by Liping Zhou, Suzanne Tilton, Linhong Yang and Jianling Wang
Solubility is an important physico-chemical parameter in drug discovery as well as development since it affects both in vitro and in vivo assay results. With the trend towards lipophilic compounds, finding an optimal strategy to assess solubility earlier as well as at different stages of drug discovery, including hit finding, lead finding and optimisation, as well as nomination of development candidates, is a challenge for today’s pharmaceutical scientists. Here we examine the advantages, limitations and use of different tools developed for solubility-driven successful drug discovery.
Solubility, widely accepted by US Food & Drug Administration (FDA) and in pharmaceutical development for drug evaluation, is defined as the quantity of one substance in another when they are mixed together and allowed to reach equilibrium. Solubility of a drug substance is a thermodynamic parameter, which depends not only on the characteristics of the drug substance, but also the assay medium and the environment under which the two are mixed. Due to its direct impact on successful drug discovery, different tools for estimating solubility have been developed to cope with the demands at different stages in research.
The accuracy of solubility data and their relevance to pharmacokinetic (PK)/pharmacodynamic (PD) events rely greatly on the proper assessment and utility of the results. This is an evolving process and requires continuous education of end users in drug discovery and development projects. For instance, it is not rare to question the necessity of solubility data in discovery, as the drug solubility varies depending on formulation and solid phase characters, both of which are addressed in development. Another misconception is that solubility issues may be deferred until later stages for a “formulation fix”. These wrong beliefs can be dangerous as the formulation approach, albeit compromising solubility burdens in some events, is quite costly and does not always lead to success. In reality, insufficient solubility is one of the leading causes for development failures. This should not be surprising as statistics reveal that the compound collections synthesised in discovery phases are shifting towards higher lipophilicity/permeability and lower solubility, these being typical Class II compounds in the well known Biopharmaceutical Classification System (BCS).
It is worthwhile emphasising that addressing solubility during the discovery process is particularly critical as insufficient solubility can cause artifacts in biological, PK, PD and toxicological assays, masking liabilities which will become unpleasant surprises for downstream development teams. In fact most of these assays start with DMSO stock solutions and retain a low percentage of DMSO in the final assay medium. For example, while thermodynamic solubility is widely utilised to project in vivo dissolution of drug substance (drug exposure), kinetic solubility is recommended for qualifying in vitro assays and related data interpretation. Therefore, instead of arguing about its necessity, one should select the right solubility tools to tackle PK/PD and pharmacological issues at the right drug discovery stage.
IN VITRO SOLUBILITY DETERMINATION
Thermodynamic solubility
Solubility is widely utilised in the development of a drug substance, especially for determining suitable formulations for dosing in animal studies. The generally established approach accepted by the FDA is the saturation shake-flask or titration method, referred to as the “gold standard”. Using the saturation shake-flask approach, the test compound is added to a standard buffer solution and shaken until the equilibration between the solute and solve nt is reached (typically within 12 to 72 hours). The two phases are separated by either filtration or centrifugation, after which the concentration of the substance in the liquid phase is quantified using HPLC with UV, mass spectrometry or nitrogen detection.
Intrinsic solubility and solubility pH profile
The potentiometric titration approach can be applied to ionisable compounds to assess the solubility of a drug as the neutral species. The data may be valuable in accounting for the net impact of drug properties preferentially involving its neutral species, such as membrane permeability. The titration curve of a drug’s ionisation constant(s) shifts due to test compound precipitation from solution. The curve shift is used to calculate the intrinsic solubility of the test compound, and the method also offers a solubility pH profile based on the compound’s ionisation and precipitation behaviours.
In comparison to the conventional shake-flask approach, the titration method is advantageous as it eliminates the separation procedure of the filtrate from the precipitate. It should be noted that the solubility products (e.g. salt and ionised species) extrapolated by the current approach at the same pH may significantly deviate from those derived by conventional shake-flask gold standard, in particular for low-soluble compounds. Both approaches are quite labour-intensive and compound-consuming processes and thus are not practical for addressing broad solubility issues in the discovery phase where assays are required with high-throughput (HT), fast turn-around time, minimal sample consumption and, most importantly, acceptable quality and predictive values to target solubility issues.
HT Equilibrium solubility
The equilibrium solubility may be determined in the drug discovery phase with a higher throughput. To overcome the challenging solid dispensing step required for the large volume of compound submissions at this stage, the solubility can be measured from solid samples reconstituted from compound DMSO solutions. To minimise DMSO-induced solubility over-estimation (up to 6 fold with 0.5-5% DMSO), the effective removal of DMSO from the reconstituted samples is essential to the quality of the data. The assay utilises a miniaturised shake-flask approach and streamlined HPLC analysis, providing an excellent correlation with the thermodynamic solubility results from the gold standard method. It should be noted that since the solid characterisation is not preserved as a result of sample reconstitution, the data are comparable to, but not necessarily always identical to those from thermodynamic solubility assays. Nonetheless, this may not be a major concern in the discovery phase as most discovery compounds are amorphous. As a whole, it seems that the HT-equilibrium solubility assay offers an optimal balance between quality and speed in drug discovery.
HT-kinetic solubility
Kinetic solubility features the direct introduction of a compound DMSO solution (0.5 - 5%) into the designated media, followed by an immediate readout (with incubation times of approximately 15-90 minutes) for the concentration at which the test compound is saturated or precipitated. In addition to HPLC, the readout in the plate reader format (e.g. turbidimetry, nephelometry, laser flow cytometry or direct-UV) offers a higher throughput and is widely adopted by pharmaceutical companies. The HT kinetic solubility has the benefits of fast turn-around time, cost-effectiveness and relatively universal detection modes. It also shows some limitations, including the interference of low soluble impurities with the readout, unless an analytical separation technique such as HPLC is involved.
IN SILICO SOLUBILITY ESTIMATION
Several in silico approaches have been developed to accommodate the needs of early discovery and negotiate the capacity of in vitro solubility assays, including models based on octanol-water partition coefficient (LogP), molecule fragments, solvation properties or a combination of these. The association between compound solubility and lipophilicity was realised some time ago, making solubility approximation by LogP the earliest practice. The melting point is added to the original linear relationship to account for the contributions by packing energy at different polymorphs. In fragment-based simulation, the solubility is a sum of the contributions from each fragment in the molecule; therefore it depends on the existence of a large and accurate set of experimental solubility values. The molecule is assumed to be rigid without interference from neighbouring substructures. To overcome the lack of consideration of intra-molecule interactions and the solute-solvent interactions in the fragment based approach, a solvation model is utilised. Here, physical properties of the molecule such as polarity, molar refraction, hydrogen bond acidity/basicity and molecular volume are integrated to enrich the model. However, obtaining these values is itself challenging.
While the speed and low cost of computational approaches are advantageous, in silico solubility models are limited as many of the descriptors used are lacking or less accurately computed during the early discovery process. In addition, as many models are built on commercially available resources, the predicting power can only be improved with larger and more discovery-like compound datasets. Thus far, there is no completely satisfactory model available and successive experimental solubility determination is still necessary.
SOLUBILITY-DRIVEN DRUG DISCOVERY STRATEGY
A general strategy to employ solubility assessment into drug discovery is presented in figure 1, with a balance of speed, cost, quality and predictive value. Typically, in silico models are used during hit validation and lead finding, whereas HT kinetic or equilibrium solubility assays are more popular in lead finding, lead prioritisation and lead optimisation processes. Key compounds are always checked with the gold standard methods during the phase transition period before being handed over to development.
Kinetic solubility assays were the first to become available in HT mode and have become a useful tool to evaluate other in vitro assays. Meanwhile, high to medium throughput equilibrium measurements are most applicable for solubility optimisation. Enhanced sample miniaturisation, throughput and dynamic range of in vitro assays should translate into more accurate in silico models as larger discovery
compound training sets are gathered.
While chemists may be successful in establishing structure property relationships with equilibrium solubility, the challenge to understand the impact of solubility on in vivo exposure remains. In this case, it is necessary to again use solubility data collectively with other in vitro profiling information such as permeability, clearance, as well as dose and formulation. Many in vivo studies are conducted with solution formulations. However, to assess the impact of solubility on exposure, a recent trend is to concomitantly evaluate PK data from suspension formulation. In this case, compounds with poor absorption as a result of precipitation under gastrointestinal conditions may be better differentiated.
In addition, due to their pharmacological relevance, gastric and intestinal fluid media are more frequently used in solubility determination to predict the in vivo exposure. The historically limited utility of these media is attributed to their complexity and lack of uniformity across the industry. However, the trend is towards uniform components of such media and their use at the discovery stage to generate more meaningful data and better in vitro-in vivo correlation, which may be developed with improved dosing strategies and the use of physiological media.
Overall, the best practice for understanding the impact of solubility in drug discovery requires the collective utilisation of in vitro data under appropriate pH, equilibrium or kinetic and physiological conditions with other profiling data.
REFERENCES
1. Yalkowsky SH. Solubility and solubilization in aqueous media. American Chemical society: Washington, D.C. and Oxford: New York, Oxford. 1999.
2. Avdeef A. Absorption and Drug Development, Wiley-Interscience: Hoboken, New Jersey. 2003.
3. Zhou L, Yang L, Tilton S, Wang J. Development of a High Throughput Equilibrium Solubility Assay Using Miniaturized Shake-Flask Method in Early Drug Discovery. J Pharma Sci 2007; 96: 3052-3071.
4. Faller B, Ertl P. Computational Approaches to Determine Drug Solubility. Adv Drug Delivery Rev 2007; 59: 533-545.
THE AUTHORS
Liping Zhou, Suzanne Tilton, Linhong Yang and Jianling Wang,
Novartis Institutes for BioMedical Research
Cambridge, MA, USA
