High throughput determination log Po/w/pKa/log Do/w of drugs by combination of UHPLC and CE methods

In 1997 Valkó et al. developed a generic fast gradient HPLC method, based on the calculation of the Chromatographic Hydrophobicity Index (CHI) from the gradient retention times, in order to measure lipophilicity. We have employed the correlations between CHI and log Po/w and adapted the rapid gradient HPLC method to UHPLC obtaining excellent resolution and repeatability in a short analysis time (< 4min). log Po/w values can be easily obtained from these CHI measurements but, unfortunately, these correlations are only valid for non-ionized compounds. Consequently, in order to determine the effective log Po/w value at a particular pH, a fast high-throughput method for pKa determination was required. The IS-CE method, based on the use of internal standards (IS) and capillary electrophoresis (CE), is a fast and attractive alternative to other methods for pKa determination, since it offers multiple advantages compared to them: low amounts of test compounds and reagents are needed, high purity is not required, specific interactions between test compounds and buffers are corrected, etc. In addition, it allows the determination of a pKa value in less than 5 minutes. Both CHI and IS-CE have been combined in order to describe a high throughput alternative in the determination of the lipophilicity profiles of bioactive compounds.


Introduction
The drug discovery and development process requires the high-throughput determination of physicochemical parameters of drug candidates.Parameters measuring lipophilicity, acidity, solubility or protein interaction must be determined in a fast way for a high number of compounds of potential interest to select the most promising compounds for further development.
One of the most important physical properties affecting the biological activity of substances is lipophilicity, traditionally expressed as the logarithm of the octanol-water partition coefficient (log P o/w ).In the common case of drug candidates with acid-base properties, the effective partition rate at a particular pH is noted as log D o/w .Although being these the most widely used liphophilicity indexes, experimental reference procedures are usually time consuming and require a high purity and a relatively high amount of sample.In order to decrease the analysis time and overcome these limitations, instrumental analytical methods have been developed to determine lipophilicity, and the values therefrom obtained in a particular lipophilicity scale have been correlated to log P o/w or log D o/w scale.About twenty year ago Klára Valkó and Péter Slégel [1] proposed a new hydrophobicity index measured by a reversed-phase HPLC method, named ϕ 0 , defined as the concentration of organic modifier in the mobile phase (methanol or acetonitrile) which is required to obtain a retention time (t R ) double than the dead time (t 0 ).In these conditions, the logarithm of the retention factor equals to zero (log k'=log ((t R -t 0 )/t 0 )=0) and, therefore, the compound is equally distributed between the mobile and the stationary phases.The higher the value, the more hydrophobic is the compound.This index is characteristic of a compound and independent of the column dimensions, the mobile phase composition and flow-rate.It only depends on the stationary phase, the particular organic modifier employed, the temperature and, for acidic or basic compounds, the pH.However, ϕ 0 values were experimentally determined in isocratic mode using several mobile phases containing low fractions of methanol or acetonitrile, which involved long run times for hydrophobic substances.A few years later, Valkó and co-workers [2] proposed a new highthroughput Chromatographic Hydrophobicity Index (CHI), based on ϕ 0 but derived from retention times observed in a fast gradient reversed-phase HPLC method.Compounds of previously determined CHI values can be used as calibration set, and the retention times of new compounds can be correlated to their corresponding CHI values.In that work the C 18 column employed was an ODS2-IK5 Inertsil with the dimensions of 150 × 4.6 mm, and the gradient program lasted for about 15 min.In a paper published in 2001 in collaboration with Michael H. Abraham [3] a much shorter column was used, a 50 x 4.6 mm Phenomenex Luna C-18(2), and therefore gradient times were reduced to 5 minutes.In addition, correlations between CHI and log P o/w were improved by means of hydrogen-bond acidity descriptors, since it was found that the major difference between these scales is their sensitivity towards the hydrogen bond acidity of the compounds.However, in the case of compounds with acid/base properties, poor correlations between CHI and log D o/w have been obtained so far.In 2002 Stephen F. Donovan and Mark C. Pescatore [4] used a 20 x 4.2 mm octadecyl-poly(vinyl alcohol cartridge and a fast gradient methanolic elution in order to estimate log P o/w from chromatographic data.Due to the pH stability of this polymeric column, the pH of the mobile phase could be adjusted to the desired value to ensure the ionizable analyte to be in its neutral form.Nowadays, UHPLC technology presents also an excellent stability of C 18 columns over a wide range pH, besides higher resolution and sensitivity, and shorter equilibration times within consecutive runs in gradient elution.
Since mostly of the potential drug candidates are acids or bases, log D o/w is a fundamental parameter to be determined in the drug discovery process in order to estimate the pharmacokinetics of a compound of interest.Especially relevant is the log D o/w at pH 7.4, the physiological value, which indicates the lipophilicity of a drug in the blood plasma.As long as this effective partition rate at any particular pH can be estimated from the measured log P o/w for the neutral substance and the pK a of the compound, we propose a highthroughput methodology based on the combination of UHPLC for CHI measurements and CE for pK a determination in order to define the log D o/w profile of drug candidates.
In a series of previous works [5][6][7], we have proposed a high-throughput method to determine acidity constants of weak acids and bases by capillary zone electrophoresis.This method is based on the use of an internal standard (IS) of similar nature and acidity constant as the test compound, which is injected into the capillary just after the sample compound and analyzed in the same electrophoretic run.This method requires only two electrophoretic runs: a first one at a pH where both the test compound and the IS are totally ionized, and a second one at another pH at which both are partially ionized.Thus, from the mobilities of the compounds at these two pH values and the pK a value of the IS, the pK a of the analyte can be easily calculated [5,8].Besides its rapidness, the main advantages of this method are the compensation of systematic errors due to simultaneous analysis of analyte and IS, and the unnecessity of an accurate pH measurement.

Apparatus
For UHPLC measurements, a Shimadzu (Kyoto, Japan) Nexera UHPLC system was used.The system was equipped with two LC-30AD high-pressure pumps, a DGU-20A 5 online degasser, a CTO-10ASvp oven thermostatized at 25 o C, a SIL-30AC autosampler, a SPD-M20A diode array detector and a CBM-20Alite controller.Retention data were obtained from a Waters (Milford, MA, USA) Acquity BEH C 18 column, 50 mm × 2.1 mm, which allowed the study of basic drugs because of their extended working pH range (up to 12).Injected samples were prepared in DMSO at a concentration of 0.5 mg/mL and thus, due to the melting point of DMSO (19 o C), the autosampler temperature was set at 20 o C. Aqueous buffers were, in all cases, 50 mM ammonium acetate at the desired pH value.Acidic buffers were prepared from glacial acetic acid and the pH was adjusted with small volumes of concentrated ammonia (25%), and basic buffers were prepared inversely.Medium acidic and basic buffers were obtained by solving the salt and adjusting the pH with concentrated ammonia or glacial acetic acid.
Capillary electrophoresis experiments were performed using a P/ACE MDQ Beckman instrument (Palo Alto, CA, USA), equipped with a diode array detector.Capillary was made of fused silica 50 µm I.D., 375 µm O.D., 35.2 cm length (25 cm to the detector) and purchased from Composite Metal Services Ltd (Shipley, West Yorkshire, UK).The temperature of the capillary was 25.0±0.1 o C and the applied voltage during separation was 20 kV.IS and analyte samples were sequentially injected at 0.5 psi for 3 s and analyzed in the same run.Additional 0.5 psi of hydrodynamic flow during separation was applied.The capillary was initially conditioned with 1 M NaOH (2.0 min), water (0.5 min), and running buffer (2.0 min).Between replicates with the same buffer the capillary was not rinsed.20 psi of pressure was applied during rinsing processes.Running buffers were prepared at different pH values and 50 mM ionic strength as described elsewhere [8].Injected samples were prepared in methanol/water mixtures at a concentration of 0.1 mg/mL, containing DMSO as neutral marker.Appropriate IS were selected from the set proposed in previous works [8].All samples and buffers were filtered through a nylon mesh 0.45 µm porous size (Whatman, Maidstone, UK) before use.pH measurements were performed with a combined Crison (Alella, Spain) 5014 electrode in a Crison GLP22 pH meter.The electrode system was standardized with ordinary aqueous buffers of pH 4.01, 7.00 and 9.21.

Measurement of CHI MeCN values by UHPLC
The HPLC fast gradient method developed by Valkó and co-workers [3] was transferred to our UHPLC system taking into consideration the column and particle sizes, the flow rate, the injection volume, and the dwell time of the UHPLC instrument.The final UHPLC conditions, together with the original HPLC ones, are shown in Table 1.DMSO was used as sample solvent because its ability to solve hydrophobic compounds sparingly soluble in MeCN/water mixtures.In the present work MeCN was used as organic modifier as long as this solvent allows better correlations between log P o/w and CHI than methanol.
Several substances covering a wide range of CHI MeCN and log P o/w values were tested as calibration standards.Finally, the representative compounds shown in Table 2 were selected.They are mainly unionizable substances that can be used at any pH value (acetanilide and phenones), but 4-hydroxybenzyl alcohol is a phenol that should be used in its neutral form below pH 8. When using aqueous buffers above this pH value, caffeine was used instead of 4-hydroxybenzyl alcohol.A representative chromatogram of the calibration set is shown in Figure 1.After plotting the CHI MeCN values against the retention times in the fast gradient elution, a quadratic calibration curve is obtained with a typical coefficient of determination (r 2 ) higher than 0.998 and a root mean square error (RMSE) lower than 1.5.0.0 min -> 0% 0.5 min -> 0% 0.4 min -> 0% 3.0 min -> 100% 2.5 min -> 100% 3.5 min -> 100% 2.9 min -> 100% 3.7 min -> 0% 3.1 min -> 0% 4.5 min -> 0% 3.8 min -> 0% Injection volume and solvent 3 μL -MeCN/aqueous buffer 0.2 μL -DMSO Similarly to log D o/w scale, CHI values of acidic or basic compounds depend on the ionization degree.The lower the ionization, the higher the CHI value [9,10].Therefore, acidic substances present the highest CHI when aqueous buffers of low pH are used, and consequently basic compounds exhibit the reversed trend.On the other hand, non-ionizable analytes show nearly the same CHI values independently of the pH.In the present work all studied analytes were injected at three different pH values (3.0, 7.4 and 11.0), and a representative collection supporting the above statements is presented in Table 3.
With the aim of validating the UHPLC method proposed, CHI MeCN were measured for several nonionizable, acidic and basic compounds in the log P o/w range between -0.07 and 4.45.All analytes were injected using aqueous mobile phases of pH 3.0, 7.4, and 11.0, and the highest value was considered for log P o/w determination.These 41 compounds are listed in Table 4, together with the CHI MeCN values of neutral species determined in the present work, the A solute descriptors obtained from the ACD/Percepta platform [15], and the log P o/w determined from CHI MeCN and A according to Eq. ( 1).There is a good correlation between literature log P o/w values (Mlog P values compiled in Bio-Loom database [12]) and the ones measured in the present work (slope: 0.99(±0.06),intercept: -0.15(±0.15),r: 0.940, RMSE: 0.45).Basic solutes correlate slightly worse than non-ionizable and acidic compounds.Nevertheless, from the 16 studied bases only 6 presented deviations from literature values higher than 0.7 log P o/w units.In four of those cases the determined log P o/w values were lower than the bibliographic ones (compounds 14, 16, 17, and 19 in Table 4), and in two cases (solutes 21 and 29) the reverse trend was shown.
As previously commented, poor correlations are obtained between log D o/w and CHI MeCN values for ionized compounds.However, the effective distribution rate of a drug at a particular pH value can be estimated from its acidity constant and the partition coefficients of the neutral and fully ionized species [16].The current state of the art does not allow the direct measurement of the partition coefficient of ions but it can be assumed that a typical value is 3.15 units lower than that of the neutral species, according to the work published by Donovan and Pescatore [4] Consequently, the measurement of aqueous acidity constants was required in order to estimate log D o/w values and the IS-CE fast method was presented as a very appropriate option.This method allows an accurate determination of acidity constants, which normally present an excellent match with reference values found in literature.In the present work, the correlation between the measured pK a values for some of the studied acidic and basic compounds (Table 4) and the ones found in literature was remarkable (slope: 1.01(±0.01),intercept: -0.07(±0.09),r: 0.998, RMSE: 0.11).
Table 4 shows the log D o/w values at pH 7.4 resulting from log P o/w determinations and pK a measurements (Eqs.(2-3)) and those obtained from reference methods [13,14].As shown in Figure 2, there is a good agreement between both sets of values, with the only exception of lidocaine, which presented yet one of the worst matches in log P o/w determination.4) and those obtained from log P o/w and pK a measurements (Eqs.(2)(3)).Linear regression (continuous lines) and the statistics are also shown, together with the lines corresponding to ±2•RMSE (dashed lines).
Figure 3 shows the log D o/w profile obtained in the present work for vanillin and clonidine, which present acidic and basic nature, respectively.For the sake of comparison, the curve calculated by ACD/Labs software is also presented, together with their corresponding calculated and measured pK a values.In the case of vanillin, the ACD/Labs calculated partition ratio (1.32) is moderately higher than the measured one (0.94), whereas both pK a values, estimated and experimental, match very well (7.30 and 7.36, respectively).Thus, the ACD/Labs estimated curve shifts up from the experimental one.For clonidine, the ACD/Labs calculated log P o/w (2.05) is higher than the CHI MeCN measured one (1.44) by 0.6 units, whereas the pK a (7.90) is lower than the IS-CE determined one (8.10).Therefore, the ACD/Labs estimated curve moves up and left from the measured one, being the values determined in the present work more accurate according to literature (8.05 and 1.43, for pK a and log P o/w , respectively [12]).

Conclusions
In the present work the combination of two high-throughput methods have been proposed in order to estimate the log D o/w of an ionizable compound at any desired pH.On the one hand, a fast gradient reversed-phase methodology was successfully transferred into UHPLC in order to determine log P o/w from chromatographic retention times, based on the formerly developed CHI lipophilicity index.After calibration of the system, a single fast-gradient run performed within 4 minutes is enough to determine the CHI MeCN value of a compound, and then the value can be transformed into log P o/w lipophilicity scale by means of the established equation.In the case of analytes with acid/base properties, runs performed at 3 different pH values (3.0, 7.4 and 11) are recommended, and the highest CHI MeCN is selected.On the other hand, the IS-CE allows the determination of a pK a value from only two runs, lasting less than 5 minutes (about 2 min each).Finally, from both the log P o/w of the neutral species (UHPLC-CHI procedure) and the pK a value (IS-CE method), the effective distribution rate log D o/w can be estimated at any particular pH value.

Figure 2 .
Figure 2. Correlation between reference log D o/w values at pH 7.4 (Table4) and those obtained from log P o/w and pK a measurements (Eqs.(2)(3)).Linear regression (continuous lines) and the statistics are also shown, together with the lines corresponding to ±2•RMSE (dashed lines).

Figure 3 .
Figure 3. log D o/w profile for vanillin and clonidine according to Eqs. (2-3) (continuous line) and calculated profile using ACD/Labs software (dotted lines), together with their corresponding measured and calculated pK a values.

Table 1 .
Comparison between HPLC and UHPLC conditions to measure CHI values
b)Used above pH 8.Figure 1. Chromatogram of the calibration set after subtraction of the DMSO blank.UHPLC conditions according toTable 1, 50 mM ammonium acetate pH 7.4 as aqueous buffer.

0 pH 7.4 pH 11.0
As mentioned in the introduction, Valkó and co-workers proposed an equation to correlate CHI MeCN and log P o/w indexes, which involved the Abraham's solute hydrogen-bond acidity descriptor (A, also noted in the original paper as Σα 2 [3]3]: 

Table 4 .
Validation set used for the determination of log P o/w and log D o/w .
. Thus, log D o/w values of monoprotic compounds can be calculated from the following expressions at any desired pH value: