Acyclic cucurbit [ n ] uril-type molecular containers : in fl uence of glycoluril oligomer length on their function as solubilizing agents †

We present the synthesis of a series of six new glycoluril derived molecular clips and acyclic CB[n]-type molecular containers (1–3) that all feature SO3 − solubilizing groups but differ in the number of glycoluril rings between the two terminal dialkoxyaromatic sidewalls. We report the X-ray crystal structure of 3b which shows that its dialkoxynaphthalene sidewalls actively define a hydrophobic cavity with high potential to engage in π–π interactions with insoluble aromatic guests. Compounds 1–3 possess very good solubility characteristics (≥38 mM) and undergo only very weak self-association (Ks < 92 M) in water. The weak self-association is attributed to unfavorable SO3 ⋯SO3 electrostatic interactions in the putative dimers 12–42. Accordingly, we created phase solubility diagrams to study their ability to act as solubilizing agents for four water insoluble drugs (PBS-1086, camptothecin, β-estradiol, and ziprasidone). We find that the containers 3a and 3b which feature three glycoluril rings between the terminal dialkoxy-o-xylylene and dialkoxynaphthalene sidewalls are less efficient solubilizing agents than 4a and 4b because of their smaller hydrophobic cavities. Containers 1 and 2 behave as molecular clip type receptors and therefore possess the ability to bind to and thereby solubilize aromatic drugs like camptothecin, ziprasidone, and PBS-1086.


Introduction
A ubiquitous problem facing the pharmaceutical industry is that an estimated 40-70% of new drug candidates are so poorly soluble that they cannot be formulated on their own. 1 Accordingly, there is a real need for the development of new tools that improve the solubility of poorly soluble drugs that enable their formulation.To date, a wide variety of methods that enhance the rate and extent of dissolution have been developed including solid dispersions and nanocrystalline solid forms of the drug. 2 Other methods that have been demonstrated to improve solubility include the preparation of salts, higher solubility prodrugs, dendrimer-drug systems, and designed co-crystalline forms of the drug. 3However, the most attractive approach to improving solubility of insoluble drugs from the point of view of supramolecular chemists relies on the use of cyclodextrin molecular containers (e.g.HP-β-CD, SBE-β-CD, Fig. 1) as solubilizing excipients. 4Cyclodextrin derivatives are currently used to formulate a number of drugs that are administered to humans.
In recent years, we and others, have been actively investigating the synthesis, molecular recognition properties, and applications of an alternative class of molecular container compounds known as cucurbit[n]urils (CB[n], n = 5, 6, 7, 8, 10, 14; Fig. 1). 5 The great interest in CB[n] compounds arises from the availability of a homologous series of hosts that display both high affinity (K a up to 10 17 M −1 ) and high selectivity toward their guests in water. 6In addition, the inherent stimuli responsiveness ( pH, chemical, electrochemical, photochemical) of CB [n] guest complexes have made CB [n] popular components to create functional systems including molecular machines, chemical sensors, affinity capture phases, and gas purification materials. 7Accordingly, workers in the CB[n] field envisioned that CB[n] compoundsparticularly CB [7] with its good water solubilitymight be good substitutes for cyclodextrins in pharmaceutical applications.In recent years, the macrocyclic CB [n] compounds were demonstrated to have low in vitro and in vivo toxicity 8 and have been used to solubilize, protect, activate, and deliver pharmaceutical agents. 9CB [8]  has been used as a glue to create polymer hydrogels and nanocapsules based on ternary complex formation. 10Derivatives of macrocyclic CB [n] have even been used for targeted drug delivery applications. 11ver the years, both the Isaacs and Sindelar groups have worked toward an improved understanding of the mechanism of CB[n] formationespecially with regard to the intermediate methylene bridged glycoluril dimers and higher oligomers 12 and the preparation of new members of the cucurbit[n]uril family.These studies have resulted in the preparation of glycoluril dimer based molecular clips, 13 CB[n] analogues, 14 inverted CB[n], 15 nor-seco-CB[n], 16 bambus[n]urils, 17 CB[n] derivatives, 11a,12e,18 CB[n] dimers, 19 and acyclic CB[n]-type receptors.12d,g,j,20 Recently, the Isaacs group used acyclic CB[n] containers as solubilizing excipients 21 for insoluble drugs and reported the influence of the nature of the solubilizing groups 22 and aromatic sidewalls 23 on their function.In this paper we explore the influence of the length of the central glycoluril oligomer on their ability of acyclic CB[n]-type receptors 1-4 to solubilize four insoluble drugs (Fig. 2).

Results and discussion
This results and discussion section is organized as follows.First, we present the synthesis of six new containers (1-3), studies of their inherent solubility, and studies of their tendency toward self-association.Next, we present the X-ray crystal structure of 3b and compare it to the X-ray crystal structures of previously reported glycoluril monomer, dimer, and tetramer derived containers.Next, we report the ability of 1-3 to act as solubilizing agents for four insoluble drugs.Finally, we compare the trends in the solubilization data for 1-3 with that of previously reported 4a and 4b.

Synthesis of Containers 1-3
Scheme 1 shows the synthesis of new compounds 1-3.For the preparation of molecular clips 1a and 1b we followed the path blazed by Nolte and co-workers and allowed glycoluril bis(cyclic ether) 1CE to react with the corresponding dialkoxyaromatic walls (5 and 6) at 70 °C in TFA-Ac 2 O to deliver molecular clips 1a and 1b in high yield. 21,24To prepare molecular clips 2a and 2b we first transformed the known glycoluril dimer 2NH 25 into bis(cyclic ether) 2CE by treatment with paraformaldehyde in TFA at reflux (40%).Subsequently, we reacted 2CE with 5 or 6 in TFA/Ac 2 O at 70 °C to yield molecular clips 2a and 2b in good yield.For the preparation of acyclic CB[n] compounds 3a and 3b we first needed to prepare the glycoluril trimer 3CE.The preparation of 3CE was accomplished by reaction of glycoluril (7) with 1CE in methanesulfonic acid with careful temperature control (10 °C for 2 h and then 23 °C for 2 h).Subsequently, we allowed 3CE to react with 5 or 6 in TFA/ Ac 2 O at 70 °C to give 3a (48%) and 3b (59%) in good yields.
We were fortunate to obtain the X-ray crystal structure of acyclic CB[n]-type receptor 3b (Fig. 3).Fig. 3 also shows the X-ray crystal structure of previously prepared molecular clips 1ester and 2ester (Fig. 4) and acyclic CB[n]-type receptor 4b to illustrate the geometrical change that occurs across the homologous series of receptors 1-4 upon elongation of the glycoluril oligomer backbone.As can be readily seen, as one increases the number of glycoluril rings the receptors change from molecular clips with divergent to nearly parallel aromatic sidewalls (1 and 2) to containers with a well defined hydrophobic cavity despite their acyclic nature (3 and 4).Overall, as a result of the glycoluril trimer backbone, 3b is C-shaped and the two naphthalene rings help to define a cavity that is occupied by a molecule of acetone in the crystal.The mean planes of the aromatic rings of 3b are oriented at an average angle of  feature aromatic rings that are close to parallel to one another (18.3°interplanarangle).In contrast, the X-ray crystal structure of 4b reported previously shows an 113.3°angle between the naphthalene rings and direct CH-π interactions which helps to define a hydrophobic box.
The packing of 3b in the crystal is intriguing (Fig. 5).Analogous to what is observed for molecular clips based on glycoluril or glycoluril dimers, 28 the individual molecules of 3b interact with one another via CH-π and π-π interactions between their naphthalene sidewalls in a head-to-tail fashion to yield linear tape like assemblies.The distance between the   mean planes of the naphthalene ring averages 3.66 Å (range 3.60 to 3.78 Å).These tapes extend along both the b-axis and the c-axis; they are alternately arranged in a criss-cross fashion reminiscent of building method of a log cabin home as one extends along the a-axis.The channels defined by the packing of the linear assemblies are filled with the solubilizing arms and solvating water molecules.

Inherent solubility of 1-4
Given our interest in assessing the performance of 1-3 as solubilizing excipients for insoluble drugs we first sought to measure their inherent solubility.For this purpose, we stir an excess of container with 20 mM phosphate buffered D 2 O and readjust the pD to 7.4 until equilibrium is reached and then remove excess insoluble container by centrifugation and filtration.An aliquot of the filtrate is diluted by a known factor and combined with a solution of benzene-1,3,5-tricarboxylic acid as internal standard (1.00 mM); the relative integrals of diagnostic resonances in the 1 H NMR spectrum for the container (1a: 6.87 ppm, 1b: 7.56 ppm, 2a: 6.99 ppm, 2b: 7.84 ppm, 3a: 7.02 ppm, 3b: 7.95 ppm) versus those of the internal standard (8.38 ppm) were used to determine the inherent solubility of each container.The results are summarized in Table 1.As can be readily seen, the solubility of all the containers bearing dialkoxy-o-xylylene sidewalls is higher than 100 mM which is advantageous for their use as solubilizing excipients.In contrast, the solubility of containers with naphthalene sidewalls are variable.Containers 2b and 3b display high solubility (>300 mM) whereas 1b and 4b are modest (38 and 14 mM, respectively).

Containers 1-3 do not self-associate strongly
In order for molecular containers to have high potential for use as solubilizing excipients they must not be strongly selfassociated which would reduce their ability to form container drug complexes.Accordingly, we studied the self-association of 1-3 by preparing solutions of different concentrations of 1-3 and monitoring the changes in the 1 H NMR chemical shifts.
The chemical shifts of the protons on the aromatic sidewall were particularly sensitive to self-association.Fig. 6 shows the changes in chemical shift of H a , H b , and H c as a function of [2a].The changes in chemical shift were fitted to the standard two-fold self-association model 21,29 implemented within Scientist 3.0™ (ESI †) which yielded a self-association constant (K s ) of 12 M −1 .Table 1 gives the self-association constants measured for 1-3 (ESI †) and also presents the known values for 4 from the literature 21 for comparison.As expected, the K s values measured for acyclic CB[n]-type receptors 3a and 3b are quite low (<50 M −1 ) which is advantageous for their use as solubilizing agents.Given the well-known propensity for glycoluril and glycoluril dimer derived molecular clips to undergo self-association, 13a,26,28a,d,30 we were surprised that the K s values for 1 and 2 were low (<100 M −1 ).For purposes of comparison, the structures and self-association constants measured previously for compounds 1acid (K s = 1840 M −1 ) 26 and 2acid (K s = 41 700 M −1 ) 13a which differ from 1a and 2a in the spatial orientation of their solubilizing groups are shown in Fig. 4. Because the CO 2 − solubilizing groups of 1acid and 2acid are on their convex face, unfavorable carboxylatecarboxylate electrostatic interactions are avoided upon formation of dimers 1acid 2 and 2acid 2 .In contrast, the SO 3 − groups of 1-4 are directed toward each other within the putative dimeric species 1 2 -4 2 which results in unfavorable electrostatic sulfonate-sulfonate interactions which decreases the propensity of 1-4 to dimerize.The self-association constant of   1a (2a) is 61-fold (3500-fold) lower than that measured for 1acid (2acid) which amounts to a destabilization of the dimer by 2.4 (4.8) kcal mol −1 which we attribute to electrostatics.The recognition of the importance of electrostatic interactions that discourage self-association of 1-4 provides a rationale for the superior binding constants observed previously for 4a and 4c relative to 4d 12g,j which lacks the anionic solubilizing groups which translates into the superior solubilizing abilities of 4a and 4b. 23nstruction of phase solubility diagrams for 1-3 with insoluble drugs In order to assess the ability of 1-3 as solubilizing agents for insoluble drugs we created phase solubility diagrams (PSD) 29,31 for the six new containers 1-3.Phase solubility diagrams are plots of concentration of container on the x-axis versus concentration of solubilized poorly soluble drug on the y-axis.Several types of PSDs are possible (A L , A P , A N ; Fig. 7) although linear (A L -type) PSDs are most common for cyclodextrin molecular containers.Containers that display A L -type PSDs behave according to eqn (1) where S 0 is the inherent solubility of the drug, K a is the binding constant for the container drug complex, and slope is the slope of the PSD. 29In this paper, we generally observed A L -type PSDs except for 2a and 2b with ziprasidone and 2a with PBS-1086 which displayed A P -type behaviour.We consider PSDs with slopes ≥0.5 (e.g. a 50 mM solution of container solubilizes 25 mM drug) as indication that a given container is a very good solubilizing agent for a given drug.If we substitute slope = 0.5 into eqn (1) then it is easy to show that K a × S 0 = 1. 23Alternatively, if we want to solubilize a drug with inherent solubility of 1 × 10 −5 M (1 × 10 −6 M) then the container must display a binding constant of 1 × 10 5 M −1 (1 × 10 6 M −1 ) to achieve a slope of 0.5.
In this section we report the phase solubility behavior of 1-3 with four insoluble drugs (camptothecin, PBS-1086, β-estradiol, and ziprasidone) which we have previously studied with 4a and 4b. 21,23To create the PSDs we prepare a solution of container of known concentration and stir it with an excess of solid insoluble drug.After equilibrium is reached, the excess of insoluble drug is removed by centrifugation and filtration.
The concentration of drug in the filtrate is determined by 1 H NMR spectroscopy after addition of benzene-1,3,5-tricarb-oxylic acid (1.00 mM) as a non-binding internal standard using the relative integrals of diagnostic resonances of the drug versus those of the internal standard.Fig. 8a shows the PSDs for solutions of containers 1-4 with the insoluble drug β-estradiol.As can be readily seen most of the PSDs are of the A L -type and it is therefore appropriate to perform a linear fitting of the data to determine the value of the slope.The slope values range from 0 to 0.92 for 4b which indicates that the affinity of the various containers toward β-estradiol differ significantly.Next, we substituted the measured slope values and the known inherent solubilities (S 0 ) 23 into eqn (1) to determine the affinity constant (K a , M −1 ) for the interaction of each container β-estradiol complex (Table 2).In Table 2, the values of slope and K a and their uncertainty are given to two significant figures which reflects the limitations of phase solubility measurements.Similar experiments were performed for camptothecin with containers 1-3 and compared with the data previously reported for 4 (Fig. 8b,c and Table 2). 21,23Fig. 8c shows that the PSD for 4a and camptothecin displays a plateau region above [4a] = 10 mM which reflects the limited solubility  of the 4a camptothecin complex.The data points shown in red (Fig. 8c) are excluded from the linear fitting used to calculate slope and K a .Using the same procedures, we generated PSDs for systems comprising containers 1-3 with ziprasidone and the developmental anti-cancer agent PBS-1086 32 and calculated K a values (Table 1 and ESI †) for the container drug complexes.

Interpretation of the trends in the K a values and solubilization efficiency
This section discusses the trends in the K a values as a function of the structural variables of the containers and the guests.
Influence of the number of glycoluril rings.As shown in Fig. 3, the crystal structures of 1-4 demonstrate that compounds 1 and 2 are molecular clips with divergent and slightly convergent aromatic sidewalls, respectively.In contrast, containers 3 and 4 are more properly defined as acyclic CB[n]-type receptors because they feature more fully formed ureidyl CvO portals and closed cavities that are fully defined by the glycoluril rings and the aromatic sidewalls.Accordingly, it is appropriate to compare the solubilization properties of 1 and 2 and separately, 3 and 4. We find that container 1a is uniformly unsuccessful as a solubilizing agent, whereas container 2a is able to solubilize camptothecin, ziprasidone, and PBS-1086.We attribute the better solubilizing ability of 2a to the more nearly parallel alignment of its aromatic walls (centroid to centroid distance = 7.40 Å) which preorganizes it to engage in π-π interactions with drugs that contain aromatic rings.In contrast, the behavior of containers 1b and 2b toward the four drugs is drug dependent.For example, container 1b performs better for camptothecin and β-estradiol but 2b performs better for ziprasidone and PBS-1086.In this regard it is noteworthy that the PSDs for 2a and 2b with ziprasidone are non-linear which is indicative of higher order container n •drug complexes.
The behavior of acyclic CB[n]-type containers 3 and 4 toward the four drugs exhibit clear trends.For example, container 3a binds less strongly (K a ≈ 10 3 − 10 4 M −1 ) to all four drugs than 4a does (2.9 × 10 3 < K a < 5.5 × 10 5 M −1 ).We attribute this result to the larger hydrophobic cavity of 4a relative to 3a and the more fully formed ureidyl CvO portals of 4a which results in stronger ion-dipole interactions with cationic drugs (e.g.ziprasidone).Similarly, container 3b (2.0 × 10 3 < K a < 3.6 × 10 4 M −1 does not perform as well as 4b (1.2 × 10 4 < K a < 1.9 × 10 6 M −1 ) as a solubilizing agent for these four drugs.Accordingly, we conclude that container 4 is a more efficient solubilizing agent than container 3 although the relative in vivo efficacy of drugs formulated by the different containers will depend on the ability of the drugs to be released by dilution and competition from endogenous cationic small molecules (e.g.spermine). 33nfluence of aromatic sidewall.The influence of the nature of the aromatic sidewall (e.g.benzene versus naphthalene) is clear cut for containers 1, 3, 4. In most cases the containers with the longer aromatic sidewall (e.g.naphthalene) display higher K a values toward the drugs.For example, container 1b binds nicely (K a = 2.2 × 10 4 and 2.6 × 10 3 M −1 ) toward camptothecin and β-estradiol whereas the binding of 1a to these drugs could not be detected.Similarly, the ratio of the K a values of 3b versus 3a toward camptothecin (2.3), ziprasidone (8.2), PBS-1086 (4.9), and β-estradiol (9.3) indicate that 3b is a significantly better host than 3a.Containers 4b and 4a display a similar trend with the ratio of K a values for 4b versus 4a toward camptothecin (>6.4),PBS-1086 (3.5), and β-estradiol (21).Interestingly, ziprasidone which is both narrow and cationic at pD 7.4 prefers to bind to the smaller cavity of container 4a relative to 4b.We conclude that the containers with substituted naphthalene sidewalls generally outperform those with o-xylylene sidewalls because they possess larger hydrophobic cavities which should increase the number of solvating a NL = non-linear PSD; -= could not be determined because the slope is too small to measure.TL = too large to be determined from PSD.
b Lit.data. 21,23 2 O molecules expelled upon binding 6d,34 and result in the formation of π-π interactions between larger π-surfaces particularly for insoluble drugs that contain aromatic rings.

Conclusions
In summary, we have reported the synthesis of a series of molecular container compounds (1-3) that differ in the number of glycoluril rings that connect the two terminal substituted o-xylylene or naphthalene sidewalls.Compounds 1-3 display very good aqueous solubility (>38 mM) and somewhat surprisingly do not undergo significant self-association in aqueous solution (K s ≤ 92 M −1 ).We trace the low values of K s observed for 1 and 2 to unfavorable sulfonate-sulfonate electrostatic interactions (2.4 to 4.8 kcal mol −1 ) that would occur in the putative dimeric complexes 1 2 and 2 2 .The X-ray crystal structure of acyclic CB[n] type receptor 3b shows a cavity shaped by its two naphthalene walls that do not engage in π-π or CH⋯π interactions.PSDs were created for 1-3 with four insoluble drugs (camptothecin, ziprasidone, PBS-1086, and β-estradiol) and compared with the PSDs measured previously for 4a and 4b.We find that the acyclic CB[n]-type containers containing larger numbers of glycoluril rings (e.g. 4) and the larger naphthalene sidewalls (e.g.4a versus 4b) are generally superior solubilizing agents for insoluble drugs.For the molecular clip receptors 1 and 2 the results are less clear cut although the naphthalene walled compounds often display higher K a values than the o-xylylene walled analogues and more often exhibit the desirable A L -type PSDs.When combined with our previous studies on the influence of the nature of the solubilizing group (e.g.anionic versus neutral versus cationic) and the aromatic sidewalls (benzene, two isomeric naphthalenes, alkylated variants), this study allows us to conclude that compounds 4a and 4b are most efficient as solubilizing agents for insoluble drugs and are best positioned for further development as solubilizing excipients for real world pharmaceutical applications.In this regard the higher values of K a exhibited by acyclic CB[n]-type receptors compared to HP-β-cyclodextrin promises to broaden the range of drugs that can be efficiently solubilized using molecular container technology and to do so at lower concentrations of container.

Fig. 2
Fig. 2 Structures of insoluble drugs used in this study.

Fig. 6
Fig. 6 Plot of chemical shift versus [2a].The solid line represents the best global fit of the data to a two-fold self-association model with K s = 12 M −1 .Conditions: 20 mM sodium phosphate buffered D 2 O, pD 7.4, room temperature.Key: H a , ■; H b , •; H c , ▲.

Fig. 5
Fig. 5 Illustration of the packing of molecules of 3b into linear assemblies along the b-axis.Color code: C, gray; H, white; N, blue; O, red; S, yellow.

Table 1
Inherent solubility values of 1-4 and their self-association constant K s determined in 20 mM sodium phosphate buffer, pD 7.4 21Values from the literature.21