Menthol Stereoisomers Exhibit Different Effects on α4β2 nAChR Upregulation and Dopamine Neuron Spontaneous Firing

Abstract Menthol contributes to poor cessation rates among smokers, in part because menthol enhances nicotine reward and reinforcement. Mentholated tobacco products contain (−)-menthol and (+)-menthol, in varying proportions. We examined these two menthol stereoisomers for their ability to upregulate α4β2 nAChRs and to alter dopamine neuron firing frequency using long-term, low-dose (≤500 nm) exposure that is pharmacologically relevant to smoking. We found that (−)-menthol upregulates α4β2 nAChRs while (+)-menthol does not. We also found that (−)-menthol decreases dopamine neuron baseline firing and dopamine neuron excitability, while (+)-menthol exhibits no effect. We then examined both stereoisomers for their ability to inhibit α4β2 nAChR function at higher concentrations (>10 µm) using the Xenopus oocyte expression system. To probe for the potential binding site of menthol, we conducted flooding simulations and site-directed mutagenesis. We found that menthol likely binds to the 9´ position on the TM2 (transmembrane M2) helix. We found that menthol inhibition is dependent on the end-to-end distance of the side chain at the 9´ residue. Additionally, we have found that (−)-menthol is only modestly (∼25%) more potent than (+)-menthol at inhibiting wild-type α4β2 nAChRs and a series of L9´ mutant nAChRs. These data reveal that menthol exhibits a stereoselective effect on nAChRs and that the stereochemical effect is much greater for long-term, submicromolar exposure in mice than for acute, higher-level exposure. We hypothesize that of the two menthol stereoisomers, only (−)-menthol plays a role in enhancing nicotine reward through nAChRs on dopamine neurons.


Introduction
Menthol cigarettes are used by a third of smokers and by the majority of African-American smokers (McCarthy et al., 1995). Smokers of menthol cigarettes are less likely to quit when compared with smokers of nonmenthol cigarettes (Ahijevych and Garrett, 2010). Youth smokers of menthol cigarettes are twice as likely to become lifelong smokers compared with youth smokers of nonmenthol cigarettes (D'Silva et al., 2012). Menthol is present in much of the electronic nicotine delivery system (ENDS) market (Centers for Disease Control and Prevention, 2016). Some countries, as well as some municipalities in the United States, ban menthol in combustible tobacco products. Apart from these exceptions, menthol holds a unique position as a tobacco flavoring: it is the only flavor allowed in combustible cigarettes in the United States following the 2009 Family Smoking Prevention and Tobacco Control Act.
Eight stereoisomers of menthol exist. Plants produce mainly the stereoisomer [1R, 2S, 5R or (Ϫ)-menthol; Chen et al., 2011); but small quantities of additional stereoisomers occur in various types of mint leaves or in the extracted oil). As with many popular natural products, demand for menthol greatly exceeds the supply that can be obtained from plants. In the past, the primary synthetic method for producing menthol used hydrogenation of pulegone, producing menthol as a racemate [(ϩ)-menthol and (Ϫ)-menthol; Sell, 2006]. However, Ohkuma et al. (2000) developed an asymmetric synthesis method that provides pure (Ϫ)-menthol. Thus, while (Ϯ)-menthol and (ϩ)-menthol are easily produced synthetically and have been found in menthol tobacco products (Heck, 2010), (Ϫ)-menthol has now become the primary stereoisomer in tobacco products (Chen et al., 2011).
Previous data show that menthol by itself upregulates nAChRs (Alsharari et al., 2015; and alters midbrain dopamine (DA) neuron firing . Because both (ϩ)-menthol and (Ϫ)menthol may be present in menthol cigarettes, we investigated the potential differences that these two stereoisomers may exhibit on nAChR upregulation and dopamine neuron firing. We show that only (Ϫ)-menthol is responsible for the upregulated ␣4␤2 nAChRs and decreased dopamine neuron firing frequency found when neurons are treated with (Ϯ)-menthol. We also found that (ϩ)-menthol is 25% less potent at inhibiting ␣4␤2 nAChRs. The concentrations of menthol that inhibit ␣4␤2 nAChRs are similar to the concentrations required to activate the TRPM8 (transient receptor potential cation channel subfamily M member 8) channel, which is mainly responsible for the cooling sensation elicited by menthol (Ha et al., 2015). We also investigated a putative binding site for menthol on ␣4␤2 nAChRs using computational modeling and site-directed mutagenesis. Here we determined that menthol likely binds to the 9´site of the transmembrane M2 (TM2) helix of ␣4 nAChR subunits, and its binding interaction depends on steric interactions. Together, these data show that menthol exhibits different actions when applied in the long term or acutely. Longterm exposure to menthol exhibits a stereospecific effect on nAChR upregulation and dopamine neuron excitability: (Ϫ)-menthol produces a robust effect, and (ϩ)-menthol seems to be inert. This suggests that (Ϫ)-menthol, and not (ϩ)-menthol, possesses the potential risk of enhancing nicotine reward and reinforcement.

Mice and cultured midbrain neurons
In electrophysiological experiments, we used pregnant female Gene Expression Nervous System Atlas tyrosine hydroxylase (TH)-eGFP mice (Henderson et al., , 2017) that were purchased from the Mutant Mouse Regional Resource Center (Chapel Hill, NC). All experiments were conducted in accordance with the guidelines for the care and use of animals provided by the National Institutes of Health, and protocols were approved by the Institutional Animal Care and Use Committee at the California Institute of Technology. To culture primary mid-brain neurons from the TH-eGFP mice, we used embryonic day 14 embryos. We did not determine the sex of the embryos; but, given the mixed population of males and females in litters, the resulting midbrain cultures were composed of neurons from both sexes. (without determining their sex). Ventral midbrain tissue was dissected and cultured using previously published procedures . Following digestion with papain (15 min at 37°C), cells were separated by DNase treatment and trituration, and were plated at a density of 100,000 cells/dish in poly-L-ornithine-and laminin-coated imaging dishes containing Neurobasal medium supplemented with B27, Glutamax, 1% hyclone equine serum, and 100 M ascorbate. For drug treatments, control medium (control), 500 nM (Ϫ)-menthol, or 500 nM (ϩ)-menthol were added to the cultures for 10 d. Culture medium was changed at ϳ3 d intervals.

Neuro-2a cell culture and transient transfections
Neuro-2a cells were cultured using standard techniques . For imaging, cells were plated by adding 90,000 cells to poly-D-lysine-coated 35 mm glass-bottom imaging dishes (MatTek) and cultured in a humidified incubator (37°C, 95% air, 5% CO 2) . Cells were transfected as described previously (Henderson et al., 2014). Similar to previous assays , drug treatments (control or 500 nM menthol stereoisomers) were applied for 24 h. Culture medium containing menthol was removed 1 h before the total internal reflection fluorescence microscopy (TIRFM) assays and replaced with extracellular solution (ECS), identical to methods previously described (Henderson et al., , 2017.

Total internal reflection fluorescence microscopy
TIRFM enables the visualization of fluorescently labeled intracellular molecules on or near the plasma membrane (within ϳ250 nm of the cell-coverslip interface). Cultured neuro-2a cells were imaged live at 37°C in a stagemounted culture dish incubator (Warner Instruments) using methods and instrumentation reported previously (Richards et al., 2011;Henderson et al., 2014). For each TIRFM session, image analysis was started 25 h after the beginning of drug treatment (24 h for drug treatment, 1 h for washout with ECS). This ensured that each session recorded cells at the same time point. For each session and each drug condition, 30 -40 cells were imaged. Each condition was measured in at least three separate imaging sessions.

Patch-clamp electrophysiology
Neurons were visualized with an upright microscope (BX50WI, Olympus) using near-infrared or blue illumination (the latter for visualizing GFP fluorescence). Wholecell patch-clamp techniques were used to record electrophysiological signals. Data were sampled at 10 kHz and filtered at 2 kHz. Patch electrodes had a resistance of 4 -8 M⍀, when filled with intracellular solution (in mM: 135 K gluconate, 5 KCl, 5 EGTA, 0.5 CaCl 2 , 10 HEPES, 2 Mg-ATP, and 0.1 GTP). The ECS was composed as follows (in mM): 140 NaCl, 5 KCl, 2 CaCl 2 , 1 MgCl 2, 10 HEPES, and 10 glucose. The pH of these solutions was adjusted to 7.2 with Tris-base, and their osmolarity was adjusted to 300 mOsm with sucrose.
All recordings were performed at a mean temperature of 32 Ϯ 1°C. ACh (300 M) was puffed onto neurons using a Picospritzer III for a duration of 300 ms (holding levels: voltage clamp, Ϫ65 mV; current clamp, 0 pA). Neurons were selected by the presence or absence of TH-eGFP fluorescence and labeled as putative DA or GABA neurons. Firing frequency and action potential duration were recorded (Ͼ1 min) to further verify GFP or non-GFP distinctions of DA or GABA neurons. Drug treatments for cultured midbrain neurons were performed for 10 d (control or 500 nM either menthol stereoisomer; as described above). At the beginning of a recording session, dishes containing cultured midbrain neurons were rinsed with ECS to remove any remaining drug treatment. Cultured neurons were placed on the recording stage and perfused with ECS for 1 h to ensure that all remaining menthol from chronic treatments was removed. For all recording sessions, perfusion rates were maintained at constant flow so that changes in agonist washout did not complicate data interpretation.

Oocyte preparation and injection
Rat ␣4 and ␤2 nAChR subunits were in pGEMhe vectors. The mRNAs were prepared from linearized DNA, using a T7 mMessage mMachine kit (Ambion) and were purified with the RNeasy Mini Kit (Qiagen). Xenopus laevis stage V and VI oocytes were harvested via standard protocols (Marotta et al., 2014). The ␣4 and ␤2 mRNAs were mixed in a 1:10 ratio by mass to obtain the (␣4) 2 (␤2) 3 or in a 10:1 ratio to obtain the (␣4) 3 (␤2) 2 receptor. For unnatural amino acid incorporation, a TAG codon was incorporated into the site of interest, and 40 ng of unnatural amino acid-coupled THG73 tRNA was added to the injection solution. The ␣4 and ␤2 mRNAs were mixed in a 100:1 ratio by mass to obtain the (␣4) 3 (␤2) 2 receptor for these unnatural amino acid experiments. A total of 50 nl of the RNA mixture were injected into each oocyte, delivering an mRNA mass total of 22 ng. After injection, the oocytes were incubated at 18°C in ND96 medium (see below) enriched with theophylline, sodium pyruvate, and gentamicin for 48 h before recording.

Oocyte electrophysiology
The OpusXpress 6000A (Molecular Devices) in twoelectrode voltage-clamp mode was used for all electrophysiological recordings. ACh was dissolved to 1 M stock solutions in ND96 Ca 2ϩ -free buffer (96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , and 5 mM HEPES at pH 7.5). The holding potential was set to Ϫ60 mV, and the running buffer used was ND96 Ca 2ϩ -free solution for all experiments. All drugs were applied as a 1 ml application in ND96 Ca 2ϩ buffer. Drug applications used 1 ml of drug solution applied over 15 s followed by a 5 min buffer wash at a rate of 3 ml/min. Data were sampled at 50 Hz and then low-pass filtered at 5 Hz. Averaged and normalized data were fit to one Hill term to generate EC 50 , IC 50 , and Hill coefficient (n H ) values. All currents for the activity testing were normalized to the maximum current pro-duced by the oocyte for a given compound (I max ) from ACh potentiation. The efficacy of compounds was measured as the ratio of the I max of the compound divided by the I max of the ACh. Error bars represent SEM values.

Molecular dynamics simulations
To characterize putative binding sites of menthol to the human ␣4␤2 nicotinic receptor, flooding simulations were performed using NAMD 2.12 (Phillips et al., 2005). Flooding simulations are molecular dynamics (MD) simulations in which many identical molecules are initially placed in solution and allowed to diffuse and interact with the various environments such as water, membrane, and proteins.
For the protein structure, we used the x-ray crystal structure of the human ␣4␤2 nicotinic receptor (Morales-Perez et al., 2016; Protein Data Bank code, 5KXI). To prepare the protein for the simulation, we first used Mod-Web (Sánchez and Sali, 1998), an automated comparative protein structure modeling web server, to build the missing intracellular loops (which connect M3 and M4 helices) for both ␣4 and ␤2 subunits. Then, CHARMM-GUI (Jo et al., 2008) was used to embed the protein into a 120 ϫ 120 Å POPC lipid bilayer in the x-y-plane. The x-ray crystal structure captures the receptor in its desensitized state with two nicotine molecules bound to the extracellular domain of the receptor. In our simulation, we did not include the two bound nicotine molecules. To preserve the crystallographically captured state of the protein, we imposed harmonic restraints on backbone atoms over the course of the simulation. The SOLVATE and AUTOIONIZE plugins in VMD (Humphrey et al., 1996) were used to solvate and to ionize the system to an NaCl concentration of 150 mM, respectively.
We started the flooding simulation with 178 menthol molecules (corresponding to a concentration of ϳ165 mM) randomly placed in the solution. In our preliminary simulations, the majority of menthol molecules aggregated in the solution before getting a chance to interact with the lipid head groups or the protein, an expected outcome given the highly hydrophobic nature of menthol (logD ϭ 2.7). To overcome this problem, nonbonded repulsive energy terms were introduced between the two oxygen atoms of each pair of menthol molecules (using NBFix corrections). We chose to decrease the well depth of the Lennard-Jones potential to Ϫ0.10 kcal/mol instead of its original value of Ϫ0.1921 kcal/mol. We also increased the minimum interparticle distance between each pair of oxygen atoms from two menthol molecules from 3.53 to 12.00 Å. As a result, pairs of menthol molecules were prevented from aggregating. Given the NAMD cutoff of 12 Å for nonbonded interactions, by using these NBFix corrections, we basically turned off the attraction between any two oxygen atoms that belong to separate menthol molecules. The production run was performed for 800 ns. After the first 350 ns, the majority of menthol molecules partitioned into the membrane, allowing us to turn off the NBFix corrections so that multiple menthol molecules were not prevented from approaching the same binding site/region on the protein.
The simulations were performed under periodic boundary conditions. Long-range electrostatic interactions were calculated by the PME (Particle mesh Ewald method; Darden et al., 1993;Essmann et al., 1995) with one grid point density per angstrom. Except for the very first phase of melting the lipid tails, where the NVT (constant temperature, constant volume) ensemble is preferred, the NPT (constant temperature, constant pressure) ensemble was used. The constant temperature of 310 K for the system was provided by coupling it to a heat bath and through Langevin dynamics for all nonhydrogen atoms with a Langevin damping coefficient of 5 ps Ϫ1 . For the NPT simulation, the pressure was kept constant at 1 atmosphere through coupling to a Nosé-Hoover Langevin piston, with a period of 200 fs and a damping timescale of 50 fs. Time steps of 2 fs were chosen for the simulation. Code and software availability NAMD 2.12 was used for all of the molecular dynamics simulations, and VMD was used for visualization and analysis of the simulation results. Both of these software packages are free and available for all users at https:// www.ks.uiuc.edu/Research/namd/ and at https://www.ks.uiuc.edu/Research/vmd/, respectively.

Experimental design and statistical analysis
All results are presented as the mean Ϯ SEM, and all statistical analyses were performed using GraphPad Prism. For microscopy assays ( Fig. 1) and electrophysiology using cultured neurons (Figs. 2, 3), we used a power analysis (G‫ء‬Power software, www.gpower.hhu.de) to determine appropriate sample sizes. In experiments that involved three or more group comparisons (Figs. 1-3), data were analyzed using a one-way or a two-way ANOVA. The choice of a one-way or two-way ANOVA is listed with the corresponding experiment in the Results section. When effects were shown to be significant, a post hoc analysis (Tukey's test) was performed to compare the individual group means.

Long-term treatment with menthol stereoisomers causes different effects on dopamine neuron firing frequency
We previously used whole-cell patch-clamp electrophysiology to report that long-term treatment of (Ϯ)menthol decreases the baseline firing of midbrain dopamine neurons in both mouse brain slices and cultured mouse midbrain neurons . To further characterize different effects of menthol stereoisomers, we conducted similar studies with cultured mouse midbrain dopamine neurons treated for 10 d with (ϩ)-menthol or (Ϫ)-menthol (500 nM; Figs. 2, 3). To facilitate the identification of midbrain dopamine neurons, we cultured neurons from TH-eGFP mice (Fig. 2A1-B3; Henderson et al., 2017). The presence of hyperpolarization-activated inwardly rectifying nonspecific cation current (I h ) and spike duration was also assessed to aid in the identification of dopamine neurons. For all TH-eGFP-positive neurons (Fig. 2A1-A3), we detected the presence of I h and spike durations Ͼ4 ms.
We examined cultured dopamine neurons for changes in baseline firing frequency following 10 d of treatment with either 500 nM (ϩ)-menthol or 500 nM (Ϫ)-menthol (Fig.  2C,D). Control neurons exhibited a mean firing frequency  (Fig. 2C). Long-term treatment with (Ϫ)menthol produced a significant decrease in dopamine neuron mean firing frequency to 1.5 Ϯ 0.3 Hz (p ϭ 0.04, one-way ANOVA with Tukey's test). This change in firing frequency is similar to the effect we reported following long-term treatment with (Ϯ)-menthol . Long-term treatment with (ϩ)-menthol did not produce a significant change in dopamine neuron firing frequency (Fig. 2C,D). We did observe that 20% of neurons treated with (ϩ)-menthol exhibited an increase in firing  frequency. All neurons exhibiting an increase in firing frequency were positive for TH-eGFP fluorescence, I h , and spike duration Ͼ4 ms. Therefore, we are confident that these neurons are indeed dopamine neurons and not GABA neurons, which are also present in these cultured midbrain neuron preparations. Further examination showed that a few of these neurons (10% of the 20%) exhibited spontaneous variations in baseline firing frequency ( Fig. 2E1-E3).

Acute applications of menthol stereoisomers exhibit similar pharmacologic profiles
Our long-term applications of menthol have studied submicromolar concentrations that we perceive to be pharmacologically relevant . Pioneering observations showed that menthol at higher concentrations (100-to 1000-fold greater than the concentration we use in our long-term studies) acts as a negative allosteric modulator of ␣4␤2 nAChRs (Hans et al., 2012), a noncompetitive antagonist of ␣3␤4 nAChRs (Ton et al., 2015), and a noncompetitive antagonist of ␣7 nAChRs (Ashoor et al., 2013b). Although the concentrations used in these studies may not be pharmacologically relevant in some cases, this profile suggests that menthol directly interacts with nAChRs and most likely at an allosteric site. Using X. laevis oocytes and two-electrode voltage-clamp electrophysiology, we used concentration-response assays to examine how acute applications of (ϩ)-menthol and (Ϫ)-menthol differed in their ability to act as inhibitors of ␣4␤2 nAChR function (Fig. 4).

Flooding simulations suggest a putative menthol binding site on ␣4␤2 nAChRs
As computational resources become more efficient, more studies are using molecular dynamics to guide experimental design (Ashoor et al., 2013b; Dalton et al., 2014;Arcario et al., 2017;Bottaro and Lindorff-Larsen, 2018). We used "flooding" simulations in which copies of menthol were added to the solution phase of a simulation system containing an ␣4␤2 nAChR (Brannigan et al., 2010). Most of the menthol molecules partition to the lipid membrane, leaving just a small percentage of the total number of menthol molecules in the aqueous phase. The protein remained stable throughout the simulation and predicted binding sites were deemed notable if menthol remained within 5 Å of the protein for Ͼ50 ns. Many of these predicted binding sites were on the periphery of the protein; however, one was inside the receptor pore (Fig. 5). Specifically, the site in the pore was the ring of leucines at the 9´position on TM2 (Fig. 6). This position is important for modulating the sensitivity of the protein to agonists without affecting the agonist-binding site (Labarca et al., 1995;Kearney et al., 1996;Kosolapov et al., 2000). Based on the simulation and the importance of the 9´site, we probed this prediction further with oocyte expression experiments (Tables 4, 5). We also probed many of the other predicted binding sites on the periphery of the protein, but our results showed that mutating these sites had no meaningful influence on menthol potency (Tables 6, 7) F328W]) 2 because inhibition with (Ϫ)-menthol was too weak for systematic fitting to the Hill equation. We probed only a subset of residues with (ϩ)-menthol for stereoselectivity comparison experiments.

The L9´residue on M2 is important for menthol inhibition
Based on the simulation results, we made additional mutations to (␣4) 2 (␤2) 3 nAChRs at the L9´site and measured dose-response relations for (Ϫ)-menthol using twoelectrode voltage-clamp electrophysiology (Fig. 7). The data confirmed that menthol block depends on the residue at the 9´position on the M2 helix in nAChR subunits (Fig. 7). Further, mutating this position to different amino acids elicited a range of potency for menthol inhibition. The greatest effect on the ability of menthol to inhibit ␣4␤2 nAChRs was observed with the leucine to alanine (L9´A) mutation (Fig.  7A1,A2, Tables 4, 5). The ␣4[L9´A]␤2 receptor is ϳ100-fold less sensitive to menthol block than wild type (WT; Fig. 7B). At the other extreme of sensitivity, ␣4[L9´M]␤2 is more sensitive (ϳ10-fold). Based on these results, menthol-induced inhibition depends strongly on the side chain at the 9´position.
A possible complication is that we tested menthol using much lower ACh for measurements on the (␣4[L9´A]) 3 (␤2) 2 receptor than on the WT receptor. This choice was driven by our desire to test menthol effects at roughly equally effective [ACh] for each individual receptor combination; the (␣4[L9´A]) 3 (␤2) 2 receptor has 10 4 -fold lower EC 50 values for ACh when compared with the wild-type receptor (Labarca et al., 1995;Kearney et al., 1996;Kosolapov et al., 2000;Dash et al., 2014). Nonetheless, one asks whether one can treat the menthol block as truly independent of events of the agonist binding site. To address this issue, we performed additional experiments using 1 M ACh to activate the WT and mutant receptors (Fig. 7B). The partially activated WT receptor is blocked ϳ50% by ϳ30 M menthol; but the fully activated (␣4[L9´A]) 3 (␤2) 2 receptor remains almost completely unblocked by [menthol] Ͻ1 mM. The data therefore preserve the concept that menthol exerts a "noncompetitive" block (presumably at the 9´site in the transmembrane pore), and this block can be measured over a wide range of [ACh].
Next, we determined the minimum number of 9´mutations, within an nAChR pentamer, required to perturb the activity of menthol. Because the L9´A mutation most drastically reduced menthol sensitivity, we varied the number of L9´A mutations within the receptor pentamer. Since nAChRs are pseudosymmetric with fivefold symmetry, we can make up to five 9´A mutations. We made mutants with one, two, and three 9´A mutations since (␣4[L9´A]) 3 (␤2) 2 proved to be unaffected by menthol (Fig. 8A). We were also interested in determining whether this was a subunit-dependent effect, so we incorporated 9´A mutations in either ␤2 or ␣5 nAChR subunits. In all the mutants tested, as long as there was at  least one 9´A mutation in ␣4, ␣5, or ␤2 nAChR subunits, inhibition by 100 M (Ϫ)-menthol was significantly reduced (unpaired t test, p ϭ 0.0001; Fig. 8A). The reduction in the ability of menthol to inhibit nAChRs that harbor 9´A mutations suggests that menthol binds to this site in TM2.

One bound menthol molecule is sufficient for ␣4␤2 inhibition
Hill coefficients, which describe the overall steepness of the concentration-response relation, are often used to gain an understanding of ligand-receptor binding stoichiometry (Prinz, 2010). We investigated the minimum number of bound menthol molecules necessary for channel inhibition. To test this, we calculated the Hill coefficients for both stereoisomers of menthol on all of the mutants tested (Fig. 8B). Since the average Hill coefficient is not significantly different from unity for either (Ϫ)-menthol or (ϩ)-menthol, we concluded that the presence of a single bound menthol molecule is sufficient for nAChR inhibition.

The activity of (؉)-menthol and (؊)-menthol depends on the end-to-end length of the 9´residue
Which properties of the side chain at 9´are important for the inhibitory activity of menthol on ␣4␤2 nAChRs? We examined the IC 40 or IC 30 values for (Ϫ)-menthol and (ϩ)-menthol, respectively, against the 9´residue volume, end-to-end length, and other parameters in the reduced amino acid index (AA index) (Fig. 9, Fig. 9 -1, and Fig. 9-2;Zamyatnin, 1972;Kibinge et al., 2014). We calculated an additional parameter, end-to-end residue length, using a 6-31G‫ء‬ basis set Hartree-Fock calculation in SPARTAN '14 (Wavefunction). We choose to use IC 40 or IC 30 so that the values calculated remained within the limits of our experimental data (Fig. 7A1,A2) because many of our dose-response studied failed to reach 100% inhibition at 1 mM menthol. We obtained a satisfactory relation when we compared the end-to-end length of residues with the IC 40 of (Ϫ)-menthol; the correlation coefficient was 0.69 (Fig. 9A1). For (ϩ)-menthol (IC 30 ), we observed a correlation coefficient of 0.74 (Fig. 9A2). Thus, the potency of either stereoisomer of menthol exhibited a strong correlation with residue end-to-end length. We observed similar correlations when we compared the IC 40 or IC 30 of (Ϫ)-menthol and (ϩ)-menthol, respectively, to residue volume (Fig. 9B1,B2).
Interestingly, previous data show that the EC 50 of agonists depends strongly on polarity, a different property of the side chain at the 9´position (Kearney et al., 1996;Kosolapov et al., 2000). To test whether polarity is a factor for menthol-induced inhibition, we incorporated O-methyl-threonine (O-Me-Thr) at the 9´position. The end-to-end length of O-Me-Thr is just 0.41 Å longer than isoleucine (3.55 vs 3.14 Å), but because of the oxygen instead of a methylene unit in the side chain, O-Me-Thr is substantially more polar than isoleucine. If polarity at the 9´position is important for menthol inhibition, we would expect to measure a meaningful difference in the concentration-response curve for menthol inhibition at these two mutants. We observed that the concentration-response relationships overlapped when comparing ␣4[L9´I]␤2 to its O-Me-Thr variant (Fig. 9C). This suggests that menthol inhibition does not depend on the polarity of the 9´side chain.
Additionally, we fitted the IC 40 or IC 30 values for (Ϫ)menthol and (ϩ)-menthol, respectively, against the reduced AA index (Kibinge et al., 2014). The reduced AA index is a set of eight indices describing the variability of amino acids based on experimental results. The reduced AA index is a subset of the AA index after the elimination of redundancies. We found that the end-toend residue length and residue volume had more meaningful correlations than any parameter in the reduced AA index.
The two stereoisomers of menthol inhibit ␣7 nAChRs with similar IC 50 values, but differences have been observed on the GABA A receptors, biomembrane models, and mouse odorant receptors (Corvalán et al., 2009;Ashoor et al., 2013b;Takai and Touhara, 2015;Gusain et al., 2017). To further compare the potency of (Ϫ)menthol and (ϩ)-menthol on ␣4␤2 nAChRs, we compared the percentage inhibition by 100 M menthol along with the IC 50 and pIC 30 values on the 10 (␣4[L9´X]) 2 (␤2) 3 nAChRs and wild-type (␣4) 3 (␤2) 2 nAChR (Fig. 7C,D1,D2). These scatterplots show either a comparison of the inhibition data (Fig. 7B) or comparison to a logarithmic transformation (Fig. 7D1,D2). We excluded the IC 50 values for (␣4[L9´A]) 2 (␤2) 3 because we failed to observe inhibition of 50% of the control response. The data extend the experiments on wild-type nAChRs (Fig. 4), which showed modestly less inhibition by the (ϩ) stereoisomer: (ϩ)-menthol was on average 25% less potent than (Ϫ)-menthol (Fig.  7D1,D2). WT-LS, Wild-type ␣4␤2 receptor with the low sensitivity stoichiometry; WT-HS, wild-type ␣4␤2 receptor with the high sensitivity stoichiometry; WT-␣5, wildtype ␣4␤2 receptor with ␣5 in the auxiliary position; 9´residue, a mutant at the 9´position to probe the effects of mutations at this site; GABA, analogous site for the menthol binding site in the GABA receptor; MD, molecular dynamics predicted position; TRPM8, analogous site for the menthol binding site on the TRPM8 receptor; ANDFLE, a mutant that is found in ADNFLE patients; Docking, position predicted by preliminary docking studies; #L9´A, mutant made to probe the effects that the number of 9´A mutations has on the effects of menthol; ACh binding site, the putative ACh binding site; Intra-subunit, site between subunits in the transmembrane domain of ␣4␤2; Near Cys loop, site that is near the Cys loop of ␣4␤2; Channel pore, site that is in the channel pore of ␣4␤2; Interfacial, a site on the transmembrane domain of a single subunit that faces toward the other transmembrane domain helices on the same subunit.

Discussion
Throughout the history of menthol-flavored tobacco products, various methods have been used to obtain the menthol used in cigarette production. Isolations from natural sources (mint leaves and oils) primarily produce (Ϫ)menthol; most synthetic methods produce a racemic mixture of both (ϩ)-menthol and (Ϫ)-menthol. The exception is the asymmetric synthesis developed by Ohkuma et al. (2000). We found that (ϩ)-menthol and (Ϫ)-menthol exhibit different effects on ␣4␤2 nAChR upregulation (Fig.  1). We also observed differences among the menthol stereoisomers in their ability to alter dopamine neuron firing frequency and excitability (Figs. 2, 3). Our previous investigations into the actions menthol used (Ϯ)-menthol (Henderson et al., 2016, 2017). Previous reports observed that (Ϯ)-menthol by itself upregulated nAChRs, altered dopamine neuron firing, enhanced the upregulation of nicotine-induced nAChR upregulation, and enhanced nicotine reward-related behavior. Because the present investigation shows that (ϩ)-menthol has no effect on nAChR upregulation and dopamine neuron excitability, we conclude that previous observations of long-term menthol exposure and its effect on nAChR upregulation and dopamine neuron excitability were likely caused primarily by (Ϫ)-menthol only.
Because of the contrast between the high concentration of menthol required for acute inhibition of nAChRs (Ͼ30 M) and the submicromolar effects on upregulation and dopamine neuron firing (Յ 500 nM), it is likely that menthol acts via multiple targets. The simulations and Figure 7. ␣4L9´mutations probe the putative binding site of menthol. A 1 , A 2 , Concentration-response curves for menthol against (␣4[L9´X]) 2 (␤2) 3 , where X is any amino acid. Each receptor is activated by its respective EC 50 dose of ACh (A 1 ) (Ϫ)-menthol and (A 2 ) (ϩ)-menthol. B, Concentration-response curves for (␣4[L9´X]) 3 (␤2) 2 and (␣4) 3 (␤2) 2 using 1 M ACh. C, Comparing the percentage maximum current induced when the oocyte is exposed to 100 M menthol. ‫ء‬p Ͻ 0.05; ‫‪p‬ءء‬ Ͻ 0.01; ‫‪p‬ءءء‬ Ͻ 0.005; ‫‪p‬ءءءء‬ Ͻ 0.001. Exact p values are provided in extended data ( Fig. 7-1). D 1 , Comparing the IC 50 values. In this plot, (␣4[L9´A]) 2 (␤2) 3 is omitted because at no tested concentration of menthol was the receptor inhibited 50%. D 2 , Plot of IC 30 values comparing (ϩ)-menthol and (Ϫ)-menthol; n ϭ 6 -18 oocytes. data in this study strongly suggest that the acute inhibitory effect of menthol is mediated through its direct interaction with nAChRs at the 9´leucine site within the TM2 region. In previous electrophysiological experiments, well characterized nAChR blockers have permanent positive charges (e.g., QX-222) or can be protonated (e.g., mecamylamine) and bind near this site (at the 6´and 10ṕ ositions) by approaching through the open pore (Charnet et al., 1990;Lester, 1992;Papke et al., 2013). However, other experiments suggest that uncharged, membrane-permeant blockers, such as deprotonated procaine, can bind to the 6á nd 10´residues in closed channels by approaching through the membrane or through the receptor protein (Adams, 1977;Lester, 1992). The simulations in this study reinforce the concept that menthol approaches the 9´site in closed ␣4␤2 nAChR channels by approaching from the protein. Our simulations are uninformative about a possible approach from the membrane phase, because the parameters were adjusted to minimize accumulation in the membrane phase (see Materials and Methods).
The experiments also show that (Ϫ)-menthol is only slightly (ϳ25%) more effective than (ϩ)-menthol at inhibiting WT receptors (Fig. 4). This difference would be neglected if the dataset contained only WT receptors, but it is consistent and significant when assessed over a series of 9´mutations that encompass a nearly 1000-fold range of sensitivities to menthol block (Fig. 7D1,D2). S(ϩ)mecamylamine is a slightly more potent blocker than R(Ϫ)mecamylamine over a series of nAChR subtypes (Papke et al., 2013). These differences between stereoisomers as nAChR blockers are markedly less than the threefold to fourfold greater potency of (Ϫ)menthol versus (ϩ)-menthol at TRPM8 channels (Sherkheli et al., 2010).
The submicromolar chronic effects of menthol may result from a non-nAChR target. In addition to its actions on nAChRs, menthol acts as a positive modulator on GABA A and glycine receptors and a noncompetitive antagonist of 5-HT 3 receptors (Hall et al., 2004;Ashoor et al., 2013a). Menthol has well studied actions on TRP family proteins, especially TRPM8 and TRPA1 . Menthol also acts on other targets . Therefore, while menthol has acute actions on nAChRs (noncompet-itive inhibition) through an allosteric interaction at the 9ś ite, its effects on nAChR upregulation and dopamine neuron excitability may be caused by another protein target. Previous investigators suggested that menthol acts as a chemical chaperone , in part because menthol enhanced the ER export of nAChRs. This designation has both analogies and differences with the pharmacological chaperoning of nAChRs by nicotine . Menthol may indeed be a chemical chaperone for nAChRs; but, given its ability to bind to many proteins of interest, we must consider that the enhancement of ER export by menthol may arise from its actions on a non-nAChR target related to anterograde trafficking of proteins or membranes. As discussed earlier, menthol enhances nicotine reward (Henderson et al., 2017) and nicotine reinforcement (Wang et al., 2014;Biswas et al., 2016) in rodent models and produces poorer cessation rates in human smokers (Food and Drug Administration, 2012). Although the predominant form present in current tobacco products is (Ϫ)-menthol, the composition of menthol used in many tobacco products is presently neither disclosed nor regulated. For this reason, it is important to understand how both (ϩ)-menthol and (Ϫ)-menthol alter the nAChRs on dopamine neurons that are involved in nicotine reward and reinforcement. This is especially important given that the ENDS market will become subject to systematic regulatory control by the Food and Drug Administration. Our data suggest that menthol exhibits a stereospecific effect only with low-dose, long-term exposure. We hypothesize that of the two primary menthol stereoisomers, only (Ϫ)menthol plays a role in enhancing nicotine reward through nAChRs on dopamine neurons. We do not yet suggest that (ϩ)-menthol lacks any role in nicotine reward. GABA neurons in the VTA and SNr play an important role in nicotine reward as their disinhibition plays a critical role in the ability of nicotine to drive enhancements, dopamine neuron excitability, and dopamine release (Mansvelder et al., 2002;Nashmi et al., 2007). Menthol does act as a positive allosteric modulator of GABA A receptors and (ϩ)menthol is more potent than (Ϫ)-menthol on these recep- Error bars represent the SEM; n ϭ 7-33 oocytes. B, Average Hill coefficients for wild-type and all mutant ␣4␤2 mutants tested (p ϭ 0.07). Error bars represent the SD; n ϭ 7-39 mutants tors (Hall et al., 2004;Corvalán et al., 2009). Given that stereospecific effects of menthol have been observed with nAChRs and GABA receptors (both important targets for nicotine reward), there is a continued need to understand their distinct pharmacology and how it alters the actions of nicotine on midbrain neurons.
Given the reduced effect (ϩ)-menthol exhibits on dopamine neurons, would public health benefit from availability of combustible products in which (ϩ)-menthol replaces (Ϫ)-menthol? Because (ϩ)-menthol is also less effective at activating TRPM8, (ϩ)mentholated products might also less effectively provide the characteristic cooling sensation. Smokers might therefore continue to prefer tobacco flavored with (Ϫ)-menthol (Sherkheli et al., 2010). We also have no information about stereospecificity of TRPA1 activation by either menthol or nicotine (Talavera et al., 2009). This report, and others (Alsharari et al., 2015; Biswas  Fan et al., 2016;Henderson et al., , 2017, provide evidence both that (Ϫ)-menthol plays a role in enhancing the addiction to nicotine, and also that simply substituting (ϩ)-menthol for (Ϫ)-menthol in combustible tobacco products may not be the correct strategy for harm reduction.