Combinatorial chemistry is now widely used to synthesize

Transcription

1 High-Throughput Measurement of Drug pk a Values for ADME Screening John Comer, Karl Box* Keywords: ionization constant, drug absorption, ADME Abstract Combinatorial chemistry is now widely used to synthesize large numbers of new chemical entities. New methods have been developed for early-stage pharmacokinetic tests, to help decide which molecules have the properties required to make them efficacious and successful in the marketplace. Of the various physicochemical tests that are used to assess ADME, pk a (aqueous ionization constant) is the key to understanding how other properties will be affected by ph. A new method is described for rapid measurement of pk a of samples drawn from 96 well DMSO stock solutions. Introduction - Why pk a is Important It is widely recognized that new molecules require appropriate ADME (absorption, distribution, metabolism, and excretion) properties if they are to be developed into successful drugs. The ADME of ionizable drugs is greatly influenced by their pk a values, which influence their transport properties (absorption and distribution), and the excretion of metabolites. The pk a values are often measured during a process of physicochemical profiling 1 that occurs during the selection of newly discovered molecules for further development. They are also measured to a high standard of accuracy and GLP during later-stage development and for regulatory purposes. At least 60-70% of drugs undergo acid-base ionization in aqueous solution, 2 which influences their binding characteristics, and has a profound effect on their other physicochemical properties including lipophilicity, permeability, and solubility. During drug discovery, a molecule may be found which has a biological effect related to a particular structural feature. Chemists will then try to enhance the biological effect by modifying the structure, often by substituting ionizable groups. During this process of modification, the structural changes will lead to changes in solubility, lipophilicity, and permeability. Table 1 describes some important physicochemical parameters, the values of which are all pk a -dependent for ionizable drugs. The following case studies further illustrate the importance of pk a values in ADME. Many basic drugs are absorbed in the intestine, where the ph may vary between 5 and 8 depending on which part of the intestine is considered, and whether the patient is in the fed or fasted state (food raises the ph of the stomach, but reduces the ph of the intestine). Itraconazole (marketed as Sporanox) is an antifungal drug that may be taken orally. Patients are advised to take itraconazole with meals, when the ph of the intestine contents may go down to 5 or below. The wall of the intestine is shielded from the intestine contents by a layer of mucus. The ph at the wall itself is around 6, and not greatly affected by the ph of the intestine contents. Itraconazole is a base with a pk a of 3.7, and log P greater than 5. At high ph it is not ionized and is almost insoluble. It ionizes at low ph, and is rather more soluble at ph 5 than at ph 6. At ph 5 some compound can dissolve in the intestine contents. This enables it to approach the wall, where it will be transformed into the neutral form that is more easily absorbed. Sildenafil is a vasodilator, and is marketed under the name Viagra. This molecule is an ampholyte, meaning that it has two pk a s, one acidic and one basic. The acidic pk a is 9.12, and the basic pk a is While this molecule is charged at low and high ph, it is neutral at mid-range ph, and therefore exists in its most lipophilic form (logp = 3.18) at intestinal ph; this enhances its absorbance in the GI tract. MEASURING pk a VALUES The aqueous ionization constant of a molecule is denoted by its pk a value/s, where pk a is equivalent to the ph at which a given ionizable group on the molecule is half-ionized. Traditionally, it has required a lot of skill to measure pk a values. First, methods are required for monitoring the state of ionization of the molecule during conditions of changing ph. Then, methods must be devised to interpret the experimental data, taking full account of the theory of chemical equilibria, and the complications introduced by poor sample solubility and overlapping pk a values. *Correspondence Sirius Analytical Instruments Ltd. Riverside, Forest Row Business Park Forest Row, East Sussex RH18 5DW United Kingdom Phone: john.comer@sirius-analytical.com

2 A classical method for measuring pk a values is potentiometric titration with results calculated from the shape of the titration curve. Another method is to measure UV absorbance at each ph point of the titration curve, and calculate results from the change in UV absorption as a function of ph. 4 Both these methods call for ph to be measured at between different values throughout the titration. A third method is capillary electrophoresis in several different buffer solutions with results calculated from the relative mobilities at different ph. 5 While powerful, none of these methods is particularly fast. Figure 1. Photograph of Sirius ProfilerSGA instrument for rapid measurement of pk a values. Sirius, a world leader in instrumentation for pk a measurement, has recently developed a new automated instrument (ProfilerSGA - Figure 1) for rapid measurement of pk a values of samples supplied in 10mM solution in DMSO in 96-well microtiter plates. ProfilerSGA uses a new technique called ph Gradient Titration, 6, 7 in which two multi-component buffer solutions are mixed together to create a ph gradient flowing through a tube. The buffer components have been chosen such that their UV absorbance is negligible above 250nm, and they have been formulated such that the ph change in the gradient is linear with time. The buffers used for the validation study shown Table 1. Some physicochemical properties used in ADME studies. in Figure 3 contained various concentrations of boric acid, tris (hydroxymethyl) methylamine, butylamine, citric acid, and potassium dihydrogen phosphate. Samples are injected at constant speed into the flowing ph gradient, and the solution passed through a flow cell in a diode-array UV spectrophotometer, as depicted in Figure 2. In a typical assay using the Sirius ProfilerSGA, the user will create a spreadsheet listing the well position of each compound in the run, along with its ID and the expected number of pk a values, if known. This number may be obtained by inspecting the structure manually, or by using a predictive software package. The spreadsheet is used by the instrument control software to set up the assay. The instrument takes approximately seven hours to run a tray containing 96 samples. Typically, 20µL of sample is diluted to 2mL, and the diluted sample is injected into an up gradient (ph 2 to 12) and a down gradient (ph 12 to 2). Samples can be diluted automatically by the instrument, or diluted samples can be prepared off line. After the assays have been run, the data are analyzed by the refinement software. Spectra from up and down titrations should be identical for samples that remain in solution at all phs. For poorly soluble samples, spectra must be used only from titrations in which samples started in solution. Results for poorly soluble basic samples should therefore be determined from titrations starting at low ph, at which they will be ionized; acidic samples should be determined from titrations starting at high ph. A comparison between the methods described above is given in Table 2.

3 Figure 3. Results of a study of 75 drugs with one, two or three pk a s (112 in total) comparing pk a values measured by the ph gradient technique with measured pk a values reported in the literature. start of gradient generation. The standard gradient is run between ph2 and ph12 in 120 seconds. Figure 3 compares 112 pk a values measured using this instrument with measured values given in the literature, with an R 2 of Results for standards and monoprotic samples (i.e. those with only one pk a ) with large absorbance changes were calculated by an automated first derivative method (AFD). Results for samples with small absorbance changes or overlapping pk a s were calculated by target factor analysis (TFA). Examples of these calculation types are explained below for nitrofurantoin and labetalol. Figure 4 shows the ionization of nitrofurantoin, an acid for which the ionizable group is close to a chromophore. The disposition of the electrons in the chromophore changes when the molecule ionizes (i.e. gains or loses an H + ); as a result, the UV absorbances of the ionized and neutral species are different. As CALCULATING PK a RESULTS FROM GRADIENT UV DATA For a pk a to be measured successfully by this instrument, the UV absorbance of the sample must change as a function of ionization. The pk a values of the samples are calculated from the change in multiwavelength UV absorbance above 250nm as a function of ph. The method is fast because ph is not measured (a slow process), but inferred from the time elapsed from the Figure4. Structure of nitrofurantoin showing ionized and neutral species and identifying the ionizing group. Figure 5. The graph in 5a shows UV absorbance of nitrofurantoin at 30 wavelengths between 250 and 400 nm. The graph in 5b shows the first derivative (FD) of the absorbance value. The pk a of nitrofurantoin is equivalent to the mean position of the peak maxima and minima on the ph axis.

4 Figure 6. Structure of labetalol showing ionized and neutral species and identifying the ionizing groups. shown in Figure 5, the pk a of nitrofurantoin is easily calculated by differentiating the absorbance with respect to ph, and is equivalent to the position of the peak maxima and minima on the ph scale. If the ionizable group is a long way from a chromophore, the change in UV absorbance as a function of ph is likely to be small. Nonetheless, sufficient absorbance change to calculate a pk a can often be observed in multiwavelength UV data. Figure 6 shows the structure of labetalol, which has two ionizable groups. The acidic phenol is attached directly to a benzene ring, which is strong chromophore. However the basic nitrogen is three bond-lengths away from the nearest region of conjugation. It might be expected that the UV absorbance of the basic group will not be affected by ionization. In fact a small absorbance change with ionization occurs, perhaps because the molecule s flexibility allows the nitrogen to approach the benzene ring and thus slightly influence the disposition of electrons in the ring. The 3-D spectrum of labetalol is shown in Figure 7a. It may be deconvoluted by TFA into a graph of molar absorptivity as a function of wavelength (7b), and a graph of % species vs. ph (7c); the pk a s correspond to the intercepts on the distribution of species graph. Any absorbance which cannot be allocated to these two graphs is shown in a 3-D graph of residual absorbance (7d). The mathematical basis of TFA for determining pk a values has been described, 4 and is summarized in the caption for Figure 7. Data for a 96-well tray is processed automatically by both AFD and TFA in under 30 minutes. Afterwards, results for individual samples can be inspected and recalculated manually, if required. Reports can be created automatically in HTML or MS Word format. Conclusion Finding the right equilibrium between the physicochemical properties and the biological activity of a drug candidate is a dif- Figure 7. Illustration of pk a determination by target factor analysis (TFA). a. 3-D graph of raw absorption data. In the TFA calculation, the raw data is expressed as a matrix A of measured absorbance values as a function of wavelength and ph. The software constructs two other matrices; normalized concentration of species as a function of ph and ionized form (the C matrix), and absorptivity as a function of ionized form and wavelength (the E matrix). Initially these matrices are seeded with estimated values for concentration and absorptivity. By invoking Beer s Law, the matrices are related by the equation A = C x E. Different values are proposed for concentration and absorptivity in the C and E matrices until A - (C x E) reaches a minimum value. Since the concentration of species is derived from the pk a value/s, the pk a s can be determined from the minimized solution. b, c. Raw data is resolved into two 2-D graphs of molar absorbance vs. wavelength (equivalent to the E matrix), and distribution of species vs. ph (equivalent to the C matrix). In the distribution of species graph, the points represent the theoretical distribution of species corresponding to the measured pk a values, while the solid lines correspond to the experimental data. The better the solution found by the TFA calculation, the closer the points and curves coincide. d. Residual absorbance, which cannot be assigned in graphs b and c. Note that the absorbance scale has been exaggerated to illustrate the concept.

5 ficult, challenging and expensive process. The Sirius ProfilerSGA can play a valuable part in this process. The Sirius ProfilerSGA is the solution for rapid pka measurement: measure pk a rapidly multi-wavelength spectrophotometric technique open access option automated data handling and processing full computer control micro-plate format measures below 10-5 M concentration 300 assays in 24 hours Sirius ProfilerSGA uses multi-wavelength UV spectroscopy and a linear 2-12 ph unit gradient for truly rapid pka determination. Taking samples direct from stock solutions and working on a microtitre tray format, ProfilerSGA automatically prepares the samples by diluting to the required concentration. Each solution is then injected directly into a ph gradient stream, mixed and passed through the quartz flow cell, allowing pk a s to be measured rapidly. References 1. Kerns, E.H. high-throughput physicochemical profiling for drug discovery. J. Pharm. Sci. 90(11), 1838, Comer, J.; Tam, K.; Lipophilicity Profiles: Theory and Measurement, in Pharmacokinetic Optimization in Drug Research: Biological, Physicochemical and Computational Strategies, E: Testa, B.; van de Waterbeemd, H.; Folkers, G.; Guy, R.; VHCA: Zurich, 275, Gobry, V., Bouchard, G., Carrupt, P.A., Testa, B., Girault, H. H. Physicochemical characterization of sildenafil: Ionization, lipophilicity behavior, and ionic-partition dia gram studied by two-phase titration and electrochemistry. Helv. Chim. Acta, 83(7), 1465, Allen, R.I., Box, K.J,. Comer, J.E.A., Peake, C., Tam, K.Y. Multiwavelength Spectrophotometric Determination of Acid Dissociation Constants of Ionizable Drugs. J. Pharm. Biomed. Anal. 17, 699, Kibbey, C. E., Poole, S.K., Robinson, B., Jackson, J D., Durham, D. An integrated process for measuring the physicochemical properties of drug candidates in a preclini cal discovery environment. J. Pharm. Sci. 90(8), 1164, Box, K., Bevan, C., Comer, J. Hill, A., Allen, R., Reynolds, D. high-throughput Measurement of pka values in a mixedbuffer linear ph gradient system. Anal. Chem. 2003, in press. 7. Hill, A., Bevan, C., Reynolds, D. Analytical method and apparatus therefor. 1999, International Patent Application No. WO99/ Avdeef, A. Assessment of Distributon-pH Profiles, in Lipophilicity in Drug Action, E: Pliska, V. Testa, B. Van de Waterbeemd, H. VCH (Weinhem), 109, Valkó, K. Measurements of physical properties for drug design in industry, in Handbook of Analytical Separations, Volume 1. Separation Methods in Drug Synthesis and Purification, E: Smith, R.M. Elsevier (Amsterdam), 539, Mirrlees, M.S., Moulton, S.J., Murphy, C.T., Taylor, P.J. Direct measurement of octanol-water partition coefficients by high-pressure liquid chromatography. J. Med. Chem. 19, 615, Lombardo, F., Shalaeva, M.Y., Tupper, K A., Gao, F., Abraham, M.H. ElogP(oct): A tool for lipophilicity determination in drug discovery. J. Med. Chem. 43(15), 2922, Avdeef, A. High-Throughput Measurements of Solubility Profiles, in Pharmacokinetic Optimization in Drug Research. Biological, Physicochemical, and Computational Strategies, E: Testa, B., van de Waterbeemd, H., Guy, F.R., VHCA: Zurich, 304, Avdeef, A. ph-metric solubility. 1. Solubility-pH profiles from Bjerrum plots. Gibbs buffer and pk a in the solid state. Pharm. Pharmacol. Commun. 4, 165, Bevan, C.D., Lloyd, R.S. A high-throughput screening method for the determination of aqueous drug solubility using laser nephelometry in microtiter plates. Anal. Chem. 72(8), 1781, Lipinski, C.A., Lombardo, F., Dominy, B.W., Feeney, P.J. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Advanced Drug Delivery Reviews, 46(1-3), 3, Kansy, M., Senner, F., Gubernator, K. Physicochemical highthroughput screening: Parallel artificial membrane permeation assay in the description of passive absorption processes. J Med Chem. 41(7), 1007, 1998.