The solid-state conformation of the topical antifungal agent O-naphthalen-2-yl N-methyl-N-(3-methylphenyl)carbamothioate

Large flat colorless needles of tolnaftate have been obtained from Tinactin, and have been shown to contain a 50:50 mixture of the (+ap,−sp,+ac,−ac) and (−ap,+sp,−ac,+ac) conformers. This finding fills a long-standing void in the structural chemistry of this classic antifungal compound, and suggests that prior theoretical models used to explain the spectroscopy, bioactivity, and mode of binding of tolnaftate to squalene epoxidase are probably inappropriate or suspect.


Introduction
O-Naphthalen-2-yl N-methyl-N-(3-methylphenyl)carbamothioate, (I), is a synthetic thiocarbamate from the 1960s with antimycotic activity (Noguchi et al., 1961(Noguchi et al., , 1963. It is perhaps most readily recognized by the generic name tolnaftate and is primarily used to treat fungal skin infections, such as athlete's foot (tinea pedis), jock itch (tinea cruris), and ringworm (tinea capitis and tinea corporis). Tolnaftate is a squalene epoxidase inhibitor used to disrupt the biosynthesis of ergosterol, resulting in a toxic accumulation of squalene and ultimately fungal cell death (Morita & Nozawa, 1985;Ryder et al., 1986;Barrett-Bee et al., 1986). It was launched for human use in 1965 by Schering Corporation as the active pharmaceutical ingredient (API) in Tinactin (Sittig, 1988). Schering Corporation subsequently merged with Plough in 1971 and Merck in 2009. Today, Tinactin is marketed by Bayer which acquired Merck's consumer care products in 2014. Tolnaftate is present in numerous antifungal products worldwide either as the sole active ingredient or in combination with one or more other APIs.
Historically, tolnaftate was introduced after griseofulvin, a natural product isolated from the mycelium of Penicillium griseofulvum (Oxford et al., 1939). Griseofulvin was launched ISSN 2053ISSN -2296 in 1959 by McNeil, Schering and Ayerst (Sittig, 1988) and is widely held to have been the first globally successful commercial antifungal agent. Griseofulvin is administered orally (being topically ineffective) and adverse side effects, e.g. photosensitivity, nausea, headaches, insomnia and so on, while infrequent, have been noted. In contrast, tolnaftate is administered topically (being orally ineffective) with little to no side effects and holds the distinction of being the first globally successful synthetic topical antifungal agent (Robinson & Raskin, 1964). Other popular topical antifungal compounds would be launched years later, e.g. clotrimazole, miconazole nitrate, terbinafine hydrochloride, and butenafine hydrochloride in 1973, 1974, 1991, and 1992, respectively (Sittig, 1988;Newman & Cragg, 2016). That tolnaftate has maintained a presence in the global over-the-counter marketplace in spite of the development of these newer APIs is rather remarkable.
Crystallographically, griseofulvin, clotrimazole, miconazole nitrate, and the hydrochloride salts of terbinafine and butenafine have all been structurally characterized. For example, the Cambridge Structural Database (CSD; Groom et al., 2016) lists 13 entries for griseofulvin alone, with the two most recent studies having been published earlier this year (Mahieu et al., 2018;Su et al., 2018). However, the crystal structure of tolnaftate, an equally classic compound as griseofulvin in the antifungal arena, is nowhere to be found in the CSD or the scientific literature. We therefore felt compelled to remedy this long-standing oversight of this historic compound that has provided relief to so many of us over the past half century.

Isolation and crystallization
A small vial was charged with 2 ml of Bayer Tinactin Liquid Spray followed by 2 ml of water and sealed. Upon standing at room temperature, extremely tiny colorless needles of tolnaftate formed and were harvested. These were redissolved in a minimum amount of 1:1 (v/v) acetone-water and the capped vial of the resultant solution then placed in a freezer to effect supersaturation and nucleation. As soon as crystals were noted, the vial was removed from the freezer and allowed to warm to room temperature, at which point the vial cap was loosened and the acetone-water allowed to evaporate slowly further to yield large flat colorless needles suitable for a singlecrystal X-ray diffraction experiment. No attempt was made to optimize the acetone-water ratio or to try other solvents since the task of obtaining crystallographic quality crystals had been achieved.

Refinement
Crystal data, data collection and structure refinement details are summarized in Table 1. A riding model was used for the H atoms, with the C-H distances constrained to 0.95 and 0.98 Å for the aryl and methyl moieties, respectively, and the U iso (H) values set at 1.2U eq (C) and 1.5U eq (C) for the aryl and methyl H atoms, respectively. The m-tolyl methyl group was treated as rotationally disordered over two orientations. The refined site-occupancy factors were 0.76 (2) and 0.24 (2) for the major and minor components of that disorder, respectively.

Results and discussion
Some readers may have already surmised from x1 and x2.1, that this study is a spin-off from a STEM outreach project for informal chemical and crystallographic education, i.e. for grades 6-12 pre-college students, homeschoolers, hobbyists, and amateur scientists. Chemistry is often introduced to this audience in the digestible and relatable form of common molecules and common household chemicals. One of the design criteria for our outreach project was to base it on a less commonly recognized common molecule. We believe that tolnaftate fits that criterion. It has been found in numerous households for over 50 years, yet most individuals have no notion of its structural identity and make-up. Another design criterion was cost, i.e. the chemical source was required to be relatively inexpensive and readily available to the targeted audience. The Bayer Tinactin Liquid Spray used in this study was purchased from a retail store for less than 6 US dollars.
Other generic sprays can also be purchased for 3-4 US dollars.
A single can of Tinactin can provide 2 ml aliquots for a class of roughly 75 students at a cost of 8 cents per student (or 150 students at 4 cents per student, if they work in pairs). We hope that the results and discussion provided below will be educational and simulating for those interested in STEM, and meaningful and entertaining for our academic and industrial colleagues as well.
The three substituents attached to the CNOS core exhibit the expected structural metrics. The aromatic rings are flat, with the r.m.s. deviations for the planes defined by atoms C1-C4/C4A/C5-C8/C8A and C13-C18 both being 0.0106 Å . The C2-O9-C10 angle is 119.25 (11) versus 120.0 for an idealized Osp 2 atom, the C10/N12/C13/C20 moiety and the CNOS core are both planar, with r.m.s. deviations of 0.0060 and 0.0053 Å for the fitted atoms defining each plane, and the C2, C13, and C20 atoms are 0.091 (2), 0.009 (2), and À0.074 (3) Å off of the CNOS plane, respectively. The CNOS and C 3 N moieties are also nearly coplanar with each other, with the angle between their normals being 2.24 (11) . These observations suggest that delocalization of -electron density over the entire CNOS unit is not geometrically disallowed or, at least, that the core C-N bond possesses partial double-bond character.
As depicted in the Scheme and Fig. 1, to a first approximation, the tolnaftate molecule is present in an E conformation in the solid state. The four most closely related N,N-disubstituted thiocarbamate structures in the CSD are GEHSAO (Mugnoli et al., 2006), JOXQIW (Sakamoto et al., 1998), MESHAY (Bowman et al., 2007), and YEDRAA (Vovk et al., 1992). In these, the methyl group is replaced by a C( X)R group, with X being either an O or S atom, and the aryl substitutent is either a tolyl or a phenyl group. The molecules in these prior structures are also present as E conformers. A wider comparison involving all relevant monosubstituted N-aryl thiocarbamates, i.e. with the methyl group replaced by an H atom, yields 46 such entries in the CSD with the ratio of E:Z stereochemistries being 40:6. Obviously, the N,N-disubstituted comparison suffers from both steric and electronic factors, e.g. the C( X)R groups are significantly larger and more polar than methyl, and the monosubstituted comparison suffers from N-H being significantly more prone to hydrogen-bonding effects than N-CH 3 . Nevertheless, prior studies would seem to suggest that an E conformation is preferred, and that is indeed what is found for tolnaftate as well.
For casual readers, this first approximation for describing the tolnaftate molecule is more than adequate. For others, additional stereodescriptors are required. Readers in the latter group will point out that Fig. 1 also clearly shows that the naphthyl moiety is cis to the S atom, i.e. s-cis with respect to the C-O core bond, and that a more precise description of the tolnaftate molecule is that it has an E,Z or trans,cis conformation. While justifiably superior to the E-only description, this second approximation using two stereodescriptors rather than one also falls short of being fully descriptive. For example, the tolnaftate molecule shown is three-dimensional (3D) and chiral in the solid state, i.e. the conformer in Fig. 1 and its enantiomer are present in our crystal as a 50:50 racemic mixture, and therein lies the problem. While the inverted molecule is indeed nonsuperposable on the conformer in Fig. 1, that enantiomer would be assigned the exact same stereodescriptors, i.e. it too is an E,Z or trans,cis conformer. Hence, one cannot differentiate between the two enantiomers with these stereodescriptors because the descriptors themselves are invariant on reflection in a mirror. The molecular structure of (I), showing the atom-labeling scheme. Displacement ellipsoids are drawn at the 50% probability level. The minor component of the disordered tolyl methyl group is drawn with dashed circles for the H atoms and dashed C-H bonds.
A third approach for specifying the conformation is to use clinal and periplanar descriptors, i.e. https://doi.org/10.1351/ goldbook.T06406 (Klyne & Prelog, 1960;Moss, 1996). These offer two significant advantages over the Z/E and cis/trans nomenclature, i.e. (a) they divide torsional space into six 60 regions rather than two 180 semicircular sections, and (b) they are signed + or À. We will apply a few nonstandard conventions and further subdivide the +30 to À30 region into 0 to +30 and 0 to À30 and assign the descriptors +sp and Àsp to them. Similarly, the +150 to À150 zone will be subdivided into 180 to +150 and 180 to À150 and assigned the descriptors +ap and Àap, respectively. This subdivision of torsional space into eight regions rather than six provides an even greater ability to distinguish one conformation from another. Lastly, we will expand the bonds of interest to be the core C-N, core C-O, N-C tolyl , and O-C naphthyl bonds, and assign descriptors to each in that order. Thus, the tolnaftate molecule shown in Fig. 1 is the (+ap,Àsp,+ac,Àac) conformer, while its enantiomer would be uniquely described as the (Àap,+sp,Àac,+ac) conformer.
The other two intermolecular interactions correspond to naphthyl-to-naphthyl Csp 2 -HÁ Á Á hydrogen bonds, i.e. C3-H3Á Á ÁX1 and C8-H8Á Á ÁX2. The first is an offset face-to-face hydrogen bond typically observed instacking, while the second is an edge-to-face hydrogen bond. The H3Á Á ÁX1 distance is 3.43 Å and is comparable to the value of 3.5 Å expected for a face-to-face hydrogen bond, while H8Á Á ÁX2 and C8-H8Á Á ÁX2 are 2.62 Å and 142 , and are in agreement with the values of 2.73 (13) Å and 148 (11) expected for edge-toface HÁ Á Á and Csp 2 -HÁ Á Á, respectively (Takahashi et al., 2001(Takahashi et al., , 2010. As shown in Fig. 2, the Csp 2 -HÁ Á ÁS C interactions form two-dimensional (2D) networks of hydrogen bonds that are parallel to the (h00) family of planes at x = 0.2 and x = 0.8. Similarly, the Csp 2 -HÁ Á Á interactions form a separate 2D network of hydrogen bonds also parallel to the (h00) family of planes but at x = 0.5. The end result is a packing structure reminiscent of an interdigitated lipid bilayer with the heteroatoms and polar bonds positioned near the outer surfaces and the nonpolar naphthyl substituents sandwiched in the interior of the bilayer.

Web theoretical conformations
While the CSD mentioned above is the world's repository for small-molecule crystal structures with over 900,000 curated entries, its database of experimental 3D coordinates is miniscule compared to databases providing theoretical 3D coordinates. For example, the PubChem database currently contains 96,396,575 compounds and 3D coordinates for over 88.    compounds, 3DChem is not as comprehensive as PubChem, Mol-Instincts or ATB, but it did list tolnaftate among its April 2017 Molecules of the Month showcasing of antifungal agents. A visual comparison of online theoretical tolnaftate models versus our X-ray structure is shown in Fig. 3, and selected distances and angles are provided in Table 3. The conformations for the tolnaftate molecules in Fig. 3 are (+sc,+sc,+sp,+sc), (+sp,+sp,+ac,Àac), (+sp,+sp,+ac,Àac), and (+ac,+ac,Àsp,+ap) for the 3DChem, ATB, Mol-Instincts, and PubChem models, respectively, while the experimentally observed conformation for (I) is (+ap,Àsp,+ac,Àac). The ATB and Mol-Instincts models are visually similar, but are strikingly different from the 3DChem and PubChem models. None of the theoretical models are a match to our X-ray structure. This is not unexpected since most theoretical models ignore intermolecular interactions and packing forces such as those described above in x3.2. That having been said, the mismatch among the theoretical models themselves is probably of greater concern than their mismatch to (I).
The distances and angles in Table 3 reveal the discrepancies in the online theoretical structures. The 3DChem C-N bond at 1.468 Å is a significant outlier. The PubChem C-O bond at 1.432 Å is unusually long. The 3DChem C-S bond is uncomfortably short at 1.595 Å , while the opposite is true for the Mol-Instincts C-S bond at a lengthy 1.712 Å . The 3DChem N-C-O angle is alarmingly acute at 99.0 and its O-C-S angle is alarmingly obtuse at 131.1 . The Mol-Instincts N-C-O, N-C-S, and O-C-S angles are all 120.0 , a highly improbable occurrence, suggesting that that model was likely minimized with constraints. The PubChem O-C-S angle is also an outlier at a meager 115.3 . These nonsensical distances and angles for just the CNOS cores alone suggest that the 3DChem, Mol-Instincts and PubChem models are somewhat suspect. This is not to say that these models are invalid, as they may represent local minima on the tolnaftate potential energy landscape, but the unreasonableness of their CNOS core geometries suggests that attempting to rank them on a common relative energy scale is not worth the effort. Rather, we will simply say that the 3D coordinates from ATB (Malde et al., 2011) appear to be the most robust set among this small sampling of theoretical tolnaftate models.
Taken as a whole, all of these observations suggest that the current standards and guidelines for online theoretical 3D models and coordinates are rather loose, and that the validation methods used by website providers for assessing the structural reasonableness of their optimized molecules are less than fully adequate. Individuals in our targeted audience of nonscience professionals should therefore consider any 3D model or coordinates that they download from the web to be potentially suspect unless clearly demonstrated otherwise, i.e. we encourage them to critically examine the geometrical attributes (distances, angles etc.) of those models or seek help from others to do so, if need be.

Peer-reviewed theoretical conformations
For completeness, we are aware of two additional theoretical tolnaftate models. The first of these was published by Joe and co-workers as part of a vibrational analysis of the tolnaftate IR and Raman spectra (Dhas et al., 2011). A simulation of their model is shown in Fig. 4. The conformation depicted in Fig. 4 was generated with the freeware molecular editor Avogadro 1.2.0n, downloaded from https://avogadro.cc/ (Hanwell et al., 2012), and adjusted until there was a reason-research papers Acta Cryst. (2018). C74, 1495-1501 Ho and Zdilla The solid-state conformation of a topical antifungal agent 1499 Table 3 Selected distances and angles (Å , ) for a sampling of online theoretical tolnaftate models versus the experimental X-ray structure (I).    Hydrogen-bond geometry (Å , ) for (I).
X1 and X2 are points on neighboring naphthalene planes from which normals are drawn to H3 and H8, respectively (Takahashi et al., 2001). Symmetry codes: (i) x, Ày + 3 2 , z + 1 able match to Fig. 1 in the Dhas 2011 publication. The Dhas model is clearly a variant of the ATB and Mol-Instincts theoretical models, i.e. (+sp,+sp,+ac,Àac), and not a match to our experimentally observed (+ap,Àsp,+ac,Àac) conformation. Our X-ray results are particularly relevant to this 2011 paper since their IR spectrum was taken on a solid-state tolnaftate sample in KBr. Since our study unequivocally establishes that the solid-state conformation of tolnaftate is (+ap,Àsp,+ac,Àac) and not (+sp,+sp,+ac,Àac), their vibrational analysis based on the latter conformation is unlikely to be valid. For their published results to be valid, their sample would have to be a tolnaftate polymorph with the molecules in the (+sp,+sp,+ac,Àac) conformation, which is highly improbable. To our knowledge, there is no powder diffraction or DSC evidence that tolnaftate polymorphs exist. Significant variations in powder pattern peak intensities have been observed, but such observations are completely attributable to preferred orientation effects without the need to invoke polymorphism. Hence, the calculated wavenumbers and all other computed quantities based on their use of a (+sp,+sp,+ac,Àac) model should be considered suspect.

D-HÁ
The second peer-reviewed theoretical tolnaftate model that we are aware of is that published by Sun and Liu and coworkers as part of a study on the in silico docking of tolnaftate into the active site of squalene epoxidase (Sun et al., 2017). A simulation of their model is shown in Fig. 5. As with the 2011 paper above, no 3D coordinates were provided, so we used Avogadro 1.2.0n to approximate the theoretical tolnaftate model in Fig. 4 of the Sun 2017 publication. The Sun model appears to be a variant of the 3DChem model and is also not a match to our experimentally observed X-ray structure. Further, the CNOS core depicted in their Fig. 4 exhibits significant abnormalities, e.g. their tolnaftate C-S and C-N bonds seem unrealistically long, the C-O bond too short and the N-C-S angle overly obtuse. Their tolnaftate N atom having a pyramidal geometry is also highly unprecedented. Moreover, their Fig. 4 also indicates that the binding of liranaftate, a related and sterically bulkier thiocarbamate, to squalene epoxidase occurs without comparable distortion to its CNOS core. Unfortunately, even their liranaftate model is not without peculiarities, e.g. the pyramidalization of the aromatic C4a and C8a atoms in the O-5,6,7,8-tetrahydronaphthalen-2-yl moiety is chemically and structurally unrealistic. All of these observations suggest that both of their theoretical models, i.e. for tolnaftate and for liranaftate, are probably indefensible and that a follow-up study may be needed to ascertain whether the findings and conclusions of their 2017 study are valid or not.
What this comparison of our experimental model to peerreviewed theoretical models reveals is that the standards and guidelines for publishing theoretical structures are currently less rigorous than those for publishing crystal structures. Acta Crystallographica, where Density Functional Theory (DFT) and various other theoretical results are increasingly being showcased, should consider implementing policies that will safeguard against questionable theoretical models reaching print. Insisting that 3D coordinate files must be submitted as supplementary material for theoretical models (and not just for X-ray structures) would greatly facilitate the peer-review process by referees and editors, and any subsequent scientific inquiries by readers.

Summary and closing comments
The take-home message for amateur scientists and science enthusiasts is that opportunities for scientific adventures and discoveries in a home setting do indeed exist and may even be publishable. All of the wet chemistry presented in this paper were done in a common household kitchen and all of the structure solution, refinement, and manuscript preparation were carried out with readily available freeware using public computers in local libraries. Exploring science in the 21st Century can be that simple. Also, be willing to investigate opportunities at local colleges and universities where specialized equipment like an X-ray diffractometer might be available through various outreach programs. Adult supervision and guidance are, of course, encouraged for any projects involving minors or other pre-college individuals.