Na v and K v 3 currents could result in a disruption of their interplay and significantly reduce AP frequency, 57 as we have observed.Pharmacological inhibition of K v 3 channels has already been shown to reduce fast spiking in CGCs and other neurons. 61,62We, therefore, hypothesize that the MLT-mediated reduction in CGC excitability is likely a result of the inhibition of different voltage-gated channels.Furthermore, we also noticed a reduction in AHP under the influence of MLT.Considering that K v 3.1 is known for its repolarizing effects in CGCs, 63 this further supports our hypothesis of a MLT-mediated K v 3.1 inhibition.However, we cannot rule out that MLT inhibits more types of K v channels such as K v 1.3.Varga et al. 23 previously demonstrated that MLT inhibits K v 1.3 channels in T-lymphocytes by directly interacting and binding to them with an IC 50 (1.5 mM) that is in a similar range as in our study.5][66] In summary, we conclude that pharmacological MLT concentrations caused mature CGCs to fire significantly less APs, but reduced the stimulus necessary in eliciting APs slightly.
MLT-mediated reduction in AP frequency might not be linked to just Na v and K v channels.It is possible that the observed MLT-mediated inhibition of Ca v channels could lead to a reduced activation of Ca 2+ -activated K + channels, such as BK.In turn, the reduced BK activity would lead to slower AHP, thus reducing AP frequency. 67,68Intriguingly, a recent publication presents an MT1-dependent mechanism in which physiological concentrations of MLT activate BK through MT1 and the G βγ subunit of the G protein. 25 On the other hand, 100 μM MLT strongly inhibited BK currents in MCF-7 cells. 22astly, we showed that MLT likely possesses the highest affinity to Ca v channels, inhibiting them already at low micromolar concentrations.Since the activity of Ca v channels was measured mostly indirectly through Ca + imaging, slight differences compared with direct current measurements are possible.Similar to the inhibition of Na v s and K v s, the inhibition of Ca v s was quick, reversible and independent of MT1/2 receptors and of G βγ signaling.The fact that MLT probably inhibited diverse Ca v channels, including P-, Q-, and L-type currents, further hints towards a receptorindependent effect of MLT on these channels.A mechanism involving signaling cascades would be expected to be slower and more specific to certain types of Ca v channels.Similar to our study, micro-and millimolar concentrations of MLT quickly and reversibly inhibit all types of Ca v currents in PC12 cells, and it has been suggested that Ca v s directly interact with MLT in these cells. 21n the future, electrophysiological recordings with cerebellar slices could uncover how MLT-mediated Ca vinhibition would affect synaptic transmission and plasticity.Inhibition of Ca v channels by MLT might reduce excitatory postsynaptic potentials (EPSPs) by reducing the amount of neurotransmitter release, hence decreasing fast synaptic transmission. 69,70ur report of a broad inhibition of diverse voltagegated channels could contribute to understand MLT's antiepileptic properties.Although MLT has been successfully used in experimental models of epilepsy and to treat epilepsy in patients as well, 13,71,72 its underlying mode of action remains unknown.Given the importance of many Ca v , Na v , and K v channels for the pathophysiology of seizures, 73 an inhibition of voltage-gated channels by pharmacological MLT concentrations could alter the biophysical properties of neurons in such a way as to reduce epileptogenic potentials.The therapeutic effect of MLT on other neuropsychiatric disorders, such as depression and insomnia, might also be in part explained by our observations. 74n the other hand, we expect that such a dramatic reduction of the excitability of a very abundant neuron may have primarily adverse effects.In this respect, our observations could contribute to understanding the short-term side effects of MLT.The National Institutes of Health lists possible mild side effects such as dizziness, drowsiness and nausea. 75Rarer adverse effects include paradoxical agitation and even an increase in aggressive behavior. 76,77Drastic changes in the electrophysiological profile of CNS neurons might be linked to these effects.
Finally, the broad inhibition of voltage-gated channels could contribute to MLT's presumed anticancer properties against a variety of malignant entities, such as glioblastoma, leukemia and Ewing sarcoma, [78][79][80] which are mainly based on in vitro experiments using cell lines.These properties are mostly independent of the MT1/2 receptors and are poorly understood.Some publications point towards the antioxidant quality of MLT being key in understanding its oncostatic and cytotoxic effects. 81owever, given the broad spectrum of tumor entities that are affected by MLT, other mechanisms are likely involved as well.In recent years, voltage-gated channels have been located in almost all tumors, even though they consist of mostly non-excitable tissue.This conserved and often high expression suggests a role in cancer.3][84] Therefore, a potential inhibition of voltage-gated channels by MLT could be another mode of action through which MLT exerts its effects on cancer cells.
To conclude, we report that pharmacological concentrations of MLT induced a quick, broad and reversible inhibition of Ca v , K v , and Na v channels, with decreasing order of affinity.This inhibition is most likely the result of a receptor-independent interaction between MLT and voltage-gated channels, which might be mediated by the plasma membrane.The result was a strong alteration in the excitability of mature CGCs.To our knowledge, the present study is the first to demonstrate such a multifaceted effect of MLT on voltage-gated channels in neurons, specifically in CGCs.Our results may help to understand the therapeutic properties of MLT, for instance its antiepileptic and anticancer effects.They may also assist in illuminating possible neurological side effects in patients treated with high MLT concentrations.

AUTHOR CONTRIBUTIONS
Karolos-Philippos Pissas performed all experiments, except that CGCs were isolated together with Maria Schilling.Karolos-Philippos Pissas, Ahmet Korkmaz, Yuemin Tian, and Stefan Gründer designed the study.Ahmet Korkmaz, Yuemin Tian, and Stefan Gründer supervised the study.Karolos-Philippos Pissas wrote the original draft and Stefan Gründer reviewed and edited the original draft with important input from all other authors.

F
I G U R E 1 Melatonin (MLT) changes the firing properties of mouse cerebellar granule cells (CGCs).(A) Representative traces of action potentials (APs) induced by a rheobase current before (control, ctrl), during (MLT) and after (wash) application of 3 mM MLT in three different cells.MLT was preincubated for 30 s. (B) Relative proportion of the three types of CGCs.(C) Number of APs elicited with or without MLT (Type 1, n = 7; Type IIa, n = 5; Type IIb, n = 4).Graphs summarizing spikes/s (n = 9, D), threshold potential (n = 16, E), AP amplitude (n = 16, F), amplitude of afterhyperpolarization (n = 16, G), and rheobase (n = 16, H) of CGCs treated with or without 3 mM MLT. Results are shown as box plots (C) or as bar graphs representing mean ± SD (D-H) and were obtained from two independent CGC cultures.Statistical analysis was performed with paired Student's t tests.n.s., no statistical significance; *p < .05;**p < .01;***p < .001.

F
I G U R E 2 Melatonin (MLT) inhibits voltage-gated sodium channels (Na v s) in mouse cerebellar granule cells (CGCs) independently of MT1/MT2.(A) Representative traces of isolated Na v currents in a single cell treated with the following three conditions: ctrl, 3 mM MLT, 3 mM MLT plus 1 μM luzindole (luz).(B) Left, current-voltage curves (I-V curves).Right, summary of current amplitudes at −10 mV (n = 4).(C) Left, I-V curves of isolated Na v currents in cells treated with different MLT concentrations.Right, summary of current amplitudes at −10 mV (n = 7).MLT and luz were preincubated for 30 and 60 s, respectively.I-V curves represent the mean of results obtained from two independent CGC cultures and results are shown as box plots (B) or as bar graphs representing mean ± SD (C).Statistical analysis was performed with a Friedman test (B) or a repeatedmeasures one-way analysis of variance (C).n.s., no statistical significance; *p < .05.

F
I G U R E 3 Melatonin (MLT) inhibits an outwardly rectifying K + current (ORK current) in mouse cerebellar granule cells (CGCs).(A) Left, representative traces of the ORK current before (ctrl), during (MLT), and after (wash) application of 3 mM MLT in the same cell.Center, I-V curves of the ORK current in cells treated with 1 μM ω-Agatoxin IVA (ω-Aga IVA), 3 mM MLT, or 3 mM MLT plus 1 μM ω-Aga IVA.Right, summary of current amplitudes at +50 mV (n = 5).(B) Left, I-V curves of an ORK current recorded under control conditions (ctrl) or with Cs + in the pipette solution.Right, summary of current amplitudes at +50 mV (n = 6).(C) Left, I-V curves of the ORK current in cells treated with 100 nM iberiotoxin (IbTx) or a Ca 2+ -free bath solution.Right, summary of the current amplitudes at +50 mV (n = 6).MLT and IbTx were preincubated for 30 and 60 s, respectively.For (A), cells were preincubated with 100 nM ω-Aga IVA for 10 min before stimulation with 1 μM ω-Aga IVA.I-V curves represent the mean of results obtained from two independent CGC cultures and results are summarized as box plots (A) or as bar graphs representing mean ± SD (B, C).Statistical analysis was performed with a Friedman test (A), a paired Student's t test (B) or with a repeated-measures one-way analysis of variance (C).n.s., no statistical significance; **p < .01;***p < .001.

F
I G U R E 4 Melatonin (MLT) inhibits voltage-gated potassium channels (K v channels) in mouse cerebellar granule cells (CGCs).(A) Left, I-V curves of the outwardly rectifying K + (ORK) current in cells treated with 3 mM MLT or 20 μM nimodipine (nimo).Right, summary of current amplitudes at +50 mV (n = 6).(B) As in A, but in cells treated with the general K v channel inhibitors quinidine (100 μM), verapamil (100 μM), tetrapentylammoniumchlorid (TPA; 100 mM), or 3 mM MLT (n = 6).(C) As in (A), but in cells treated with 100 nM α-Dendrotoxin (α-DTx) (n = 6).(D) As in (A), but in cells treated with different MLT concentrations (n = 6).MLT, nimo and α-DTx were preincubated for 30 s, 120 s and 300 s, respectively.Quinidine, verapamil, and TPA were preincubated for 60 s each.I-V curves represent the mean of results obtained from two independent CGC cultures and results are shown as box plots (A, B) or as bar graphs representing mean ± SD (C, D).Statistical analysis was performed with Friedman tests (A, B), a paired Student's t test (C) or with a repeatedmeasures one-way analysis of variance (D).*p < .05,**p < .01versus ctrl (A-D).

F
I G U R E 5 Melatonin (MLT) inhibits voltage-gated calcium channels (Ca v channels) in mouse cerebellar granule cells (CGCs).(A) Representative traces of isolated Ca v currents of the same cell in the absence or presence of 3 mM MLT. (B) Left, I-V curves of isolated Ca v currents of cells in the absence or presence of 3 mM MLT. Right, summary of current amplitudes at 0 mV (n = 6).(C) Effects of different voltage prepulses on isolated Ca v currents at 0 mV in the presence of 10 μM MLT using an inactivation voltage-step protocol (n = 6).MLT was preincubated for 30 s. I-V curves represent the mean of results obtained from two independent CGC cultures and bar graphs the mean ± SD.Statistical analysis was performed with a paired Student's t test (B).**p < .01.

F
I G U R E 6 Melatonin (MLT) inhibits the increase in intracellular Ca 2+ concentration [Ca 2+ ] i induced by high K + in mouse cerebellar granule cells (CGCs).(A) Trace illustrating the average [Ca 2+ ] i of all cells treated with 100 nM and 3 mM MLT (n = 53).(B) Left, trace illustrating the average [Ca 2+ ] i of all cells repeatedly depolarized with 30 mM K + .Right, summary data (n = 229).(C) Top, pseudo-colored images of representative experiments showing relative levels of Ca 2+ concentrations as visualized by Fura-2 340/380 nm fluorescence.Bottom left, trace illustrating the average [Ca 2+ ] i of all cells treated with a co-application of 30 mM K + plus different MLT concentrations.Bottom right, summary data (n = 202).Ionomycin (Iono) was used as a positive control.MLT was not preincubated.Results are shown as median with box plots and obtained from at least two independent CGC cultures.The whiskers show the minimum and maximum values.Statistical analysis was performed with Friedman tests (B, C). ****p < .0001.n.s., no statistical significance.

F
I G U R E 7 Melatonin (MLT) inhibits Ca v channels independently of MT1, MT2, and G β/γ in mouse cerebellar granule cells (CGCs).(A) Left, average Ca 2+ trace of all cells treated with co-application of 30 mM K + , 1 μM luzindole (luz) plus different MLT concentrations.Right, summary data (n = 196).(B) Concentration-response curves of the inhibitory effect of MLT on Ca v currents with and without 1 μM luz.Data for MLT without luz are taken from Figure 6.(C) Left, average Ca 2+ trace of all cells treated with a co-application of 30 mM K + , 3 mM MLT plus 50 μM luz.Right, summary data (n = 137).(D) As in C, but with a co-application of 30 mM K + , 3 mM MLT plus 10 μM gallein (n = 260).MLT was not preincubated.For A and B, luz was preincubated for at least 120 s and was present throughout the entire recordings.For C, luz was preincubated for 120 s.For D, gallein was preincubated for 180 s.Results are shown as box plots and were obtained from at least two independent CGC cultures.Statistical analysis was performed with a Wilcoxon signed-rank test (C) and Friedman tests (A, D). ****p < .0001versus ctrl (A, C). n.s., no statistical significance.

F
I G U R E 8 Melatonin (MLT) inhibits several types of Ca v channels in mouse cerebellar granule cells (CGCs).(A) Summary of normalized Ca v signals in all cells treated with a co-application of 3 mM MLT plus 30 mM K + plus 1 μM ω-agatoxin IVA (ω-Aga IVA) (n = 237) or 1 μM niodipine (nimo; n = 473).Before co-application, the cells were pre-incubated with 100 nM ω-Aga IVA for 10 min.(B) Approximate Ca v channel subunit distribution in CGCs based on the effects of ω-Aga IVA and nimo.(C) Concentration-response curves of the inhibitory effect of MLT on K v /Na v currents (left y axis; data taken from whole-cell patch clamp experiments) and on Ca v signals (right y axis; data taken from Ca 2+ imaging experiments).MLT was not preincubated.For A, cells were preincubated with 100 nM ω-Aga IVA for 10 min before stimulation with 1 μM ω-Aga IVA.Nimo was preincubated for 60 s.Results are shown as box plots were and obtained from at least two independent CGC cultures.Statistical analysis was performed with Wilcoxon signed-rank tests (A).****p < .0001.