Peimine, an Anti-Inflammatory Compound from Chinese Herbal Extracts, Modulates Muscle-Type Nicotinic Receptors

Fritillaria bulbs are used in Traditional Chinese Medicine to treat several illnesses. Peimine (Pm), an anti-inflammatory compound from Fritillaria, is known to inhibit some voltage-dependent ion channels and muscarinic receptors, but its interaction with ligand-gated ion channels remains unexplored. We have studied if Pm affects nicotinic acetylcholine receptors (nAChRs), since they play broad functional roles, both in the nervous system and non-neuronal tissues. Muscle-type nAChRs were incorporated to Xenopus oocytes and the action of Pm on the membrane currents elicited by ACh (IAChs) was assessed. Functional studies were combined with virtual docking and molecular dynamics assays. Co-application of ACh and Pm reversibly blocked IACh, with an IC50 in the low micromolar range. Pm inhibited nAChR by: (i) open-channel blockade, evidenced by the voltage-dependent inhibition of IAch, (ii) enhancement of nAChR desensitization, revealed by both an accelerated IACh decay and a decelerated IACh deactivation, and (iii) resting-nAChR blockade, deduced from the IACh inhibition elicited by Pm when applied before ACh superfusion. In good concordance, virtual docking and molecular dynamics assays demonstrated that Pm binds to different sites at the nAChR, mostly at the transmembrane domain. Thus, Pm from Fritillaria bulbs, considered therapeutic herbs, targets nAChRs with high affinity, which might account for its anti-inflammatory actions.


Introduction
The scientific interest in Traditional Chinese medicines (TCMs) has bloomed in the last decades, as they provide a broad source of compounds of putative clinical relevance. The active ingredients of TCM plants include: (i) alkaloids, such as peimine (Pm), found in Fritillaria bulbs (Fb), which are commonly used to treat cough and asthma [1,2], (ii) terpenoids, as ginsenoides from Panax ginseng, which might reverse multidrug resistance of certain chemotherapeutic drugs [3], (iii) phenols and flavonoids (polyphenolic compounds), which are common in medicinal plants and own strong antioxidant and anti-inflammatory activities [4], tannic acid-related gallotannins and polyphenols, which are relatively abundant in green tea, inhibit TMEM16A, a calcium-activated chloride channel [5] and some flavonoids, as quercetin or genistein, act as positive allosteric modulators of α7 nicotinic receptors [6], and (iv) other compounds, as cinnamaldehyde, obtained from cinnamon cortex, which is known to activate TMEM16A [7].
Pm, also known as verticine, has been related to diverse therapeutic actions [8], including: (i) anti-inflammatory and analgesic, (ii) antitumor, inhibiting proliferation of human Correlation of structural data, concerning the specific binding sites of Pm at the nAChR, with the functional effects elicited by this molecule should contribute to expand our understanding of the modulation of nAChRs by molecules of therapeutic relevance.

I ACh Blockade by Pm
The membrane conductance of oocytes either uninjected or bearing microtransplanted nAChRs was unaffected by bathing the cell with Pm (up to 100 µM) while holding the membrane potential at −60 mV. By contrast, co-application of ACh (10 µM) together with Pm (0.02-100 µM) to microinjected oocytes reversibly reduced the peak-amplitude (I p ) of I ACh , in a dose-dependent manner ( Figure 1B), following a sigmoid function ( Figure 1C). At Pm concentrations over 0.1 µM, the extent of I ACh inhibition measured 20 s after I p (I ss ) was greater than that corresponding to I p values. In this way, the IC 50 and n H values (see equation 1) for the I p were 2.9 µM (confidence interval (CI), 2.0-4.3 µM; n = 5-16, N = 3-8) and 0.7 ± 0.1, respectively, whereas the dose-inhibition curve for the I ss displayed a lower IC 50 (1.2 µM; CI 0.9-1.5 µM) but a similar slope (0.8 ± 0.1; the same cells and donor frogs as above; Figure 1C). Most likely, this lower IC 50 for I ss is because of the enhancement of nAChR desensitization by Pm (see below).
The specificity of Pm effects on muscle-type nAChR blockade was assessed by testing its effects on GABA subtype A receptors (GABA A Rs), which belong to the same Cys-loop family. For these experiments, oocytes were microinjected with rat brain synaptosomal membranes, which allowed the incorporation of GABA A Rs into the oocyte membrane [21]. These cells were later challenged with GABA (1 mM) either alone or together with Pm (up to 100 µM). Both I GABA amplitude and kinetics were rather unaffected by the presence of Pm (Supplementary Figure S1), in contrast to the marked inhibition of muscle-type nAChRs by Pm.

Competition Assays
The pharmacological profile of nAChR inhibition by Pm was determined by superfusing oocytes with ACh at different concentrations (3, 10, 30, 100 µM, and 1 mM) either alone or together with Pm at a concentration close to its IC 50 (3 µM). Co-application of 10 µM ACh with 3 µM Pm halved I p , as expected from the computed IC 50 ( Figure 1C). Similar percentages of I p inhibition were found when Pm was co-applied with different ACh concentrations (Figure 2A-C), suggesting that Pm is acting as a non-competitive blocker. Notably, the percentages of I ss inhibition attained by co-applying Pm (3 µM) with different ACh concentrations were higher than those corresponding to the I p inhibition (Figure 2A,C). Furthermore, the percentage of I ss inhibition increased significantly as the ACh concentration augmented ( Figure 2C). This ACh concentration-dependence of I ss inhibition might be related to the enhancement of nAChR desensitization by Pm (see below), since the rate of desensitization is known to be dependent on ACh concentration [22,23].   10 µM ACh either alone (Ctr) or co-applied with different Pm concentrations, as stated on the right. Note that Pm accelerates I ACh decay when applied at concentrations of 0.1 µM or above. Hereafter, unless otherwise stated, the holding potential was −60 mV, downward deflections represent inward currents and the bars above recordings indicate the timing of drug application. (C) Pm concentration-I ACh inhibition relationship. I ACh amplitudes at their peak (I p ; filled symbols) and at their steady state (I ss , measured 20 s after the peak; open symbols) were normalized to the I ACh evoked by ACh alone and plotted against the logarithm of Pm concentration. Solid and dashed lines are sigmoid curves fitted to I p and I ss data, respectively. Error bars indicate SEM. Each point is the average of 5-16 oocytes from 3-8 frogs.  (*) indicates significant differences between I p and I ss inhibition, for each ACh concentration (p < 0.05, paired t-test). (#) indicates significant differences among I ss inhibition at 10 µM ACh and other concentrations (p < 0.05, ANOVA followed by Bonferroni t-test). Each point of panels B and C is the average of 4-14 cells from 1-2 donors.

Enhancement of nAChR Desensitization by Pm
As previously demonstrated, co-application of ACh with Pm at concentrations over 0.1 µM accelerated I ACh decay ( Figure 1B), which, in turn, resulted in a large percentage of I ss inhibition, as compared to I p ( Figure 1B,C). This acceleration of I ACh decay by Pm might be due to: (i) slow I ACh blockade, (ii) enhancement of nAChR desensitization, and (iii) a combination of both. Aiming to unravel the involvement of each one of these options in boosting I ACh decay we conducted several complementary experimental protocols. First, ACh (10 µM) was co-applied with different Pm concentrations (0.05-100 µM) and both the time to reach I p (aTtP) and the I ACh decay kinetics were determined ( Figure 3).  . Pm accelerates I ACh decay and shortens the time to reach I p . (A 1 ,A 2 ) Superimposed I ACh s elicited by 10 µM ACh either alone (black and grey recordings) or together with different Pm concentrations (shown at right). I ACh s were scaled to the same I p amplitude to better compare the differences in time to reach I p (aTtP; A 1 ) and kinetics of I ACh decay after I p (A 2 ). (B 1 ,B 2 ) Column graphs displaying Pm effects on aTtP (B 1 ) and τ-values of I ACh -decay (B 2 ). (*) indicates significant differences among I ACh s in presence of Pm (colored columns; same color code as in (A 1 ,A 2 )) and their control values (Ctr, black column; p < 0.05, ANOVA and Bonferroni t-test). Note that post-control values (after Pm applications; grey column) were similar to control ones. Each point is the average of 4-24 cells (N = 3-12).
Acceleration of I ACh decay was dependent on Pm concentration and significantly increased at concentrations over 0.5 µM Pm ( Figure 3A 1 ,A 2 ,B 2 ). Actually, the I ACh decay time-constant (τ) decreased from roughly 30 s for control I ACh s to less than 2 s when ACh and 100 µM Pm were co-applied ( Figure 3B 2 ). Furthermore, Pm elicited changes in the aTtP ( Figure 3B 1 ), which paralleled fairly well the acceleration of I ACh decay ( Figure 3B 2 ). Noticeably, a shortened aTtP has been previously related to enhancement of nAChR desensitization, as reported for the action of both lidocaine and 2,6-dimethylaniline (DMA) on this receptor [24,25].
Moreover, the kinetics of I ACh tails (deactivation) differed depending on the presence or absence of Pm while rinsing ACh out. In these experiments, 100 µM ACh, which elicits large nAChR desensitization, was bathed alone or together with either 1 or 5 µM Pm for 32 s. Afterwards, ACh was removed, keeping the cell superfused with normal Ringer with atropine (ANR) either alone or together with 1 or 5 µM Pm ( Figure 4). As previously demonstrated, 1 µM Pm accelerated I ACh decay and this effect was more pronounced when the cell was bathed with 5 µM Pm ( Figure 4A 1 ,A 2 ,B 1 ). Thus, the ratio of I ACh decay time constant values in the presence of 1 or 5 µM Pm versus those in ACh alone were significantly smaller than 1 (0.67 ± 0.04 and 0.48 ± 0.05, respectively; same cells in both groups; n = 9, N = 3; p < 0.001, one-sample t-test). Furthermore, 5 µM Pm elicited a greater nAChR blockade and faster I ACh decay as compared to 1 µM Pm (1.18 ± 0.05 s against 0.83 ± 0.06 s for I ACh decay at 1 and 5 µM Pm, respectively; p = 0.002, paired t-test, same cells as above). As previously reported [26], deactivation of control I ACh s, elicited by ACh washout, followed an exponential function with a time course of roughly 1.5 s (this value likely limited by the solution exchange kinetics [26]).  . Pm superfusion remained 12 s after ACh washout (as indicated by the application bars). These recordings were normalized to either the same I p , to better compare their I ACh decay (A 2 ), or the same I ss , to facilitate comparisons of deactivation kinetics (A 3 ). (B 1 ,B 2 ) Column bar plots displaying the effect of 1 (orange) or 5 µM (purple) Pm on the I ACh decay time-constant (τ Desensitization ; (B 1 )) and the deactivation kinetics (τ Deactivation ; (B 2 )), as compared to control I ACh s (in the presence of ACh alone; black). (*) indicates significant differences with the control group (p < 0.05, paired t-test) and (#) indicates differences between 1 and 5 µM Pm groups (n = 9, N = 3; the same cells for all comparisons; p < 0.05, paired t-test). Notice that Pm accelerated the desensitization rate and slowed down the deactivation kinetics.
Noticeably, the presence of Pm in the ANR decelerated I ACh deactivation in a dosedependent manner (τ Deactivation of 2.1 ± 0.2 s and 2.9 ± 0.3 s for I ACh s in the presence of 1 µM and 5 µM Pm, respectively; p = 0.044, same cells as above; Figure 4B 2 ). In fact, the ratios of τ Deactivation of I ACh s in the presence and the absence of Pm were 148 ± 17% and 206 ± 30%, for 1 and 5 µM Pm, respectively. These later percentages most likely are underestimated since the actual ratios are conditioned by the apparent kinetics of control I ACh deactivation. Remarkably, if Pm actually enhances nAChR desensitization, a deceleration of I ACh deactivation is expected, since desensitized nAChRs display a higher affinity for ACh [22,27].

nAChRs Blockade by Pm Is Voltage-Dependent
Pm contains a protonable amine group (see Figure 1A), being its strongest basic pK a 10.56 (data from Chemicalize, https://chemicalize.com/; accessed on 30 March 2017). Consequently, more than 99.9% of Pm molecules are in a charged form at the recording pH. To unravel if I ACh inhibition exerted by Pm is voltage-dependent, voltage jumps (from −120 to +60 mV, in 20 mV steps) were imposed to oocytes superfused with ANR or during the I ACh plateau elicited by 10 µM ACh, either alone or co-applied with 1 or 5 µM Pm ( Figure 5A).
The i/v curves of net I ACh s elicited in the presence of Pm demonstrated that nAChR blockade by Pm was voltage-dependent, the more hyperpolarized the cell membrane, the larger the I ACh blockade ( Figure 5). Furthermore, i/v curves displayed a reversal potential close to −5 mV, indicating that channel-permeability properties were unaffected by the presence of Pm. Noteworthy, 1 µM Pm only inhibited I ACh at negative potentials, suggesting that Pm is electrostatically blocking the channel pore and, thus, causing an open-channel blockade. The relationship between the percentages of I ACh remnant and voltage displayed a slight slope at negative potentials ( Figure 5C), suggesting that Pm binds into the channel pore at a shallow site from the extracellular side. The increase of Pm concentration to 5 µM enhanced I ACh inhibition and the blockade was evident both at positive and negative potentials ( Figure 5B,C). Thus, at +60 mV, the I ACh remnant in the presence of 5 µM Pm was 0.66 ± 0.02, n = 5, N = 2 (p < 0.05, one sample t-test) whereas I ACh was unaffected by 1 µM Pm (I ACh remnant 0.96 ± 0.05, n = 11, N = 3; p > 0.05, one sample t-test). Therefore, Pm, at 5 µM, should bind to additional sites beyond those involved in open-channel blockade.

Kinetics of nAChR Blockade by Pm and Its Recovery
To better characterize the open-channel blockade of nAChRs by Pm, we tested the effect of a 20 s pulse of either 1 or 5 µM Pm applied during the I ACh plateau elicited by a 50 s pulse of 10 µM ACh ( Figure 6A,B). Pm addition to the bathing solution evoked a marked I ACh inhibition, decreasing 42 ± 2% (n = 10, N = 5) and 71 ± 1% (n = 8, N = 5) by 1 and 5 µM Pm, respectively (p < 0.05, t-test). The τ values for the fast blockade phase, estimated by fitting a single exponential function to the recordings, were 5.3 ± 0.2 and 3.1 ± 0.3 s for 1 and 5 µM Pm, respectively (same cells as above; p < 0.05, t-test; Figure 6C).
These τ values were rather slow, as compared to those previously reported, under similar experimental conditions, for I ss inhibition elicited by different local anesthetics (LAs), as tetracaine or benzocaine, at their IC 50 . In fact, for these latter compounds, τ values were faster than the solution exchange kinetics (roughly 1.5 s [26,28]). Furthermore, the voltage-dependent blockade of nAChRs by Pm allowed us to determine more accurately the rate of open-channel blockade, by jumping the membrane potential during the I ACh elicited in the presence of Pm. Thus, during the I ACh plateau elicited, at −60 mV, by 10 µM ACh either alone or in the presence of 1 or 5 µM Pm, the membrane potential was stepped to +40 mV for 2 s, to remove the open-channel blockade elicited by Pm. Following this pulse, the membrane potential returned to −60 mV, to determine the kinetics of the openchannel blockade by Pm ( Figure 7A 1 ,B 1 ). The I ACh blockade by Pm followed an exponential function ( Figure 7A 2 ,B 2 ) with τ values of 2.25 ± 0.12 s (n = 10, N = 3) and 1.01 ± 0.08 s (n = 7; N = 2) for 1 and 5 µM Pm, respectively. As expected, the kinetics of the voltage-dependent blockade of I ACh accelerated by increasing Pm concentration (p < 0.05, t-test; Figure 7C). These τ values were shorter than those obtained by direct Pm superfusion during the I ACh plateau ( Figure 6), most likely because they were not affected by the kinetics of solution exchange.   )). Notice that 1 µM Pm, in contrast to 5 µM, did not significantly decrease I ACh at +60 mV.  Pm, applied at the I ACh plateau. I p s were normalized to the same amplitude to facilitate the kinetics comparisons. The kinetics of I ACh inhibition and its recovery from blockade followed exponential functions (green traces (A,B)). (C) Column graph of the τ values found for I ACh blockade onset ("On" columns) when 1 (orange) or 5 µM (purple) Pm was co-applied with 10 µM ACh. The "Off" columns correspond to the kinetics of recovery (τ) from blockade, following Pm removal (same color code). (*) indicates significant differences of τ values between Pm concentrations for either "On" or "Off" data (p < 0.05, ANOVA and Bonferroni t-test). (#) denotes differences between "On" and "Off" values for either 1 or 5 µM Pm (p < 0.05, paired t-test). Data are for 10 and 8 oocytes (N = 5) for 1 µM and 5 µM Pm, respectively.  The recovery of nAChR from Pm blockade was also rather slow, in the range of several seconds, and followed a single exponential function ( Figure 6A-C). However, in contrast to the development of nAChR blockade, the recovery decelerated by raising Pm concentration ( Figure 6C).

Effect of Pm-Application Timing and Holding Potential on nAChR Blockade
As aforementioned, when ACh is co-applied with Pm, up to 1 µM, I ACh is mostly inhibited by open-channel blockade. This blockade is voltage-dependent and requires the channel opening by the agonist. However, increasing Pm to 5 µM, unraveled an additional blocking mechanism, which could not be precluded by applying positive-voltage pulses ( Figure 5B At 1 µM Pm, direct co-application of Pm and ACh decreased I ACh roughly 40% when the membrane potential was held at −60 mV, but it had almost no effect at +40 mV ( Figure 8A1,C). When Pm was pre-applied before superfusing ACh to the cell, there was a slight I ACh inhibition (roughly 10%), but only at the negative potential ( Figure 8A 2 ), indicating that some Pm molecules bound at the channel pore pathway. Accordingly, Pm pre-application followed by its co-application with ACh blocked I ACh roughly the same degree as merely its co-application with the agonist (compared to Figure 8 panels A 1 and A 3 ). When Pm concentration increased to 5 µM, its co-application with ACh, at −60 mV, significantly enhanced I ACh inhibition, as compared to that elicited by 1 µM Pm (Figure 8 B 1 ,C) and, interestingly, it also blocked nAChRs at +40 mV ( Figure 8B 1 ,C). Just pre-application of 5 µM Pm inhibited I ACh both at positive and negative potentials, although the I ACh blockade at +40 mV was roughly one third of that elicited at −60 mV (20.4 ± 1.7%, n = 8, versus 7.0 ± 2.2%, n = 6, for −60 and +40 mV, respectively; p < 0.01, t-test; Figure 8C). Thus, Pm could also block resting nAChRs and their recovery, after Pm washed out, was very slow, requiring over 30 s for a full recuperation (see Figure 8A 2 ,B 2 ). When 5 µM Pm was pre-applied and then co-applied with ACh, the I ACh inhibition, at either −60 or +40 mV, was quite similar to that evoked by solely Pm and ACh co-application at the respective potential, (Figure 8B 1 ,B 3 ,C), likewise as when applying 1 µM Pm (see above).

Molecular Docking and Dynamics Simulations of Pm-nAChR Complexes
Several docking solutions were found when performing 999 Pm runs for each nAChR conformation. The choice of the best solutions was based on the combination of two criteria: the best calculated binding energy (∆G ≤ −9.3 kcal/mol) and the most crowded clusters. Most Pm clusters were located at the TMD, in both the open (1-13) and the closed (1-10) conformations ( Figure 9A,B and Supplementary Table S1). TMD clusters were located at either intra-or inter-subunit crevices, without a clear preference for any subunit (see Supplementary Table S1). Noticeably, three solutions situated Pm interactions within the channel pore; two of them were in the open conformation (Pm 1 and Pm 2; Figure 9A,C), whereas the other was found in the closed-state (Pm 1, Figure 9B,D and Supplementary  Table S1). Although Pm is a highly hydrophobic molecule ( Figure 1A), and therefore it should preferentially distribute within the lipid bilayer, several clusters (sites 14-17 and 11-16 in the open and closed structures, respectively) localized at ECD. Some of these clusters were sited at the interface of the α-γ and α-δ subunits in both states, close to the orthosteric binding sites ( Figure 9A,B and Supplementary Table S1).  (C) Column graph shows the percentages of I p inhibition by Pm when applied as indicated in panels (A 1 -A 3 ,B 1 -B 3 ), at −60 mV (on the left) and +40 mV (on the right). (*) indicates significant differences between I p inhibition elicited by 1 and 5 µM Pm (p < 0.05, t-test). (#) denotes significant differences, for each Pm concentration, among the percentages of I p inhibition elicited by ACh and Pm co-application and other Pm-application protocols, at the same holding potential (p < 0.05, ANOVA and Bonferroni t-test). Each column is the average of 5-11 and 5-17 oocytes, for −60 mV and +40 mV, respectively. Molecular dynamics simulations for each Pm cluster bound to the TMD and ECD of lipid-reconstituted nAChR is presented in Figure 10, for the open (panel A) and closed (panel B) conformations. Each plot displays the trajectory of a Pm molecule from the selected clusters through a 100 ns simulation. Root mean square deviation (RMSD) values lower than 10 Å correspond to rearrangements of the Pm into their binding site through the simulation period, since Pm molecule is circa 12.5 Å long. Remarkably, all Pm molecules, except Pm 16 of the closed-state ( Figure 10B), remained steadily bound at their initial binding site on the nAChR through the 100 ns simulation.    To further verify the stability of the Pm-nAChR complexes, we carried out an evaluation of the MM/PBSA, which estimates the free energy of the binding of small ligands to biological macromolecules. MM|PBSA average binding energies were computed for Pm bound to both the open ( Figure 10C,D, for TMD and ECD, respectively) and closed ( Figure 10E,F for TMD and ECD, respectively) conformations. Several aspects should be highlighted regarding these results: (i) for all Pm molecules, the average values calculated for the last 60 ns are very close to those obtained for the last 30 ns, which suggests that the system stabilizes during the first 40 ns of simulation, (ii) the binding free energy values for the different Pm molecules are quite similar, regardless of whether we consider the open-or closed-states ( Figure 10C-F), (iii) Pm binding to ECD, both for the closed-and open-states, is commonly weaker than its binding to TMD, and (iv) high MM|PBSA values (as much as 70 kcal/mol) were reached for some Pm binding sites (see Figure 10C,E).
Channel pore hydration analysis displays a continuous water column when the nAChR is in the open-state, whereas it is interrupted in the upper half of the M2 transmembrane segment in the closed-state, indicating the presence of a hydrophobic gate [29]. This partial dehydration of the extracellular half of the channel pore interrupts the connection between the extra-and intracellular media and is a consequence of the conformational changes of the M2 helices [29]. Consequently, the number of water molecules into the extracellular half of the pore region might be considered a gating indicator. The effect of Pm binding to different sites of the nAChR on both the empty volume of the channel-pore and the number of water molecules within this region are displayed in Figure 11, for both the open and the closed-states. In the absence of Pm (i.e., control conditions ( Figure 11A,B)), the average of water molecules located within the hydrophobic gate was 48 for the openstate whereas for the closed-state was only 4. Even more, in the closed-state, most of the time there were no water molecules within this channel-pore region. The empty volume of the hydrophobic gate region was roughly 2900 Å 3 Figure 11C,D, which displays the structural changes elicited when Pm binds to three representative sites: Pm 1 (located shallow within the channel pore), Pm 6 (sited at TMD, in a crevice of the β subunit), and Pm 14 (bound to the ECD, at the interface between α-δ subunits; see Figure 9C). Noticeably, all of these binding sites decreased the number of water molecules to almost none ( Figure 11D), and both Pm 6 and Pm 14 markedly decreased the empty volume of the pore ( Figure 11C). By contrast, Pm 1, located within the hydrophobic gate, did not reduce the empty volume of this zone (≈3100 Å 3 ; Figure 11C), most likely because Pm bound to this channel-pore region precluded its narrowing.
Regarding the closed-state, Pm 1 (located inside the pore), Pm 2 (docked to the TMD at crevices in the interface between β-δ subunits) and Pm 13 (sited at the ECD, at the interface between α-γ subunits) are representative examples of the structural effects mediated by Pm when binding to different nAChR regions (see Figure 9D). As compared to the control situation, Pm 1 neither affects the number of water molecules at the hydrophobic gate region nor the empty volume of this zone, which remained close to 1800 Å 3 ( Figure 11E,F). Interestingly, the effects of Pm 2 and Pm 13 on closed nAChRs were quite similar ( Figure 11E,F), even though their binding sites were located far away from each other. In fact, both Pm 2 and Pm 13 decreased the empty volume of the pore and prevented the presence of water molecules at the hydrophobic gate.

Discussion
A broad number of medicinal plants have been used for centuries as therapeutic tools in TCM. However, neither the active compounds of many of these plants nor their mechanisms of action are yet well-understood. We have now studied the effect of Pm, an isosteroidal alkaloid considered one of the main bioactive molecules of Fb, on nAChRs.
Remarkably, Pm decreased I ACh in a dose-dependent manner, with an IC 50 in the low micromolar range (circa 3 and 1 µM for I p and I ss , respectively; Figure 1). These IC 50 s are markedly lower than the values previously reported for Pm blockade of voltage-dependent potassium channels. Thus, Pm IC 50 for blocking Kv1.2 was 472 µM, it was 354 µM for Kv1.3 (142 µM if measured 150 ms after the current peak), and even much higher for Kv1.4 to Kv1.8 channels [10]; additionally, Pm inhibited the potassium channel hERG, with an IC 50 of 44 µM, most likely by enhancing its inactivation [11]. Pm also blocked the Nav1.7 channel (IC 50 , 47 µM), demonstrating use-dependent inhibition, like the blocking mechanism of lidocaine on this channel [10]. Pm effects on hERG channels are particularly remarkable, since these channels play a key role in myocardial repolarization and, therefore, their inhibition might cause serious cardiac arrhythmias. Despite this, humans have used Fb as a therapeutic herb for centuries, being considered safe for consumption. Consequently, intake of Fb should not markedly affect the activity of hERG channels, despite that Pm, one of its main bioactive compounds, blocks these channels with an IC 50 of 44 µM [11]. Nevertheless, Pm has a very low oral bioavailability [9] and thus its plasma concentration after Fb intake should be fairly low. Actually, Pm content in bulbs of Fritillaria ussuriensis and thunbergii ranged from 0.58 to 1.2 mg/g and oral administration of powder from these Fritillaria plants to dogs (1 g/kg) raised Pm plasma concentration to a maximum of 100-200 nM [30]. Accordingly, a similar Pm bioavailability was reported after oral administration of Fritillaria thunbergii extracts in rats, with peak plasma concentrations of Pm of roughly 100 nM [31]. These Pm concentrations are several orders of magnitude lower than the IC 50 s reported for sodium or potassium channels (including hERG), muscarinic receptors, or acethylcholinesterase [8]. Interestingly, submicromolar Pm concentrations elicit a significant inhibition of nAChRs (roughly 20%; Figure 1) and therefore this family of LGIC might be relevant targets of its actions. Of note, we have assessed Pm actions on muscle-type nAChRs, because they are the prototype member of this family of receptors, but Pm might have different affinities for related receptors, as the homomeric α7. In fact, α7 nAChRs are largely expressed in non-neuronal tissues, including macrophages, and exert powerful anti-inflammatory actions [16]. Moreover, other nAChR subtypes, such as α4β2 and/or α9α10, may also play a role in modulating inflammatory processes and even in chronic pain [15]. Noticeably, Pm displayed a differential affinity for different LGICs, even of the same family. Thus, whereas Pm inhibits muscle-type nAChRs with and IC 50 close to 1 µM, GABA A receptors were almost not affected by Pm at concentrations up to 100 µM.
Pm exerted a non-competitive inhibition on muscle-type nAChRs, since I ACh s halved when co-applying Pm, at its IC 50 , with different ACh concentrations (Figure 2A,B). Several blockade mechanisms seem involved in this non-competitive inhibition of nAChRs by Pm, which is coherent with the multiple binding sites predicted by the docking simulations. First, open-channel blockade, as I ACh inhibition by Pm was voltage-dependent, the more hyperpolarized the cell, the more pronounced the blockade ( Figure 5). Actually, this is what would be expected for a positively charged molecule plugging the channel pore. In agreement with this, the docking assays predicted some Pm clusters located within the channel pore, both in the open and closed conformations (Figure 9). Thus, the open-channel blockade of nAChRs mediated by Pm resembles that mediated by some LAs, such as lidocaine [24] or tetracaine [26]. Nevertheless, the kinetics of open-channel blockade of nAChR elicited by Pm was rather slow. Thus, at −60 mV, the τ of open channel blockade elicited by 1 µM Pm (close to its IC 50 ) was over 2 s and even by 5 µM Pm (eliciting roughly 70% of I ss blockade) the time constant was above 1 s (Figure 7). In contrast, at the same membrane potential, 0.7 µM tetracaine (close to its IC 50 ) blocked open nAChRs with a time course of roughly 300 ms [26]. These differences in time constants between Pm and tetracaine are most likely related to their distinct molecular sizes (Pm molecular weight is over 60% greater than that of tetracaine). Second, Pm enhanced nAChR desensitization, as evidenced by: (i) acceleration of I ACh decay when co-applying ACh with Pm at concentrations of 0.5 µM, or above ( Figure 3A 2 ,B 2 ), and (ii) shortening of the I ACh aTtp, which correlated with the acceleration of I ACh decay ( Figure 3A 1 ,B 1 ). Likewise, lidocaine decreased the I ACh aTtp only at concentrations that enhanced I ACh decay [24]. Furthermore, DMA, a lidocaine analog, both sped up I ACh decay and shortened the aTtP [25]. By contrast, diethylamine (DEA), a lidocaine analog that mainly blocks nAChR by open channel blockade neither accelerates I ACh decay nor decreases aTtP [21], and (iii) deceleration of I ACh deactivation, which was dependent on Pm concentration and displayed a good correlation with the rate of I ACh decay ( Figure 4A, B). The deceleration of I ACh deactivation when Pm remained in the solution strongly supports that Pm enhanced nAChR desensitization, since desensitized nAChRs display higher affinity for the agonist [22,26,27]. Third, Pm elicited the blockade of resting (closed) nAChRs. This effect was unraveled by applying Pm before challenging the cells with ACh alone. This protocol, which allowed Pm to act only on resting (closed) nAChRs, elicited a mild nAChR blockade, mostly at negative potentials. Actually, at positive potentials, I ACh only decreased by Pm pre-application when rising its concentration to 5 µM (Figure 8). In agreement with this, I ACh inhibition by 5 µM Pm was slightly higher when Pm was pre-applied and then co-applied with ACh than when solely co-applied with ACh ( Figure 8C).
Our virtual docking and MD simulations used as a template the structure of Torpedo nAChRs in the open and closed conformations released by Unwin's group [32,33]. However, a new structural model of the nAChR from Torpedo, at higher resolution and stabilized in the closed conformation by α-bungarotoxin, has been recently disclosed [34]. The new structural model (pdb entry 6UWZ) share large similarities with the Unwin's model for the closed conformation, though there are certain differences between them. Mostly, they differ in the upper portion of the pore, which is more constricted in the new model, and in the δ subunit arrangement [35]. It seems that these discrepancies arise because of differences in the lipid matrix surrounding the nAChR. Actually, cholesterol interactions with the nAChR are apparently essential for stabilizing its structure and the absence of cholesterol (as in the model of Rahman et al. [34]) leads to a more compact arrangement of TM helices (displacement of helices circa 1-3 Å; [35]). Noticeably, the major lipid present in electroplax membranes rich in nAChRs is cholesterol [36] and purified nAChRs from T. marmorata and E. electricus interact preferentially with cholesterol rather than with either phospholipid monolayers or other sterols [37]. Moreover, nAChRs in native electroplax membranes are arranged as dimers, linked by their δ-subunits. This interaction between neighboring nAChRs might account for the differences in the δ-subunit between both structural models since dimers were reduced to monomeric receptors in the Rahman's model. In fact, we chose for our structural studies Unwin's models because of: (i) the structures for the open and closed conformations are available, (ii) the nAChR is present in their original membrane, and (iii) we have significant experience correlating structural and functional results using these commonly accepted models; actually, Unwin's models have so far provided a coherent correlation with our functional results [21,25,26,28].
The virtual docking assays predicted Pm binding to the nAChR at different sites of the TMD and ECD in the open conformation. Most Pm clusters were located at the TMD, at inter-and intra-subunit crevices, although some of them located into the channel pore ( Figure 9A,C). The binding energies estimated for these clusters were rather high (from −9.3 to −12.87 kcal/mol; see Supplementary Table S1), pointing out that Pm has a high affinity for the nAChR. Remarkably, MD simulations of nAChRs in the open conformation indicate that Pm binding to the nAChR either at the TMD (i.e., cluster 6) or at the ECD (i.e., cluster 14) markedly decreased both the volume and the number of water molecules at the hydrophobic gate region of the channel pore ( Figure 11). Furthermore, virtual docking and MD assays pointed out that Pm also interacts with the nAChR in the resting conformation, binding to residues located at both the TMD and the ECD ( Figure 9B,D and Figure 11E,F). Interestingly, the structural changes of the nAChR induced by Pm, as predicted by docking and MD assays, are in good agreement with the functional changes elicited by Pm on this receptor. Actually, the structural and functional results can be correlated as follows: (i) the high binding energies computed accounted for the high potency of Pm blocking nAChRs, (ii) Pm interaction with residues located within the channel pore should trigger the openchannel blockade, (iii) Pm binding to different sites at the nAChR might explain both its heterogeneity of actions on nAChRs and the effect-dependence on Pm concentration, and (iv) Pm binding to the nAChR in the closed conformation might underlie the blockade of resting nAChR. Consequently, the good correlation between structural simulations and electrophysiological results strongly suggests that Pm actually blocks nAChRs by the different aforementioned mechanisms.
Since Fb has been used as therapeutic herb for thousands of years, there is a strong support for its beneficial effects and its weak (or lack) of toxicity. However, neither the identity of all its bioactive compounds nor their mechanisms of action are yet well known. Now, we report here that Pm, considered one of the main bioactive compounds from Fritillaria, exerts a powerful inhibition of muscle-type nAChRs, which, as far as we know, is the first report demonstrating that Pm might modulate LGICs, besides acting on other targets as voltage-dependent channels or metabotropic receptors. It remains to be unraveled if Pm might modulate other nAChRs, including the homomeric α7, which is broadly expressed in immune cells and has been related to powerful anti-inflammatory actions [16]. Furthermore, both mecamylamine, a non-competitive antagonist of α7 nAChRs, and 1-ethyl-4-(3-(bromo)phenyl)piperazine, which promotes α7 desensitization, reduce pro-inflammatory responses [38]. We have now demonstrated that Pm inhibits muscle-type nAChRs and it can be hypothesized that it might modulate, though in a different way or with a different potency, either α7 or other nAChRs. In this regard, lidocaine exerted similar inhibitory actions on muscle-type nAChRs [24] and on neuronal nAChRs expressed in autonomic-ganglia neurons [39]. Alternatively, it could be that different bioactive compounds from Fb accounted for its anti-inflammatory actions, as anthocyanin pigments, which are flavonoids present in Fb [40]. Noticeably, some flavonoids act as positive allosteric modulators of α7 nAChRs, although without affecting desensitization [6], and their enhancement of α7 activity has been proposed as a therapeutic strategy for inflammatory disorders [41].

Oocyte Microinjection with Either Proteoliposomes Bearing nAChRs or Rat-Brain Synaptosomal-Enriched Membranes Bearing GABA A Rs
Adult female Xenopus laevis (purchased from Centre National de la Recherche Scientifique, Montpellier, France and European Xenopus Resource Centre (EXRC), Portsmouth, UK) were immersed in cold 0.17% tricaine methanesulfonate (MS-222) for 20 min and a piece of ovary was drawn out aseptically. Animal handling was carried out in accordance with the guidelines for the care and use of experimental animals adopted by the European Union Torpedo marmorata nAChRs were purified and reconstituted in asolectin lipids, at a final protein concentration of 0.3-1.2 mg/mL, as previously reported [42]. Oocytes were microinjected with 100 nL of an aliquot of reconstituted nAChRs. In some experiments, gamma-aminobutyric acid GABA A Rs were microtransplanted to the Xenopus oocyte membrane from rat-brain synaptosomal-enriched membranes, as previously described [21].

Experimental Design
Experimental procedures were similar to those previously used to study the effects of either acetylcholinesterase inhibitors [44,45] or LAs on nAChRs [24,26,28]. Briefly, Pm concentration-I ACh inhibition relationship was determined by measuring I ACh s evoked by 10 µM ACh alone or together with different Pm concentrations. For competition assays, ACh concentration-versus I ACh amplitude curves were obtained by bathing nAChRbearing oocytes with increasing ACh concentrations either alone or together with Pm at a concentration close to its IC 50 . I ACh s were normalized to the maximum I ACh evoked by ACh alone, and a sigmoid curve was fitted to these values (see equation 2 below). To allow nAChRs to recover from desensitization, the interval between consecutive ACh applications was, at least, 5 min. Blockade of resting nAChRs by Pm was assessed by the pre-application of Pm (1-5 µM) for 12 s before challenging the cell with ACh alone. The voltage dependence of the I ACh inhibition by Pm was determined by giving series of 800-1200 ms voltage pulses (from −120 to +60 mV, in 20 mV steps) to the oocyte before ligand superfusion and during the I ACh plateau elicited by 10 µM ACh, either alone or co-applied with either 1 or 5 µM Pm. The time course of nAChR blockade by Pm and its recovery were assessed by applying Pm (1-5 µM) for 20 s during the I ACh plateau elicited by a 50 s pulse of 10 µM ACh. Furthermore, to better resolve the time course of open-channel blockade of nAChR by Pm, a 2 s voltage pulse, from −60 to +40 mV, was applied during the I ACh plateau elicited by 10 µM ACh either alone or co-applied with 1 or 5 µM Pm and the kinetics of I ACh blockade by Pm after jumping back from +40 to −60 mV was computed.
Oocytes previously injected with synaptosomal membranes bearing GABA A R were superfused with 1 mM GABA alone or together with Pm (10-100 µM) to assess the Pm effects on GABA elicited currents (I GABA ).

Data Analysis and Statistical Procedures
Inhibition curves were determined by measuring the I ACh evoked by 10 µM ACh in the presence of different concentrations of Pm. The I ACh s (both at the peak, I p , and 20 s later, I ss ) elicited in the presence of Pm were normalized to the I ACh evoked by ACh alone. A logistic curve was fitted to the data with the OriginPro 8 software (OriginLab Corp., Northampton, MA, USA), using the following Equation (1): where I ACh+ Pm is the I ACh amplitude elicited by co-application of 10 µM ACh with Pm at a given concentration ([Pm]), I ACh max and I ACh min are the maximum and minimum I ACh s recorded, respectively, IC 50 is the Pm concentration required to halve the I ACh max, and n H is the Hill coefficient. The rate of desensitization (I ACh decay) was determined by fitting to a single exponential function the I ACh elicited by 10 or 100 µM ACh, either alone or co-applied with different Pm concentrations [26]. The apparent time-to-peak (aTtP) was determined as the time elapsed from I ACh onset to the I p elicited by ACh either alone or with Pm. We have called this parameter as "apparent" to indicate that these values do not necessarily reflect "real" time-to-peak values of nAChR activation, but those observed in our experimental conditions.
To characterize the pharmacological profile of nAChR blockade by Pm, nAChRs were activated by different concentrations of ACh alone, or co-applied with Pm at roughly its IC 50 . The following form of the Hill Equation (2) was used to fit dose-response data: where I is the I ACh amplitude elicited at a given concentration of ACh ([ACh]) applied either alone, or together with Pm, EC 50 is the agonist concentration required to halve the maximum I ACh , and I ACh max and n H are as in Equation (1). Net i/v curves for I ACh were computed by subtracting, for each voltage, the steady-state currents attained in ANR (measured at the last 100 ms of the pulse) from the corresponding currents recorded in the presence of 10 µM ACh either alone or together with Pm. These net I ACh values were normalized, for each oocyte, to the I ACh at −60 mV elicited by ACh alone.
To determine the rate of open-channel blockade by Pm, the oocyte was superfused with either 1 or 5 µM Pm at the plateau of the I ACh elicited by 10 µM ACh. A single exponential function was fitted to the I ACh decrease elicited in the presence of Pm. The time constant (τ) of the I ACh decay was computed by using the OriginPro 8 software. The same procedure was followed to determine the kinetics of I ACh recovery upon Pm withdrawal. A similar fitting method was used to estimate the kinetics of I ACh deactivation (i.e., the time course followed by the I ACh to return to the baseline level after removal of 100 µM ACh either alone or together with 1 or 5 µM Pm).
Unless otherwise specified, values presented are the mean ± standard error of the mean (SEM), "n" indicates the number of oocytes, and "N" is the number of oocytedonor frogs from which the data were obtained. When comparing two-group means of normally distributed values, the Student's t-test was used; otherwise, the Mann-Whitney rank-sum test was applied. Among-group differences were determined by the analysis of variance (ANOVA), and mean differences for each pair of groups were determined with the Bonferroni t-test. The one-sample t-test was used to compare the mean of an experimental group with a specified value. A significance level of p < 0.05 was considered in all cases.

Pm-nAChR Docking Simulations
The structures of the nAChR (Torpedo marmorata), both for the tense (closed, 2BG9 [32]) and for the more relaxed (open, 4AQ9 [33]) conformations, were obtained from the Research Collaboratory for Structural Bioinformatics (RCSB). Molecular docking simulations of Pm on the surface of nAChR have been carried as described elsewhere [21,25,46]. A total of 999 flexible docking runs were set and clustered (i.e., two docked compounds were considered to belong to different clusters when the ligand root-mean-square deviation of their atomic positions was greater than 7 Å around certain hot spot conformations). The YASARA software calculated the Gibbs free energy variation (∆G, kcal/mol) with more negative values, indicating stronger binding. The Pm molecule for each cluster with a more negative ∆G value (stronger binding) was used in the molecular dynamics simulations.

Molecular Dynamics Simulations (MD) of Lipid Reconstituted nAChR-Pm Complex
Pm molecules with the best binding to the nAChR of each cluster, were used as the starting point for independent MD simulations. Before starting each simulation, the Pm-AChR complex was reconstituted into a lipid bilayer (phosphatidyl-choline:phosphatidylserine, 80:20 for both bottom and top membrane side) using the YASARA macro md_ runmembrane.mcr. The simulation cell was allowed to include 20 Å surrounding the protein, which were filled with water at a density of 0.997 g/mL. Initial energy minimization was carried out under relaxed constraints using the steepest descent minimization.
Simulations were performed in water at constant conditions of pressure (1 bar) and temperature (25 • C). To mimic physiological conditions, counter ions were added to neutralize the system, Na + or Cl − were added as replacement of water to give a total NaCl concentration of 0.9%, and pH was maintained at 7.4. The pK a was computed for each residue according to the Ewald method [47]. All simulation steps were run by a preinstalled macro (md_run.mcr) within the YASARA suite. Data were collected every 100 ps for 100 ns. Each simulation was carried out in the high-performance computing Linux cluster of the Centro de Computación Científica (CCC-UAM). The measurement of the number of water molecules in the upper half of the channel pore was carried out using the macro md_analyze.mcr from YASARA. The Molecular Mechanics/Poisson-Boltzmann surface area (MM/PBSA) was implemented with the YASARA macro md_analyzebindenergy.mcr, to calculate the binding free energy with the solvation of the ligand, complex, and free protein, as previously described [48,49].
Measurements of cavity volume inside the nAChR were performed with 3V Voss Volume Voxelator [50] installed locally, and UCSF Chimera v1.15, build 42,209 [51]. The volumes were set with 3V by transforming the protein pdb coordinates to xyz format with the "pdb_to_xyzr" program. Then, the identification of the main, largest cavity was performed with the "AllChannel.exe" program, using an external probe radius of 10 Å, an internal probe radius of 3.3 Å, and restricting the number of channels to 1. This provided an mrc map containing a three-dimensional grid of voxels, each with a value corresponding to cavity volume. Molecular graphics Chimera software, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, was used to measure volumes using the commands "vop zone" and "measure volumes". The sub-volumes were measured by selecting the residues of each subunit that outline the upper part of the pore (α1: 255-261, β: 261-267, δ: 269-275, α2: 255-261, and γ: 263-269; see inset of Figure 11). Values were provided in Å 3 .

Drugs
ACh, atropine sulphate, Pm, GABA, MS-222, DMSO, penicillin, and streptomycin were from Sigma-Aldrich-Merck (Darmstadt, Germany). HEPES was obtained from Acros Organics (Geel, Belgium). Other reagents of general use were purchased from Scharlau Chemie SA (Barcelona, Spain). Pm solutions were prepared from a 10 mM stock solution in DMSO. All solutions were made in ANR just before each application.

Conclusions
A plethora of therapeutic herbs are used in TCM to treat different illnesses. Fritillaria plants are included among them, as they are often used to treat asthma, to alleviate cough, and as anti-inflammatory and analgesic therapy. We have now assessed the effect of Pm, an isosteroidal alkaloid found in Fb, on purified and reconstituted nAChRs microtransplanted to the Xenopus oocyte membrane. Pm exerted a powerful inhibition of muscle-type nAChRs acting by different mechanisms: (i) open-channel blockade, as evidenced by its voltagedependent inhibition of I ACh , (ii) enhancement of nAChR desensitization, as supported by three observables: first, the acceleration of the I ACh decay, second, the slowing down of I ACh deactivation in the presence of Pm, and third, the shortening of the I ACh aTtP; and (iii) closed (resting) channel blockade, which was demonstrated by the I ACh inhibition elicited by Pm when it was pre-applied before challenging the cell with ACh alone (thus Pm action was restricted to resting nAChRs). Furthermore, virtual docking and MD assays of Pm interactions on nAChRs, both in the open and closed conformations, predicted that Pm interacts with high affinity at different sites on this receptor, which seems consistent with the variety of functional effects observed. Moreover, the observed binding sites support fairly well the functional effects of Pm on nAChRs.
As far as we know, this is the first report demonstrating that Pm modulates LGICs, and moreover, its effects seem quite selective on nAChRs, since other receptors of the same family, such as GABA A R, were unaffected by Pm. Considering that Pm plasma levels after oral administration of Fb are rather low (in the submicromolar range) and that nAChRs constitute a high affinity target for Pm, it turns out that its anti-inflammatory and analgesic effects could be mediated through this interaction, given that nAChRs mediate potent anti-inflammatory effects. MM|PBSA