Host-Guest Systems for the SAMPL9 Blinded Prediction Challenge: Phenothiazine as a Privileged Scaffold for Binding to Cyclodextrins

Abstract

Model systems are widely used in biology and chemistry to gain insight into more complex systems. In the field of computational chemistry, researchers use host-guest systems, relatively simple exemplars of noncovalent binding, to train and test the computational methods used in drug discovery. Indeed, host-guest systems have been developed to support the community-wide blinded SAMPL prediction challenges for over a decade. While seeking new host-guest systems for the recent SAMPL9 binding prediction challenge, which is the focus of the present PCCP Themed Collection, we identified phenothiazine as a privileged scaffold for guests of β cyclodextrin (βCD) and its derivatives. Building on this observation, we used calorimetry and NMR spectroscopy to characterize the noncovalent association of native βCD and three methylated derivatives of βCD with five phenothiazine drugs. The strongest association observed, that of thioridazine and one of the methyl derivatives, exceeds the well-known high affinity of rimantidine with βCD. Intriguingly, however, methylation of βCD at the 3 position abolished detectible binding for all of the drugs studied. The dataset has a clear pattern of entropy-enthalpy compensation. The NMR data show that all of the drugs position at least one aromatic proton at the secondary face of the CD, and most also show evidence of deep penetration of the binding site. The results of this study were used in the SAMPL9 blinded binding affinity-prediction challenge, which are detailed in accompanying papers of the present Themed Collection. These data also open the phenothiazines and, potentially, chemically similar drugs, such as the tricyclic antidepressants, as relatively potent binders of βCD, setting the stage for future SAMPL challenge datasets and for possible applications as drug reversal agents.

2. Introduction

Community-wide blinded prediction challenges are a powerful and rigorous way to evaluate computational approaches to a range of technical problems. They allow a direct comparison of methods, because all participating groups work on the same systems or data, and they avoid bias that might otherwise result from the availability of the “true” result to the predictors. Accordingly, blinded challenges are used by multiple communities, including protein structure prediction (CASP(1,2)), protein function prediction (CAFA(3)), various problems in machine learning (Kaggle), and protein-ligand pose and affinity prediction (D3R(4)). The SAMPL blinded prediction exercises challenge computational chemists to predict various experimental observables related to computer-aided drug design(5–8), posing problems that range from full-on prediction of protein-ligand binding affinities down to the prediction of much simpler physical properties, such as the hydration free energies of small molecules.

These simpler properties may have little direct biomedical or economic importance, but they are informative because they can isolate and test the ability of a computational method to predict properties that play a role in important problems. For example, the hydration free energies of organic molecules are relevant because protein-ligand binding causes partial dehydration of both the protein and the ligand. If the free energy changes associated with these dehydration events is erroneous, it will be difficult to compute an accurate binding affinity. Simple problems also require less conformational sampling, by e.g., molecular dynamics and Monte Carlo methods, so they maximize the precision of the results and are relatively inexpensive to compute, enabling broader participation and testing of more methodologies.

In the spirit of assigning simple but informative computational challenges, SAMPL has, since 2012, challenged computational chemists to predict the affinities of aqueous host-guest systems(6,8), for hosts that include cucurbiturils(9–13), deep cavity cavitands(14,15), cyclodextrins(16–18) (CDs), and derivatives thereof. Host-guest binding is both driven and opposed by the same basic forces the determine the affinities of protein-ligand systems, but they can be chosen to avoid ambiguities in protonation states that frequently arise in protein-ligand modeling, and they avoid the difficulties of ensuring adequate numerical convergence that are endemic in protein simulations. As a consequence, they can report clearly on the quality of the potential function, or force field, used in the calculations. Accordingly, recent studies have demonstrated the use of host-guest binding data to test and adjust the parameters of force fields, in order to enhance the accuracy of binding affinity calculations(19–21).

However, new host-guest binding data that are suitable for the SAMPL challenges are relatively sparse in the scientific literature, and most new host-guest data generated by experimental labs are published straightaway, rather than being reserved long enough that they can be used in a blinded challenge. As a computationally focused research group, we are keenly aware of the value of new data that can be held back long enough to run a blinded challenge, so we have developed the ability to synthesize and characterize new host-guest systems and have generated new host-guest systems with experimental binding data for use in SAMPL(8). When possible, we provide not only a binding constant that may be compared with computation, but also NMR data that bear on the conformation of the bound host-guest complexes, as these, too, can be used to assess the computational results.

We have focused on CDs and their derivatives, because it is reasonably straightforward to generate diverse derivatives from commercially available starting materials, and because CD have a strong history of pharmaceutical applications, originally in drug formulations(22,23) and more recently as therapeutics. For example, the γCD derivative sugammadex, which is widely used to reverse the action of the paralytic drugs rocuronium and vecuronium post-surgery, works by binding directly to the paralytic agents and thus reducing their free concentration(24). There is also promise in the use of cyclodextrin derivatives in the treatment of Niemann-Pick disease(25,26) and atherosclerosis(27,28). Thus, new data for CD-guest binding may have pharmaceutical implications beyond SAMPL.

In developing a dataset for SAMPL9, which is the focus of a themed Collection of Physical Chemistry chemical Physics, we sought to collect guests having drug-like structures that would survey structure-activity relationships (SARs) analogous to those typically developed by medicinal chemists engaged in targeted drug discovery projects, and that would include some with reasonably strong affinities for the CDs. This is not a trivial task, as CDs bind most guests in a rather narrow band of rather modest affinities(18,29). One survey of over 1200 available experimental CD-guest binding free energies reported a mean of −14.6 kJ/mol with standard deviation 5.9 kJ/mol(29). Because the adamantyl derivatives rimantidine and amantadine are known to bind βCD with good affinity(30), we initially tested purchasable adamantane derivatives, but these were not sufficiently water-soluble. Other compound series that also possess a hydrophobic moiety to drive binding and a polar moiety to drive solubility also failed to combine both attributes. We then found that βCD bound a phenothiazine drug with good affinity, so we procured and tested phenothiazines and chemically similar drugs and found that virtually all have favorable solubilities and CD affinities. These observations, combined with the ongoing clinical relevance of the phenothiazines and the fact that the phenothiazines would be new to SAMPL, motivated us to use them as the basis of the present challenge. The phenothiazines chosen here are listed in Table 1.

Table 1.

Names, abbreviations, and structures of the five phenothiazine drugs studied here, along with computational estimates of selected chemical properties from Drugbank (31). Protonation states shown are those expected at pH 7.4.

In a dataset that we previously provided for SAMPL7, we used CDs monosubstituted at the secondary face with large substituents of varied polarity (e.g., a steroid, a zwitterionic amino acid, and a sugar). However, these substitutions led to only modest shifts in affinity. We have since found that heptakis substitution of specific hydroxyls of βCD with even small substituents can modulate binding more substantially. Accordingly, we generated the current dataset by testing our selected phenothiazines against native βCD and derivatives modified by heptakis methylation at the three classes of hydroxyls: per-6-O-methyl, per-2,6-di-O-methyl, and per-2,3,6-tri-O-methyl (Figure 1). Combining these with the phenothiazine family of guests results in a dataset that systematically probes chemical variations of both the phenothiazine guests and the CD hosts, producing trends in binding thermodynamics that make for a systematic and informative SAMPL challenge. We used a combination of isothermal titration calorimetry (ITC) and NMR spectroscopy to explore how three patterns of βCD methylation influence the binding of five phenothiazine drugs (Table 1). The resulting data were used in the SAMPL9 blinded binding affinity-prediction challenge (https://github.com/samplchallenges/SAMPL9), as reflected in the accompanying articles. Note that our core purpose is not to provide a theoretical or computational analysis of the present experimental results, but rather to provide the data needed to support the SAMPL9 studies, while also highlighting features of interest. Because the drugs studied here are in clinical use for various chiefly psychiatric indications and carry significant toxicity at high doses, the present data also raise the possibility of developing CD derivatives with higher affinities for these drugs, for possible use as overdose reversal agents, much as sugammadex is used to reverse paralysis by rocuronium and vecuronium.

Figure 1.

Structures of βCD, H6M-βCD, H26DM-βCD, and HTM-βCD.

3. Materials and Methods

3.1. Materials

The phenothiazines, HTM-βCD, and H26DM-βCD were purchased from Sigma Life Sciences; βCD was obtained from Alfa Aesar; and H6M-βCD was purchased from The Cyclodextrin Shop. All compounds were used without further purification. All solutions were prepared by weight in buffer or D₂O for calorimetry and NMR, respectively.

3.2. Isothermal titration calorimetry (ITC) experiments

All ITC measurements were run in triplicate on a Microcal ITC200 in 25mM sodium phosphate buffer, pH 7.4, at 25 °C with a 20-point binding isotherm (1-1.5μL injections). A single CD stock solution was drawn from, but a fresh drug solution was prepared for each replicate.

In each case, the guest was in the cell and the host in the syringe. The concentrations of the solutions were tailored for each complex, based on binding affinity, to obtain appropriate C values(32). Host concentrations ranged between 7 and 10 mM, and guest concentrations ranged between 200 and 500 μM, below their respective critical micellar concentrations(33). The raw data were analyzed with Microcal Data Analyzer in Origin 7.

3.3. NMR

¹H- and 2D NOESY NMR spectra of all phenothiazine-CD complexes were collected at 298 K on a 600 MHz Bruker Avance III spectrometer fitted with a 1.7 mm triple resonance cryoprobe with z-axis gradients. All NMR studies were run in D₂O. ¹H-NMR was collected with F1 presaturation of the water peak with 16 scans, and 2D NOESY spectra were performed using water suppression and excitation sculpting with gradient and States-TPPI acquisition mode. One-dimensional ¹H NMR was used to determine the chemical shift perturbations of the complexed molecules and to assign NOE signals. These spectra and the peak assignments are provided in the Supporting Information.

4. Results

4.1. Binding Thermodynamics

We used ITC to determine the binding free energy, enthalpy, and entropy, of five phenothiazine-based drugs with βCD and two methylated derivatives, H6M-βCD and H26DM-βCD, under uniform conditions of temperature and buffer. Interestingly, no binding was detected between any of the phenothiazines and the third derivative, HTM-βCD. The remaining results, averaged over the triplicate runs, are summarized in Tables 2–4 and Figure 2, while the Supporting Information provides enthalpograms, results of individual titrations, and concentrations used.

Table 2.

Binding measurements for βCD with phenothiazine drugs. N: fitted host-guest binding stoichiometry. Ka: association constant. Kd: dissociation constant = 1/Ka. ΔG: standard free energy of binding (-RT ln Ka). ΔH: standard enthalpy of binding. ΔS: standard entropy of binding. Units as listed. All results are means of triplicate measurements. Uncertainties are standard errors of the mean across the triplicate measurements.

Guest	N	Ka (M−1)	Kd (μM)	ΔG (kJ/mol)	ΔH (kJ/mol)	ΔS (kJ/K/mol)
PMZ	1.1 ±0.01	4440 ±90	225 ± 5	−20.8 ±0.1	−24.8 ±0.3	−13.2 ±1
PMT	0.94 ±0.03	1900 ±90	525 ± 30	−18.7 ±0.1	−16.5 ±0.8	7.6 ±3
CPZ	0.77 ±0.01	9390 ±190	107 ± 2	−22.7 ±0.1	−26.0 ±0.5	−11.1 ±3
TDZ	1.1 ±0.1	15100 ±400	66 ± 2	−23.9 ±0.1	−20.6 ±0.5	10.8 ±2
TFP	1.2 ±0.1	5100 ±300	196 ± 10	−21.2 ±0.1	−16.3 ±1.0	16.3 ±4

Table 4.

Binding measurements for H26DM-βCD with phenothiazine drugs. See Table 2 for details.

Guest	N	Ka (M−1)	Kd (μM)	ΔG (kJ/mol)	ΔH (kJ/mol)	ΔS (kJ/K/mol)
PMZ	0.99 ±0.01	5080 ±90	197 ± 3	−21.2 ± 0.05	−21.4 ± 0.1	−0.9 ±0.3
PMT	1.1 ±0.03	8420 ±90	119 ± 1	−22.4 ±0.03	−17.0 ± 0.1	18.2 ±0.2
CPZ	0.77 ±0.01	9100 ±300	110 ± 3	−22.6 ±0.08	−25.0 ± 0.4	−7.9 ±0.3
TDZ	0.87 ±0.01	54900 ±2000	18 ± 0.6	−27.1 ±0.09	−38.3 ± 0.2	−37.8 ± 0.8
TFP	0.56 ±0.04	11500 ±400	87 ± 3	−23.2 ±0.09	−31.1 ± 2	−26. ± 7

Figure 2.

Thermodynamic signatures of binding for βCD (top), H6M-βCD (middle), and H26DM-βCD (bottom). See Table 1 for guest name abbreviations.

4.1.1. Binding Affinities

Across all systems studied, the measured dissociation constants range between 18 μM (H26DM-βCD with TDZ and 820 μM (βCD with PMT). On average, the binding free energies are 1.3 standard deviations below the mean (higher affinity) of 1,257 previously compiled CD-guest binding free energies(29), and TDZ/H26DM-βCD is 2.1 standard deviations below the mean. In fact, TDZ binds each three hosts more tightly than any of the other guest molecules. Overall, methylation at the 2 and 6 positions has only modest effects on affinity, as the maximum changes in K_d from native βCD are only about four-fold across both CD derivatives and all guests. In contrast, as noted above, 3-methylation destroys binding in all cases. Going from native βCD to H6M-βCD produces small, mixed effects on affinity; the greatest drop in K_d, which occurs for TDZ, is only two-fold. Going from native βCD to H26DM-βCD leads in all cases to either an improvement in affinity or an essentially unchanged affinity, with a change in the mean binding free energy from −21.5 to −23.3 kJ/mol, and the greatest improvement, ~4 kJ/mol, observed for PMT. Although promazine and chlorpromazine differ only in the replacement of a hydrogen by a chlorine, chlorpromazine binds more tightly than promazine in all cases, with a maximal improvement, relative to promazine, of 4.6x in the case of H6M-βCD. Finally, we note that our result for the binding free energy of CPZ with βCD, −22.7 kJ/mol, agrees closely with a prior ITC measurement (−22.4 kJ/mol)(34).

Rimantidine is known as one of the tighter-binding guests of βCD, with a measured dissociation constant of ~30 μM(30,35). Thus, it is of interest that TDZ with βCD comes close to this affinity (66 μM), and that TDZ with H26DM-βCD surpasses it (18 μM). Indeed, two of the TDZ complexes have binding free energies over 2 standard deviations better than the mean in an overview of CD-guest binding(29).

In our prior study of rimantidine and methylcylohexanol binding to βCD and βCDs mono-derivatized at the 3 and 6 hydroxyls, we observed that derivatization at 3 always weakened binding, while derivatization at the 6 position largely preserved it(35). This result is broadly consistent with our present observation that per-substitution at 3 eliminates binding, while derivatization at 6 leaves it largely unchanged. Although there is no guarantee that similar results will be observed for other guests or CD substituents, we have observed similar results with other guests as well (unpublished results). Thus, we suggest that efforts aimed at boosting the affinity of CDs for various guests will be more productive if they avoid modifying the 3 hydroxyl site.

4.1.2. Binding Enthalpies and Entropies

The standard binding free energies across all host-guest systems with detectable binding range over 10 kJ/mol, but the binding enthalpies range much more widely, from −38.3 kJ/mol to −12.6 kJ/mol. The change in enthalpy favors binding in all cases, though. The small range of the binding free energy and the large range of the binding enthalpy imply the presence of entropy-enthalpy compensation within the dataset. This results, frequently observed in many other systems but, arguably, not yet fully understood(36–41), is illustrated in Figure 3, where the present data are displayed along with those from a prior ITC study from our lab.

Figure 3.

Entropy-enthalpy compensation plot. Each point is one host-guest pair. Colored points are new data from this study and are colored according to the host compound. Gray-scale points, provided for context, are data from our prior study (35), which studied the guests rimantidine and trans-methyl-cyclohexanol across multiple βCD derivatives; these are shaded by guest compound. The radius of each data point is proportional to the measured standard free energy of binding.

4.2. Host-Guest Interactions via 2D NOESY NMR

Both the guests and the hosts studied here are asymmetric, so their bound complexes might adopt any of a number of conformations, defined, for example, by the positioning of the guests’ bulky, tricyclic phenothiazine moieties and their cationic side chains within the binding site. We gathered information on these conformations by using two-dimensional NOESY NMR studies of the host-guest complexes to identify pairs of host and guest atoms that reside close enough (less than about 6Å) to be detectible as an NOE cross-peak. We note, however, that the modest solubility of the CDs studied led to limited resolution, and some NOEs may not have been identified due to the challenge of assigning certain chemical shifts in the 1D spectra. Note that we did not seek to resolve intramolecular NOEs. Interested readers may find the 1D and 2D spectra in the Supporting Information. The resulting NMR information is not sufficient to define the bound poses of the guests within the CD hosts, but they can be used as a check of the complexes generated with computer models.

Every complex studied here, except for one where no NOEs at all were observed, has NOEs between at least one aromatic guest proton (G1-G9 in Figure 4. Numbering of atoms used to identify NOEs in Tables 5–7) and the hydrogen of the CD’s H3 carbon (H3 in Figure 4. Numbering of atoms used to identify NOEs in Tables 5–7.) Thus, in all cases, at least part of the phenothiazine moiety is at the secondary face of the CD. In most complexes, there are also NOEs between aromatic guest protons and the CD’s H5, signifying that part of the phenothiazine moiety penetrates deep into the host’s cavity. The drugs TDZ and TFP show additional NOEs that involve the cationic ends of their side-chains. These NOEs consistently involve H6 or HMe1 on the host, meaning that the cationic side chains lie at the primary face of the hosts. We conjecture that the cationic side chains of all of the phenothiazines studied here reside at the primary face, but that only the relatively bulky side-chains of TDZ and TFP are sufficiently locked in to generate observable NOEs.

Figure 4.

Numbering of atoms used to identify NOEs in Tables 5–7.

Table 5.

Observed NOEs in phenothiazine-βCD complexes. Atom notation is provided in Figure 4.

Guest	NOEs Observed
PMZ	3.423-7.313 (H3-G8 & 2)
PMT	3.619 – 7.265 (H5 – G7) 3.619 – 7.045 (H5 – G9) 3.759 – 7.044 (H3 – G9)
CPZ	3.806- 7.401-7.343( H5-G8, 6, 4, & 2) 3.806- 7.219-7.209( H5-G3 & 7) 3.853 – 4.378- H3-G11a 3.806 – 4.282- H3-G11b
TDZ	3.712- 7.357-7.331 ( H5-G2 & 3) 3.712- 7.357-7.331 ( H5-G2 & 3) 3.759 – 6.928 ((H3 – G9) 3.712 – 6.900 (H5 – G7) 3.853 – 2.468 (H6 – GMeS
TFP	3.622 - 7.479 ( H5-G2) 3.763- 7.480 (H3-G2) 3.853 – 2.468 (H6 – G20)

Table 7.

Observed NOEs phenothiazine-H26DM-βCD complexes.

Guest	NOEs Observed
PMZ	3.712 – 7.270 – 7.259( H3 – G6 &4) 3.572 – 7.096 ( HMe2 – G3&7)
PMT	No NOE Observed
CPZ	3.723- 7.228 (H3 – G6) 3.723- 7.004 (H3 – G9)
TDZ	3.712- 7.276 (H3 – G6) 3.712 – 6.625 (H3 -G9) 3.385 - 2.480 (HMe1 – GMeS)
TFP	3.590 - 7.104 ( H3-G7 & 9) 3.590 – 2.935 (H6 – G20)

5. Discussion

Phenothiazine is a privileged bioactive scaffold, made up of a tricyclic aromatic unit with a thiazine center(42,43). It is present in multiple antipsychotic drugs, including promethazine (Phenergan) chlorpromazine (Thorazine), thioridazine (Mellaril), and trifluoperazine (Stelazine), and prochlorperazine (Compro). In addition, the phenothiazine drugs are toxic at high dose and there is no specific antidote for overdoses. CDs that bind phenothiazines well could thus be of interest as antidotes. However, little is known about how even the most straightforward CD derivatives interact with this class of drugs. The present study clearly shows that phenothiazine is also a privileged scaffold for tight-binding CD guests, as every guest studied here shows good affinity for native βCD and two derivatives thereof. Intriguingly, though, none of these drugs bound a derivative permethylated at the 3 position. Notable additional observations include:

1. Permethylation at the 2 and 6 positions leads to up to 4-fold stronger binding of some of these guest molecules.

2. TDZ is consistently the strongest binder, with a dissociation constant as low as 18 μM in the case of H26DM-βCD, surpassing that of rimantidine with native βCD.

3. Chlorpromazine binds consistently more strongly than promazine, with a Kd nearly 5-fold lower in the case of H6M-βCD.

4. The enthalpy of binding is strongly favorable in all cases, and the thermodynamic data fall on a classic entropy-enthalpy compensation line.

5. 2D NMR studies show that all guests position at least one aromatic proton at the secondary face of the CD in all cases, and most also have NOEs to H5, indicating deep penetration of the binding site.

6. TDZ and TFP have additional NOEs between their cationic side-chains and protons at the primary face.

It is of interest to consider the possible causes of these structure-activity relationships. Small molecule crystallographic studies of pure phenothiazines(44,45) show that the aromatic rings of neighboring phenothiazines can form various stacked conformations that are probably energetically favorable(46). However, CDs lack aromatic groups, so such interactions are not available to stabilize the present host-guest systems. It has also been argued, based on crystallographic data, that the aromatic protons of phenothiazines can donate hydrogen bonds to, e.g., nitro oxgyens(47). Here, in principle, these protons could donate hydrogen bonds to the hydroxyl oxygens of the CDs. However, aromatic hydrogens are not very good hydrogen bond donors, so any such interaction is likely weak and, in the aqueous environments studied here, would likely be outcompeted by strong hydrogen bonds from water molecules. All of the phenothiazine drugs studied here possess an amine side chain that is expected to be fully protonated at the experimental pH, and we expect that the ammonium groups spend most of their time at locations where they remain well hydrated -- i.e., outside the hydrophobic binding cavities of the CDs -- because of the high free energy cost of stripping water from charged groups. Thus, the ammonium groups could, in principle, form stabilizing interactions with the hydroxyl groups of the CDs, since these are at the well-hydrated portals of the host molecules . However, if ammonium-hydroxyl interactions contributed much to binding, we would expect the drugs to bind H26DM-βCD with less affinity than native βCD, because 14 of the hydroxyls in H26DM-βCD have been converted to ethers in H26DM-βCD, and ether oxygen atoms are considerably worse hydrogen bond acceptors than hydroxyl oxygens(48), but we find instead that H26DM-βCD binds the drugs at least as well as βCD does. Thus, we doubt that ammonium-hydroxyl interactions contribute much to stability. That said, it remains possible that the 3-hydroxyls, which remain in H26DM-βCD, do form stabilizing interactions with the ammonium groups of the phenothiazine drugs. Ultimately, though, we surmise that the most important single contribution to binding in these systems is the hydrophobic effect, given that the largely hydrophobic phenothiazine ring systems are sized well to fit into the hydrophobic CD binding cavity. This view is consistent with the trend to preserved and even improved binding by H26DM-βCD, which is more hydrophobic than native βCD. We conjecture that the sharp , across-the-board drop in affinity on going to the even more hydrophobic HTM-βCD traces to steric obstruction by the newly added methyl groups at the 3-hydroxyls.

The present data set, which was designed to challenge computational methods of predicting binding affinities in the SAMPL9 challenge, also highlights the promise of βCD and its derivatives as potent binders of phenothiazine drugs and provides intriguing observations, such as the strong, unfavorable influence of the 3 position of βCD on affinity. These data are expected to be of ongoing use for the evaluation and improvement of computational methods.

Supplementary Material

zip file containing details of the experimental results

1. SAMPL9-Andrade-Chen-Gilson-ITC-complexes.xlsx: Results of analysis of ITC data for triplicate runs for each host-guest system, Host and guest SMILES strings, and binding thermodynamic summary.

2. SAMPL9-Andrade-Chen-Gilson-ITC-Raw-Data.zip: compressed file with numerical data from the ITC runs.

3. SAMPL9-Andrade-Chen-Gilson-Enthalpograms.pptx: Images of combined enthalpogram for each CD-phenothiazine pair.

4. SAMPL9-Andrade-Chen-Gilson-2DNMR-BCD.pptx: 2D NMR spectra of βCD and phenothiazine complexes

5. SAMPL9-Andrade-Chen-Gilson-2DNMR-H6MBCD.pptx: 2D NMR Heptakis-(2,6-di-methyl)-β-CD and phenothiazine complexes

6. SAMPL9-Andrade-Chen-Gilson-2DNMR-H26DMBCD.pptx: 2D NMR Heptakis-(2,6-di-methyl)-β-CD and phenothiazine complexes

7. SAMPL9-Andrade-Chen-Gilson-1HNMR-Phenothiazines-CDDerivatives.pptx: 1D NMR spectra of the bound complexes.

Table 3.

Binding measurements for H6M-βCD with phenothiazine drugs. See 2 for details.

Guest	N	Ka (M−1)	Kd (μM)	ΔG (kJ/mol)	ΔH (kJ/mol)	ΔS (kJ/K/mol)
PMZ	1.2 ±0.1	2700 ±400	380 ± 50	−19.9 ±0.2	−20.2 ±2	−1. ±8
PMT	0.79 ±0.05	1220 ±50	820 ± 40	−17.6 ±0.1	−24.7 ±2	−24 ±8
CPZ	0.62 ±0.02	12100 ±700	83 ± 5	−23.3 ±0.2	−12.6 ±0.6	36 ±2
TDZ	1.2 ±0.03	31200 ±1000	32 ± 1	−25.7 ±0.1	−29.4 ±0.5	−13 ±2
TFP	1.0 ±0.06	7700 ±600	130 ± 10	−22.2 ±0.2	−22.8 ±1	−2 ±5

Table 6.

Observed NOEs in phenothiazine-H6M-βCD complexes.

Guest	NOEs Observed
PMZ	3.712 – 7.313 (H5 – G4 & 6) 3.712 – 7.152 (H5 – G1 & 9)
PMT	3.776 – 7.377 (H3 – G3 &7) 3.776 – 7.339 (H3 – G6 & 4) 3.776 – 7.201 (H3 – G1 & 9)
CPZ	3.745 – 7.375 (H3 – G3) 3.745 – 7.217 (H3 – G1) 3.698 – 7.217 (H5 – G1) 3.698 – 7.270 (H5 – G6) 3.698 – 7.146 (H5 – G4) 3.698 – 7.051 (H5 – G9)
TDZ	3.700 – 7.236 (H5 – G1) 3.700 – 6.902 (H5 – G7) 3.746 – 7.356 (H3 – G4 &6) 3.746 – 7.270 (H3 – G6) 3.746 – 6.915 (H3 – G7) 3.746 – 6.870 (H3 – G9)
TFP	3.722 – 7.263(H5 – G1) 3.722 – 7.492 (H5 – G2) 3.722 – 7.276(H5– G6) 3.722 – 7.158(H5 – G9) 3.769 – 7.170 (H3 – G9) 3.769 – 4.204 (H3 – G11)