Liquid Metastable Precursors of Ibuprofen as Aqueous Nucleation Intermediates

Abstract The nucleation mechanism of crystals of small organic molecules, postulated based on computer simulations, still lacks experimental evidence. In this study we designed an experimental approach to monitor the early stages of the crystallization of ibuprofen as a model system for small organic molecules. Ibuprofen undergoes liquid–liquid phase separation prior to nucleation. The binodal and spinodal limits of the corresponding liquid–liquid miscibility gap were analyzed and confirmed. An increase in viscosity sustains the kinetic stability of the dense liquid intermediate. Since the distances between ibuprofen molecules within the dense liquid phase are similar to those in the crystal forms, this dense liquid phase is identified as a precursor phase in the nucleation of ibuprofen, in which densification is followed by generation of structural order. This discovery may make it possible to enrich poorly soluble pharmaceuticals beyond classical solubility limitations in aqueous environments.


Introduction
Crystallization, anatural phenomenon observed in everyday life,isc rucial to many processes occurring in nature and in the chemical, pharmaceutical, and food industries.T he majority of agrochemical and pharmaceutical products undergo many crystallization steps during their development and manufacture,w here corresponding processes serve as versatile techniques in separation, purification, and product design. [1,2] More than 90 %o fa ll active pharmaceutical ingredients (APIs) are small organic molecules in the crystalline state. [2] Notably,t he selection of suitable polymorphs of drug components plays ak ey role in formulation and manufacturing design, and thus it is crucial for achieving the desired solubility and stability properties. [3] Since solution crystallization starts with nucleation, early events of nucleation play ad ecisive role in the generation of the crystal structure and size distribution of generated particles. [4] Hence, understanding the fundamentals of nucleation and precursor phases to the final crystal is vital to control the properties of the final product.
Nevertheless,t he mechanistic understanding of phase separation and the formation of solid particles or liquid intermediates of the final crystalline systems is rather limited, especially for small organic molecules.D ue to the analytical simplicity of classical nucleation theory (CNT), researchers have applied it extensively to solution crystallization. CNT considers that clusters of critical size form stochastically in supersaturated solutions due to the reversible addition of single molecules to unstable precritical clusters.O nce they reach the critical size,t heir growth is thermodynamically favorable and as aresult of this,crystal growth proceeds up to the final crystal. [5,6] Thegeneral assumption in CNT (capillary assumption) is that the properties of the nucleus can be represented by those of the bulk, for example,m acroscopic surface tension.
On the other hand, mechanisms of nonclassical nucleation were explored in simulations by ten Wolde and Frenkel. [7] Further experiments showed evidence for atwo-step process for protein crystal nucleation in which the separation of ad ense,l iquid phase is followed by formation of crystalline order inside the liquid precursor. [8][9][10][11] Computer simulations [12,13] and experimental studies of protein nucleation in solution [14,15] suggested that nucleation involves at least two stages.However,recent work also challenged the applicability of the model of two-step nucleation to the field of protein crystallization. [16] Thep renucleation cluster pathway pioneered by Gebauer et al. [17] for CaCO 3 involves the formation of prenucleation clusters in solution, decrease in their dynamics and densification as the key step for phase separation, formation of al iquid phase,s olidification, and finally crystallization. [18] While both two-step nucleation and the prenucleation cluster pathway consider liquid intermediates,the former mechanism relies on the formation of critical nuclei within the liquid intermediate for crystallization, while the latter is based on the aggregation and dehydration of larger entities that cannot be accounted for in such classical notions of nucleation. [19,20] However,f or small organic molecules the experimental evidence for metastable liquid phases within at wo-step nucleation or aprenucleation cluster pathway is rather poor. Supersaturated glycine solutions underwent laser-induced nucleation at rates much faster than those of control solutions. [21] This was explained by the electric-field-induced alignment of the molecules in existing prenucleation clusters in solution. Thee ntropic barrier for the formation of an ordered lattice is supposed to be reduced in this way. Nakamura et al. [22] described the formation of ad ense liquid phase mediated by surface interactions with carbon nanohorns.I nasecond step,h eterogeneous nucleation was observed as the precursor phase was attached to the nanostructured surface.A nother study reported by Rybtchinski et al. [23] corroborated at wo-step nucleation pathway for perylene diimides as simple aromatic compounds.H ere,t he final crystallinity was observed by cryo-TEM (cryogenic transmission electron microscope) imaging to gradually develop in amorphous precursors rather than at the nucleation stage.M olecular dynamics computer simulations at increasing levels of supersaturation showed that liquid-liquid phase separation occurs before nucleation, independent of the present solvent. [24] Furthermore,l inks between the selfassembly of organic molecules in solution and crystal structure after nucleation were found to depend on the specific solvent used for the nucleation of the organic crystal. [25,26] Thepoorly water-soluble compound ibuprofen is aprominent model system and one of the most frequently prescribed essential analgesics in the world according to WHO. [27] Commercially it is employed as ar acemic crystal while the API is the S-enantiomer. [28] Thefinal crystal structure formed from aqueous solutions differs depending on whether the racemate or solely the S-enantiomer is employed. [29,30] Due to the acid functionality of the molecule,the solubility depends strongly on the protonation state,and its supersaturation can be adjusted via the pH value and calculated based on the law of mass action.
Here,w ea im to elucidate the underlying nucleation pathway of ibuprofen with emphasis on al iquid precursor phase in aqueous solution. We monitored the early stages of nucleation with ap otentiometric titration assay taking advantage of the protolysis equilibrium of ibuprofen in combination with turbidimetry.F urthermore,w ec haracterized the physicochemical properties of this precursor phase such as its molecular dynamics and viscosity.B yq uantifying the amount of ibuprofen bound within the dense liquid phase via 1 HNMR spectroscopy,w el ocated the binodal as well as the spinodal limits of the corresponding liquid-liquid miscibility gap at room temperature.Furthermore,wemeasured the interproton distances in the dense liquid phase by 2D 1 H-1 HN OESY NMR experiments and compared them with interproton distances in the final crystal. This reveals that the liquid intermediates are dynamic species with intermolecular distances similar to those in crystals,where the concentration of the poorly soluble compound is significantly increased.
This inherent property might provide an ew means to improve the bioavailability of poorly water-soluble compounds when liquid intermediates can be stabilized in pharmaceutical formulations.

Potentiometric Titration Experiments
Thes upersaturation of protonated ibuprofen (IbuH) was increased by adding dilute hydrochloric acid solution to an ibuprofen sodium solution and decreasing the pH (Scheme 1).
Another dosing unit was used to titrate ibuprofen sodium solution in order to maintain aconstant overall concentration of ibuprofen throughout the whole experiment. In Figure 1 (left) it can be seen that after ac ertain point the turbidity increased steadily and reproducibly while HCl was added to the solution. Stopping the addition of HCl (Figure 1, right) after the observed increase in turbidity resulted in aconstant level of turbidity (blue line). After acertain stirring time,the pH rose and crystals formed, which were characterized by Xray diffraction ( Figure S1, Supporting Information). Thetime point of nucleation could be determined as as light decrease of turbidity (red arrow), caused by the incorporation of ibuprofen molecules into the forming crystals and-related to this phenomenon-due to the associated withdrawal of protons from solution (Scheme 2). After crystallization and ashort period of equilibration, the pH value rose to aconstant value which reflects the solubility limit of crystalline ibuprofen in aqueous solution at this pH value.

Determination of the Locus of the Liquid-Liquid Binodal Limit
Fort he determination of the locus of the liquid-liquid binodal limit, samples were drawn from titrations performed with the double-dosing method in 5vol. %D 2 Oand 1 HNMR spectra were recorded subsequently.W hile signals of the ibuprofen molecule were present in samples taken before phase separation, additional peaks were present in those samples,inw hich liquid-liquid phase separation had already occurred. Interestingly,a ll of the recorded signals of ibuprofen exhibit secondary signals due to the separated phase, shifted by about 0.3 ppm (for aliphatic protons) to about 0.5 ppm (for aromatic protons) ( Figure 2). Considering the observed turbidity increase,w es uppose that the chemical environment of ibuprofen in the dense phase is different and hence results in the difference in chemical shift. Also,t hese signals show alower intensity because of the relatively small amount of ibuprofen present in the dense liquid phase compared to the overall amount of ibuprofen in solution. During the course of titration in the regime of liquid-liquid phase separation, samples were drawn at specific pH values. These samples were investigated in terms of their signal integrals for peaks (a) and (a*) arising from molecules in the mother phase and dense liquid phase,respectively,while their sum always equals the total invariant ibuprofen concentration of 3mm.The decrease in volume of the mother phase is likely minor and negligible for calculating the concentration of IbuH in this phase ( Figure 3). With decreasing pH, the amount of ibuprofen present in the dense liquid phase (n*(Ibu)/V total )i ncreases with ac oncurrent decrease of the amount of ibuprofen in the mother phase (n(IbuH) L1 /V total ), whereby the total amount of ibuprofen remains constant; thus,t he amount in both phases correlates linearly.E xtrapolation of the corresponding linear fit yields the lowest concentration at which liquid-liquid phase separation can occur.
Hence,t he y-offset of the linear fit corresponds to the binodal limit of liquid-liquid miscibility gap.O nt he other hand, asaturation threshold can be identified at much higher supersaturations.T his implies that the composition of the dense liquid phase does not change upon further increase in the concentration of IbuH, and thus the saturation threshold corresponds to the locus of the liquid-liquid spinodal limit.

Translational and Rotational Motion in the Dense Liquid Phase
In order to study the molecular dynamics of ibuprofen in the dense liquid phase,t he pure S-enantiomer of ibuprofen was employed since it exhibits ahigher solubility than racemic ibuprofen. [31] Forsamples drawn from titration, 1 HPFG-STE diffusion NMR was employed as an on-invasive,h ighly precise technique for the determination of diffusion coefficients. [32,33] Thediffusion coefficient of ibuprofen molecules in both phases was determined according to Equation (6) (see the Supporting Information) and linear fitting of the corre-  sponding data ( Figure S2 in the Supporting Information). This procedure yielded the same diffusion coefficient D for ibuprofen in the mother phase for several different samples drawn in the binodal regime (Figure 4, left). However,t he diffusion coefficient of the ibuprofen molecules in the dense liquid phase (labeled as D*) decreases with an increase in the amount of ibuprofen in the dense liquid phase.While the ratio of diffusion coefficients (D/D*) is 34 upon crossing the binodal limit, it increases to aratio of 239, advancing further into the binodal regime (Table 1). This shows that with an increasing amount of ibuprofen bound in the dense liquid phase,t he translational diffusion of molecules in this phase becomes significantly slower.M oreover,N MR provides evidence for the liquid character of the separated phase. Further information on molecular dynamics can be obtained from the rotational correlation time of the ibuprofen molecules in the distinct phases by determining the T 1 and T 2 relaxation times associated with the peaks (a) and (a*) in the one-dimensional 1 HNMR spectrum ( Figure S2, Supporting information). According to Carper et al. [34] [Eq. (9), Supporting Information), this yields the rotational correlation times t c (Figure 4, right), which strongly increase by afactor of 5-7 in the dense liquid phase,w hereas t c does not change significantly in the mother phase.
Theo verall results show that the molecular dynamics, both in terms of translation and rotation, are slowed down in the dense liquid phase.W hile D/D*i ncreases by af actor of 34-239, rotational correlation times t c increase only by afactor of 5-7 in the dense liquid phase.
Consequently,rotational and translational motions of the molecules in the dense liquid phase do not sense the same    [35] for solutions of proteins in the presence of solubilized polymers. We assume that the microviscosity experienced by ibuprofen molecules in the dense liquid phase,a sa ssessed by 1 HNMR spectroscopy,differs from the bulk viscosity,and thus does not strictly obey Stokes laws.N evertheless,i ti sc lear that the viscosity in the dense liquid phase is much higher than in the mother phase (D/D*3 4-239), which may explain why the liquid phase is kinetically stabilized for asignificant period of time before it crystallizes.V iscosity was proposed to play ak ey role in the formation of the ordered nucleus within adense liquid intermediate by Vekilov et al. [8] Determination of Intermolecular Distances In order to gain quantitative information about inter-v s. intramolecular distances of ibuprofen molecules in the dense liquid phase,t wo-dimensional homonuclear 1 H-1 HNMR NOESY methodology (nuclear Overhauser effect spectroscopy) was employed. Solely the protons (a), (b), (c) are relevant here,b ecause they cannot come closer due to the rigid aromatic ring structure and hence,they have aintermolecular distance higher than 5 in any conformation. Owing to the general limitations of NOESY methodology,t he maximum separation that leads to cross peaks is about 5 . [36] In samples drawn after liquid-liquid phase separation, no NOE signal can be detected between the protons of the amethyl group (a) and those of the isopropyl group (b,c )f or the ibuprofen molecules present in the mother phase (Figure 5, inset, dotted line mode). Thedistance from (a) to (b) in the molecule itself is at their closest distance % 7 .T his implies that they are too far away from each other to contribute to aN OE cross signal, and as expected, no NOE signal of (a) to (b) could be observed in the 1 HNMR NOESY spectrum ( Figure 5).
Interestingly,t here are NOE signals observed for the proton pairs (a*)-(b*) and (a*)-(c*) in the dense liquid phase, although the intramolecular distance remains about 7 . Hence,t his NOE signal necessarily arises from an intermolecular NOE with protons of other neighboring ibuprofen molecules closer than 5 .I norder to obtain am ore precise value for the intermolecular proton distance,the NOE signals can be used to determine interproton distances with high accuracy. [37] As the NOE signal intensity scales with distance, d À6 ,t he intermolecular proton distances of the ibuprofen molecules in the same phase can be determined according to [Eq. (1)]: Figure 5. Two-dimensional 1 H-1 HN OESY NMR spectrum of S-ibuprofen solution after liquid-liquid phase separation (c tot = 3mm,p H2.31) in the binodal regime. Red letters indicate the assignmentt othe proton signals according to Figure 2, whereas the asterisk (*) labels the corresponding proton signal in the dense liquid phase. Intermolecular NOE signals that are present in the dense liquid phase are indicated by red lines, whereas missing intermolecular NOE signals in the mother phase are indicated by black dotted lines. Signals present in the indirect dimension close to the diagonal signals of (a), (a*), and (b) are due to t 1 noise. The signal labeled with ah ash (#) originates from trace EtOH.  ). Ther esults of this NOE interproton distance calculation are summarized in Table 2.
These as-determined distances can be contrasted with those of the protons in the crystal structures of both S-IbuH and the racemic RS-IbuH (Table 3) [29,30] Since these values correspond to the intermolecular proton-proton distances in the dense liquid phase,i tc an be concluded that the ibuprofen molecules in the dense liquid phase are roughly as close to each other as in the ibuprofen crystal.
However,ithas to be mentioned that within the (a*)-(c*) NOE signal there are also contributions of the (a*)-(c*) intermolecular NOE signals.But as the NOE signal intensity scales with d À6 ,t heir contributions are supposed to be minimal and thus can be neglected. Besides,the NOE signals originating from the close distance to other ibuprofen molecules are the sum of different interproton distances.A s their individual NOE signal intensities again scale with d À6 , closer distances are the major contributors to the total NOE signal. Thus,the calculated values represent fixed interproton distances for ibuprofen when the molecules are organized in clusters with discrete interproton distances.S ince it is not known whether this situation is representative for the real scenario,the data provides at least aqualitative measure of an increased aggregation of ibuprofen molecules in the dense liquid phase when compared to the mother phase.

Conclusion
In summary,n ucleation of ibuprofen in aqueous solution proceeds via liquid-liquid phase separation as aprecursor to crystals.The loci of the binodal (0.43 AE 0.01 mm)and spinodal limits (0.71 AE 0.01 mm)o ft he corresponding liquid-liquid miscibility gap were determined by means of 1 HNMR spectroscopy due to the different chemical environment of the ibuprofen molecules in the two phases.T he evaluation of molecular dynamics suggests that both translational and rotational diffusion of ibuprofen in the dense liquid phase are strongly hindered compared to the ibuprofen molecules in the mother phase.This shows that the ibuprofen molecules in the dense liquid phase,w hich is more concentrated than the mother phase,are in astate that is kinetically stabilized. Since rotational and translational diffusion processes are slowed down significantly,i ta ffects the crystalline structure generation within the dense liquid phase and accounts for its metastability towards nucleation. Thei ntermolecular distances in the dense liquid phase were found to be similar to those in the final crystal according to the measured interproton NOE signals.C onsequently,t he dense liquid phase can be regarded as adensified precursor phase for the nucleation of ac rystalline phase.I nt he case of ibuprofen, these results show that densification is followed by structure generation and therefore implies an onclassical nucleation pathway where the dense liquid phase is kinetically stabilized by the slow rotational and translational diffusion.
To the best of our knowledge this is the first time that molecular dynamics and intermolecular distances in am etastable precursor phase have been characterized directly by NMR spectroscopy.The application of the methods presented in this work to other compounds will further contribute to ab etter understanding of the nucleation process of small organic molecules from solution. Moreover,t he existence of this miscibility gap can be exploited in order to reach drug concentrations far above the solubility limit of the crystalline compounds.T herefore,t hese results have the potential to pave the way towards the fabrication of liquid-phase drug formulations with an enhanced bioavailability.  Table 3: Intramolecular and intermolecular distances of (a)-(b) and (b)-(c) protons in the crystal structures of S-IbuH [30] and racemic IbuH (rac IbuH). [29] (a)-(b) rac IbuH